A constrained maximum likelihood approach to evaluate the impact of dose metric on cancer risk assessment: application to ß-chloroprene.

ß-Chloroprene (2-chloro-1,3-butadiene, CD) is used in the manufacture of polychloroprene rubber. Chronic inhalation studies have demonstrated that CD is carcinogenic in B6C3F1 mice and Fischer 344 rats. However, epidemiological studies do not provide compelling evidence for an increased risk of mortality from total cancers of the lung. Differences between the responses observed in animals and humans may be related to differences in toxicokinetics, the metabolism and detoxification of potentially active metabolites, as well as species differences in sensitivity. The purpose of this study was to develop and apply a novel method that combines the results from available physiologically based kinetic (PBK) models for chloroprene with a statistical maximum likelihood approach to test commonality of low-dose risk across species. This method allows for the combined evaluation of human and animal cancer study results to evaluate the difference between predicted risks using both external and internal dose metrics. The method applied to mouse and human CD data supports the hypothesis that a PBK-based metric reconciles the differences in mouse and human low-dose risk estimates and further suggests that, after PBK metric exposure adjustment, humans are equally or less sensitive than mice to low levels of CD exposure.

Use of mode of action data to inform a dose-response assessment for bladder cancer following exposure to inorganic arsenic.

In the recent National Research Council report on conducting a dose-response assessment for inorganic arsenic, the committee remarked that mode of action data should be used, to the extent possible, to extrapolate below the observed range for epidemiological studies to inform the shape of the dose-response curve. Recent in vitro mode of action studies focused on understanding the development of bladder cancer following exposure to inorganic arsenic provide data to inform the dose-response curve. These in vitro data, combined with results of bladder cancer epidemiology studies, inform the dose-response curve in the low-dose region, and include values for both pharmacokinetic and pharmacodynamic variability. Integration of these data provides evidence of a range of concentrations of arsenic for which no effect on the bladder would be expected. Specifically, integration of these results suggest that arsenic exposures in the range of 7-43 ppb in drinking water are exceedingly unlikely to elicit changes leading to key events in the development of cancer or noncancer effects in bladder tissue. These findings are consistent with the lack of evidence for bladder cancer following chronic ingestion of arsenic water concentrations <100 ppb in epidemiological studies.

Comparison of cancer risk estimates for vinyl chloride using animal and human data with a PBPK model.

Vinyl chloride (VC) is a trans-species carcinogen, producing tumors in a variety of tissues, from both inhalation and oral exposures, across a number of species. In particular, exposure to VC has been associated with a rare tumor, liver angiosarcoma, in a large number of studies in mice, rats, and humans. The mode of action for the carcinogenicity of VC appears to be a relatively straightforward example of DNA adduct formation by a reactive metabolite, leading to mutation, mistranscription, and neoplasia. The objective of the present analysis was to investigate the comparative potency of a classic genotoxic carcinogen across species, by performing a quantitative comparison of the carcinogenic potency of VC using data from inhalation and oral rodent bioassays as well as from human epidemiological studies. A physiologically-based pharmacokinetic (PBPK) model for VC was developed to support the target tissue dosimetry for the cancer risk assessment. Unlike previous models, the initial metabolism of VC was described as occurring via two saturable pathways, one representing low capacity-high affinity oxidation by CYP2E1 and the other (in the rodent) representing higher capacity-lower affinity oxidation by other isozymes of P450, producing in both cases chloroethylene oxide (CEO) and chloroacetaldehyde (CAA) as intermediate reactive products. Depletion of glutathione by reaction with CEO and CAA was also described. Animal-based risk estimates for human inhalation exposure to VC using total metabolism estimates from the PBPK model were consistent with risk estimates based on human epidemiological data, and were lower than those currently used in environmental decision-making by a factor of 80.

Development of a physiologically based pharmacokinetic model of trichloroethylene and its metabolites for use in risk assessment.

