National Academies Press: OpenBook
« Previous: IV. Uncertainties: Integration with Risk Assessment and Resources
Suggested Citation:"V. Poster Session." 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 251
Suggested Citation:"V. Poster Session." 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 252
Suggested Citation:"V. Poster Session." 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 253
Suggested Citation:"V. Poster Session." 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 254
Suggested Citation:"V. Poster Session." 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 255
Suggested Citation:"V. Poster Session." 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 256
Suggested Citation:"V. Poster Session." 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 257
Suggested Citation:"V. Poster Session." 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 258
Suggested Citation:"V. Poster Session." 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 259
Suggested Citation:"V. Poster Session." 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 260
Suggested Citation:"V. Poster Session." 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 261
Suggested Citation:"V. Poster Session." 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 262
Suggested Citation:"V. Poster Session." 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 263
Suggested Citation:"V. Poster Session." 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 264
Suggested Citation:"V. Poster Session." 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 265
Suggested Citation:"V. Poster Session." 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 266
Suggested Citation:"V. Poster Session." 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 267
Suggested Citation:"V. Poster Session." 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 268
Suggested Citation:"V. Poster Session." 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 269
Suggested Citation:"V. Poster Session." 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 270
Suggested Citation:"V. Poster Session." 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 271
Suggested Citation:"V. Poster Session." 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 272
Suggested Citation:"V. Poster Session." 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 273
Suggested Citation:"V. Poster Session." 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 274
Suggested Citation:"V. Poster Session." 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 275
Suggested Citation:"V. Poster Session." 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 276
Suggested Citation:"V. Poster Session." 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 277
Suggested Citation:"V. Poster Session." 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 278
Suggested Citation:"V. Poster Session." 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 279
Suggested Citation:"V. Poster Session." 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 280
Suggested Citation:"V. Poster Session." 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 281
Suggested Citation:"V. Poster Session." 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 282
Suggested Citation:"V. Poster Session." 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 283
Suggested Citation:"V. Poster Session." 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 284
Suggested Citation:"V. Poster Session." 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 285
Suggested Citation:"V. Poster Session." 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 286
Suggested Citation:"V. Poster Session." 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 287
Suggested Citation:"V. Poster Session." 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 288
Suggested Citation:"V. Poster Session." 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 289
Suggested Citation:"V. Poster Session." 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 290
Suggested Citation:"V. Poster Session." 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 291
Suggested Citation:"V. Poster Session." 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 292
Suggested Citation:"V. Poster Session." 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 293
Suggested Citation:"V. Poster Session." 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 294
Suggested Citation:"V. Poster Session." 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 295
Suggested Citation:"V. Poster Session." 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 296
Suggested Citation:"V. Poster Session." 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 297
Suggested Citation:"V. Poster Session." 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 298
Suggested Citation:"V. Poster Session." 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 299
Suggested Citation:"V. Poster Session." 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 300
Suggested Citation:"V. Poster Session." 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 301
Suggested Citation:"V. Poster Session." 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 302
Suggested Citation:"V. Poster Session." 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 303
Suggested Citation:"V. Poster Session." 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 304
Suggested Citation:"V. Poster Session." 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 305
Suggested Citation:"V. Poster Session." 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 306
Suggested Citation:"V. Poster Session." 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 307
Suggested Citation:"V. Poster Session." 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 308
Suggested Citation:"V. Poster Session." 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 309
Suggested Citation:"V. Poster Session." 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 310
Suggested Citation:"V. Poster Session." 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 311
Suggested Citation:"V. Poster Session." 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 312
Suggested Citation:"V. Poster Session." 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 313
Suggested Citation:"V. Poster Session." 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 314
Suggested Citation:"V. Poster Session." 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 315
Suggested Citation:"V. Poster Session." 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 316
Suggested Citation:"V. Poster Session." 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 317
Suggested Citation:"V. Poster Session." 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 318
Suggested Citation:"V. Poster Session." 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 319
Suggested Citation:"V. Poster Session." 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 320
Suggested Citation:"V. Poster Session." 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 321
Suggested Citation:"V. Poster Session." 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 322
Suggested Citation:"V. Poster Session." 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 323
Suggested Citation:"V. Poster Session." 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 324
Suggested Citation:"V. Poster Session." 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 325
Suggested Citation:"V. Poster Session." 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 326
Suggested Citation:"V. Poster Session." 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 327
Suggested Citation:"V. Poster Session." 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 328
Suggested Citation:"V. Poster Session." 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 329
Suggested Citation:"V. Poster Session." 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 330
Suggested Citation:"V. Poster Session." 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 331
Suggested Citation:"V. Poster Session." 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 332
Suggested Citation:"V. Poster Session." 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 333
Suggested Citation:"V. Poster Session." 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 334
Suggested Citation:"V. Poster Session." 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 335
Suggested Citation:"V. Poster Session." 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 336
Suggested Citation:"V. Poster Session." 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 337
Suggested Citation:"V. Poster Session." 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 338
Suggested Citation:"V. Poster Session." 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 339
Suggested Citation:"V. Poster Session." 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 340
Suggested Citation:"V. Poster Session." 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 341
Suggested Citation:"V. Poster Session." 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 342
Suggested Citation:"V. Poster Session." 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 343
Suggested Citation:"V. Poster Session." 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 344
Suggested Citation:"V. Poster Session." 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 345
Suggested Citation:"V. Poster Session." 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 346
Suggested Citation:"V. Poster Session." 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 347
Suggested Citation:"V. Poster Session." 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 348
Suggested Citation:"V. Poster Session." 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 349
Suggested Citation:"V. Poster Session." 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 350

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 V Poster Session

Introduction Robert L. Dedrick During organization of the workshop on which this volume is based, it became clear that additional relevant work and points of view existed that could not be well represented in a limited number of primarily didactic presentations. To enrich the program, a poster session was added in which the presenters were also asked to introduce themselves in the plenary session, give 1- or 2-minute summaries of their work, and then invite general discussion of their topic by workshop participants. 253

Route-to-Route Extrapolation of DichIoromethane Exposure Using a Physiological Pharmacokinetic Mode! Michael ]. Angelo and Alan B. Pritchard BACKGROU N D Physiologically based mathematical models are useful devices for ex- ploring the information contained in pharmacokinetic data. Most of the physiological pharmacokinetic models that have appeared in the literature have been developed for pharmacological research. A number of these applications have been reviewed by Gerlowski and Jain (1983), who also present some of the fundamental principles that are used to derive this type of mechanistic model. There has been increasing interest within the field of toxicology in the use of models to process pharmacokinetic information, especially with regard to exposure assessment. Their value has been recognized for a number of reasons, including their ability to increase a scientist's under- standing of physiological factors that control chemical disposition, quan- tification of in vivo metabolism rates, and the extrapolation of pha~m- acokinetic predictions across dose levels and mammalian species. This paper illustrates how we have used a previously developed phys- iological model for the disposition of dichloromethane (DCM; methylene chloride) to compare the pharmacokinetic patterns that result from inhal- ation and oral administration of DCM in rats. With information generated by computer simulations, we determined the correlations between inhal- ation doses and the corresponding oral doses of DCM that produce the same level of pharmacokinetic impact. By examining the pharmacokinetic information in this way, we were able to infer how dosing route depen- dencies can influence the calculation and interpretation of DCM exposure. 254

Route Extrapolation Using a PK Model 255 RESULTS AND DISCUSSION The schematic diagram in Figure 1 shows the structure of the physio- logically based model that was used to describe the pharmacokinetics of DCM following inhalation and oral administrations. The mathematical details of the model, including parameter values, are given in the Appendix to this paper. Table 1 contains the physiological constants for a 250-g rat that are available in the literature, and Table 2 contains the pharmaco- kinetic parameters that were determined by using numerical optimization techniques. The kinetics of DCM biotransformation via pathways depen- dent on the mixed-function oxidase (MFO) system and glutathione (GSH) has been presented recently by Gargas et al. (1986), and it has been incorporated into the mass balance equations. The model was verified for oral administrations by using data that were collected at the Huntingdon Research Centre in England (Angelo et al., 19861. As an example of this, Figure 2 shows the actual blood concen- trations of DCM and model simulations following single oral doses of 50 and 200 mg/kg to rats on days 1 and 14 of a daily gavage dosing regimen. In developing the correlation between inhalation and oral doses, we first determined that a continuous oral infusion of DCM into the gut lumen DCM ~ ~ ALVE~R ~ =2 [~ ~~Q CO CO, ~ ~ . ~ . ·~_ · LIVER ~ GUT ~L: 1. ~14 1 ~ -| CARCASS | ~ FIGURE 1 Physiological pharmacokinetic model for dichloromethane (DCM).

256 MICHAEL J. ANGELO AND AWN B. PRITCHARD TABLE 1 Physiological Parameters for a 250-g Rat _ _ _ Parameter Abbreviation Value Compartment flow rates (ml/min) Lunga QL 32.3 Liver I 16.0 Kidney QK 10.0 Gut QG 9 1 Carcassb Qc 6.3 Compartment volumes (ml) Whole animal (g) BW 250 Blood VB 22.5 Lung V'g 2.9 Liver Via 10.0 Kidney VK 2.3 Gut VG 10.1 Gut lumen VG! 13.6 CarcassC Vc 183 Alveolar space VE 6.25 Alveolar ventilation (ml air/min) I 90 Blood/air distribution ratio A 12.8 Red blood cell/plasma distribution ratio Rrbc 5.4 aSum of venous flows from other compartments. Volumetric average of muscle, skin, and fat. CSum of muscle, skin, and fat. compartment of the model would produce pharmacokinetic profiles similar to those observed during an inhalation exposure. This was done to dem- onstrate that a constant oral input mimicked the distribution patterns that were achieved during a continuous input to the pulmonary system. There- fore, the correlations between the two methods of administration depended upon the dosing route and not upon the disposition factors that were influenced by the rate of administration. Using the data of McKenna and coworkers (1982), Figure 3 shows that the model simulations of DCM blood concentration during a 6-h oral infusion agree well with the inhal- ation data. Inhalation simulations at 50, 500, and 1,500 ppm for 6 h produced curves that were virtually identical to those obtained with con- tinuous oral input, indicating that steady infusions of DCM into the lung and gut compartments produced the same blood distribution profiles. To obtain the inhalation-to-oral dose correlations, we first evaluated delivered doses to target tissues as areas under concentration-time curves (AUCs) for a series of inhalation exposures. Metabolized doses were also

Route Extrapolation Using a PK Model 257 TABLE 2 Pharmacokinetic Parameters for a 250-g Rat _ . Parameter Abbreviation Value Tissue/plasma distribution ratio Blood RB 2.8 Lung REg 1 Liver R! 1 Kidney RK 1 Gut RG 4 5 Carcass RC 2.4 Gut contents/water distribution ratio Metabolism parameters Vm,~,` (nmol/ml/min) Km (nmol/ml) k2 (min ~ ) A a 1 A2a fco Alveolar permeability x area product (PA; ml/min) Carbon monoxide clearance (CLCo; ml/min) Binding constants for 14Co to Hb kot (nmol/ml) kp (nmol/ml) Gut absorption (ka; min - i) RGI~ 40 535 0.016 0.120 0.056 0.60 0.26 600 40 0.029 aReitz et al. (1986). E 5000- 0' E 1000 l c . 0- . 100- . z z 8 10- o o m c, \ \200 mg/kg 50 mg/kg \ DAY 1 ~ \ I _. ¢\ DAY 14 \ \ ~ 50 mg/kg\ ~ \ 200 mglkg \ ~ ~ 60 120 180 240 300 360 ~ __ 0 60 120 180 240 300 360 0 TIME (min) FIGURE 2 Blood concentrations of dichloromethane (DCM) on days 1 and 14 of a repeated oral dosing schedule. Daily gavage doses were administered to rats at SO (~) and 200 mg/kg (~) in a water vehicle. Data represent the mean + standard error of the mean for six animals; lines are the predictions from the pharmacokinetic model.

258 MICHAEL].ANGELO AND ALAN B.PRITCHARD 10,000 E - E 1,000_ - z in al in o o 0 1- m o 100- tO- 0.1 - - _ 1 ~ 50ppm _ 500 ppm ~1500ppm ~1 :4 0 60 120 180 240 300 360 TIME(min) FIGURE 3 Physiological model predictions of dichloromethane (DCM) in blood of rats during 6-h oral infusions. Simulations are compared to data from 6-h inhalation exposures of DCM at 50, 500, and 1,500 ppm. SOURCE: McKenna et al. (1982). determined in terms of the amount of DCM that was metabolized by each of two biotransformation pathways. A numerical search was then used to find the oral doses that produced the equivalent delivered or metabolized doses to those obtained in the inhalation simulations. Figure 4 shows the correlations between inhalation exposures and equiv- alent oral doses of DCM in rats for the lung, liver, and blood compart- ments. The correlations indicate that at 1,500 ppm, the equivalent oral dose of DCM is between 200 and 300 mg/kg for each tissue. Figure 5 shows the correlations that were based on the equivalent amounts of metabolized DCM by the two biotransformation pathways. The figure indicates that the correlation for the GSH-dependent pathway is a linear relationship between the oral and inhalation exposures, whereas the MFO pathway exhibited a nonlinear pattern. At 1,500 ppm, the equivalent oral dose was approximately 200 mg/kg for the glutathione pathway and about 100 mg/kg for the MFO pathway, although the latter reached this level between 500 and 750 ppm. The retention of DCM during steady-state inhalation was less than 15% of the exposure concentration in the 50- to 1,500-ppm treatment range. This was determined by using Equation 1 lb from the Appendix. Values for the alveolar concentration of DCM (Ca) that were used in the calcu-

Route Extrapolation Using a PK Model 259 ~ 300- x E 200- ~n o o V 100 - O ~ LUND BLOOD - LIVER ~ I ~ . I 0 250 500 750 ~ 000 1250 1500 DCM INHALATION EXPOSURE (ppm for oh) FIGURE 4 Correlations based on equivalent tissue AUC in rats between oral doses (infusions for 6 h) and 6-h inhalation exposures of dichloromethane (DCM). ~ 300- ~D v, A Cal ~ 200- In o ~ 100- ~r o o 0 250 500 750 1000 1250 1500 DCM INHALAT1014 EXPOSURE (ppm for Oh) FIGURE S Correlations based on equivalent amounts of metabolized dichloromethane (DCM) by MFO and GSH pathways between oral doses (infusions for 6 h) and 6-h inhalation exposures In rats. rations were obtained as a numerical output from the simulations of in- halation exposures. Previous studies have shown that the retention of DCM during steady- state inhalation is less than 100% (Fiserova-Bergerova, 19831. In these situations, calculations that are based upon complete absorption of an

260 MICHAEL3.ANGELO AND ALAN B.PRITCHARD inhaled dose overestimate the amount of material that enters the systemic portion of the body. Similarly, a sizable hepato-pulmonary first-pass effect during oral absorption of DCM reduces the systemic bioavailability of the compound, even though the retention of an oral dose by the gastrointestinal compartment is usually considered to be 100%. Therefore, quantitative relationships that are based upon 100% retention of the inhaled dose and/ or complete absorption of an applied oral dose are misleading if internal challenges to the target tissues or to the metabolic systems are not used as the basis for the correlations. By their very nature, physiologically based models are designed to describe the physiological phenomena that control the disposition of ab- sorbed chemicals. With respect to DCM, this means that the first-pass effects and incomplete pulmonary retention of an inhaled dose are ac- counted for in the mass-balance equations that comprise the model. Be- cause of this, the model establishes a theoretical basis for quantifying internal doses of DCM that should be used in the development of route- to-route pharrnacokinetic comparisons. CONCLUSIONS 1. We have established that a physiologically based pharmacokinetic model can be used to compare organ-specific doses between inhalation and oral administrations of dichloromethane. 2. We demonstrated that internal exposures, as measured by the deliv- ered doses to target tissues and by metabolic challenges, are different for a given administered dose. 3. We determined that the retention of DCM during steady-state in- halation was less than 15% over the concentration range tested. 4. The dose of DCM delivered to target organs is overestimated if total absorption of an inhaled dose is assumed and/or the hepato-pulmonary first-pass effect is ignored. 5. The model provides an effective tool for exposure assessment by quantifying the internal doses of DCM that are the appropriate measures to use in route-to-route pharrnacokinetic comparisons. REFERENCES Angelo, M. J., and A. B. Pritchard. 1984. Simulations of methylene chloride pharmaco- kinetics using a physiologically based model. Reg. Toxicol. Pharmacol. 4:329-339. Angelo, M. J., A. B. Pritchard, D. R. Hawkins, E. R. Walter, and A. Roberts. 1986. The pharmacokinetics of dichloromethane. II. Disposition in Fischer 344 rats following intravenous and oral administration. Food Chem. Toxicol. 24:975-980. Fiserova-Bergerova, V. 1983. Pp. 101-132 in Modeling of Inhalation Exposure to Vapors: Uptake, Distribution and Elimination, Vol. I. Boca Raton, Fla.: CRC Press.

Route Extrapolation Using a PK Model 261 Gargas, M. L., H. J. Clewell III, and M. E. Andersen. 1986. Metabolism of inhaled dihalomethanes ~n vivo: Differentiation of kinetic constants for two independent pathways. Toxicol. Appl. Pharmacol. 82:21 1-223. Gerlowski, L. E., and R. K. Jain. 1983. Physiologically based pharmacokinetic modeling: Pnnciples and applications. J. Pharm. Sci. 72:1103-1127. McKenna, M. J., J. A. Zempel, and W. H. Braun. 1982. The pharmacokinetics of inhaled methylene chlonde in rats. Toxicol. Appl. Pharmacol. 65:1-10. Reitz, R. H., F. A. Smith, M. L. Gargas, H. J. Clewell, and M. E. Andersen. 1986. Physiological modeling and nsk assessment: Example, methylene chlonde. Toxicologist 6:7. APPENDIX The following are the mass-balance equations for each compartment in the physiological pharmacokinetic model for dichloromethane. ·iver Gut Lung VL d = QL (CB,art RL ) QG(RGCG CB,artJ VLL(1—A~)rlL + r2L], where Vmax R L K + CL RL ,~, rlL = ~, and r2L = k2 RL (1) VG = Q G (CB,art — —CG) + D ~ where dt RG (2) D = a exp(— at) RGI RGI V Lg dt = QLg ( CB,ven R CLg )

262 MICHAEL].ANGELO AND ALAN B.PRITCHARD RLgA g) VLg(A ~ r 1 Lg + A2r2Lg), where (3) V, r ~& Rmax rlLg = Cg ~ and Km + g RLg r2Lg = k2 RLg Alveolar Space Carcass Kidney Arterial Blood Venous Blood Va dt = PA(R AC~g Vc dtC = QC(\ ~( / VK d— = QK( CB,art Ca) + Qa(Cinh Ca) (4) R C ) RBC ) VB,art dt = QLg(R CLg — CB art) VB ,,e,7 dt = QL (R CL) + QK + QCt~ (5) (6) (7) / R B C ) \R c c) Q Lg CB,ven (8)

Route Extrapolation Using a PK Model 263 14Co Balance i4CO2 Balance ~ ]t = f.~1 —A~)v,rlr + A~V~grl,g] VB dt —CLCof 14CO]f, where [l4CO]t = [l4CO]f + [Hb14CO], and Hbl4CO~ = ok, ti4CO]f Cumulative 14CO2 expired DCM Balance 1 ~ -— J {( 1 —f) [(1 —Al) VL r1 L + AlVLgrl Lg] (1O) + VL r2L + VLgr2Lg}dt Cumulative 14C-DCM I= expired curing = Jo QaCadt oral exposure Retention of DCM during steady-state inhalation (1 la) 3 ( C,,,h \) (l lb) The following nomenclature was used in the model equations. Compartment symbols Cat Concentration in compartment 1 (nmol/ml) Qua Blood flow rate compartment 1 (ml/min) Rat Tissue/plasma distribution ratio Vat Volume of compartment 1 (ml) Subscripts for compartments: B. blood; art, arterial; yen, venous; L, liver; Lg. lung; G. gut; Gl, gut lumen; C, carcass; K, kidney; a, alveolar space. Other symbols Al Distribution of MFO activity between lung and liver A2 Distribution of GSH activity between lung and liver Cinh Inhaled OCM concentration (nmol/ml of air) CL Pulmonary clearance of }4Co (ml/min) ti4Co]t Total concentration of i4Co in blood (nmol/ml)

264 MICHAEL3.ANGELO AND ALAN B.PRITCHARD Km ka k2 k,~,k' M PA Ha RA rl r2 Imp [l4CO]f Free concentration of 14CO in blood (nmol/ml) f Stochiometric fraction of CO produced from 1 mol of DCM Michaelis constant for reaction rl (nmol/ml) Gastrointestinal (GI) absorption rate constant (min-l) Metabolism rate constant for reaction r2 (min-l) Binding constants for Hbl4CO (nmol/ml) Mass of DCM in oral dose (nmol) Alveolar permeability x area product (ml/min) Alveolar ventilation (ml air/min) Blood/air partition coefficient Rate of DCM metabolism via the MFO pathway Rate of DCM metabolism via the GSH pathway Maximum reaction velocity for rl (nmol/ml/min)

Sensitivity Analysis in Pharmacokinetic Modeling Murray S. Cohn INTRODUCTION On July 15, 1985, R. H. Reitz and M. E. Andersen released a draft paper entitled "Physiologically-Based Pharmacokinetics and the Risk As- sessment Process for Methylene Chloride," which was provided to the federal health and safety regulatory agencies. The model used various assumptions and experimental data to predict internal concentrations of "toxiphors," a term used by these authors to describe presumed toxic chemical species resulting from biodegradation of methylene chloride by certain biochemical pathways. The results of the modeling were used to demonstrate that risk assessments done without consideration of the phar- macokinetic approach overestimate predicted risk to humans. The phar- macokinetic model was applied to estimate internal concentrations of toxiphors in the lungs and livers of mice, rats, hamsters, and humans, because these were the two sites of response in the National Toxicology Program's (NTP) inhalation bioassay of methylene chloride in mice (NTP, 19861. Of the many input parameters to the model, some of the most important determinations critical to its output are the values of the kinetic The material found within this paper contains the view of the author, who is an employee of the U.S. Consumer Product Safety Commission, and does not necessarily reflect official opinions or policies of the U.S. Consumer Product Safety Commission. This paper was prepared in the course of the official duties of the author as an employee of the U.S. Consumer Product Safety Commission. It is in the public domain and may be freely copied and reproduced. 265

266 MURRAY S. COHN constants (Kf, Vma,~, Km) for the two proposed pathways. According to Reitz and Andersen, Kf is the constant relating to a toxic nonsaturable pathway, that involving glutathione S-transferase (GST). VmaX and Km' on the other hand, are standard Michaelis-Menten kinetic constants that apply to a nontoxic, saturable oxidative pathway involving monooxygenase ac- tivity. In the original July 15, 1985, paper, these kinetic constants were based on the specific activities of GST and monooxygenase measured in samples of lung and liver tissue from all four species by Lorenz et al. (1984) using 1-chloro-2,4-dinitrobenzene (for GST) and 7-ethoxycoumarin (for mono- oxygenase) as substrates. Because these critical constants were not based upon experimentation using methylene chloride itself as a substrate, the assumption that the data from the study by Lorentz et al. (1984) served as an appropriate substitute led to one source of uncertainty in the use of the Reitz-Andersen model. The use of the data of Lorenz et al. (1984), however, was restricted by Reitz and Andersen in a Republication version of the same paper dated January 19, 1986 (Andersen et al., in press), for the purpose of apportioning the amount of total monooxygenase and GST between lung and liver in the four species. In this new version, a curve-fitting exercise was used instead to obtain the whole-body values of the three kinetic constants in animals. With all other physiological and biological parameters kept constant, the values of Km' V,~, and Kf were varied in an intricate computer optimization procedure, and the pharmacokinetic model was used to get the best approx- imations of actual experimental chamber data in which disappearance of methylene chloride over time was monitored. The chamber data were ob- tained for rats, mice, and hamsters by using five initial concentrations and recording the decrease in the chamber concentration of methylene chloride over a period of up to 6 h. The values of the three kinetic constants giving the best approximations to these five experimental concentration versus time curves for the three species were then used in other versions of the phar- macokinetic model to predict toxiphor levels. For humans, values of Km (0.58 mg/liter) and Vma,~ (119 mg/h) for the oxidative pathway were estimated from the human experimental data of Nolan and McKenna (in press) in which methylene chloride concentrations in expired air were measured following exposure by inhalation. For the critical GST pathway, however, no human data are available upon which an estimate of Kf can be based. Reitz and Andersen noted, however, that allometric scaling based on the concept of clearance served to adequately relate the mouse, hamster, and rat data, using a factor of body weight to the 0.7 power. Using this scaling procedure, they obtained a Kf value of 0.53 h- ' for humans, which is similar to the value of Kf for humans used by them in their original 1985 paper. They further noted that the human

Sensitivity Analysis in PK Modeling 267 values of Km and Vm~,,~ given above, when compared with the optimized animal values, also scale allometrically. This exercise examines the sensitivity of the Reitz-Andersen optimi- zation procedure to see if Kf, Km' and Vm`~ can be varied to produce an alternative, yet reasonable, fit in the experimental concentration versus time curves. As the mouse was the species that responded in the NTP's inhalation bioassay on methylene chloride at the sites modeled by Reitz and Andersen (lung and liver), the data on mice were chosen for this analysis. RESU LTS The optimized values of Km, V,,~, and Kf for the mice in the chamber experiments were 0.369 mg/liter, 0.8884 mg/h, and 4.3238 hat, respec- tively. With all other parameters, as determined by Reitz and Andersen in their original experiment and Andersen et al. (in press), held constant, it was decided to set the value of Kf equal to 1.00 h - ~ and vary the other two to obtain the best approximation of the experimental data (the value of 1.00 h- ~ was chosen to replace 4.3238 h- ~ because use of the phar- macokinetic model as described by Reitz and Andersen makes an ap- proximate fourfold difference in effective human dose). The Reitz-Andersen model was programmed for use on a microcomputer using the Livermore Solver for Ordinary Differential Equations (LSODE; Hindmarsh, 1980~. The microcomputer version used was verified by using modeled output data furnished by Reitz and Andersen. Using trial and error and the value of 1.00 ho for Kf, values of 0.7 mg/liter and 1.15 mg/h were found for Km and Vmax, respectively, which led to the curves labeled "Variant" in Figures 1 to 5. In these figures "RA" refers to the Reitz-Andersen optimized fit, and "Chamber" refers to the actual laboratory data. Each figure displays the time course of disappearance of methylene chloride (in parts per million) versus time (in hours); the five graphs represent five different initial concentrations (490, 960, 2,000, 3,200, and 10,000 ppm in Figures 1 to 5, respectively). The alternative (variant) choices for the three kinetic constants leads to curves which approximate the actual chamber data about as well as the Reitz- Andersen optimized choices, with the exception of the 3,200-ppm initial concentration chamber experiment (Figure 4~. This is not disturbing, how- ever, because there was a transient increase in methylene chloride con- centration between 2.5 and 4.5 h, and before this transient increase occurred, the chamber data followed the variant curve. A numerical description of the correlation of the variant fit to the chamber data, as compared with the RA fit, was not determined because of the transient increase in the 3,200-ppm data (Figure 41.

2663 MURRAY S. COHN 500 _ ~ 400 - C] tar: ~ 300- an - 100 - "LOW 'LOW O- 1 1 1 1 1 - ~ O Chamber Variant RA -Do 1 1 1 1 0 02 on 0.6 0.8 '.2 1.. 1.6 1.8 i 2i HOURS FIGURE l Modeled versus experimental concentrations (490 ppm at time zero). 1000- . 800 - Cal cY 53 600- c~ oo- 200 - \O "of "'me "I\ "en` "mu` 0 - "if: O Chamber Variant — RA %~0W --~-~b _ 1 1 i 1 1 1 0 0.5 1 1.5 2 2.5 HOURS FIGURE 2 Modeled versus experimental concentrations (960 ppm at time zero).

Sensitivity Analysis in PK Modeling 269 2000 f 1500 - IN 1000- CL 500 - O -.W An\ On,,\ O Chamber Variant — RA ~u~ _ 1 1 1 1 1 1 1 _ 1 o.s 1 '.s 2 2.5 3 3.5 ~ ..s s s.s HOURS FIGURE 3 Modeled versus experimental concentrations (2,000 ppm at time zero). 3s00 ~ O Chamber ~ ~ ~ Variant 3000- l RA ,,` 2500- ~ ~ A ~ 2000- Dow ~= ~0~ -""it BOO - O- 1 1 _ Coo 0 1 2 3 ~ 5 6 7 HOURS FIGURE 4 Modeled versus experimental concentrations (3,200 ppm at time zero).

270 MURRAY S. COHN loooo~ r ennn- c, t' 7000- 6000- on S000- 4000 - 3000- O t 2 ~0-~ l 5 6 7 3 4 HOURS FIGURE 5 Modeled versus experimental concentrations (10,000 ppm at time zero). The scaling approach of Reitz and Andersen, based on the computer optimization fit to the experimental animal chamber data, is shown in Table 1. By the Reitz-Andersen approach, if Kf scaled allometrically among the three animal species, the intrinsic clearance values should be similar. This is the reasoning for setting human intrinsic clearance equal to 60 and back- calculating. If 1.00 is used instead of 4.32 for the Kf for mice, however, the clearance for mice becomes 1.08, and the intrinsic clearance becomes 13.5 (Table 11. When compared with 33.5 for hamsters and 57. 1 for rats, the mouse no longer scales allometrically. Therefore, the value of a key input parameter, the human Kf, that is needed for application of the pharmacokinetic model to humans is yet to be determined. CONCLUSIONS It is concluded that although the Reitz-Andersen optimization procedure may lead to kinetic constants that give an optimal fit, by use of the pharmacokinetic model, to the experimental chamber data, there are al- ternative combinations (one of which was determined in this analysis) of

271 C) · - — Cal a: Ct :: o ._ Ct o o AS Ct o so ._ Ct so: ~_ ._ m ~ V~ _ O ~ . . . . . 00 ~ ~ O ~ ~_ - ~, ~ ~ ~ ~D Ct ~ 00 0 Ct - . . . . ~ _ ~ O _% ._ ~ ~ O ~ ;- ~ ~ ~ C~ O ~ 000~0 C~ C~ ._ C) CD V) o o ._ 3 ;^ o _ ., ~:s C) C5 Ct C) C) C~ C5 C) C5 C~ Ct C~ ._ C) C~ o Ct O a~ C) C) ._ ._ s~ s:: ._ s: Ct 8 _ o C~ o . . o _ . ~` C~ Ct ~ J ~ 00 0 C~ 00 ~ O ~ ~ O Ox .> ~ .O ~ ~ ,0 c<5 . _ .- X ~ X r: ._ Ct - - C~ ~: C~ Ct Ct o o - Ct C) C) C) Ct Ct C) C) ._ cn ._ s~ ._ ._ C~ ~> Ct - - Ct C) C~ ._ · C<S 3 o

272 MURRAY S. COHN the three kinetic constants that also can reasonably fit these data. This may be due, perhaps, to the fact that the optimum fit may have correlation coefficients that only trivially exceed those from another fit that has sub- stantial differences in the underlying constants. In other words, there are multiple adequate solutions to the problem (redundancy). On this basis, the scaling approach suggested by Reitz and Andersen for the determi- nation of a human Kf value, a most critical value according to these authors with regard to the prediction of methylene chloride carcinogenic risk to humans, is subject to uncertainty. Thus, the scaling approach should still be considered preliminary, and more research and validation is required before it can be accepted as a reasonable basis for pharrnacokinetic mod- eling in the case of methylene chloride. Most phannacokinetic modeling is dependent upon the values of many input parameters. Such modeling can also rely upon the process of solving for those input parameters that are unable to be determined otherwise. The purpose of the exercise provided here is to encourage those who employ pharrnacokinetic models as a tool to improve quantitative risk assessment to examine the sensitivity of such models. Small changes in input data or in the goodness of fit to a set of experimental data can greatly alter the effect that inclusion of pharmacokinetic modeling can have on estimates of human risk. REFERENCES Andersen, M. E., H. J. Clewell III, M. L. Gargas, F. A. Smith, and R. H. Reitz. In press. Physiologically-based pharmacokinetics and the risk assessment process for methylene chloride. Toxicol. Appl. Pharmacol. Hindmarsh, A. C. 1980. LSODE and LSODI, two new initial value ordinary differential equationssolvers.ACM-SignumNewsletterlS(4):10-11. Lorenz, J., H. R. Glatt, R. Fleischmann, R. Ferlinz, and F. Oesch. 1984. Drug metabolism in man and its relationship to that in three rodent species: Monooxygenase, epoxide hydrolase, and glutathione S-transferase activities in subcellular fractions of lung and liver. Biochem. Med. 32:43-56. National Toxicology Program (NTP). 1986. Toxicology and Carcinogenesis Studies of Dichloromethane (Methylene Chloride) (CAS No. 75-09-2) in F344/N Rats and B6C3F~ Mice (Inhalation Studies). Technical Report Series No. 306, NIH Publication No. 86- 2562. Washington, D.C.: U.S. Department of Health and Human Services. Nolan, R. J., and M. J. McKenna. In press. Pharmacokinetics of methylene chloride in human volunteers.

Mutation Accumulation: A Biologically Based Mathematical Mode' of Chronic Cytotoxicant Exposure Rory B. Conolly, Richard H. Reitz, and Melvin E. Andersen I NTRODUCTION This report describes a computer-based biological model suggesting why nongenotoxic chemicals such as chloroform test positively when carcinogenicity bioassays use doses that may cause transient cytotoxicity (Heywood et al., 1979; Jorgensen et al., 1985; National Cancer Institute, 1976; Palmer et al., 1979; Roe et al., 19791. The model is roughly iso- morphic with actual mammalian physiological and biochemical systems. It describes the mathematical relationships among: 1. In vivo cytotoxicant exposure and tissue dose of parent compound and metabolites. 2. Tissue dose and cell death. 3. Cell death and regenerative hyperplasia. 4. Regenerative hyperplasia and normally scheduled replication and the accumulation of genetic mutations. The model described here is a combination of three discrete models describing pharmacokinetics, cytotoxicity, and carcinogenesis. These dis- crete models are (1) the physiologically based pharmacokinetic (PB-PK) model of Ramsey and Andersen (1984), (2) a cytotoxicity model, and (3) the two-stage carcinogenicity model of Moolgavkar and Knudson (19811. The PB-PK model was used as a pharmacokinetic driver for the cyto- toxicity model to provide accurate linkage between in viva toxicant ex- posure and tissue doses of parent compounds and reactive metabolites. 273

274 RORY B. CONOLLY ET AL. This PB-PK model has been validated for several volatile halogenated compounds and solvents (Clewell and Andersen, 1985; Gargas et al., 1986; Ramsey and Andersen, 19841. The cytotoxicity component of the model contains equations describing depletion of a generic macromolecule (MM) by a reactive metabolite of the parent compound. The MM is subsequently resynthesized by a feed- back-controlled enzymatic system. Depletion of MM leads to cell death, which is followed by regenerative hyperplasia. The description of MM depletion was based on a model of hepatic glutathione depletion and resynthesis that has been previously described (Andersen et al., 19861. A model developed by Reitz (1986) of cytotoxicity mediated by parent chem- ical bioactivation and macromolecule depletion also contributed to this work. In Moolgavkar and Knudson's (1981) carcinogenicity model, cellular replication incurs a risk of error leading to mutations in daughter cells. We have used cytotoxicity, as described above, to force regenerative hyperplasia and thereby increase the rate of mutation accumulation. Sim- ulated data are presented showing how cells containing one and two mu- tations accumulate as a function of time and toxicant exposure. M ETHODS Computer Hardware and Software The mathematical model was written in ACSL (Advanced Continuous Simulation Language; Mitchell & Gauthier Associates, Inc., Concord, Mass.) and run on an IBM-AT (IBM Co., Boca Raton, Flab.. Modeling of Cytotoxicity The liver in Ramsey and Andersen's (1984) PB-PK model was used as the target organ for cytotoxic effects (Figure 11. This choice is not important per se for the results and conclusions reported here and was used only as a matter of convenience. Model parameters were scaled for a 250-g rat, and the liver was modeled as a population of 108 cells with basal death and birth rates of 10-3/h. Liver cells contained a generic MM, which is essential for cell viability. (In a model validated for a specific cytotoxicant whose mechanism of action involved macromolecular bind- ing, the generic MM would be replaced with a description of the actual macromoleculets) involved. Validation experiments using chloroform as the model cytotoxicant are planned.) The rate of hepatic metabolism of parent compound was dependent on the rate of toxicant delivery to the liver and the Michaelis constants for metabolism. Depletion of MM was

Cytotoxicity and Mutation Accumulation 275 (I n hal ation ) Aiveola r Space Lung Blood Fat Tissue Grou p Muscle Tissue Group l Richly Perfused Group l . l Liver · Biotransformation · Macromolecule depletion · Hepatocyte death · H epatocyte hi rth with linked mutation rate (Exhalation) (Cardiac Output) . .. FIGURE 1 Schematic of the biologically based pharmacokinetic model of Ramsey and Andersen ( 1984) as adapted to describe hepatic cytotoxicity and mutation accumulation. Each compartment is defined by biologically realistic parameters. modeled by describing a second-order reaction of the parent compound metabolite with MM. Cell death was lined to MM depletion by a normal curve (Hastings, 1955), in which >99.999% of the area under the curve lay between the maximum and minimum possible levels of an MM. The cumulative percentage of the area under the curve, corresponding to the degree of MM depletion, defined the number of hepatocytes committed

276 RORY B. CONOLLY ET AL. to die. Actual death took place 1 h later. The liver responded to unsched- uled cell death, i.e., cell death advanced by the toxicant, by increasing the rate of cell division, until the steady-state level of 108 cells was regained. This compensatory increase in birth rate occurred 8 h after the unscheduled death. Modeling of mutation accumulation in hepatocytes is described below. Hepatocytes with no mutations (N1 cells) and those with one mutation (N2 cells) were modeled as having the same basal death and birth rates, as being equally susceptible to cytotoxicant and as responding to unsched- uled cell death, i.e., cytotoxicity, in the same manner. Cells with two mutations (N3 cells) were considered to be insensitive to cytotoxicant. The cytotoxicity model did not describe the maximum amount of cell death that could be tolerated without death of the animal. This could easily be done but was not considered necessary for this exercise. Figures 1, 2, and 3 illustrate various aspects of the cytotoxicity model. Figure 4 is useful for studying model function. " Liver" 108CelIs Norma ~ Rate Ce! ~ Death 1 0~3 H R-4 O Normal Rate Cell Birth 10-3 H R-, O Chance of Mutation Per Cell Birth 10-7 Toxicant (Biotransformation)> Metabo~ite (Km, VMAX) As4tMetabolite],+ [Macromolecule] As + [Macromolecule],+ Probability of Cell Death* As ~ Cell Death, + Population of Normal (N 1 ) Cells As + N 1, 4` B i rth Rate U nti ~ Back to 1 08 Cells. * Sensitivity of cells to macromolecule depletion is normally distributed about a set level of macromolecule. FIGURE 2 Detail for the cytotoxicity component of the model. The choice of liver was one of convenience. This description of cytotoxicity could be applied equally well to other tissues.

Cytotoxicib and Mutation Accumulation 277 , | Scheduled | | Unscheduled | L Death | L Death l _~ l A L T Normal Cells (N1) , | Scheduled | | Unscheduled | I | Death | | Death I ~ T T :1 Mut al Cells (N2) Scheduled Unscheduled Birth Birth | Successful | | Replication During Replication , ~ Scheduled Unscheduled Birth Birth l _ T ~ 1 r 4/ l Successf ul Mutation Replication During I Replication 1 1 1 2 Mutation Cells (N3) - FIGURE 3 Adaptation of the Moolgavkar and Knudson (1981) two-stage model for carcino- genicity to the pharrnacokinetic/cytotoxicity model. N1 and N2 cells have the same basal birth and death rates. N3 cells are immortal. Unscheduled events are those due to cytotoxicity. Modeling the Relationship of Cell Birth to Mutation Accumulation A premise of the model is that regenerative hyperplasia consequent to cytotoxicity increases the rate at which mutations accumulate in the target organ and that some of these mutations influence tumor development. This premise is based on the assumption that a small fraction of cellular replications "go wrong," resulting in mutant progeny. Normal hepato- cytes (N1 cells) were modeled as having a 10-7 chance of suffering a mutation during replication. Cells containing one mutation (N2 cells) were modeled as replicating with the same kinetics as N1 cells and also as having a 10-7 chance of suffering a mutation during replication. Cells with two mutations (N3 cells) were assumed to be tumorigenic without further mutation. These relationships are illustrated in Figure 3. Generic Nature of Cytotoxicity Mode' We know of no body of data that is suitable for mathematical modeling of the relationships among reactive metabolite production, depletion of critical cellular macromolecules, rates of cell death and regeneration, and the accumulation of mutations. We therefore defined a generic MM and estimated reasonable quantitative relationships among the amount of parent compound metabolized, depletion of MM, cell death and birth, and mu- tation accumulation. Because the definition of MM is not linked to a

278 RORY B. CONOLLY ET AL. (C) 1.10 N1 x 108 0.55- 0.00 {b) 1.00— MMx ~ 04 050_ 0.00 - ~a' 1.00 (mg/~) 0 50 0.00 r\ - - T \_~ ,/ r I , I \ 1 \ ~ 0.00 18.0 36.0 Hours 54.0 72.0 FIGURE 4 Temporal relationships of in viva toxicant exposure, macromolecule depletion, and number of normal (N1) cells. In (a) daily 6-h exposure to parent toxicant is tracked by the spikes in CA. In (b) depletion of MM by a metabolite of the parent and resynthesis of MM as CA declines are shown. In (c) N1 death as MM is depleted and consequent regenerative hyperplasia are illustrated.