A physiologically based pharmacokinetic (PBPK) model was developed that provides a comprehensive description of the kinetics of trichloroethylene (TCE) and its metabolites, trichloroethanol (TCOH), trichloroacetic acid (TCA), and dichloroacetic acid (DCA), in the mouse, rat, and human for both oral and inhalation exposure. The model includes descriptions of the three principal target tissues for cancer identified in animal bioassays: liver, lung, and kidney. Cancer dose metrics provided in the model include the area under the concentration curve (AUC) for TCA and DCA in the plasma, the peak concentration and AUC for chloral in the tracheobronchial region of the lung, and the production of a thioacetylating intermediate from dichlorovinylcysteine in the kidney. Additional dose metrics provided for noncancer risk assessment include the peak concentrations and AUCs for TCE and TCOH in the blood, as well as the total metabolism of TCE divided by the body weight. Sensitivity and uncertainty analyses were performed on the model to evaluate its suitability for use in a pharmacokinetic risk assessment for TCE. Model predictions of TCE, TCA, DCA, and TCOH concentrations in rodents and humans are in good agreement with a variety of experimental data, suggesting that the model should provide a useful basis for evaluating cross-species differences in pharmacokinetics for these chemicals. In the case of the lung and kidney target tissues, however, only limited data are available for establishing cross-species pharmacokinetics. As a result, PBPK model calculations of target tissue dose for lung and kidney should be used with caution.

Evaluating noncancer effects of trichloroethylene: dosimetry, mode of action, and risk assessment.

Alternatives for developing chronic exposure limits for noncancer effects of trichloroethylene (TCE) were evaluated. These alternatives were organized within a framework for dose-response assessment--exposure:dosimetry (pharmacokinetics):mode of action (pharmacodynamics): response. This framework provides a consistent structure within which to make scientific judgments about available information, its interpretation, and use. These judgments occur in the selection of critical studies, internal dose metrics, pharmacokinetic models, approaches for interspecies extrapolation of pharmacodynamics, and uncertainty factors. Potentially limiting end points included developmental eye malformations, liver effects, immunotoxicity, and kidney toxicity from oral exposure and neurological, liver, and kidney effects by inhalation. Each end point was evaluated quantitatively using several methods. Default analyses used the traditional no-observed adverse effect level divided by uncertainty factors and the benchmark dose divided by uncertainty factors methods. Subsequently, mode-of-action and pharmacokinetic information were incorporated. Internal dose metrics were estimated using a physiologically based pharmacokinetic (PBPK) model for TCE and its major metabolites. This approach was notably useful with neurological and kidney toxicities. The human PBPK model provided estimates of human exposure doses for the internal dose metrics. Pharmacodynamic data or default assumptions were used for interspecies extrapolation. For liver and neurological effects, humans appear no more sensitive than rodents when internal dose metrics were considered. Therefore, the interspecies uncertainty factor was reduced, illustrating that uncertainty factors are a semiquantitative approach fitting into the organizational framework. Incorporation of pharmacokinetics and pharmacodynamics can result in values that differ significantly from those obtained with the default methods.

Evaluation of the uncertainty in an oral reference dose for methylmercury due to interindividual variability in pharmacokinetics.

An analysis of the uncertainty in guidelines for the ingestion of methylmercury (MeHg) due to human pharmacokinetic variability was conducted using a physiologically based pharmacokinetic (PBPK) model that describes MeHg kinetics in the pregnant human and fetus. Two alternative derivations of an ingestion guideline for MeHg were considered: the U.S. Environmental Protection Agency reference dose (RfD) of 0.1 microgram/kg/day derived from studies of an Iraqi grain poisoning episode, and the Agency for Toxic Substances and Disease Registry chronic oral minimal risk level (MRL) of 0.5 microgram/kg/day based on studies of a fish-eating population in the Seychelles Islands. Calculation of an ingestion guideline for MeHg from either of these epidemiological studies requires calculation of a dose conversion factor (DCF) relating a hair mercury concentration to a chronic MeHg ingestion rate. To evaluate the uncertainty in this DCF across the population of U.S. women of child-bearing age, Monte Carlo analyses were performed in which distributions for each of the parameters in the PBPK model were randomly sampled 1000 times. The 1st and 5th percentiles of the resulting distribution of DCFs were a factor of 1.8 and 1.5 below the median, respectively. This estimate of variability is consistent with, but somewhat less than, previous analyses performed with empirical, one-compartment pharmacokinetic models. The use of a consistent factor in both guidelines of 1.5 for pharmacokinetic variability in the DCF, and keeping all other aspects of the derivations unchanged, would result in an RfD of 0.2 microgram/kg/day and an MRL of 0.3 microgram/kg/day.