Cytotoxicity and Mutation Accumulation 279 biochemical mechanism of toxicity for a particular chemical, the model's descriptions of cytotoxicity and mutation accumulation are also generic. For these reasons, even though the PB-PK model is well-validated for a number of compounds, we present the simulations of cytotoxicity and mutation accumulation in terms of exposure to a model cytotoxicant rather than to any particular chemical. (For the curious, the model cytotoxicant used has the pharmacokinetic behavior of 1,2-dichloroethane.) RESU LTS Figure 4 illustrates the qualitative relationships among cytotoxicant ex- posure, depletion of MM, cell death, and regenerative hyperplasia. Sim- ulation of repeated, 6-in/day inhalation exposure to parent compound results in corresponding spikes in its arterial concentration (CA). AS CA increases, MM starts to fall. This fall in MM is due to its destruction by the cytotoxic metabolite of the parent compound. As the concentration of cytotoxic metabolite falls, MM is resynthesized and its concentration increases. The model simulates an overshoot or rebound effect in which the level of MM rises above its preexposure concentration. This type of rebound effect is seen after toxicant-induced depletion of hepatic glutathione (GSH) (Won" and Klaassen, 1981), which we have modeled for related projects. A1- though not shown in Figure 4, the MM concentration would return to the preexposure basal level between 24 and 48 h after a single 6-h exposure. A decreasing concentration of MM is followed by a corresponding decline in N1, the number of normal hepatocytes. N1 returns to its basal level (approximately 108 cells) after MM is resynthesized. In all, Figure 4 illustrates several cycles of toxicant-driven cell death and regenerative hyperplasia. Figure 5 simulates the accumulation of cells with one mutation (N2 cells) over time in toxicant-exposed and control groups. Note the linear nature of N2 accumulation with time and the demonstration that toxicant exposure can greatly increase the rate of N2 accumulation. Simulated rates of accumulation of cells with two mutations (N3 cells) are shown in Figure 6. Notice that Figure 6 uses log-log scaling and that the lines depicting N3 accumulation with time are described by a power function. In Figure 7, in which arithmetic scaling is used, the numbers of N1, N2, and N3 cells occurring during the simulation of 1 year of 6-in/day toxicant exposure are shown. The model calculates a cumulative expectation describing the prevalence of N2 and N3 cells. Figure 6 indicates that about 10-4 of one N3 cell was produced during 90 days of simulated exposure, causing 20% cell

2~30 RORY B. CONOLLY ET AL. 2.25 - ~ .69 - N2 x 1 o2 1.12— 0.56 - ,' Toxicant Exposed Control 0.00 0.54 1.08 1.62 2.1 6 Hoursx 103 FIGURE 5 Simulated 90-day accumulation of cells with one mutation (N2) in control and toxicant-exposed groups. The daily fluctuation in the number of N2 cells in the toxicant-exposed group reflects the cycle of cell death and regenerative hyperplasia caused by each 6-h exposure to toxicant. death/day. In this instance the 10-4 actually represents the average prev- alence for the population of individuals at risk at that point in time. DISCUSSION The model described here incorporates some straightforward assump- tions about relationships among the tissue dose of a toxicant, depletion of a critical cellular macromolecule, the likelihood of cell death, and the probability of mutations occurring during regenerative hyperplasia. De- pletion of hepatic GSH is known to be linked to increased probability of cell death (Docks and Krishna, 1976; Mitchell et al., 1973; Wells et al., 19801. The use of a cumulative probability function to describe this re- lationship means that the marginal increase in the probability of cell death as MM is depleted is not monotonically increasing, as might be expected. The correct mathematical description of this relationship is not known at

Cytotoxicity and Mutation Accumulation 28 ~ 10-4 10-5 - N3 10-6- 10-7- 1o-8 10-9- /~nt/Ex/po/se/d ,r 1 1 I I 41 11111 ~ I 1o2 103 Hours 1ol FIGURE 6 Simulated 90-day accumulation of cells with two mutations (N3) in control and toxicant-exposed groups. The toxicant-exposed group suffered about 20~o cell death/day, which was completely replaced within 24 h by regenerative hyperplasia. This regenerative hyperplasia leads to the difference between the control and toxicant-exposed groups. In both the control and exposed groups, N3 accumulation is proportional to timer. this time. However, over the range of MM depletion used in the simu- lations reported here (O to 20%), the cumulative probability did increase monotonically. A number of workers have studied mutation rates in replicating cell systems, and it is generally accepted that mutation does occur as a con- sequence of mistakes during cellular replication (Crawford et al., 1983; Tsutsui et al., 1978, 19811. The model suggests that repeated episodes of cytotoxicity increase the rate at which mutations accumulate in a target tissue. Moreover, cells containing two mutations and that would presumably be more likely to be tumorigenic (Moolgavkar and Knudson, 1981) accumulate with the square of time, whereas the accumulation of cells with one mutation is linear.

282 RORY B. CONOLLY ET AL. 2.0— 1.5—.75— 0 1.0— 1.0— 0.5— 0.0— 4.0— 3.0— .25— 1.0— .00— 0.0- _= ,/ / 0.00 2.1 9 4.38 Hours x 103 6.57 8.76 (one year) FIGURE 7 Simulation of 1 year of daily 6-h inhalation exposure to a model cytotoxicant. Daily fluctuations in N1 and N2 cells are due to cytotoxicity. Note that the rate of accumulation of cells with one mutation (N2) is linear with time, whereas accumulation of cells with two mutations (N3) depends on timer. A significant advantage of this model is that the tissue dose of cytotoxic metabolite is related to in vivo parent compound exposure by the well- validated PB-PK model (Clewell and Andersen, 1985; Gargas et al., 1986; Ramsey and Andersen, 19841. Modeling of cytotoxicity and other target organ consequences of in viva exposure has little meaning if the tissue dose of the ultimate toxicant is not well-characterized. Limitations of the Mode' What Is Modeled and What Is Not Several processes that would affect cytotoxicity and mutation accu- mulation have not been specifically described in this model. For example, the model was exercised repetitively in order to simulate chronic exposures of up to 1 year (Figures 5, 6, and 71. For these long-term simulations it may be necessary to model the aging process in the rat. Changes in the size of the fat compartment and in the capacity for parent compound metabolism with age would affect pharmacodynamic behavior. The pos- sibility that N2 cells have growth advantages, i.e., altered death and birth rates and different sensitivity to MM depletion relative to N1 cells, may be important and could give rise to N2 birth rates different from those

Cytotoxicity and Mutation Accumulation 283 simulated here. Immune surveillance of phenotypically altered cells prob- ably affects the N2 and N3 populations, and modeling of this parameter may be necessary for a comprehensive description. The model does not estimate the biological significance of the simulated mutations. Because not all mutations would be expected to predispose the cell to clonal growth, the accumulation of N3 cells is probably more rapid than the accumulation of tumorigenic cells. These factors may all be important for accurate simulation of rates of cytotoxicity and mutation accumulation. However, the relative shapes of the curves describing the change in N1, N2, and N3 populations with time (Figure 7) should not differ between a model specifically describing these factors and the one described here. Mode' Validation As noted above, the PB-PK model of Ramsey and Andersen (1984) which is used as a pharmacokinetic driver for the cytotoxicity model has been well-validated. We are in the process of obtaining actual data on the quantitative relationships among CHC13 biotransformation, cell death, cell regeneration, and the consequent accumulation of mutations. These data will be used to develop a CHC13 cytotoxicity model that will provide a strong test of the hypothesis that CHC13 cytotoxicity is responsible in whole or in part for the positive outcomes seen in CHC13 carcinogenicity bioassays (Heywood et al., 1979; Jorgensen et al., 1985; National Cancer Institute, 1976; Palmer et al., 1979; Roe et al., 19791. At the present stage of its development, the generic model (see "Methods") can be used to examine aspects of the quantitative relationships among depletion of critical cellular macromolecules, cytotoxicity, cell death and replication, and the accumulation of mutations. It should be noted that the model does not describe the accumulation of genetic mutations caused by direct interactions of parent compound or metabolites with DNA. However, the model does provide a framework which could readily be modified to describe these factors. In conclusion, this presentation illustrates the potential utility of using computer-based models of biological systems to study the relationships of cytotoxicity, cell replication, and the accumulation of genetic damage. A particularly interesting result generated by the model described here is the dependence of the birth rate of cells containing two mutations on the square of time. ACKNOWLEDGMENTS Our thanks to Maj. Harvey Clewell and Dr. Thomas B. Starr for their help.

284 RORY B. CONOLLY ET AL. REFERENCES Andersen, M. E., H. J. Clewell III, M. L. Gargas, and R. B. Conolly. 1986. A physio- logical pharmacokinetic model for hepatic glutathione (GSH) depletion by inhaled hal- ogenated hydrocarbons. The Toxicologist 6:148. (Abstract 598.) Clewell, H. J., III, and M. E. Andersen. 1985. Risk assessment extrapolations and phys- iological modeling. Toxicol. Ind. Health 1:111-131. Crawford, B. D., J. C. Barrett, and P. O. P. Ts'o. 1983. Neoplastic conversion of pre- neoplastic Syrian hamster cells: Rate estimation by fluctuation analysis. Mol. Cell. Biol. 3:931-945. Docks, E. L., and G. Krishna. 1976. The role of glutathione in chloroform-induced hep- atotoxicity. Exp. Mol. Pathol. 24:13-22. Gargas, M. L., H. J. Clewell III, and M. E. Andersen. 1986. Metabolism of inhaled dichloromethanes in vivo: Differentiation of kinetic constants for two independent path- ways. Toxicol. Appl. Pharmacol. 82:211-223. Hastings, C. 1955. P. 26 in Approximations for Digital Computers. Princeton, N.J.: Prince- ton University Press. Heywood, R., R. J. Sortwood, P. R. B. Noel, A. E. Street, D. E. Prentice, F. J. C. Roe, P. F. Wadsworth, A. N. Worden, and N. J. Van Abbe. 1979. Safety evaluation of toothpaste containing chloroform. III. Long term studies in beagle dogs. J. Environ. Pathol. Toxicol. 2:835-851. Jorgensen, T. A., E. F. Meierhenry, C. J. Rushbrook, R. J. Bull, and M. Robinson. 1985. Carcinogenicity of chloroform in drinking water to male Osborne Mendel rats and female B6C3F1 mice. Fund. Appl. Toxicol. 5:760-769. Mitchell, J. R., D. J. Jollow, W. Z. Potter, J. R. Gillette, and B. B. Brodie. 1973. Ac- etaminophen-induced hepatic necrosis. IV. Protective role of glutathione. J. Pharmacol. Exo. Ther. 187:211-217. Moolgavkar, S. H., and A. G. Knudson, Jr. 1981. Mutation and cancer: A model for human carcinogenesis. J. Natl. Cancer Inst. 66:1037-1052. National Cancer Institute. 1976. Carcinogenesis bioassay of chloroform. NTIS No. PB246018/ AS. Bethesda, Md.: National Cancer Institute. Palmer, A. K., A. E. Street, F. J. C. Roe, A. N. Worden, and N. J. Van Abbe. 1979. Safety evaluation of toothpaste containing chloroform. II. Long term studies in rats. J. Environ. Pathol. Toxicol. 2:821-833. Ramsey, J. C., and M. E. Andersen. 1984. A physiologically-based description of the inhalation pharmacokinetics of styrene in rats and humans. Toxicol. Appl. Pharmacol. 73: 159-175. Reitz, R. H. 1987. The role of cytotoxicity in the carcinogenic process. In Nongenotoxic Mechanisms in Carcinogenesis, Banbury Report 25. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory. Roe, F. J. C., A. K. Palmer, A. N. Worden, and N. J. Van Abbe. 1979. Safety evaluation of toothpaste containing chloroform. I. Long term studies in mice. J. Environ. Pathol. Toxicol. 2:799-819. Tsutsui, T., J. C. Barrett, and P. O. P. Ts'o. 1978. Induction of 6-thioguanine- and oubain- resistant mutations in synchronized Syrian hamster cell cultures during different periods of the S phase. Mutat. Res. 52:255-264. Tsutsui, T., B. D. Crawford, and P. O. P. Ts'o. 1981. Comparison between mutagenesis in normal and transformed Syrian hamster fibroblasts. Difference in the temporal order of HPRT gene replication. Mutat. Res. 80:357-371.

Cytotoxicity and Mutation Accumulation 285 Wells, P. G., R. C. Boerth, J. A. Dates, and R. D. Harbison. 1980. Toxicologic en- hancement by a combination of drugs which deplete hepatic glutathione: Acetaminophen and doxorubicin (adriamycin). Toxicol. Appl. Pharmacol. 54:197-209. Wong, K.-L., and C. D. Klaassen. 1981. Relationship between liver and kidney levels of glutathione and metallothionein in rats. Toxicology 19:39-47.

Physiologically Based Pharmacokinetic Mode' for Ethylene Dichloride and Its Application in Risk Assessment Richard W. D'Souza, William R. Francis, Robert D. Bruce, and Meivin E. Andersen INTRODUCTION Ethylene dichloride (EDC; 1,2-dichloroethane) is a large-volume chem- ical extensively used in industry. EDC is presently being used as an alternative to ethylene dibromide as a fumigant for the treatment of food grains to control rodents, insects, and soil nematodes. In 1978, a study conducted by the National Cancer Institute (NCI) demonstrated treatment- related tumors in rats and mice dosed with 75 and 150 mg/kg EDC as a corn oil gavage (NCI, 19781. At about the same time the Bologna Tumor Center in Italy reported that no treatment-related tumors were seen in rats or mice chronically exposed to EDC by inhalation 7 h daily at the maximum tolerated dose of 150 ppm (Maltoni et al., 198()~. These apparently con- tradictory results of the two bioassays have not been satisfactorily resolved to date. Pharmacokinetic studies employing compartmental models have been conducted on EDC (Reitz et al., 1982; Spreafico et al., 19801. The results of these studies suggest that the differences in the two bioassays may be because of pharmacokinetic differences; that is, target organ exposure to EDC may have been greater after the gavage dosing then after a 7-h inhalation exposure. Unfortunately, the compartmental modeling em- ployed in these studies could not quantitate target organ exposures with different doses and routes of exposure, nor extrapolate the relevance of the observations to human exposure situations for assessing human risk. The objectives of this study were to develop a physiologically based pharmacokinetic (PB-PK) model for EDC and its reactive metabolites, to demonstrate the use of information derived from this model in cancer risk 286

PB-PK and EDC Risk Assessment 287 assessment, and to use the model to help explain the route-of-exposure differences observed in the two EDC bioassays. METHODS PB-PK Mode' Development PB-PK models are mathematical models that mimic the way the body handles chemicals. These models are physiologically realistic and utilize all available anatomical, physiological, and physiocochemical informa- tion. -A schematic representation of the PB-PK model developed for EDC is shown in Figure 1. For the distribution of EDC within the body, organs 1 ~ Gas Exchange _ Lung _ _ _ . . ~ Metabolism Am;/ K\`GS Liver Ql . ~ _ Metabolism Kit/ K\~` Richly-Perfused I Slowly-Perfused _ Tissues ~ - Fat ~ Qc Qs - Qf ~ GUT FIGURE 1 Schematic representation of the pharmacokinetic model developed for EDC. The biodistribution of EDC was modeled by dividing body tissues into three compartments based on their blood flow and relative ability to accumulate EDC. The liver and lung were considered the only metabolizing organs. Vm``r and Km are the rate constants for the oxidation pathway, and Kf is the rate constant for the GSH pathway. GSH ~ represents a glutathione depletion model.

288 RICHARD W. D'SOU~ ~ AL. TABLE 1A Partition Coefficients Obtained for EDC Ratio Sprague- Fisher Dawley 344 B6C3F1 Rat Rat Mouse Human Blood:air 27.6 30.4 29.7 21.1 Richly perfused:blood 1.1 1.2 1.0 Slowly perfused:blood 0.8 O. 8 O. 8 Fat:blood 12.2 11.4 12.1 were lumped together into three compartments, based on their blood flow and ability to accumulate EDC. For the purposes of this demonstration, the liver and the lung were considered to be the only metabolizing organs. All organs were connected to the circulatory system in an anatomically accurate fashion, that is, the lung was modeled as receiving the entire cardiac output, whereas other organs received a fraction of cardiac output. Because EDC is a small, lipophilic, and readily diffusible molecule, the PB-PK model was set up as a flow-limited model; that is, there were no diffusion barriers to the distribution of EDC. The parameters used in constructing this model are given below. Partition Coefficient The steady-state distribution ratio or partition coefficient of EDC was calculated by the method of Sato and Nakajima (19791. These partition coefficients were calculated for the B6C3F1 mouse and Fischer 344 rat and Sprague-Dawley rat. The brood: air partition coefficient was also ob- tained for human blood. The results are shown in Table 1A. The partition coefficients of many tissues were similar, and for modeling purposes these tissues were grouped together. The richly perfused group included such tissues as the kidney and spleen, the slowly perfused comprised muscle and skin, and the fat compartment represented body fat. It was noted that the human blood:air partition coefficient was somewhat smaller than that for the rat or the mouse. This lower value was, therefore, used in our model when scaling up from the rodent to the human. Metabolism Rate Constant EDC is metabolized by two competing pathways, a saturable pathway that involves P-450 oxidation, and by direct conjugation with glutathione (GSH). A simplified metabolism schematic, as reported by Anders and Livesey (1980), is shown in Figure 2.

PB-PK and EDC Risk Assessment 289 OH Vmax; Km Kim I Cl-CH2-CH2-CI ~ Cl-CH2-CHO _ - - Cl-CH=CH-S-G G SH GSH'` Kgs Cl-CH2-CH2-S-G Kfee Macromolecules CI-CH2-COOH etc. FIGURE 2 EDC metabolism pathway. EDC is metabolized by a saturable oxidation pathway and by conjugation with GSH. Estimates of metabolism rates for He two pathways were obtained from gas uptake measurements (Gargas et al., 19861. The results of the gas uptake chamber runs are shown in Figure 3. As with similar halogenated hydrocarbons, we observed that the oxidative pathway for EDC is saturated at relatively low concentrations and is best described by a Michaelis- Menten type equation. The GSH pathway is first-order until high EDC exposure concentrations, where we observed a shift from first-order ki- netics. This shift from first-order behavior for the GStI pathway was due to depletion of GSH, so that not enough GSH was available to react with EDC. In the recirculating chamber experiments Hat we (M.E.A.) have conducted with the halogenated hydrocarbons, this behavior has only been observed with EDC and allyl chlonde. Metabolism rate constants for the two pathways are shown in Table 1B. At exposure concentrations where GSH is not depleted, a first-order rate constant (Kf) was used in the PB- PK model. At higher exposure concentrations, as used in bioassays, this first-order constant did not provide an adequate description of EDC me- tabolism, and the model therefore underestimated EDC concentrations To account for this behavior, a GSH depletion model (D'Souza et al., 1986) had to be incorporated as part of the PB-PK description of EDC. This model kept track of liver and lung GSH levels, which were necessary to compute the amount of GC metabolite. The gas uptake studies provided rate constants for total metabolism. Metabolism rates between the liver and lung were split by using data from the literature on relative enzymatic TABLE 1B Metabolism Rate Constants Obtained for EDC EDC Van,` = 3.25 mg/h/kg Km = 0.25 mg/liter Kf = 9.0 h-t kg-' GSH model Kgs = 0.0014 h-i kg- ' Kfee = 4,500 h-t kg-t Kgsm = 0.14 h-t kg- '

290 RICHARD W.D'SOUZA ET AL. 1 0,000 - . ~ 1 000 - Q - o ._ _, a) CD o a) Q E Cal If. ~ 100 ~ 10 - ·~\ ·-\ -_~ _~ - at, 'a .~ .\ ~ _~ - -~. ~ of I ~ 0 100 200 300 400 9~ Time (Minutes) 1 1 1 FIGURE 3 Gas uptake runs for EDC. The initial chamber concentrations used were 300 (in), 500 (a), 1,000 (a), and 2,000 (o) ppm. One group of animals exposed to 2,000 ppm was pretreated with 2,3-epoxy propanol (a). The lines represent model predictions that were generated by defining a PB-PK model of rats in a closed chamber. activities between these two organs, as was accomplished for the methyl- ene chloride model (Andersen et al., 1987~. Scaleup of metabolism rates for different species was performed by using allometric scaling (Adolph, 1949; Dedrick, 1973; Lindstedt and Calder, 1981~. Physiological Parameters The physiological parameters that were used to construct the model are well established in the literature (Caster et al., 1956~. In most cases allometric equations were used to scale these parameters for different species.

PB-PK and EDC Risk Assessment 291 Mass-balance differential equations were written for each compartment of the PB-PK model, describing the inflow, outflow, binding, and me- tabolism of EDC and the formation and detoxification of the intermediate metabolites. These equations were similar to those used by Andersen et al. (1987), except for the GSH model. The ACSL (Advanced Continuous Simulations Language; Mitchell & Gauthier Associates, Inc., Concord, Mass.) computer program was used to solve this set of simultaneous differential equations by numerical integration, employing Gear's algo- r~thm for stiff systems. RESULTS The PB-PK model was validated by measuring EDC and GSH concen- trations in the rat and mouse. Because the short-lived reactive metabolites cannot be measured readily, GSH depletion was used as an indirect val- idation. Figure 4 depicts PB-PK model predictions and experimentally determined EDC blood concentrations in the rat. Model predictions were in good agreement with observed data at the dose range tested. Blood concentrations were similarly determined for the mouse and are shown in Figure 5. Again, experimental data were in close agreement with model predictions. Our predictions were also compared with data in _ ~ o2~ - 10 -. - ~0 - \t o ._ cat i_ c' a: o o m 1 o~2- co) 10-3- . ~ I · ~ ~ 0 1 2 3 4 5 6 ~ . . \ . Time Post-Dose (Hour) FIGURE 4 Model predicted ( ) and experimentally determined EDC blood concentrations in the rat after corn oil gavage doses of 1.5 (a), 25 (a), and 150 (a) mg/kg.