Investigation of the potential impact of benchmark dose and pharmacokinetic modeling in noncancer risk assessment.

There has been relatively little attention given to incorporating knowledge of mode of action or of dosimetry of active toxic chemical to target tissue sites in the calculation of noncancer exposure guidelines. One exception is the focus in the revised reference concentration (RfC) process on delivered dose adjustments for inhaled materials. The studies reported here attempt to continue in the spirit of the new RfC guidelines by incorporating both mechanistic and delivered dose information using a physiologically based pharmacokinetic (PBPK) model, along with quantitative dose-response information using the benchmark dose (BMD) method, into the noncancer risk assessment paradigm. Two examples of the use of PBPK and BMD techniques in noncancer risk assessment are described: methylene chloride, and trichloroethylene. Minimal risk levels (MRLs) based on PBPK analysis of these chemicals were generally similar to those based on the traditional process, but individual MRLs ranged from roughly 10-fold higher to more than 10-fold lower than existing MRLs that were not based on PBPK modeling. Only two MRLs were based on critical studies that presented adequate data for BMD modeling, and in these two cases the BMD models were unable to provide an acceptable fit to the overall dose-response of the data, even using pharmacokinetic dose metrics. A review of 10 additional chemicals indicated that data reporting in the toxicological literature is often inadequate to support BMD modeling. Three general observations regarding the use of PBPK and BMD modeling in noncancer risk assessment were noted. First, a full PBPK model may not be necessary to support a more accurate risk assessment; often only a simple pharmacokinetic description, or an understanding of basic pharmacokinetic principles, is needed. Second, pharmacokinetic and mode of action considerations are a crucial factor in conducting noncancer risk assessments that involve animal-to-human extrapolation. Third, to support the application of BMD modeling in noncancer risk assessment, reporting of toxicity results in the toxicological literature should include both means and standard deviations for each dose group in the case of quantitative endpoints, such as relative organ weights or testing scores, and should report the number of animals affected in the case of qualitative endpoints.

Implementation of EPA Revised Cancer Assessment Guidelines: Incorporation of Mechanistic and Pharmacokinetic Data.

A workshop entitled "Implementation of EPA Revised Cancer Assessment Guidelines: Incorporation of Mechanistic and Pharmacokinetic Data" was held in Anaheim, California, in 1996 at the 35th Annual Meeting of the Society of Toxicology (SOT). This workshop was jointly sponsored by the Carcinogenesis, Risk Assessment, and Veterinary Specialty Sections of the SOT. The thrust of the workshop was to discuss the scientific basis for the revisions to the EPA Guidelines for cancer assessment and EPA's plans for their implementation. This is the first revision to the original EPA guidelines which have been in use by EPA since 1986. The principal revisions are intended to provide a framework for an increased ability to incorporate biological data into the risk assessment process. Two cases were presented, for chloroform and triclioroethylene, that demonstrated the use of the revised guidelines for specific cancer risk assessments. Using these new guidelines, nonlinear margin of exposure analyses were proposed for these chemicals instead of the linearized multistage model previously used by the EPA as the default method. The workshop participants generally applauded the planned revisions to the EPA guidelines. For the most part, they considered that the revised guidelines represented a positive step which should allow for and encourage the use of biological information in the conduct of cancer risk assessments. Several participants cautioned however that the major problem with cancer risk assessments would continue to be the inadequacy of available data on which to conduct more scientific risk assessments.

Incorporating biological information in quantitative risk assessment: an example with methylene chloride.

The interplay between chemical risk assessment and scientific research is discussed in the context of recent attempts to improve the scientific basis for estimates of the human carcinogenic risk from methylene chloride. A combination of basic biochemical research and risk assessment oriented research, both mechanistic and pharmacokinetic, provided the initial impetus for re-evaluating the use of the default risk estimation approach. Resulting efforts to incorporate the new scientific information into the risk assessment process in turn identified specific additional research needed to reduce uncertainty in the estimated risk. This healthy interchange between the two disciplines has served both to refine the human risk estimates for methylene chloride and to more clearly identify key scientific issues for chemical risk assessment in general.

The application of physiologically based pharmacokinetic modeling in human health risk assessment of hazardous substances.

Physiologically based pharmacokinetic (PBPK) modeling is an important tool for improving the accuracy of human health risk assessments for hazardous substances in the environment. The proper use of PBPK modeling can reduce uncertainties that currently exist in risk assessment procedures by providing more scientifically credible extrapolations across species and routes of exposure, and from high experimental doses to potential environmental exposures. Current applications of PBPK models range from relatively straightforward uses for the extrapolation of chemical kinetics across species, route, and duration of exposure to much more demanding chemical risk assessment applications requiring a description of complex pharmacodynamic phenomena such as mitogenicity and hyperplasia secondary to cytotoxicity. PBPK modeling helps to identify the factors that are most important in determining the health risks associated with exposure to a chemical, and provides a means for estimating the impact of those factors both on the average risk to a population and on the specific risk to an individual. The chief challenge in the application of PBPK modeling in human health risk assessment lies in the need to generate chemical-specific data to support the development and validation of the models. Extensive use of rapidly developing in vitro and structure-activity relationship techniques is needed to provide the data required for the large number of hazardous chemicals currently contaminating the environment.

Applying simulation modeling to problems in toxicology and risk assessment--a short perspective.

The goals of this perspective have been to examine areas where quantitative simulation models may be useful in toxicology and related risk assessment fields and to offer suggestions for preparing manuscripts that describe these models. If developments in other disciplines serve as a bell-wether, the use of mathematical models in toxicology will continue to increase, partly, at least, because the new generations of scientists are being trained in an electronic environment where computation of all kinds is learned at an early age. Undoubtedly, however, the utility of these models will be directly tied to the skills of investigators in accurately describing models in their research papers. These publications should convey descriptions of both the insights obtained and the opportunities provided by these models to integrate existing data bases and suggest new and useful experiments. We hope these comments serve to facilitate the expansion of good modeling practices as applied to toxicological problems.

Sensitivity of physiologically based pharmacokinetic models to variation in model parameters: methylene chloride.

The parameters in a physiologically based pharmacokinetic (PBPK) model of methylene chloride were varied systematically, and the resulting variation in a number of model outputs was determined as a function of time for mice and humans at several exposure concentrations. The importance of the various parameters in the model was highly dependent on the conditions (concentration, species) for which the simulation was performed and the model output (dose surrogate) being considered. Model structure also had a significant impact on the results. For sensitivity analysis, particular attention must be paid to conservation equations to ensure that the variational calculations do not alter mass balance, introducing extraneous effects into the model. All of the normalized sensitivity coefficients calculated in this study ranged between -1.12 and 1, and most were much less than 1 in absolute value, indicating that individual input errors are not greatly amplified in the outputs. In addition to ranking parameters in terms of their impact on model predictions, time-dependent sensitivity analysis can also be used as an aid in the design of experiments to estimate parameters by predicting the experimental conditions and sampling points which will maximize parameter identifiability.

Incorporation of pharmacokinetics in noncancer risk assessment: example with chloropentafluorobenzene.