292 RICHARD W. D'SOUZA ET AL. - _ l 1 .o~ 1 oo at_ a) ~ 1o-l o o m 1 o-2 10-3 . it _ . ~ 1 1 1 0 1 2 3 Time Post-Dose (Hour) FIGURE 5 Model predicted ( ) and experimentally determined (a) EDC blood concentrations in the mouse after a cons oil Savage dose of 150 mg/kg. the literature and were found to be in excellent agreement. Spreafico et al. (1980) measured EDC levels in blood and tissues in the rat at several dose levels after oral, intravenous, and inhalation exposures. The results of their studies are very predictable by our model, both for blood and tissue concentrations. EDC blood concentration data of Reitz et al. (1982) after both oral and inhalation exposures in the rat were also in close agreement with model predictions. GSH concentrations for both liver and lung of the rat and mouse after oral doses of 150 mg/kg were in close agreement with model predictions, as shown in Figures 6 and 7, respectively. The strains of rats used in the studies cited from the literature and ours were different, but there were no apparent differences in the pharmacoki- netics of EDC because of this difference. Spreafico et al. (1980) used Sprague-Dawley rats, and Reitz et al. (1982) used Osborne Mendel rats.

PB-PK and EDC Risk Assessment 293 10 - - ~ 8- o E .° ~ ~ ~ - a) o <' 4- I oh 2- '~ ' / l . 2 4 6 8 10 12 Time Post-Dose (Hour) FIGURE 6 GSH concentrations in the liver (to) and lung (a) of the rat following an EDC dose of 150 mg/kg. Symbols represent mean + standard deviation of four rats. The curves are model predictions. We employed Fisher 344 rats and B6C3F1 mice in our studies. It appears that strain differences do not significantly affect the biodistribution or metabolism of compounds like EDC. From the close agreement between model predictions and observed concentrations both from our studies and data reported in the literature, it is clear that the PB-PK model developed has strong predictive powers. This has been seen with different dose levels, routes of exposure, and two species. Dose Surrogates Several reports in the literature have concluded that the GSH conjugate formed in the metabolism of halogenated hydrocarbons like EDC and ethylene dibromide (EDB) is the carcinogenic moiety and not the parent

294 RICHARD W. D'SOUZA ET AL O1 | L ~ 8 - 3 6- 4- ~ BY ,; ~ a ~ ~ 6 8 10 12 2 4 Time Post-Dose (Hour) FIGURE 7 GSH concentrations in the liver (a) and lung (a) of the mouse following an EDC dose of 150 mg/kg. Symbols for the liver represent mean + standard deviation of four mice. The lungs of four mice were pooled together and analyzed as a single sample. The curves are model predictions. compound. For instance, White et al. (1983) compared the amount of DNA single-strand breaks produced by EDB and its tetradeuterated de- rivative (EDB-d4), and found that EDB-d4 was more genotoxic than EDB. Because the deuterated compound would preferentially be metabolized via the GSH pathway, these results demonstrated that the GSH pathway and not the oxidative pathway was the source of genotoxic metaboliteks). Similarly, Storer and Conolly (1985) studied hepatic DNA damage by EDC after pretreating mice with either piperonyl butoxide to block micro- somal oxidation or diethyl maleate to deplete GSH, and found that the GSH pathway was responsible for DNA damage and not the oxidative pathway. With methylene chloride, Andersen et al. (1987) related the number of tumors at each dose level with the PB-PK model predictions of the amount of GSH and oxidative metabolite produced at that dose

PB-PK and EDC Risk Assessment 295 level and noted that tumors related quite well with GSH metabolite formed. We have, therefore, used the total amount of the parent compound-gluta- thione conjugate (GC) produced in a target tissue as a surrogate for dose. PB-PK Mode' and Risk Assessment Because of the nonlinear metabolism pathways for EDC, the relationship between administered EDC dose, either as inhalation or oral exposure, and the amount of GC metabolite produced in a target tissue is complex. Figure 8 illustrates the relationship between different oral gavage doses of- EDC and the resulting GC metabolite that would be produced in the liver for the B6C3Fl mouse. It can be seen that at low doses, because of first-order metabolism of EDC, the relationship is l:l. As the exposure dose is increased, the amount of GC produced is proportionally much 1 000 1 00 - - a) e_ 0 10 Q Ct a) ~ 1.0 CD ._ J 0.1 0.01 _ - ' _ _ '~ ' - _~' 7' i. ~ 10 100 1000 0.1 1.0 EDC Oral Exposure Dose (mg/kg) FIGURE 8 Relationship between EDC administered dose and the dose surrogate (the amount of liver GC metabolite) in the mouse. The curves represent the complex relationship predicted by the model. The dashed line is a direct extrapolation from the 150-mg/kg dose, assuming a 1:1 relationship. The curve predicted for the human is virtually superimposable on the mouse curve and is, therefore, not shown in the figure (see text for detailed explanation).

296 RICHARD W. D'SOUZA ET AL. greater than the administered dose. This nonlinearity is due to the fact that the oxidative pathway for EDC metabolism is saturated, resulting in increasing EDC levels and, therefore, increasing GC levels. At exagger- ated doses, as employed in bioassays, the capacity of the GSH pathway is also overwhelmed due to depleted GSH concentrations, and the amount of GC produced becomes proportionally lower than the administered dose. The dashed line (Figure 8) is a direct linear extrapolation of the GC metabolite from the 150-mg/kg dose to low doses, ignoring any nonlin- earities. It can be readily seen from this type of plot that by not taking into account the nonlinear metabolism of EDC, the risk of liver cancer to the mouse, based on results of a 150-mg/kg dose, can be easily overes- timated by an order of magnitude. The PB-PK model was also used to compute this relationship for the human. Liver GC in the human was predicted to be very similar to that in the mouse and is therefore not shown in Figure 8. Figure 9 depicts the relationship between administered dose and lung GC metabolite for the B6C3F1 mouse. A similar relationship is observed for the lung, except that the human GC exposure doses are about 2.5 times smaller. This suggests that by ignoring the nonlinear metabolism of EDC, lung cancer risk for the mouse can be overestimated by a factor of 10, while for the human it can be overestimated by 25-fold. Administered dose versus GC exposure plots were also constructed for inhalation ex- posure for the mouse and for oral and inhalation exposures for the rat, with similar findings (plots not shown here). PB-PK Model and Virtually Safe Doses As part of the conventional cancer risk assessment process, the number of tumors in each dose group of the bioassay is correlated with the ad- ministered dose. Statistical models are used to fit the data to extrapolate these findings to low exposure levels. These models estimate the dose required for 1 out of 1 million animals to develop a certain type of tumor during lifetime dosing. This dose is commonly called the virtually safe dose (VSD). The VSD for humans is obtained by dividing the mouse VSD by a factor of 12.7, which is the relative difference in body surface area between the mouse and man. Additional safety factors are also built into the risk assessments, including the use of confidence intervals and upper limits on estimated exposure of humans. The VSD is then compared with anticipated human exposures to prepare a risk assessment. The PB-PK model was used to generate GC amounts with different EDC exposure doses in the mouse, rat, and human (as in Figures 8 and 91. The multistage model (Crump, 1982) was then used to relate the number of tumors produced in each dose group with GC produced at that

PB-PK and EDC Risk Assessment 297 1 000 100 E - ._ o Q Cal 1.0 CD J 10 0.1 0.01 0.001 . '' /~ Mouse > Human ~ I I I I ~ I 0.1 1.0 10 100 1000 EDC Oral Exposure Dose (mg/kg) FIGURE 9 Relationship between EDC administered dose and lung GC metabolite for the mouse and human. The dashed line is a direct extrapolation from the 150-mg/kg dose, assuming a 1:1 relationship. tumor site. That is, instead of correlating tumors with the administered dose, or the external dose, the dose surrogate at the target site, or the internal dose, was employed. There was no additional justification for choosing the multistage model over several commonly used low-dose extrapolation models other than the fact that it is one of the models used most often by regulatory agencies. The VSD obtained by using this in- ternalized dose concept was converted to the corresponding EDC exposure concentration both for the mouse as well as the human employing Figures 8 and 9. That is, the extrapolation from mouse to human was made by using the PB-PK model and not a surface area correction, as would be done in a conventional risk assessment. The results of this analysis are shown in Table 2. For comparison purposes, VSDs were calculated both by the conventional method and by the PB-PK approach. Because regulatory agencies commonly use the lower confidence limit (LCL) of dose, the comparisons employed here are sim- ilarly restricted to the LCLs and not to the fitted lines. When comparing

298 RICHARD W. D'SOU~ ET AL. TABLE 2 VSDs Calculated for EDC by Employing the Conventional and PB-PK Approach Conventional PB-PK Human Ratio, Mouse EDC GC Mouse Human PB-PK/ Estimate EDC (~g/kgl Metabolite EDC EDC Conven- Organ Type (~g/kg/day) day) (~g/liter/day) (~g/kglday) (~g/kg/day) tional Liver Best estimate 3.1 0.24 150 240 230 960 95% LCL 0.46 0.036 2.3 3.8 3.6 100 Lung Best estimate 270 21 245 4,100 7,500 360 95% LCL 0.99 0.078 0.91 15 39 500 NOTE: In the conventional method tumors were correlated with EDC administered dose to obtain a VSD in the mouse (column 3). VSD for the human was obtained by dividing the mouse VSD by 12.7 (column 4). In the PB-PK approach tumors were correlated with the amount of GC metabolite produced in a target organ, and a VSD for this dose surrogate was obtained (column 5). This dose surrogate was then converted into the corresponding EDC administered dose for the mouse (column 6) and human (column 7) by employing the PB-PK model. Column 8 compares the results of these two approaches. the VSD for liver tumors by the two methods, it can be seen that the VSD for liver tumors obtained by using the pharmacokinetic approach is about 2 orders of magnitude greater than that obtained by the conventional method. VSD for lung tumors, however, shows about a 500-fold differ- ence. These data indicate that by not considering the pharmacokinetics of EDC in the mouse and human, conventional methods may overestimate cancer risk due to low-level EDC exposure by 2 or even 3 orders of magnitude. PB-PK Mode' and Route-of-Exposure Differences To address the route-of-exposure differences in the two EDC bioassays, the amount of GC metabolite produced in the liver and lung was computed for oral Savage doses of 75 and 150 mg/kg and compared with those that would be produced by a 7-h inhalation exposure at 150 ppm. These results are shown in Table 3. It is interesting to note that lower amounts of the dose surrogates are predicted at the maximum tolerated inhalation dose of 150 ppm, compared with the other two oral doses. This observation may be a possible reason that treatment-related tumors were not seen in the inhalation bioassay, but were observed both at 75 and 150 mg/kg for the oral bioassay. DISCUSSION PB-PK models are mechanistic models that attempt to account for the important processes that take place from the time a chemical is present

PB-PK and EDC Risk Assessment 299 TABLE 3 Amount of GC Metabolite Formed in the Liver and the Lung of the Rat Following Oral and Inhalation Exposures GC Metabolite Dose and Route Liver Lung Corn oil gavage 150 mg/kg 630 131 75 mg/kg 372 71 Inhalation 150 ppm (7 h) 230 64 at an absorption site for potential inhalation or oral or dermal uptake into the body, to the interaction that takes place between the chemical or its metabolites and some body tissue. The amount of chemical that is present at an absorption site may, therefore, have a very complex and indirect relationship with the amount of that chemical or its metabolites that is present at a target site and elicits a biological response. The chemical must first get absorbed; it may then bind to blood components that may limit its further distribution, it may undergo detoxification or may become more toxic, or it may be eliminated rapidly from the body. An under- standing of these processes, as well as the quantitation of the time course in the way that these processes operate and interact, may play a very important role in determining the potential risk from exposure to a certain concentration of a chemical. Once it is known whether the parent chemical or its metaboliteks) is responsible for a toxic or carcinogenic response, a PB-PK model can be developed to quantitate the amount of exposure and the time course of exposure to this moiety at a target site in laboratory animals. Using the model, this information can then be scaled up to the human to assess target organ exposure and, therefore, the risk of exposure to humans. The model does not provide insight into the mechanism of cancer, nor does it predict sensitivity of one target organ over another or one species over another. The model is simply a tool to quantitate target organ exposure to the relevant chemical species. One of the major problems in assessing toxicity or cancer risk is route- of-exposure extrapolations. Often, data are only available from one route of intake, while a risk assessment must be conducted for exposure via another route. Sometimes, like with the EDC and DCM bioassays, the results are positive from one route and negative via another route of exposure, which makes the results difficult to interpret. Unfortunately, there has been no reliable method so far for any extrapolation between routes of intake. PB-PK models, however, have the potential to quantitate the amount of chemical and its metabolites that is taken in from different exposure sites and that reaches a target tissue to elicit a response.

300 RICHARD W. D'SOU~ ET AL. Another factor that is not readily resolved without these kinds of models is species differences in the biodistnbution of chemicals. Although PB- PK models do not predict sensitivity of a biological response of one species over another, they can reliably quantitate target organ doses between species. Differences in the physiology of different species, like organ blood flow or pulmonary ventilation, are taken into account in these models. Anatomical differences, like organ sizes, are appropriately cor- rected when extrapolating results from one species to another. Also, phys- icochemical parameters of the chemical, like blood and tissue binding, are taken into consideration. The power and flexibility of these models lies in the fact Hat they are physiologically realistic and not mathematical "black boxes." As more information is gained about the behavior of a chemical, it can be added to the model without changing its basic structure, thus improving its predictability. The PB-PK model developed for EDC is in reality quite basic, because the model only computes gross tissue exposures of the relevant chemical species. Further work in this area should concentrate on cellular distr~- bution, as well as molecular dosimetry aspects like DNA binding and repair processes. All of these processes, as well as the processes that take place from the time that an unrepaired DNA adduct is formed to the time that a tumor is produced, are expected to be complex and potentially nonlinear with exposure dose. As information is gained about the rates of these processes, they can readily be added on to the PB-PK model. REFERENCES Adolph, E. H. 1949. Quantitative relationships in physiological constituents of mammals. Science 109:579-585. Anders, M. W., and J. C. Livesey. 1980. Metabolism of dihaloethanes. Pp. 331-341 in Ethylene Dioehloride: A Potential Health Risk?, Banbury Report 5, B. Ames, P. Infante, and R. Reitz, eds. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory. Andersen, M. E., H. J. Clewell III, M. L. Gargas, F. A. Smith, and R. H. Reitz. 1987. Physiologieally-based pharmacokineties and the risk assessment process for methylene chloride. Toxicol. Appl. Pharmacol. 87:185-205. Caster, W. O., J. Poncelet, A. B. Simon, and W. D. Armstrong. 1956. Tissue weights of rats. I. Normal values determined by dissection and chemical methods. Proe. Soe. Exp. Biol. Med. 91:122-126. Crump, K. S. 1982. An improved method for low dose carcinogenic risk assessment from animal data. J. Environ. Pathol. Toxicol. 5(2):675-684. Dedrick, R. L. 1973. Animal scale-up. J. Pharmacokinet. Biopharm. 1(5):435-461. D'Souza, R. W., W. R. Francis, and M. E. Andersen. 1986. A mathematical model for glutathione depletion and increased resynthesis following ethylene dichloride exposure. Pharm. Res. 3(5):137S. 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. Toxieol. Appl. Pharmacol. 82:211-223.

PB-PK and EDC Risk Assessment 301 Lindstedt, S. L., and W. A. Calder III. 1981. Body size, physiologic time, and longevity of homeothermal animals. Q. Rev. Biol. 56:1-16. Maltoni, C., L. Valgim~gli, and C. Scarnato. 1980. Long term carcinogenic bioassays of ethylene dichloride administered by inhalation to rats and mice. Pp. 3-29 in Ethylene Dichlonde: A Potential Health Risk?, Banbury Report 5, B. Ames, P. Infante, and R. Reitz, eds. Cold Spnng Harbor, N.Y.: Cold Spnng Harbor Laboratory. NCI (National Cancer Institute). 1978. Bioassay of 1,2-Oichloroethane for Possible Car- cinogenicity, NCI Carcinogenesis Technical Report, Series No. 55, DHEW Publication No. (NIH) 78-1361. Washington, D.C.: U.S. Government Printing Office. Reitz, R. H., T. R. Fox, J. C. Ramsey, J. F. Quast, P. W. Langvardt, and P. G. Watanabe. 1982. Pharmacokinetics and macromolecular interactions of ethylene dichloride in rats after inhalation or Savage. Toxicol. Appl. Pharmacol. 62:190-204. Sato, A., and T. Nakajima. 1979. Partition coefficients of some aromatic hydrocarbons and ketones in water, blood and oil. Br. J. Ind. Med. 36:231-234. Spreafico, F., E. Zuccato, M. Marcucci, M. Sironi, S. Paglialunga, M. Madonna, and E. Mussinin. 1980. Pharrnacokinetics of ethylene dichloride in rats treated by different routes and its long-term inhalatory toxicity. Pp. 107-129 in Ethylene Dichloride: A Potential Health Risk?, Banbury Report 5, B. Ames, P. Infante, and R. Reitz, eds. Cold Spnng Harbor, N.Y.: Cold Spnug Harbor Laboratory. Storer, R. D., and R. B. Conolly. 1985. An investigation of the role of microsomal oxidation metabolism in the in veto genotoxicity of 1,2-dichloroethane. Toxicol. Appl. Pharmacol. 77:36-46. White, R. D., A. J. Gandolf1, G. T. Bowden, and I. G. Sipes. 1983. Deutenum isotope effect on the metabolism and toxicity of 1,2-dibromoethane. Toxicol. Appl. Pharmacol. 69:170-178. APPENDIX: NOMENCLATURE GC Kf Kfee Kgsm Km Qc Of Ql Qlu Up Qr As vm~ Metabolite formed by conjugation of EDC and GSH. First-order rate constant for EDC metabolism via GSH path- way (ho) First-order rate constant for reaction of chloroacetaldehyde with everything else besides GSH (hat) First-order rate constant for reaction of GSH with chloro- acetaldehyde (hat) Michaelis-Menten constant for oxidation pathway (mg/liter) Cardiac output Fat blood flow Liver blood flow Lung blood flow Alveolar ventilation Blood flow to richly perfused tissues Blood flow to slowly perfused tissues Maximum capacity of oxidation pathway (mg/h)

Mathematical Modeling of Ozone Absorption in the Lower Respiratory Tract John H. Overton, Jr., Richard C. Graham, and Frederick ]. Miller INTRODUCTION Environmental toxicologists are confronted with the difficult task of interpreting the manifold results from human clinical, epidemiological, and animal studies on air pollutants and assessing their implications and relevance concerning pollutant levels to which man is exposed. To ac- complish this objective, mathematical models that predict respiratory tract dose patterns are essential. A purpose of these models is to extrapolate quantitatively effective pollutant concentrations between animals and man. In so doing, the existing animal data base will have increased direct relevance to national ambient air quality standards, and future studies can be designed to permit improved correlation with human effects. As part of a program to accomplish this objective, we have developed a mathematical dosimetry model that simulates the uptake and distribution of absorbed O3 in the respiratory tract (RT) of humans and laboratory animals (Miller et al., 19851. Originally, the dosimetry model was de- veloped to simulate the local absorption of O3 only in the lower respiratory tract (LRT). However, the present model can be used with other gaseous The research described in this article has been reviewed by the Health Effects Research Laboratory, 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. 302

Modeling Ozone Uptake in the Lower Respiratory Track 303 pollutants, such as NO2, if reactions with biological constituents can be modeled the same way as the O3 reactions. Absorption in the entire re- spiratory tract can be modeled also by adding appropriate pretracheal "generations" to a given LRT anatomical model. (Although the upper respiratory tract fURT] anatomical features can be accounted for correctly, the treatment of transport and chemical reactions, which is the same as that in the tracheobronchial region, may be inadequate; future studies are required.) The present dosimetry model includes use of LRT anatomy, ventilatory parameters, and varying airway dimensions during the breath- ing cycle. The model also accounts for the processes of transport in the lumen and air spaces and transport and irreversible chemical reactions in the liquid lining and in the underlying tissue and capillaries. Plans for future modifications to the model include (1) an improved treatment of IJRT transport and absorption; (2) taking into account me- tabolism in lung fluids and tissues, as well as a more general treatment of chemistry, including reversible reactions and reaction orders greater than 1; and (3) interfacing the RT model with pharmacokinetic models to allow a more complete description of the absorption, distribution, and metabolism of inhaled gases in the body than is presently available. This paper describes the present O3 dosimetry model and illustrates the use of the model with the results of two investigations of (1) the sensitivity of predictions to anatomical models of the guinea pig and rat, and (2) the effect of exercise on O3 absorption in man. M ETHODS Given the O3 concentration at the entrance (e.g., nose, mouth, or tra- chea), the model simulates, during one or more breathing periods, the transport and absorption of O3 in airways and alveolar air spaces of each generation or segment of a respiratory tract anatomical model. Species lung dimensions are taken into account by making use of an- atomical or airway models, as illustrated in Figure 1, a stylized diagram. In these models, airways of the LRT are represented by a sequence of sets of right circular cylinders. All cylinders or airways in a set that correspond to a particular generation are the same size. The URT region (if used) consists of pretracheal generations or sequential segments. For each generation, the simulation model requires the specification of the number of airways or segments and their diameters and lengths. Addi- tionally, for the pulmonary region, the alveolar volume and the surface area for each generation are needed. Similarly, the surface area of each URT segment is required. During the simulated breathing cycle, LRT linear dimensions are varied isotropically (linear dimensions are proportional to the one-third power of

304 JOH N H. OVERTON, ~ R., ET AL. L! l ~ . rT--I--T1 ' Or! ~ , · . r—- ~ ~ ~ ~ to ET--I--T1 ' Lt ~ i~ 1.——- - - — t rT--I--T1 L1 1 ! ~ 1-—-— - - — t rT I T1 |~ TRACH EA ~1 · URT ~ TRACHEOBRONCHIAL PULMONARY REGION /~* REGION - LRT - FIGURE 1 A stylized diagram of the type of respiratory tract anatomical models used with the dosimetry model. the ratio of the expanded volume of the generation to the volume of the generation at FRC) to account for changes in lung volume. The simulation of volume flow rate during a breath can be based on experimental data (e.g., plethysmograph studies) or on an assumed time-dependent function such as the sine function. The simulation model accounts for transport in the lumen and air spaces a well as for chemical reactions and transport in the liquid lining, tissue, and blood. Although O3 reacts with many different chemical species, for modeling purposes one effective reaction is assumed, O3 + HC ~ products where k is the effective second-order rate constant and HC represents the chemical components that react with O3. The processes of transport and chemical reactions are described in terms of partial differential equations (Figure 21. In the liquid lining, tissue, and blood compartments, where only the processes of molecular diffusion and chemical reactions are considered, the form of the equation is the same (see Miller et al., 1985, for justification of assumptions). In the lumen of the airway and in the alveolar air spaces, axial dispersion, axial convection, the loss of O3 to the liquid lining, and lung expansion and contraction are taken into account.

Modeling Ozone Uptake in the Lower Respiratory Track 305 I molecular chemical diffusion reactions BLOOD TISSUE DCa Cc- k2c[Hc~cCc L-~-Q-U-~-D---------------~-z2~~~~~~~~~~~~~~~~~~~~~~~~ AIR _ JW= kg(C~HCli) ,_ flux to liquid lining ac + u aC=D a2C S id it) t ;~> ~~, V axial axial wall expansion- convection dispersion loss contraction VC 1 "X FIGURE 2 Equations used to describe the transport and chemical reactions of O3 in the com- partments of the respiratory tract. Definition of terms includes: t = time; x = distance along airway path; z = distance within compartments; C = average cross-sectional O3 concentration; Cc = pointwise O3 concentration in compartment; D = effective dispersion coefficient of O3 in lumen; DC—molecular diffusion coefficient of O3 in a compartment; u = average air velocity in lumen; S = surface area; V = volume of airway; V = time rate of change of airway volume; Jw = radial flux to wall; kg = gas phase mass transfer coefficient; k2C = effective second-order chemical rate constants; [HC]C = concentration of components that react with 03; H = Henry's Law constant; Cal = liquid layer O3 concentration at the liquid-air interface. Given boundary conditions; initial conditions; and values for the phys- ical, chemical, and biological parameters, the equations can be solved to give simulated O3 dose and dose patterns. RESULTS AND DISCUSSION Effect of Anatomical Mode' on the Distribution of Simulated Absorbed O3 in the CRT Overton et al. (1987) investigated the effects of different anatomical models on the uptake and distribution of absorbed O3 in the LRT of guinea pigs and rats. Figure 3, which is based on results of the investigation, displays simulated net and tissue dose profiles for the two species when two anatomical models are used for each species. The net and tissue doses

Modeling Ozone Uptake in the Lower Respiratory Track 307 are plotted versus zone, order, or generation (depending on the anatomical model). Dose is defined as the quantity of O3 reacting with biological constituents per unit surface area of airway per unit time per tracheal O3 concentration. Net dose is the sum of the blood, tissue, and liquid com- partment doses. On each graph the species, source of the anatomical model, and ventilatory parameters are indicated. Regardless of the ventilatory parameters, the plots are typical for each anatomical model and illustrate qualitatively the differences and similar- ities resulting from the simulations using the different anatomical models. Generally, in the trachea the tissue dose is relatively low; the dose increases distally to a maximum in the vicinity of the first pulmonary airway unit and then rapidly decreases. Net dose is much larger in the trachea, how- ever, and decreases distally throughout the tracheobronchial region. In the pulmonary region, the net and tissue doses are essentially the same. The curves, based on the anatomical models of the guinea pig and rat of Kliment (1973) (Figures 3A and 3C), are similar. The difference be- tween the shapes of the two tissue dose curves at zones 6 and 7 is due to our choice of where the pulmonary or surfactant-lined region begins, not to differences in anatomical models. For the rat, this region begins at zone 7, and for the guinea pig it begins at zone 6. Quantitatively, the total and pulmonary uptakes for the Kliment guinea pig model (Figure 3A) are 92% and 85%, respectively. For the Kliment rat model (Figure 3C), these uptakes are 87% and 81%, respectively. Based on the results with the Kliment zone model, one might conclude that the uptake properties of rats and guinea pigs are very similar with respect to how the absorbed O3 iS distributed in the LRT. However, Figures 3B (guinea pig, Schreider and Hutchens, 1980) and 3D (rat; Yeh et al., 1979) suggest that the LRT deposition pattern for the two animals is dissimilar. For example, the net trachea dose is ~7 times the net dose of the first alveolated airways (order 10, Figure 3B) for the guinea pig. On the other hand, the net dose in the trachea of the rat model of Yeh et al. is less than the net dose of the first alveolated generation (generation 16, Figure 3D). Results with the Kliment (1973) rat and the Yeh et al. (1979) rat anatomical models were further investigated by Overton et al. (1987) by varying the breathing frequency and keeping the minute volume constant. The two anatomical models were found to result in significantly different LET total and pulmonary percent uptakes: 20%~0% (depending on breathing frequency) higher uptakes for the Kliment rat model than for the Yeh et al. rat model. Further, the Kliment rat was less sensitive to breathing frequency changes: The Kliment rat total and pulmonary percent uptakes decreased by ~6% when the breathing frequency was increased from 80 to 140 breaths per minute, whereas the percent uptakes of the Yeh et al. rat decreased by ~20% over the same frequency range.

308 JOH N H. OVERTON, J R., ET AL. Effect of Exercise on Predicted Uptake and Distribution of Absorbed O3 in the CRT of Man The effect of exercise on the uptake and the distribution of absorbed O3 in man was investigated by Miller et al. (1985~. Figure 4 and Table 1 illustrate the results. The anatomical model used was based on that of Weibel (1963) for man, but it was modified so that the lung volume at the end of exhalation was 2,650 ml (based on data obtained from William F. McDonnell, Environmental Protection Agency, Research Tnangle Park, N.C., personal communication; the data [FRC versus TLC] were extrap- olated to the TLC value that corresponded to the Weibel t1963] anatomical model). Lung expansion was not taken into account for these simulations. All calculations were based on specifying the concentration at the entrance to the trachea. The effect of exercise on predicted tissue dose is illustrated in Figure 4 for several levels of exercise. The level of exercise increases with increasing curve number, with curve 1 being for normal respiration and curve 4 for heavy exercise. The simulations show that exercise has very little effect on O3 tissue dose in the tracheobronchial region and a very large effect on this dose in the pulmonary region. In addition, the maximum tissue dose not only shifts distally from generation 17 (the first pulmonary generation) at rest to generation 20 at the heaviest exercise, but it also increases by a factor of 3.3. Table 1 illustrates the effect of exercise on the distribution of absorbed O3 in the regions and compartments of the LRT of man. For each level of exercise, Table 1 gives the percent uptalce of absorbed O3 in the regions TABLE 1 Simulated Effect of Exercise on Regional and Compartmental Uptake of O3 in Mana Tracheobronchial Total Region Uptake (%) Pulmonary VTC fc VEC Uptake Liquid Region Uptake (%)d IDb (liters) (BPM) (liters/min) (%) Lining Tissue Tissue Blood 1 0.5 15.0 7.5 89.2 18.4 8.3 60.3 2.0 2 1.0 15.0 15.0 95.8 (2.2)e 10.8 (1.2) 5.2 (1.3) 77.1 (2.6) 2.6 (2.6) 3 1.75 20.3 35.5 98.5 (5.2) 5.0 (1.3) 2.5 (1.4) 87.9 (6.9) 3.0 (7.1) 4 2.25 30.0 67.5 99.4(10.0) 2.8 (1.4) 1.4(1.5) 92.0(13.7) 3.1 (14.0) aUptake is based on the quantity of O3 inhaled at the trachea. bSimulation identification of curve number (see Figure 4). CVT = tidal volume;f = breaths per minute; VE = VT X f = minute volume. The liquid lining absorbs less than 0.3%. eValues in parentheses are the ratio of mass absorbed to the mass absorbed at normal respiration (for the first row of data, ID = 1).

Modeling Ozone Uptake in the Lower Respiratory Track 309 10-6 0 10-7 - ~ 10-8_ . _ 1 E - {51 - UJ o LLJ 10-9- lo-lo - l l l l l l l l l l l l l l 1 1 1 1 1 1 1 ~ _ ~ I~4 ~ . ~ ~ , ~ '. ~ .` t. ~ -'I'm '1, Anti' ,l\ 11:\ ~ . . \ 11 _ CURVE 1 2 4 VT f 0.5 15.0 1.0 1 5.0 1.75 20.3 2.25 30.0 VE 7.5 15.0 35.5 67.5 i_ 0 2 4 6 8 10 12 14 16 18 20 22 24 AIRWAY GENERATION TO _ I _ 1- FIGURE 4 Effect of O3 exercise on predicted tissue O3 dose in the lower respiratory tract of man (based on data from Miller et al., 1985). Dose is the quantity of O3 reacting with biological constituents per unit surface area of a zone, order, or generation per unit time per tracheal O3 concentration. The level of exercise and its corresponding tissue dose curve are indicated on the figure. VT = tidal volume (liters); I= breaths per minute; VE = minute volume; TB = tracheobronchial region; P = pulmonary region.

310 JOHN H. OVERTON, JR., ET AL. and compartments of the LRT. Also presented for each compartment and level of exercise is the ratio of the absorbed mass to the absorbed mass at normal respiration. The first column identifies the simulations with the curves in Figure 4. Although the percent total uptake does not change much (from ~89% to 99%), the percent uptake in the regions and compartments can vary considerably as the level of exercise increases. With increasing minute volume the tracheobronchial uptake decreases from ~27% to 4% whereas the pulmonary uptake increases from ~62% to 95%. A better indicator of the effect of exercise is the change in the mass of the absorbed O3. For example, the mass absorbed by the tracheobronchial region increases to ~1.4 times the amount absorbed during the same time of normal respi- ration, even though the percent uptake decreases. On the other hand, pulmonary absorption increases by a factor of about 14. A discussion of the significance of these results can be found in Miller et al. (19851. SU M MARY A brief description of a mathematical dosimetry model has been given. The model was developed to simulate the uptake and distribution of ir- reversible chemically reacting toxic gases, such as O3 and NO2, in the LRT of man and laboratory animals and takes into account species LRT anatomy and ventilatory characteristics, transport in the lumen of the airways and in alveolar air spaces, and transport and chemical reactions in the liquid lining and the underlying tissue and blood compartments. Model predictions were illustrated with the results of two investigations. In the first, two anatomical models for both the guinea pig and rat were used to investigate the effects of different anatomical models on predictions of O3 uptake. This investigation indicates the importance of reliable an- atomical models for a given animal. The second illustration of O3 dosi- metry model results was a discussion of an investigation into the effects of exercise on predicted O3 dose and distribution in the LRT of man. As exercise increased, the model predicted a moderate increase in percent total uptake, with very little effect on the amount of O3 absorbed by the tracheobronchial tissue. However, a pronounced increase in pulmonary absorbed O3 was predicted. REFERENCES Kliment, V. 1973. Similarity and dimensional analysis, evaluation of aerosol deposition in the lungs of laboratory animals and man. Folia Morphol. 21:59-64. 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

Modeling Ozone Uptake in the Lower Respiratory Track 311 ozone in the human lung to lower respiratory tract secretions and exercise. Toxicol. Appl. Pharmacol. 79:11-27. 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. Toxicol. Appl. Pharmacol. 88:418-432. Schreider, J. P., and J. O. Hutchens. 1980. Morphology of the guinea pig respiratory tract. Anat. Rec. 196:313-321. Weibel, E. R. 1963. Morphometry of the Human Lung. New York: Academic Press. Yeh, H. C., G. M. Schum, and M. T. Duggan. 1979. Anatomic models of the tracheo- bronchial and pulmonary region of the rat. Anal. Rec. 195:483-492.

Development of a Physiologically Based Pharmacokinetic Mode! for Multiday Inhalation of Carbon TetrachIoride Dennis ]. Paustenbach, Harvey ]. Clewell III, Michael L. Gargas, and Melvin E. Andersen I NTRODUCTION The extrapolation of animal data to humans has generally been consid- ered a nonquantitative exercise that is affected by numerous unknown biological factors. It has been recognized that this process, often called scaleup or extrapolation, should be influenced by differences in metab- olism, body size, and pharmacokinetics between the test species and man. In the past, because a mathematical method for describing these differences was not clearly defined, safety factors were used in an attempt to account for these differences between species (Calabrese, 1 983; Dourson and Stara, 1983; Gaylor, 1983; Krewski et al., 1984; National Research Council, 1975; Well, 1972; Zielhuis and van der Kreek, 1979 a,b). Physiologically based pharmacokinetic (PB-PK) modeling offers a promising approach for scaling up animal data to predict kinetic behavior in humans (Andersen, 1981 a,b; Dedrick and Bischoff, 1980; Himmelstein and Lutz, 19791. These models have been developed for styrene (Ramsey and Andersen, 1984), polybrominated biphenyls (Dedrick, 1973), tetra- chlorinated dibenzofurans (King et al., 1983), hepatic glutathione deple- tion (D'Souza et al., 1987), and methylene chloride (Andersen et al., 1987a). In general, these models quantitatively describe the kinetic be- havior of the parent molecule and its metabolites in the test species, and ideally, they successfully predict the metabolism and elimination of that molecule in humans (Clewell and Andersen, 19851. The underpinnings 312

PK Model for CC14 3 13 of physiological models have been reviewed by Ramsey and Andersen (19841. In this paper, we develop and apply a PB-PK model for carbon tetra- chloride (CCl44. Because CC14 has been of considerable interest in toxi- cology, there were sufficient animal and human data to construct and validate the model (Cornish, 1980; Recknagel, 19831. Specifically, the distribution and/or elimination of CC14 in the rat have been studied by Dambrauskas and Cornish (1970), Shimizu et al. (1973), Paustenbach et al. (1986a,b), and Uemitsu et al. (19851. The kinetic behavior in the mouse (Bergman, 1979), dog (Robbing, 1929), and monkey (McCollister et al., 1951) has also been evaluated. In those species studied, about 45% of the inhaled i4CCl4 was eliminated unchanged in the breath, 45% of the )4C activity was in the feces, and about 2% was in the breath as i4CO2. The remaining 14C activity, approximately 7%, was eliminated in the urine. Paustenbach et al. ~ 1986a,b) studied the effects of repeated exposure as well as unusually long periods of exposure. Because of the thoroughness of the data collection, these data were used to validate this model. In this paper, the PB-PK model was used to predict the behavior of CC14 in Sprague-Dawley rats for repeated exposure to schedules of 8 and 11.5 in/day, and the results were compared against actual laboratory data. The behavior of inhaled CC14 in humans was also predicted, and the results were compared with previously published data describing the elimination of CC14 in human volunteers. Lastly, the model was used to study the potential for day-to-day accumulation of CC14 in the adipose tissue of rats and humans following repeated inhalation exposure to 5 ppm CC14 (the current American Conference of Governmental Industrial Hygienists LAC- GIH] threshold limit value [TLV]) for periods of 8 and 12 in/day. METHODS Data Base Paustenbach et al. (1986a,b) published time course data obtained in male Sprague-Dawley rats following repeated inhalation exposure (average body weight, about 400 g) to CC14 at 100 ppm for 8 and 11.5 in/day for up to 10 of 14 days. At a number of serial sacrifices, the concentration of i4C activity in seven tissues and the blood was determined. Samples of fat, liver, adrenal, kidney, brain, heart, spleen, and blood were collected after 1, 2, 3, 4, 5, 7, and 10 days of exposure. A sufficient number of samples was collected to describe the elimination of ~4CCl4 and i4CO2 in the expired air and ~4C activity in the urine and feces (Paustenbach et al., 19831.

314 DENNIS j. PAUSTENBACH ET AL. Elimination In the study by Paustenbach et al. (1986a,b), the elimination of i4C activity was measured in the expired air, urine, and feces for up to 100 h following exposure, and various half-lives of elimination were deter- mined. Following 2 weeks of exposure to the 8-in/day schedule. i4CCl4 in the breath and ~4C activity in the feces comprised 45% and 48% of the total ~4C excreted. Following 2 weeks of exposure to the 11.5-h/day schedule, the values were 32% and 62%, indicating that repeated exposure to the longer schedule altered the elimination of CCl4. Regardless of the period of exposure, less than 8% of the inhaled i4CCl4 was excreted in the urine and less than 2% was exhaled in the breath as the SCOT metabolite (Paustenbach et al., 1986a). Humans Stewart et al. (1961) exposed human volunteers to 49 or 10 ppm for periods of 70 and 180 mins, respectively. The parent CCL4 measured in the expired air of these volunteers showed a biphasic elimination that was not unlike that collected in the rat (Paustenbach et al., 1986a), monkey (McCollister et al., 195 1), or mouse (Bergman, 19791. We estimated that the cat and ~ half-lives were approximately 50 and 240 min. respectively. Physiological Modeling The data of Paustenbach et al. (1986a,b), together with the results of gas-uptake data and handbook data on physiological parameters, were sufficient for developing a PB-PK model. The data of Stewart et al. ~ 1961 ~ provided the mechanism for validating the model's ability to scale up animal data to predict the response in humans. Ramsey and Andersen (1984) described a physiologically based model for examining the kinetic behavior of inhaled gases and vapors that are essentially nonirritating to the respiratory tract. In this approach, which is similar to that developed by Riggs (1970) and Fiserova-Bergerova and Holaday (1979), the body is lumped into tissue groups analogous to (1) highly perfused organs, excluding the liver; (2) muscle and skin; (3) fat; and (4) organs with a high capacity to metabolize the inhaled chemical. The physiological parameters of the metabolizing tissue groups are es- sentially those for the liver, although for some chemicals it could represent the kidney, lung, or skin. Blood flows and organ volumes are generally based on literature values for these parameters (Altman and Dittmer, 1979; Snyder, 19751. Organ partition coefficients and metabolic constants for each chemical are determined by appropriate experimentation (Gargas et al., 1986b). These various constants are used in the four mass-balance

PK Model for CC14 315 differential equations that describe the time-dependent changes of tissue concentrations in each of the compartments. Simulations of expected be- havior are conducted with a commercial software package (Advanced Continuous Simulations Package, Mitchell & Gauthier Associates, Inc., Cambridge, Mass.) on a CDC 6700 computer (Agin and Blau, 1982~. The PB-PK model for CC14 has been thoroughly described elsewhere (Paustenbach et al., in press). Metabolites The physiological model for volatile compounds used previously (Ram- sey and Andersen, 1984) was modified to account for the proposed me- tabolism and elimination model for CCl4. In the study by Paustenbach et al. (1986a), the chemical structures of the radioactive metabolites found in the urine and feces were not identified; therefore, kinetic constants were estimated for the model based on the laboratory data. In the schematic of the CC14 model (Figure 1), inhaled CC14 is either exhaled unchanged or converted to a metabolite. The model accounts for metabolism and ex- cretion of these metabolites via the urine, feces, and breath (as CO21. Using the actual Km and Vm~,,~ for the rat, as determined in gas-uptake studies, it was possible to predict the total quantity of metabolites that would be produced. In the metabolite portion of the model, total metabolite was apportioned to three pools: material excreted in the feces (the major portion of the metabolized CCl4), material excreted in the urine, and material eliminated as exhaled CO2. Elimination of these various metabolite pools was as- sumed to follow first-order behavior for all the material in the compart- ment. The rate constants for elimination of metabolites via urinary, fecal, and carbon dioxide metabolites were K2, K3, and Kit, respectively. The total amount of CC14 converted to a metabolite and subsequently eliminated is given the value of 1.0. This is equivalent to the i4C activity excreted in the urine (A2), as i4CO2 in the breath (Ai), and i4C in the feces (A31. To develop a model that could simulate the rat data of Paustenbach et al. (1986a) the standard physiological model of Ramsey and Andersen (1984) and Andersen et al. (1987b) was modified to account for four routes of elimination, repeated exposure, varying times of exposure, and the formation of a second or a third metabolite, which may have been eliminated in the feces and/or the urine. Mode' Parameters To develop any physiological model, data need to be either collected in the laboratory or obtained from the literature. Physiological descriptions of most test animals have been developed (Altman and Dittmer, 1979;

316 DENNIS J. PAUSTENBACH ET AL. 1 100 ppm I CC14 1 ~ 1 —| Lung ~ Fat Liver Muscle viscera Parent Chemical Physiological Model E] ,k4 etanolltes I ~ a1 ~ El 1~ my. ~ "ON ~ ~ .. 1 i | Urinary14C | 1 > | Exhaled '4CO k3 ~ | Fecal14C 1 delay I I Compartmental Based Model for Metabolites FIGURE 1 Schematic of the physiologically based pharmacokinetic (PB-PK) model used to describe the absorption, metabolism, and excretion of '4CC14 in exposed rats. a,, a2, and as describe the amounts of '4C in the elimination compartments; and K', K2, and K3, respectively, are the rate constants used to describe that elimination. In this model, transport between com- partments M2 and M, is accounted for by using the rate constant K4 and between M3 and M, by using Ks. This later transport accounts for the slow oxidative metabolism of metabolites to CO2. Dorato et al., 19831. The tissue volume, perfusion rate, ventilation rate, and percentage of body fat for any given body weight of test animal are estimated (based on literature values). Because tissue volumes do not always scale up in proportion to body weight (e.g., liver, kidney, and bone), some investigators have suggested that an alternative method in- volving uniformity of perfusion between species should be used (Box- enbaum, 1980, 19821. For humans, values for most anatomical and physiological parameters are also available (Snyder, 1975~. The partition coefficients for various tissues and the blood, as well as the kinetic con- stants for important metabolic pathways, also need to be determined or estimated for volatile chemicals; the latter are usually determined using gas-uptake procedures (Gargas et al., 1986a,b).

PK Model for CCI4 317 Partition Coefficients Partition coefficients for CC14 were determined by using a vial equili- bration technique (Sato and Nakajima, 1979) in which CC14 was added to a closed vial containing test liquid. Partitioning was determined by estimating the amount of disappearance from the headspace after 1 or 2 h of incubation at 37°C. These are reported in the legend to Figure 2. Biochemical Constants Carbon tetrachloride is metabolized via an oxidative pathway involving cytochrome P-450. Lipid peroxidation, presumably initiated by a free radical metabolite of CC14 (Butler, 1961), appears to be the most important factor in CCl4-induced liver injury (Glende, 1972; Recknagel and Glende, 19731. In this work, we estimated the values of Vma,: and Km using the gas-uptake simulation approach of Gargas et al. (1986b). This is consistent with that reported by Uemitsu (19861. Validation The model was validated by comparison with published data. Specif- ically, the model was used to predict the concentration of }4C activity in the exhaled breath, urine, feces, and adipose tissue, as well as SCOT in expired air following inhalation of 100 ppm for exposure periods of 8 and 11.5 in/day in the rat (Paustenbach et al., 1986a,b). In addition, a de- scription of the kinetic behavior following repeated exposures to both schedules, for up to 10 of 14 days, was attempted. The ability of this model to accurately describe the elimination of ~4CCl4 in the breath of humans exposed to 70 ppm for 10 min (Stewart et al., 1961) was also attempted. A more thorough description of the CC14 model is described elsewhere (Paustenbach et al., 19881. RESULTS Validity of the PB-PK Mode' We compared data describing the elimination of i4C activity in the Sprague-Dawley rat for all four excretory pathways with the kinetic be- havior (uptake, metabolism, and elimination) predicted by the CC14 model. The model (Figure 2A) accurately predicted accumulation of ~4C activity in the adipose tissue of test animals following repeated exposure, illus- trating that in the rat there is a rapid attainment of steady-state levels in fat following repeated exposure to 100 ppm for 8 in/day.

3 ~ ~ DEN N IS J. PAUSTEN BACH ET AL. 150 _ 120 3 c .o - - c 90 60 30 100 10 Q Q - o A 0 0 2 4 _ ° 0.1 _ \ tl5 _ \ a) o 0.01 0.001 I Rat Weight = 435 9. 6 8 10 12 14 Time (days) R \ \ 1 o 1 800 3600 5400 7200 Time Following Exposure (minutes) FIGURE 2 Comparison of the model predicted and actual concentrations of 14C activity (ex- pressed as CC14) in the adipose tissue of rats exposed to 100 ppm of '4CC14 for 8 hiday, S + 5 days (A). Concentration of CC14 in the exhaled breath of rats following exposure to 100 ppm of '4CC14 for l l.S in/day for 7 days (4 of 7, plus 3 of 7 days) (B). The solid line represents the computer-predicted concentration whereas the actual data are shown for the various time points ( + standard error). These simulations were based on the actual weight of the rat (as shown). Physiological parameters for the rat (0.42 kg) used in the PB-PK model were: cardiac output, 8.15 liters/h; alveolar ventilation, 7.9 liters/in. Tissue volumes (as percentage of total) were: liver, Who; fat, 25%; muscle, 7.5%; richly perfused organs, 57%. Blood flows (as percentage of total) were: liver, 25%; fat, 6%; muscle, lasso; richly perfused tissue, 54%. For the rat, the blood:air partition was 4.52, fat:blood was 79.4, liver:blood was 3.13, and muscle: blood was 2.00.

PK Model for CCI4 3~9 The model (Figure 2B) also predicted the elimination of parent i4CCl4 in the expired air of rats following repeated exposure to 100 ppm. This plot shows the actual versus model predicted elimination following the day 7 (4 + 3 days) of exposure to the 11.5-h/day schedule. These sim- ulations were especially sensitive to changes in body weight because CC14 is lipophilic and a larger animal (over 350 g) has a greater proportion of fat than does a smaller animal. This produces storage of larger amounts of CC14 in fat during exposure and prolonged metabolism of CC14 after cessation of exposure. Elimination via i4CO2 and Urine The behavior of the metabolites in the urine and breath (as i4CO2) was more difficult to describe than that of the parent compound in the breath. Interestingly, even with our simple description of the kinetics of the elim- ination of radioactivity, there was reasonably good agreement between the elimination of i4CO2 in the breath with that predicted by the model (Figure 3A). Similarly, adequate results were obtained for ]4C activity eliminated in the urine and feces (Figures 3B and C). In both cases, the biological data and model simulation accurately described exposures of 11.5 in/day for 7 of 14 days (4 of 7 days, followed by 3 of 7 days). One advantage of the PB-PK approach is that when biological behavior is not well predicted by the model (e.g., CO2 and urine), there is a suspicion that other important factors were omitted from the model de- scription. In this example, we assumed that CO2 elimination was fast and monoexponential. This model failed to accurately reproduce SCOT elim- ination because it did not account for the formation of CO2 from other compartments. Accordingly, the movement of radiolabel from the fecal and urinary compartments to the CO2 compartment was included to im- prove the description. K4 was the most important constant, and it accounted for the movement of long-lived fixed (bound) radioactivity in tissues and its release as CO2 over time. Using the published data of Stewart et al. (1961), we scaled the PB- PK model for the rat to humans. In their work, healthy volunteers were exposed to either 10 or 49 ppm for 10 or 70 min. respectively. Expired air was collected for 77 h after exposure. As shown (Figure 4A), there is good agreement between the actual human data and what was predicted by the model. This agreement is especially impressive in that the model was based on data obtained only in rats at exposures of 100 ppm and the use of a different duration of exposure.

320 DENNIS J. PAUSTENBACH ET AL. 1.0 O c' _ E 0 c 0 ¢', ~ 0.1 a! c c c-~} x - 0.01 1.0 ~~ c c' : ~ ° a, 0.1 c 0 _ ~ ~ Be c Be E ~ — Cat, ul x - 0.01 10 ._ . _ _ ~ c _ ~ 1.0 to of c 0 c _ ._ ~ LU x - 0.1 A . 0 1250 2500 3750 5000 Time Following Exposure (minutes) - B 0 1250 2500 3750 5000 Time Following Exposure (minutes) C 0 1250 2500 3750 5000 Time Following Exposure (minutes) FIGURE 3 Comparison of the actual versus predicted concentration of 14CO2 in the expired air of rats exposed to 100 ppm of CC14 for 8 in/day for 10 days (5 of 7, plus 5 of 7 days) (A), the actual versus predicted concentration of 14C activity eliminated in the urine (B), and 14C activity in the feces (C) of rats exposed to the same schedule. The model was initially unable to accurately describe the formation and elimination of CO2 for periods after 3,600 min. This was later corrected by the addition of K4 to the model.

PK Model for CC14 321 ~ 100 _ ~ _ Q _ - 6 . _ Q 10 111 . _ ° 1.0 o . _ - a) c: o 0.1 8.0 ^ 6.4 .° 4.8 - - a) o () 1 3.2 . . o A 49 ppm for 70 minutes Stewart et a/, 1961 0 100 200 300 400 Time (Minutes) - 500 600 ~ Humans - Time (Days) B 5 6 FIGURE 4 Comparison of the actual versus predicted concentration of CC14 in the expired breath of humans exposed to 49 ppm of CC14 for 70 min (A). The experimental data for humans was obtained from Stewart et al. (1961). This simulation of the predicted human response was based solely on the scale up of data collected in the Sprague-Dawley rat. The model was used to predict behavior for repeated inhalation exposure to 5 ppm of CC14 for both the rat and human (B). The PB-PK model predicted that rats exposed to 5 ppm for 8 in/day will not accumulate CC14 or its metabolites in adipose tissue with repeated exposure. In contrast, humans exposed to 5 ppm of CC14 for 8 in/day (the current ACGIH TLV concentration) would be expected to have day-to-day increases in adipose tissue levels. Physiological parameters for the human (70 kg) used in the PB-PK model were: cardiac output, 256 liters/h; alveolar ventilation, 254 liters/in. Tissue volumes (as percentage of total) were: liver, 4%; fat, 20%; muscle, 62%; richly perfused tissue, 5%. Blood flows (as percentage of total) were: liver, 25%; fat, 6%; muscle, 18%; and richly perfused organs, Slab. For humans, the blood:air partition was 2.7.

322 DENNIS J. PAUSTENBACH ET AL. A validated PB-PK model can also be used to examine the behavior of a chemical in humans under a number of different exposure scenarios (Figure 4B). In this example, the model shows that rats repeatedly exposed to 5 ppm would not demonstrate CC14 accumulation with repeated ex- posure. In contrast, when humans are repeatedly exposed to S ppm of CC14 (the current ACGIH-TLV), there is a day-to-day accumulation in adipose tissue. Due to degassing after a weekend away from exposure, it is not anticipated that this degree of accumulation poses a measurable increased risk to exposed persons. Additional development of PB-PK models for volatile chemicals should be very useful in allowing us to better understand the various margins of safety presently implicit in the more than 1,000 occupational exposure limits that have been established for workers (Cook, 1987; WHO, 1986~. DISCUSSION A PB-PK model that accurately describes the behavior of CC14 in rats was developed by using fundamental biological data and was validated by comparison with rat and human kinetic data. The model was capable of describing CC14 pharmacokinetics for a variety of exposure regimens and for concentrations between 10 and 100 ppm. We obtained an excellent description of the elimination of parent CC14 in the exhaled breath and were able to predict the concentration of CC14 in the adipose tissue of rats at any time following exposure to either an 8- or 11.5-h/day dosing reg- imen. By making some basic assumptions, we also predicted the time course of formation and elimination of CC14 metabolites in the urine, the breath (CO2), and the feces. The usefulness of applying the PB-PK ap- proach to solving industrial hygiene questions, such as the modification of exposure limits for long work shifts, has been discussed (Paustenbach, 19851. It was difficult to simulate adequately the elimination of ~4C activity in feces and CO2 by using a model involving just one metabolite. It appears likely that two or more metabolites are formed and then eliminated in the feces and that perhaps this is responsible for the latent production of SCOT seen in the exhaled breath. Whenever this occurs, plausible biological reasons for these deviations can be developed; these can be incorporated into the model (Figure 1~. We have found that one benefit of developing a PB-PK model is to better understand the potential mechanism of action of the test chemical, and from this one can identify the optimal laboratory experiments to be used to examine the kinetics of toxicity of a particular chemical. While there is always the danger of overinterpreting disposition studies which monitor only radioactivity, the combination of information on parent

PK Model for CC14 323 chemical, particular metabolites (CHCl3 and CO2), and total radioactivity, allow some conclusions to be made regarding the pathways of metabolism of CC14 and the persistence of bound metabolites derived from this chem- ical. The Vma,` of CC14 is very low, about 0.28 mg/h (1.81 ~mol/h) in a 300-g rat. In contrast, the Vm`= for trichloroethylene is 37.6 ~mol/h (An- dersen et al., 1987c). With CC14 the vast majority of the metabolized material becomes bound in compartments that lead to slow elimination in the feces and urine. Thus, al plus a2 is 93.5% of total metabolism. Only 6% is metabolized to CO2 with a rapid elimination rate constant. Paus- tenbach et al. (1986) reported that some 3% of the exhaled material trapped on charcoal was CHC13 and not CCl4. This amount of radioactivity rep- resents only 2-3% of the total metabolized dose. Thus, the sum of CO2 and CHC13 is somewhat less than 10% of the amount metabolized. The CO2 produced and eliminated with a rapid rate process is most likely a secondary product of CHC13 metabolism. Lipid peroxidation initiated by CC14 is generally regarded to represent abstraction of a proton from polyunsaturated fatty acids. This labelizes the fatty acid structure to further oxidation and produces CHCl3. From the result of modeling CC14 metabolism, this type of peroxidative process is clearly only a minor pathway for CC14 metabolism, accounting, in this study at 100 ppm, for only 0.028 mg/h of total CC14 metabolism. Importantly, we found that the PB-PK model could predict the elimi- nation~of unchanged CC14 in the breath of exposed humans based solely on the partition data collected in human blood, pharmacokinetic data collected in rats, and typical scaleup of parameters. The ability to predict the behavior of chemicals at various concentrations through models that account for the metabolism and pharmacokinetics of a chemical in lower mammalian species should be a powerful tool for assessing the human risk of exposure to these agents. Already, these models have been used to estimate better occupational exposure limits for persons who are exposed to airborne chemicals during shifts as long as 10, 12, or 14 in/day (Andersen et al., 1987b). PB-PK models have also been used to improve the quan- titative risk assessment process for methylene chloride (Andersen et al., 1987a), and this appears to be an important new use for this approach. In this model, we were successful in describing metabolism by assuming a single, time-independent, oxidative pathway for CC14 metabolism. It is well-documented, however, that CC14 can cause destruction of the cyto- chrome P-450 enzyme system under various in vivo exposure conditions. The success of this present model may be fortuitous, and there may well be time-dependent changes in Vma,~ during exposure of rats to a concen- tration of 100 ppm. In this case, the basal Vmax level would be larger than that estimated in this study, and the residual Vmax level at the end of exposure would be lower than the 0.4 mg/h/kg estimated by gas-uptake

324 DENNIS J. PAUSTENBACH ET AL. methods. Andersen et al. (1987a) have described techniques for examining suicide enzyme inhibition by reactive metabolites in viva by using venous exposure conditions in the gas-uptake chamber. One such approach is coexposure to binary mixtures of structurally related chemicals, where one chemical is very well metabolized and the other is suspected of causing destruction of the major oxidative enzymes responsible for the metabolism of the two chemicals. The chemical sus- pected of inhibiting metabolism is added first, and after a variable preex- posure period, He second well-metabolized chemical is added to He chamber atmosphere. The experiment then consists of observing the loss of me- tabolizing capacity for the second substrate. These experiments are now under way in our laboratory in Dayton, Ohio, with CC14 and CHCl3. It would be premature to use this present model for risk assessment calcu- lations until these types of inhibition experiments are completed and fully interpreted. REFERENCES Agin, G. L., and G. E. Blau. 1982. Application of DACSL (Dow Advanced Continuous Simulation Language) to the design and analysis of chemical reactors. AICHE Sym- posium Series No. 214. 78:108-118. Altman, P. L., and D. S. Dittmer, eds. 1979. P. 1582 in Biology Data Book, 2nd ed. Vol. II. Bethesda, Md.: Federation of American Societies for Experimental Biology. American Conference of Governmental Industrial Hygienists, Threshold Limits Committee. 1986. Threshold limit values for chemical substances in the workroom air adopted by ACGIH for 1986-87. Cincinnati, Ohio: American Conference of Governmental Industrial Hygienists. Andersen, M. E. 198 la. A physiologically based toxicokinetic description of the metabolism of inhaled gases and vapors: Analysis at steady state. Toxicol. Appl. Pharmacol. 60:509- 526. Andersen, M. E. 1981b. Pharmacokinetics of inhaled gases and vapors. Neurobehav. Toxicol. Teratol. 3:383-389. Andersen, M. E., H. J. Clewell III, M. L. Gargas, F. A. Smith, and R. H. Reitz. 1987a. Physiologically-based pharmacokinetics and the risk assessment process for methylene chloride. Toxicol. Appl. Pharmacol. 87:185-205. Andersen, M. E., M. G. MacNaughton, H. J. Clewell III, and D. J. Paustenbach. 1987b. Adjusting exposure limits for long and short exposure periods using a physiological pharmacokinetic model. Am. Ind. Hyg. Assoc. J. 48:335-341. Bergman, K. 1979. Whole-body autoradiography and allied tracer techniques in distribution and elimination studies of some organic solvents (including carbon tetrachloride). Scand. J. Work Environ. Health 59: 1-263. Boxenbaum, H. 1980. Interspecies variation in liver weight, hepatic blood flow, and antipyrine intrinsic clearance: Extrapolation of data to benzodiazepines and phenytoin. J. Pharmacokinet. Biopharm. 8:165-176. Boxenbaum, H. 1982. Interspecies scaling, allometry, physiological time, and the ground plan of pharmacokinetics. J. Pharmacokinet. Biopharm. 10:201-227. Butler, T. C. 1961. Reduction of carbon tetrachloride in-vivo and reduction of carbon tetrachloride and chloroform in-vitro by tissues and tissue constituents. J. Pharmacol. Exp. Ther. 134:311-319.

PK Model for CC14 325 Calabrese, E. J. 1983. Principles of Animal Extrapolation. New York: John Wiley & Sons. Clewell, H. J., III, and M. E. Andersen. 1985. Use of physiologic pharmacokinetics in risk assessment. J. Toxicol. Ind. Health 1:111-122. Cook, W. A. 1987. A Compendium of World-Wide Occupational Exposure Limits. Akron, Ohio: American Industrial Hygiene Association. Cornish, H. Hi 1980. Solvents and vapors. In Casarett and Doull's Toxicology, 2nd ea., J. Doull, C. D. Klaassen, and M. O. Amdur, eds. New York: Macmillan. Dambrauskas, T., and H. H. Cornish. 1970. Effect of pretreatment of rats with carbon tetrachloride on tolerance development. Toxicol. Appl. Pharmacol. 17:83-97. Dedrick, R. L. 1973. Animal scale-up. J. Pharmacokinet. Biopharm. 1:435-461. Dedrick, R. L., and K. B. Bischoff. 1980. Species similarities in pharmacokinetics. Fed. Proc. 39:54-59. Dorato, M. A., K. H. Carlson, and D. L. Copple. 1983. Pulmonary mechanics in conscious Fischer 344 rats: Multiple evaluations using nonsurgical techniques. Toxicol. Appl. Pharmacol. 68:344-353. Dourson, M. L., and J. F. Stara. 1983. Regulatory history and experimental support of uncertainty (safety) factors. Reg. Toxicol. Pharm. 3:224-238. D'Souza, R. W., W. R. Francis, and M. E. Andersen. 1987. A physiologic model for tissue, glutathione depletion and increased resynthesis following ethylene dichloride exposure. J. Pharmacol. Exp. Ther. Fiserova-Bergerova, V., and D. A. Holaday. 1979. Uptake and clearance of inhalation anesthetics in man. Drug Metab. Rev. 9(1):43-60. Gargas, M. L., H. J. Clewell III, and M. E. Andersen. 1986a. Metabolism of inhaled dihalomethanes in vivo: Differentiation of kinetic constants for two independent pathways. Toxicol. Appl. Pharmacol. 82:211-223. Gargas, M. L., H. J. Clewell III, and M. E. Andersen. 1986b. A physiologically-based simulation approach for determining metabolic constants from gas uptake data. Toxicol. Appl. Pharmacol. 86:341-352. Gaylor, D. W. 1983. The use of safety factors for controlling risk. J. Toxicol. Environ. Health. 11:329-336. Glende, E. A., Jr. 1972. On the mechanism of carbon tetrachloride toxicity coincidence of loss of drug metabolizing activity with perioxidation of microsomal lipid. Biochem. Pharmacol. 21:2131-2138. Himmelstein, K. J., and R. J. Lutz. 1979. A review of the ap,plications of physiologically based pharrnacokinetic modeling. J. Pharmacokinet. Biopharm. 7:127-145. King, F. G., R. L. Dedrick, J. M. Collins, H. B. Matthews, and L. S. Birnbaum. 1983. A physiological model for the pharmacokinetics of 2,3,7,8-tetrachlorodibenzofuran in several species. Toxicol. Appl. Pharmacol. 67:390-400. Krewski, D., C. Brown, and D. Murdoch. 1984. Determining "safe" levels of exposure: Safety factors or mathematical models. Fund. Appl. Toxicol. 4:S383-S384. McCollister, D. D., W. H. Beamer, G. J. Atchison, and H. C. Spencer. 1951. Absorption, distribution and elimination of radioactive carbon tetrachloride by monkeys upon exposure to low concentrations of vapor. J. Pharmacol. Exp. Ther. 102:112-120. National Research Council. 1975. Principles for Evaluating Chemicals in the Environment. Washington, D.C.: National Academy of Sciences. 454 pp. Paustenbach, D. J. 1985. Occupational exposure limits, pharmacokinetics, and unusual work schedules. Pp. 111-277 in Patty's Industrial Hygiene and Toxicology, Vol. IIIA, L. J. Cralley and L. V. Cralley, eds. New York: John Wiley & Sons. Paustenbach, D. J., G. P. Carlson, G. S. Born, J. E. Christian, and J. E. Rausch. 1983. A dynamic closed-loop recirculating inhalation chamber for conducting pharmacokinetic and short-term toxicity studies. Fund. Appl. Toxicol. 21:128-132.

326 DENNIS J. PAUSTENBACH ET AL. Paustenbach, D. J., G. P. Carlson, J. E. Christian, and G. S. Born. 1986a. A comparative study of the pharmacokinetics of carbon tetrachloride in the rat following repeated in- halation exposures of 8 and 11.5 Friday. Fund. Appl. Toxicol. 6:484-497. Paustenbach, D. J., J. E. Christian, G. P. Carlson, and G. S. Born. 1986b. The effect of an 11.5 hr/day exposure schedule on the distribution and toxicity of inhaled carbon tetrachloride in the rat. Fund. Appl. Toxicol. 6:472-483. Paustenbach, D. J., M. E. Andersen, H. J. Clewell III, and M. L. Gargas. 1988, submitted. A physiologically-based pharmacokinetic model for inhaled carbon tetrachloride in the rat. Toxicol. Appl. Pharmacol. Ramsey, J. C., and M. E. Andersen. 1984. A physiologically based description of the inhalation pharmacokinetics of styrene in rats and man. Toxicol. Appl. Pharmacol. 73: 159-175. Recknagel, R. O. 1983. A new direction in the story of carbon tetrachloride hepatotoxicity. Life Sci. 33:401-408. Recknagel, R. O., and E. N. Glende. 1973. Carbon tetrachloride hepatotoxicity: An ex- ample of lethal cleavage. Crit. Rev. Toxicol. 2:263-297. Riggs, M. 1970. Chapt. 13 in The Mathematical Approach to Physiological Problems. Cambridge, Mass.: MIT Press. Robbins, B. H. 1929. The absorption, distribution and excretion of carbon tetrachloride in dogs under various conditions. J. Pharmacol. 37:203-216. Sato, A., and T. Nakajima. 1979. Partition coefficients of some aromatic hydrocarbons and ketones in water, blood and oil. Br. J. Ind. Med. 36:231-234. Shimizu, Y., C. Nagase, and K. Kawai. 1973. Accumulation and toxicity of carbon tetrachloride after repeated inhalation in rats. Ind. Health 11:48-54. Snyder, W. S. 1975. Report of the Task Group on Reference Man. International Commission on Radiological Protection, Report No. 23. Elmsford, N.Y.: Pergamon. Stewart, R. D., H. H. Gay, D. S. Erley, C. L. Hake, and J. E. Peterson. 1961. Human exposure to carbon tetrachloride vapor relationship of expired air concentration to exposure and toxicity. J. Occup. Med. 3:586-590. Uemitsu, N. 1986. Inhalation pharmacokinetics of carbon tetrachloride in rats based on arterial blood:inhaled air concentration ratios. Toxicol. Appl. Pharmacol. 83:20-29. Uemitsu, N., Y. Minobe, and H. Nakoyoshi. 1985. Concentration-time-response relation- ship under conditions of single inhalation of carbon tetrachloride. Toxicol. Appl. Phar- macol. 77:260-266. Weil, C. S. 1972. Statistics versus safety factors and scientific judgement in the evaluation of safety for man. Toxicol. Appl. Pharmacol. 21:454-463. World Health Organization (WHO). 1987. Occupational exposure limits. Geneva, Swit- zerland. Zielhuis, R. L., and F. van der Kreek. 1979a. Calculations of a safety factor in setting health based permissible levels for occupational exposure. A proposal I. Int. Arch. Occup. Environ. Health 42: 191 -201. Zielhuis, R. L., and F. W. van der Kreek. 1979b. Calculations of a safety factor in setting health based permissible levels for occupational exposure. A proposal II, comparison of extrapolated and published permissible levels. Int. Arch. Occup. Environ. Health 42:203- 215.

The Delivered/Administered Dose Relationship and Its Impact on Formaldehyde Risk Estimates Thomas B. Starr LIMITATIONS OF CONVENTIONAL LOW-DOSE RISK EXTRAPOLATION It is now well-established that the typical chronic bioassay lacks suf- ficient power to discriminate among the different mathematical dose-re- sponse models that are commonly employed for low-dose extrapolation purposes (Krewski et al., 1983~. Indeed, even the so-called megamouse or EDGE study of 2-acetylaminofluorine, in which over 24,000 mice were used, was inadequate to this task (Brown and Hoel, 1983a, b; SOT Task Force, 19811. Generally, several dose-response models each provide an adequate fit to tumor incidence data in the observable response range, and yet the predicted risks at exposure levels below this range differ from one another by orders of magnitude. Low-dose extrapolations of risk employ- ing data from the Chemical Industry Institute of Toxicology formaldehyde bioassay (Kerns et al., 1983) provide an excellent illustration of this phenomenon (Starr and Buck, 19841. It is also known that optimization of the experimental design of chronic bioassays (in terms of the number of treatment groups, their relative size, and their placement relative to the maximum tolerated dose) so as to minimize the uncertainty in predicted risks at low exposure levels does not significantly improve the situation (Portier and Hoel, 19831. In es- sence, knowledge of the risk at high exposure levels is by itself insufficient to predict accurately the risks at low exposure levels. 327

328 THOMAS B. STARR Also clear is the fact that the various mathematical models of carci- nogenesis utilized for low-dose extrapolation were conceptualized initially and formulated in terms of interactions between the biologically active forms of chemicals agents (i.e., the pharmacokinetic delivered dose) and cellular macromolecules in target tissues (Brown, 1976; Cornfield, 1977; Crump, 1979; Hoel et al., 19831. These models lack the structure necessary to characterize the many physiologic and pharmacokinetic factors that are likely to govern the complex relationship between the dose delivered to various target tissues and the dose administered to whole animals. Fur- thermore, information on internal measures of exposure is not routinely available. Rather, the externally administered dose, e.g., milligrams per kilogram in feed, micrograms per liter in water, or parts per million in chamber air, is usually the only measure of exposure provided by a bioas- say. Consequently, a critical assumption is made when low-dose risk ex- trapolation is performed in the absence of data regarding internal measures of exposure. This assumption is that the dose administered in a bioassay is a valid linear proxy for the biologically active dose delivered to specific target tissues (EPA, 1986; Starr and Buck, 19841. This is equivalent to assuming that the kinetics of distribution and disposition of chemicals within whole animals are entirely and exactly linear. It is additionally assumed that a simple scaling of the administered dose (e.g., on a relative body weight or body surface area basis, depending on the exposure route) is all that is necessary to convert predicted risks for one species to those for another (EPA, 19861. Indeed, for exposure via inhalation, different species exposed to the same airborne concentration of a chemical for the same fraction of their lifetimes are presumed to experience the same risk (EPA, 19861. EVIDENCE THAT LINEAR PROPORTIONALITY DOES NOT HOLD FOR FORMALDEHYDE Subchronic mechanistic studies of formaldehyde have provided strong direct evidence that the dose delivered to the DNA of replicating cells in the respiratory epithelium of the rat nasal cavity is nonlinearly related to the airborne formaldehyde concentration (Casanova-Schmitz and Heck, 1985; Casanova-Schmitz et al., 19841. Specifically, as is shown in Figure 1, significantly less formaldehyde (by approximately a factor of 3) is covalently bound to respiratory mucosal DNA at low airborne concentra- tions than is predicted by downward linear extrapolation from the amounts of such binding observed at high concentrations. Related studies have demonstrated that exposure of rats to formaldehyde via inhalation induces the respiratory depression reflex (Chang et al.,

Cancer Risk Versus Delivered Dose 329 0.05 0.04 ~ 0.03 C] He m O 0.02 0.01 _ k - i1 1 , 1 1 1 0.3 2 6 10 15 HCHO AIR CONC, PPM FIGURE 1 Amount of formaldehyde (HCHO) covalently bound (COV BND) to respiratory mucosal DNA of Fischer 344 rats normalized by airborne concentration (i.e., expressed as nanomoles covalently bound/mg of DNA/ppm of HCHO) plotted versus the airborne HCHO concentration to which the animals were exposed. Adapted from Casanova-Schmitz et al. (1984). 1983), an inhibition of mucociliary clearance (Morgan, 1983; Morgan et al., 1983) and intracellular metabolism of formaldehyde (Casanova-Schmitz and Heck, 1985), and stimulation of cell proliferation (Chang et al., 1983; Swenberg et al., 1983, 1986), all as nonlinear functions of the airborne formaldehyde concentration. Because each of these phenomena appears to be an important controlling factor in the relationship between admin- istered and delivered doses (Starr and Gibson, 1985), the nonlinear re- lationship between covalent binding to UNA and airborne concentration is not unexpected. The respiratory depression reflex mediates the inhaled dose. Mucoci- liary clearance mediates the fraction of the inhaled dose that penetrates the mucus layer covering underlying epithelial cells in the nasal cavity. Intracellular metabolism mediates the fraction of formaldehyde entering these cells that remains free to bind with cellular macromolecules, in- cluding DNA. Important in this regard is the fact that formaldehyde is an

330 THOMAS B. STARR essential biochemical that is normally present in all living cells. it is thus not surprising that efficient metabolic pathways exist for its detoxication. Finally, the rate of cell replication mediates the fraction of DNA that is single stranded, and it is known that formaldehyde binds covalently only to single-stranded DNA (Lukashin et al., 1976; von Hippel and Wong, 1971). COMPARISON OF RISK ESTIMATES: ADMINISTERED VERSUS DELIVERED DOSE The implications of these studies for risk assessment have been explored and elucidated (Starr and Buck, 19841. Use of the amount of formaldehyde covalently bound to respiratory mucosal DNA (rather than airborne con- centration) as the measure of exposure leads to lower point and upper bound estimates of risk at low doses, irrespective of the mathematical dose-response model employed. For illustration, point and upper bound estimates corresponding to an airborne concentration of 1 ppm of for- maldehyde are displayed in Tables 1 and 2. It is apparent that risk as- sessments that do not explicitly incorporate the data regarding covalent binding of formaldehyde to target tissue DNA are likely to overestimate (at least relatively) the cancer risk associated with formaldehyde exposure. QUESTIONS STILL TO BE RESOLVED Related mechanistic studies have also provided strong evidence that dramatic interspecies differences in the inhaled dose are likely to exist even when the species are exposed identically to the same airborne for- maldehyde concentration (Chang et al., 19831. Such differences would presumably also be manifest in the resulting delivered dose, and the re- TABLE 1 Maximum Likelihood Risk Estimatesa for 1.0 ppm of Airborne Formaldehyde Dose Measure Probit Logit Weibull Multistage ADM 2.65(—ll)b 2.87( - 6) 5.94( - 6) 2.51( - 4) DEL 4.00( - 20) 2.15( - 8) 7.13( - 8) 4.70( - 6) Reduction factor 6.60( + 8) 133.0 83.4 53.4 aPredicted excess nasal cancer risk for rats exposed to 1 ppm of formaldehyde, 6 in/day, 5 days/ week, for up to 24 months. bNumbers in parentheses represent powers of 10.

Cancer Risk Versus Delivered Dose 33 ~ TABLE 2 Upper 95% Confidence Bounds on Risk for 1.0 ppm of Airborne Formaldehyde Dose Measure Probit ADM 2.58(—1O)a 1.24( - 5) DEL 7.09( -19) 1 .22( - 7) Logit Weibull Multistage 2.54( - 5) 3.98( - 7) Reduction factor 3.63( + 8) 101.8 63.8 1.80( - 3) 6.24( - 4) 2.9 aNumbers in parentheses represent powers of 10. markable 50-fold disparity in tumor incidence in rats and mice identically exposed to 14.3 ppm of formaldehyde in the CIIT bioassay (Kerns et al., 1983) may be attributable to this phenomenon. Clearly, interspecies differences in anatomy and physiology, metabo- lism, and the rates of cell proliferation and repair of DNA damage may each play a critical role in the accurate assessment of the risk to humans from formaldehyde exposure. If the cancer risk in two species as similar as rats and mice can differ by a factor of 50, even though they are identically exposed to the same airborne concentration, can we have much confidence in the extrapolation of risks from rodents to the very different human species under very different exposure conditions? Additional research is clearly desirable on many aspects of this problem. For example, the above-mentioned phenomena can be studied intensively in nonhuman primates so that the potential effects (on ONA binding and tumor incidence) of differences in anatomy and physiology of the upper respiratory tract can be established. These phenomena also need to be studied following more extended periods of exposure to establish whether or not the dramatic nonlinear effects observed in the short term are rep- resentative of effects that occur at later times during the course of a standard 2-year rodent bioassay. Identification of the specific DNA adducts that form as a result of interaction with formaldehyde can also be pursued. Studies that clarify the potential role of these adducts as causal agents in the carcinogenic process must also be conducted. Finally, mechanistic mathematical models that describe quantitatively dependences of these phenomena on airborne concentration must also be developed. It has been argued by some that pharmacokinetic data cannot be em- ployed in regulatory assessments of carcinogenic risk until uncertainties such as those mentioned above are resolved (cf., Cohn et al., 1985~. However, the opposite view, namely, that such data cannot in good con- science be ignored, seems reasonable. At the very least, the comparison of risk estimates based upon pharmacokinetic data with similar estimates

332 THOMAS B. STARR derived without such data can expose some of the real uncertainties that would otherwise remain hidden in the many unverified assumptions that are employed in the quantitative risk assessment process. REFERENCES Brown, C. C. 15376. Mathematical aspects of dose-response studies in carcinogenesis- the concept of thresholds. Oncology 33:62-65. Brown, K. G., and D. G. Hoel. 1983a. Modeling time-to-tumor data: Analysis of the EDIT study. Fund. Appl. Toxicol. 3:458-469. Brown, K. G., and D. G. Hoel. 1983b. Multistage prediction of cancer in serially dosed animals with application to the EDIT study. Fund. Appl. Toxicol. 3:470-477. Casanova-Schmitz, M., and H. Heck. 1985. DNA-protein cross-linking induced by for- maldehyde (FA) in the rat respiratory mucosa: Dependence on FA concentration in normal rats and in rats depleted of glutathione (GSH). Toxicologist 5:128. Casanova-Schmitz, M, T. B. Starr, and H. Heck. 1984. Differentiation between metabolic incorporation and covalent binding in the labeling of macromolecules in the rat nasal mucosa and bone marrow by inhaled ['4C]- and [3H] formaldehyde. Toxicol. Appl. Pharmacol. 76:26-44. Chang, J. C. F., E. A. Gross, J. A. Swenberg, and C. S. Barrow. 1983. Nasal cavity deposition, histopathology, and cell proliferation after single or repeated formaldehyde exposure in B6C3F1 mice and F-344 rats. Toxicol. Appl. Pharmacol. 68:161-176. Cohn, M. S., F. J. DiCarlo, and A. Turturro. 1985. Letter. Toxicol. Appl. Pharmacol. 77:365-368. Cornfield, J. 1977. Carcinogenic risk assessment. Science 198:693-699. Crump, K. S. 1979. Dose response problems in carcinogenesis. Biometrics 35:157-167. EPA. 1986. (U.S. Environmental Protection Agency). Guidelines for carcinogen risk as- sessment. Fed. Regist. 51:33992-34003. Hoel, D. G., N. L. Kaplan, and M. W. Anderson. 1983. Implication of nonlinear kinetics on risk estimation in carcinogenesis. Science 219:1032-1037. Kerns, W. D., K. L. Pavkov, D. J. Donofrio, E. J. Gralla, and J. A. Swenberg. 1983. Carcinogenicity of formaldehyde in rats and mice after long-term inhalation exposure. Cancer Res. 43:4382-4392. Krewski, D., K. S. Crump, J. Farmer, D. W. Gaylor, R. Howe, C. Portier, D. Salsburg, R. L. Sielken, and J. Van Ryzin. 1983. A comparison of statistical methods for low dose extrapolation utilizing time-to-tumor data. Fund. Appl. Toxicol. 3:140-160. Lukashin, A. V., A. V. Vologodskii, M. D. Frank-Kamenetskii, Y. L. Lyubchenko. 1976. Fluctuational opening of the double helix as revealed by theoretical and experimental study of DNA interaction with formaldehyde. J. Mol. Biol. 108:665-682. Morgan, K. T. 1983. Localization of areas of inhibition of nasal mucociliary function in rats following in viva exposure to formaldehyde. Am. Rev. Respir. Dis. 127:166. Morgan, K. T., D. L. Patterson, and E. A. Gross. 1983. Formaldehyde and the nasal mucociliary apparatus. Pp. 193-210 in Formaldehyde Toxicology, Epidemiology, and Mechanisms, J. J. Clary, J. E. Gibson, and R. S. Waritz, eds. New York: Dekker. Portier, C., and D. G. Hoel. 1983. Low-dose-rate extrapolation using the multistage model. Biometrics 39:897-906. SOT Task Force. 1981. Re-examination of the EDIT study: Overview. Fund. Appl. Toxicol. 1 :28-63.

Cancer Risk Versus Delivered Dose 333 Starr, T. B., and R. D. Buck. 1984. The importance of delivered dose in estimating low- dose cancer risk from inhalation exposure to formaldehyde. Fund. Appl. Toxicol. 4:740- 753. Starr, T. B. and J. E. Gibson. 1985. The mechanistic toxicology of formaldehyde and its implications for quantitative risk estimation. Annul Rev. Pharmacol. Toxicol. 25:745- 767. Swenberg, J. A., E. A. Gross, H. W. Randall, and C. S. Barrow. 1983. The effect of formaldehyde exposure on cytotoxicity and cell proliferation. Pp. 225-236 in Formal- dehyde Toxicology, Epidemiology, and Mechanisms, J. J. Clary, J. E. Gibson, and R. S. Waritz, eds. New York: Marcel Dekker. Swenberg, J. A., E. A. Gross, and H. W. Randall. 1986. Localization and quantitation of cell proliferation following exposure to nasal irritants. Pp. 291-300 in Toxicology of the Nasal Passages, C. S. Barrow, ed. Washington, D.C.: Hemisphere. von Hippel, P. H., and K. Y. Wong 1971. Dynamic aspects of native DNA structure: Kinetics of the formaldehyde reaction with calf thymus DNA. J. Mol. Biol. 61:587- 613.

Pharmacokinetic Simulation as an Adjunct to Experimental Data in Risk Assessment: Predicting Exposure of the Bladder Epithelium in Dogs to Urinary N-Hydroxy Metabolites of Carcinogenic Arylamines John F. Young and Fred F. Kadiubar I NTRODUCTION An analog-digital hybrid computer (Figure 1) (Pearce and Young, 1981; Young et al., 1981) and LOTUS 1-2-3 on an International Business Ma- chines (IBM)-AT personal computer was used to predict the extent of urinary bladder exposure to N-hydroxy (N-OH) arylamines, which form adducts with urothelial DNA and are believed to serve as ultimate car- cinogenic metabolites (Kadlubar et al., 19811. A three-compartment model (Figures 2 and 3) (Young and Kadlubar, 1982) was used to fit the data obtained from rats given an intraurethral instillation of N-OH-2-naphthyl- amine and its acid-hydrolyzable N-glucuronide conjugate (Oglesby et al., 19811. This rat model was then assumed to be valid for other species, and simulations were conducted under varying conditions of urinary pH and voiding intervals. These simulations were then used to aid in the data interpretation of experiments with several dogs to which were administered orally the arylamine carcinogen 4-aminobiphenyl. This predictive ap- proach is used to assess the role of urinary pH and voiding interval in the release of the free N-OH arylamines in the bladder lumen and the sub- sequent formation of arylamine-DNA adducts in the urothelium and in the levels of 4-aminobiphenyl-hemoglobin adducts in blood, the latter of which is being used as a potential biological marker of carcinogen exposure (Green et al., 19841. 334

335 : ~~ ~~ ~~ ~~ ~i.i.i. o ;^ Ct - ~4 ._ so . Ct - o C) 5 ~ ~ 3 , ~ — 1 1 Ct 1 == 1 =~ -= 1 ~ A, 1 1 0 ~ 1 ~ s:: ~ ~ — 1 't c' 1 ~ ~ ·_ Ce o c: ~ ·t I C) 3 ~ .= Ct . o OO ~3 3 ._ so V, no - an, on Cal o · Ct o <,, so C) c: C,) ·— ~ no _. ._ Us: o s no ~ C) Ct — no <: _ _ , C) o ~ _ V , ~

336 JOHN F. YOUNG and FRED F. KADLUBAR A ~ s.s Absorption _ . ,~, , . _ 10.5 _ 5.5 Body 1 _ _ ,200 ~ ~ Bladder H - : U lumen f~pH} ~ Excretion y X FIGURE 2 Three-compartment (B. T. and H) pharmacokinetic model for the distribution of N-hydroxy~ylamine (NHAA).A = NHAA to be absorbed; B = NHAA in blood; T = NHAA in tissue; H = NHAA in bladder; X = NHAA excreted; C = NHAA metabolite in blood; U = NHAA metabolite in bladder; and Y = NHAA metabolized and excreted. Values are half- lives in minutes.) M ETHODS The analog-digital hybrid computer (Pacer 500; Electronics Associates, Inc., West Long Branch, NJ.) was used to generate amount-time curves for each component of the model. A set of differential equations was written that described the model (Figure 2) and were wired directly onto the analog patch panel (Figure 11. The schematic diagram of the differ- ential equations describing the model is presented in Figure 3 (Young et al., 19811. The analog portion of the hybrid computer solved all of the differential equations simultaneously; the digital portion sampled each analog curve 100 times during the 1-s solution and created an amount- time matrix. This matrix was then used in iterative schemes for optimi- zation of parameters and statistical analyses of the data. The digital portion of the hybrid computer controlled the setting of the rate constants, opti- mization schemes, graphics, printouts, and many other special features of this hybrid configuration (Pearce and Young, 1981~. All operator inter- actions with the system were through a cathode ray tube (CRT) input/ output terminal. The spreadsheet software package LOTUS 1-2-3 was used extensively to manipulate the exposure data obtained from the hybrid computer-generated curves. Values for integral H (N-OH arylamine blad- der exposure), integral U (N-OH arylamine conjugate bladder exposure), sum X (total N-OH arylamine excreted into the urine), and sum Y (total N-OH arylamine conjugate excreted into the urine) as a function of voiding

PK Stimulation to Predict Bladder Exposure 337 ~ + IC Absorption I | L~AIl[A] b l 1 1 ! 1 1 1 1 _ _ Logic Board ( Voiding Interval Control ) Q cry ~ ~ 0~ '~10 0 ;~ ~FMstri~tlijo~n;;l r Ic it, k-H[B] Cl{~k,,H| _ _ 1c ~~ [ B] rl.7 ~ k.,rBW —=_ _~. I \'C] ~ ECU 1 1 1 Counter 1 1 Q O Holding Time I ~ _ ~ Voidi 9 Time 1 1 1 Bladder Exposure (Integrated Areas) ~ ~H; _ __ ~11 Elimination | JO j Voi~j<4ontrol ~ . US Ro <I', 1 1 FIGURE 3 Analog and logic schematic diagram of the model. The analog schematic represents the differential equations that are written directly from the model. Holding time > Voiding interval; KUH ~ controls pH.

338 JOHN F. YOUNG and FRED F. KADLUBAR interval and urinary bladder pH were obtained from the hybrid simulation and entered into the LOTUS spreadsheet. These data were plotted indi- vidually as well as with various combinations of the data to observe the relationships among the components of the model. This type of data manipulation and graphic visualization is extremely easy, fast, and ac- curate when the spreadsheet approach is used. The three-compartment model was designed and tested by using data from the bladder instillation experiments of N-OH-2-naphthylamine in rats described by Oglesby et al. (19814. The bladder lumen portion of the model was included to be consistent with physiological reality because independent in vitro studies had determined the rate of N-OH arylamine N-glucuronide hydrolysis to increase as a function of decreasing pH. The ability to control voiding interval was also included for consistency because release of the free N-OH metabolite occurs as a linear, time-dependent process. Both of these latter parts of the model were unique to this ap- plication; normally, in pharmacokinetic modeling the excretion of a chem- ical into the urine is considered an end product that is not available for recycling and is dealt with as a continuous, cumulative function. f2,2'-3H]-4-Aminobiphenyl (4-ABP; 5 mg/kg; 12 mCi/mmol) was ad- ministered orally to three male dogs. Blood and urine samples were taken at various intervals over the next 24 h. Total radioactivity was determined z Cal I o 100- 50 - o o ~ O JH - - - - X~ H~ 0 3 6 9 12 Time (hours) FIGURE 4 Hybrid computer simulation curves for model components B. H. and X and integral H (area under the concentration-time curve [AUC]) for a urine pH of 5 and a voiding interval of 3 h. Plots such as this were generated for each combination of pH and voiding interval.

PK Stimulation to Predict Bladder Exposure 339 for whole blood, plasma, and urine samples. Analysis of blood for levels of 4-ABP-hemoglobin adducts at selected intervals were provided by P. Skipper (Massachusetts Institute of Technology, Boston, Mass.~. Urine samples were subjected to high-pressure liquid chromatographic analysis to determine the metabolic profile as a function of time (Frederick et al., 19811. The experimental treatment of the dogs differed only by the way in which the urine was collected: dog 2, natural voiding intervals; dog 3, 2 h voiding intervals via catheter for the first 12 h and naturally thereafter; dog 4, continuous voiding via catheter for the first 12 h and naturally thereafter. RESU LTS AND DISCUSSION A set of amount-time curves was generated by the hybrid computer for each pH and voiding interval. This set was visually simplified by the CRT display of only those model component curves of interest. Figure 4 rep- resents a set of curves for three components (B. H. X) of the model and the integrated area for the H curve under one set of conditions (pH 5; voiding interval, 3 h). The hybrid computer was used to generate a set of data at each of five pHs (4, 5, 5.5, 6, and 7) and eight voiding intervals (1, 2, 3, 4, 5, 6, and 10 h and continuous). These data were entered into the LOTUS 1-2-3 spreadsheet program, which was then used to manipulate the data and generate the plots. Figure 5 is a plot of integral H (urinary bladder exposure to N-OH arylamine) versus voiding interval; each curve represents a different pH value. The average value taken from the literature for pH and voiding interval for four species is superimposed on the plots. The relative bladder exposure values of these species are in the same order as their relative bladder carcinogenic potential as reported in the literature (human > dog > monkey > rat) (Deichmann, 1967; Radomski, 1979; Wynder and Gold- smith, 19771. This correlation suggests that the residence time of the N- OH arylamine in the bladder lumen should be directly related to the species sensitivity to urinary bladder cancer. Another set of curves obtained from the hybrid simulation resulted in a plot of sum X (excretion of N-OH arylamine in the urine) versus voiding interval as a function of pH (Figure 61. By dividing the value of the integral H by the value for the sum X at each pH and voiding interval, a matrix of data resulted that is plotted in Figure 7. This relationship in turn can be used to predict the exposure of the urothelium to N-OH arylamine metabolites (integral H) by urinary excretion measurements of the N-OH arylamine (sum X). Therefore, a noninvasive biological marker for bladder exposure is obtained from the analysis of levels of the N-OH arylamine at each urination, given the urine pH and voiding interval.

340 JOHN F. YOUNG and FRED F. KADLUBAR 220 - 2 1 0 - 200 ~ 190 ~ 180 ~ 170 ~ 160 ~ 150 ~ 140 ~ 130 - 120 ~ 110 - 100 - 90 - 80 - 70 - 60 - 50 — 40 — So 1 20 ~ 10 - 0 2 4 O pH~4 ~ pHs5 O pH=5.5 VOIDING INTERVAL (hours) ~/~ '' - _ Human * A * Dog ~~ * Monkey 8 10 pH=6 X pH =7 FIGURE 5 Plot of integral H versus voiding interval as a function of pH. The asterisk represents the positions on the graph for each species as determined by average urine pH and voiding interval taken from the literature. Symbols: By, pH 4; +, pH 5; O. pH 5.5; L\, pH 6; x, pH 7. 60 50 x ~ 30 v' 20 o 0 2 4 O pH = 4 4- pH ~ 5 O pH = S.S VOIDING INTERVAL (hours) 8 10 pH = 6 X pH =7 FIGURE 6 Plot of sum X versus voiding interval as a function of pH. For symbol definitions, see Figure 5 legend.

PK Stimulation to Predict Bladder Exposure 341 Another relationship to estimate bladder exposure was obtained from the plot of integral U (bladder exposure to the N-OH arylamine conjugate) versus voiding interval as a function of pH (Figure 81. The data from Figures 5 and 8 were then combined to generate Figure 9, which is a plot of integral H percentages Integral H x 100/integral H + integral U)] as a function of pH and voiding interval. The data in Figure 9 can then be used to predict bladder exposure (integral H) based on total urinary recovery of the arylamine and its matabolites. To test our hypothesis that N-OH arylamine bladder exposure can be used as a biological marker for assessing carcinogenic potential, t3H]4- A-BP has thus far been given orally (5 mg/kg) to three dogs. The only differences experimentally between the three dogs is the manner in which urine was collected: dog 2, natural voiding intervals; dog 3, 2-h voiding intervals for the first 12 h via bladder catheterization and natural voiding thereafter; dog 4, continuous voiding via indwelling bladder catheter~za- tion for the first 12 h and natural voiding thereafter. To Coo 90 80 70 C5 50 UJ by 60 40 30 20 10 O ~ O pH r 4 ~ pH ~ 5 1 A /: /' ,~ r I I I I I . I I . 0 2 4 6 8 10 VOIDING INTERVAL (hours! O pH · 5.5 ~ pH r 6 X pH s7 FIGURE 7 Plot of integral H/sum X versus voiding interval as a function of pH. By knowing the amount of.N-OH arylamine excreted in the urine, the urinary pH, and the voiding interval, an estimate of the bladder exposure to N-OH arylamine can be determined (I\ symbols in Figure 13). For symbol definitions, see Figure 5 legend.

342 JOHN F. YOUNG and FRED F. KADLUBAR 2.4 - 2.2 - 2 — 1.8 - 1.6 - 1.4 - 1.2 - 0.8 0.4 - 0.2 - O pH = 4 ~ pH - 5 0 2 4 6 VOIDING INTERVAL (hours) O pH ~ 5.5 8 10 pH = 6 X pH =7 FIGURE 8 Plot of interval U versus voiding integral as a function of pH. For symbol definitions, see Figure 5 legend. 90 - . 80 70 g x c ~ 50 c 40 - by 60 30 - 10 O- 1 art 0 2 4 6 8 10 O pH = 4 ~ pH 5 5 O pH = 5.5 VOIDING INTERVAL (hours) ~ pH = 6 X pH =7 FIGURE 9 Plot of integral H percent versus voiding integral as a function of pH. By knowing the total percentage of dose excreted for a given voiding interval and urinary pH, the bladder exposure to N-OH arylamine can be estimated (O symbols in Figures 13-15). For symbol definitions, see Figure 5 legend.

PK Stimulation to Predict Bladder Exposure 343 The carcinogenic potential of 4-ABP is assumed to be directly related to the amount of DNA adducts formed in the bladder from N-OH-4-ABP, which is assumed to be proportional to the residence time of the N-OH metabolite in the bladder lumen. The DNA adduct measurements involve the use of highly sensitive and specific immunoassays and are currently In progress. At present, however, this is not a convenient or readily convertible measurement to be applied to human populations. Therefore, a biological marker is needed to assess exposure to the potentially carcinogenic aryl- amines. Hemoglobin adducts (Hb-ABP) offer a good possibility because they can be readily measured from a blood sample and may reflect ac- cumulative exposure. Figures 10-12 are the percentage of dose versus time plots for whole blood, plasma, and hemoglobin-4-ABP adducts (Hb-ABP). The data for the three dogs were very similar, regardless of the manner in which the urine was collected. This was somewhat unexpected because we had anticipated that the Hb-ABP adducts might be formed as a consequence of reabsorption of the N-OH metabolite across the bladder wall rather than hepatic metabolism/transfer or direct formation in the blood. 19 — LU o C, 6 — o O ~ 1:: Hb-ABP ~ Whole Blood ,l,~0 I I I - I I ~ I I 0 4 8 12 16 20 24 TIME (hours) O Plasma FIGURE 10 Plot of percentage of dose (5 mg/kg) versus time for dog 2. Urine samples were collected as they occurred naturally. Symbols: By, Hb-ABP; +, whole blood; O. plasma.

344 jOHN ~ YOUNG and fRED ~ K^DLUB^R 12 - 11 - 10 - e- 8 - ~ e- o / . ' ~ I I 1 i ~ : i 1 ~ 1 0 4 8 12 16 20 24 ARE Rout O Hb-^BP + WhoIe BI_ ~ PIesm" FIGURE 11 Plot of percentage of dose (5 mg/kg) versus time far dog 3. Uhne samples wed collected eve~ 2 h via a catbeter. ~r symbol de~nitions, see Figure 10 legend. o 0 ~ ^ . ' I 1 # 1 1 1 1 1 0 4 8 12 16 1 1 1 1 20 24 O Hb-^- + WhoIe BIo~ ~E ~ou~ PIasma FIGURE 12 Plot of percentage of dose (3 mg/kg) versus dme ~r dog 4. Uhne samples wem collec~d condnuously via a c~he~r. ~r symbol deAnidons, see Figum 10 k~end.

PK Stimulation to Predict Bladder Exposure 345 Through pharrnacokinetic modeling, three other potential biological markers are being examined. Plasma AUC determination requires multiple blood samples and therefore may not be useful in population studies. Unnary excretion patterns either for total excretion or specifically for the N-OH arylamine levels may be easier to obtain but do not reflect multiple exposures over extended times. Table 1 presents the experimental data and calculated exposure data from the hybrid computer simulation for dog 2. Four biological markers are presented that are potential measures of bladder exposure to the N- OH arylamine. The Hb-ABP adduct level was measured directly from the blood sample (column D). The area under the plasma concentration-time curve was calculated by use of the trapezoid rule (column H). Two separate cumulative integral H values were calculated based on total recovery of radioactivity (columns E and I-K) or on excretion of N-OH-ABP (columns F and L-N). These latter two values differed by about a factor of 2 but had the same shape as the Hb-ABP curve (Figure 131. Figures 13-15 are plots of these various exposure measurements versus time for the three dogs. The most complete set of data is for dog 2. The analysis of the rest of the data for dogs 3 and 4 is ongoing. 20 9 ~ 18 - 17 - _ 16 ~ c ,5 `~' 44 - `~: 13 ~ [r 12 - c,0 1 1 - lo: 1 0 - 9 - ~: 8 - a 7 ~ `~O 6 - OX 5 ~ 113 4 - 3 ~ 2 - 1 - 0 1 ·, . I it' I ~ -—~ T 0 4 8 - 1 1 - 1 -1 1 ~ 1 12 16 20 24 TIME (hours) O HB-ABP ~ P-AUC O Sum H (a) ~ Sum H (b) FIGURE 13 Plot of four measures of N-OH arylamine bladder exposure as a function of time for dog 2. The dose of 4-ABP was 5 mg/kg. Symbols: G. Hb-ABP; +, p-AUC; O. sum H (a); A, sum H (b). (a) Sum of integral H based on total urinary recovery. (b) Sum of integral H based on urinary recovery of N-OH arylamine.

346 Ct - Ct By Hi_ o so <Jo-. o ._ - e~ ._ V) so - o v o Cat o Ct Cal o x A - - v a: Ct - a: ~ _` X ~ I,, ~ _4 ~ .= m ~ <( ._ En ~ m x o C) Ct Ct So - ~ 1 no Ct Cal - o o Ct 3 V, Cal ~ — O I i_ - C) _ ~ _ e,.e Cal — O ~ 00 C) ~ 4. U) ~ ~ - 0~ _ _ _ ce Al 0~ LO Ce it; ~: ·- 1 2 m s~ ~ ~ O ~ _ .= ~ ~ ~o =, ,~, m ~ D ~ CC ~ Ct ,,~ ~ m 0 0 ~ _ m 3 m o 0 _ ~ ~ ~ Z . . . . O v~ _ · · · O u~ O O ~ , _ ~ ~ ~ ~ . . . . ~D _ ~ ~ ~ 0 ~ ~: 0 ~ ~ ~ . . . . O _ 0 ~ 00 . . . . O ~ 0 0 ~ 0 0 . . . . ~ ~ 0 u~ _ ~ ~ 00 00 ~ ~ 00 0 0 0 0 — ~ ~ ~ ~ 00 ~ ~ _ x ~ c: . . . . ~o c~ 0 ~ ~ 0 ~ ~ 0 0 ~ 0 0 . . . . O 0 0 0 ~ oo ~ ~ LU 0 ~ _ ~ . . . . O ~ O0 0 — — — — 00 - , 00 ~ ~ 0 0 0 0 — ~ ~ ~ ~ ~ ~ 00 00 _ ~ ~ ~ 00 0 — ~ — ~ 00 — 00 . . . . . . . . . . O 0 0 ~ ~ ~ ~ ~ — — 0 ~D 00 . . . . . . . . . 0 r~ 0 ~ ~ oo x oo oo — C~ ~ ~ ~D oo O _ _ ~ m D o o s~ Ct ~L o C~ ,, - o C) ,,~ ~s r~o U~ D 0c ._ ._ ._ ;^ D C~ (,.~2 C~ - ._ o ; - D Ct - C~ - . . ~ C~ _ _ O O _ _ ._ ._ m ~ ~ . _ . _ _ _ 0 0 D D ,' ~, + O O + _ ~ ~ ~ . ~ _ _ 1 1 O — — ~ ,1 '' ~s ~ ~ ~ .o4._ o O LL ~ C) C) ~ .= X X O O ~ LL ;^ ,, _ ~ oc - _ !, ; - . _ . _ ._ ._ _ _ O O D D O ~

PK Stimulation to Predict Bladder Exposure 347 20 — 19 — 18 — 17 — 16 — 15 - 14 — 13 - 12 — 11 - 10 - 9 _ 8 — 7 — 6 — 5 — 4 — 3 — 2 — 1 - O ~ ~ 1 _ 0 4 8 12 TIME (hours) O HB-ABP ~ P-AUC O Sum H (a) 16 20 24 FIGURE 14 Plot of three measures of N-OH arylamine bladder exposure as a function of time for dog 3. The dose of 4-ABP was 5 mg/kg. Symbols: 2, Hb-ABP; +, P-AUC; O. sum - H (a). (a) Sum of integral H based on total urinary recovery. 20 - 19 - 18 — 17 - - - a' co MU a: LU tar En to x 16 - 15 - 14 - 13 - 12 - 11 - 10 - 9 _ 8 — 7 — 6 — 4 — 3 — 2 — 1 - O ~ i/ I/ I /' ~ - - I - 1 - I - I r 7 1 ~ ~ ~ 1 4 8 12 16 20 24 TIME (hours) to O HB-ABP ~ P-AUC O Sum H (a) FIGURE 15 Plot of three measures of N-OH arylamine bladder exposure as a function of time for dog 4. The dose of 4-ABP was 5 mg/kg. For symbol definitions, see Figure 14 legend. (a) Sum of integral H based on total urinary recovery.

348 JOHN F. YOUNG and FRED F. KADLUBAR CONCLUSIONS If carcinogenic potential is related to N-OH arylamine bladder exposure and DNA adduct formation, then Hb-ABP measurements may not be indicative of this process because the level of Hb-ABP adduct formation was about the same for all 3 days under three different urinary voiding conditions. We had anticipated that dog 4 (Figure 12) would have had a much lower level of Hb-ABP adducts because the bladder was continu- ously drained via the indwelling catheter, and therefore, the potential for bladder exposure should have been greatly reduced. Nevertheless, Hb- ABP appears to be an accurate measure of the external dose and hence a potentially useful biological marker. Prediction of N-OH arylamine bladder exposure based on integral H percent calculations from total recovery in the urine also does not seem predictable across venous conditions. We would anticipate that continuous urinary excretion would allow the least amount of bladder exposure; how- ever, Figure 15 indicates a high level of exposure potential from this calculation. This might be an artifact of the simulation and data manip- ulation, or it could be an indication that the experimental conditions were not exact and that the bladder was not entirely empty at all times. Prediction of N-OH Melamine bladder exposure based on integral H calculations from excretion levels of N-OH arylamine appears to be most promising, but we must wait for the completion of the urine analyses for 4-ABP metabolites. On completion of the arylamine-DNA adduct determinations in the urothelium, the endpoint evaluation of our simulations will have more meaning, and new avenues of analysis can then be explored. ACKNOWLEDGM ENTS The authors wish to thank Mary Ann Butler, Candee Teitel, John Bailey, Ken Dooley, Paul Skipper, and Steven Tannenbaum for the use of their data in this simulation exercise. REFERENCES Deichmann, W. G. 1967. Bladder Cancer, A Symposium, K. F. Lampe, ed. Birmingham, Ala.: Aesculapius, p. 30. Frederick, C. B., J. B. Mays, and F. F. Kadlubar. 1981. A chromatographic technique for the analysis of oxidized metabolites: Application to carcinogenic N-hydroxyarylam- ines in urine. Anal. Biochem. 118:120-125.

PK Stimulation to Predict Bladder Exposure 349 Green, L. C., P. L. Skipper, R. J. Turesky, M. S. Bryant, and S. R. Tannenbaum. 1984. In vivo dosimetry of 4-aminobiphenyl in rats via a cysteine adduct in hemoglobin. Cancer Res. 44:4254-4259. Kadlubar, F. F., J. F. Anson, K. L. Dooley, and F. A. Beland. 1981. Formation of urothelial and hepatic DNA adducts from the carcinogen 2-naphthylamine. Carcinogen- esis 2:467-470. Oglesby, L. A., T. J. Flammang, D. L. Tullis, and F. F. Kadlubar. 1981. Rapid absorption, distribution, and excretion of carcinogenic N-hydroxy-arylamines after direct urethral instillation into the rat urinary bladder. Carcinogenesis 2:15-20. Pearce, B. A., and J. F. Young. A hybrid computer system for pharmacokinetic modeling. I. Software considerations. Pp. 117-121 in Proceedings of the 1981 Summer Computer Simulation Conference. La Jolla, Calif.: Simulation Council, Inc. Radomski, J. L. 1979. The primary aromatic amines: Their biological properties and structural-activity relationships. Annul Rev. Pharmacol. Toxicol. 19:129. Wynder, E. L., and R. Goldsmith, Jr. 1977. The epidemiology of bladder cancer. Cancer 40: 1 246. Young, J. F., and F. F. Kadlubar. 1982. A pharmacokinetic model to predict exposure of the bladder epithelium to urinary N-hydroxyarylamine carcinogens as a function of urine pH, voiding interval, and resorption. Drug Metab. Dispos. 10(6):641-644. Young, J. F., C. G. White, and B. A. Pearce. 1981. A hybrid computer system for pharmacokinetic modeling. II. Applications. Pp. 122-129 in Proceedings of the 1981 Summer Computer Simulation Conference. La Jolla, Calif.: Simulation Council, Inc.

Next: VI. Applications of Mathematical Modeling »
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!