Noncancer risk assessment traditionally relies on applied dose measures, such as concentration in inhaled air or in drinking water, to characterize no-effect levels or low-effect levels in animal experiments. Safety factors are then incorporated to address the uncertainties associated with extrapolating across species, dose levels, and routes of exposure, as well as to account for the potential impact of variability of human response. A risk assessment for chloropentafluorobenzene (CPFB) was performed in which a physiologically based pharmacokinetic model was employed to calculate an internal measure of effective tissue dose appropriate to each toxic endpoint. The model accurately describes the kinetics of CPFB in both rodents and primates. The model calculations of internal dose at the no-effect and low-effect levels in animals were compared with those calculated for potential human exposure scenarios. These calculations were then used in place of default interspecies and route-to-route safety factors to determine safe human exposure conditions. Estimates of the impact of model parameter uncertainty, as estimated by a Monte Carlo technique, also were incorporated into the assessment. The approach used for CPFB is recommended as a general methodology for noncancer risk assessment whenever the necessary pharmacokinetic data can be obtained.

Variability of physiologically based pharmacokinetic (PBPK) model parameters and their effects on PBPK model predictions in a risk assessment for perchloroethylene (PCE).

When used in the risk assessment process, the output from physiologically based pharmacokinetic (PBPK) models has usually been considered as an exact estimate of dose, ignoring uncertainties in the parameter values used in the model and their impact on model predictions. We have collected experimental data on the variability of key parameters in a PBPK model for tetrachloroethylene (PCE) and have used Monte Carlo analysis to estimate the resulting variability in the model predictions. Blood/air and tissue/blood partition coefficients and the interanimal variability of these data were determined for tetrachloroethylene (PCE). The mean values and variability for these and other published model parameters were incorporated into a PBPK model for PCE and a Monte Carlo analysis (n = 600) was performed to determine the effect on model predicted dose surrogates for a PCE risk assessment. For a typical dose surrogate, area under the blood time curve for metabolite in the liver (AUCLM), the coefficient of variation was 25% and the mean value for AUCLM was within a factor of two of the maximum and minimum values generated in the 600 simulations. These calculations demonstrate that parameter uncertainty is not a significant potential source of variability in the use of PBPK models in risk assessment. However, we did not in this study consider uncertainties as to metabolic pathways, mechanism of carcinogenicity, or appropriateness of dose surrogates.

Biologically motivated models for chemical risk assessment.

Assessing the risk associated with human exposure to environmental chemicals depends to a large extent on the ability to extrapolate from a particular range of exposure conditions in the test animal species to a very different range of exposure conditions in the human. One of the more promising tools for accomplishing this extrapolation is the biologically motivated pharmacokinetic/pharmacodynamic model. In a biologically motivated model, the structure is based on the physiological and biochemical structure of the animal system being described. This paper provides an overview of the biologically motivated modeling approach. Examples of models for styrene and methylene chloride are discussed in relation to their ability to predict human kinetics for these chemicals and their use in estimating the risk of chemicals to exposed humans. Finally, the use of a biologically motivated model to analyze the mechanistic basis of chemical carcinogenesis is discussed.

Physiologically based pharmacokinetics and the risk assessment process for methylene chloride.

Methylene chloride (dichloromethane, DCM) is metabolized by two pathways: one dependent on oxidation by mixed function oxidases (MFO) and the other dependent on glutathione S-transferases (GST). A physiologically based pharmacokinetic (PB-PK) model based on knowledge of these pathways was used to describe the metabolism of DCM in four mammalian species (mouse, rat, hamster, and humans). Kinetic constants for the model were derived from in vivo experiments or the literature. The model was constructed to distinguish contributions from the two pathways of metabolism in lung and liver tissue, and to permit extrapolation from rodents to humans. Model validation was conducted by comparing predicted blood concentration time-course data in rats, mice, and humans with experimental data from these species. The tumor incidence in two chronic studies of DCM toxicity in mice was correlated with various measures of target tissue dose calculated with the PB-PK model. Tumor incidence correlated well with tissue AUC (area under the concentration/time curve) and amount of DCM metabolized by the GST pathway. However, tumor incidence did not correlate with the amount of DCM metabolized by the MFO pathway. Because of its low chemical reactivity, DCM is unlikely to be directly involved in carcinogenesis. Consequently, metabolism of DCM by GST appears to be important in carcinogenesis. The PB-PK model was used to estimate target doses of presumed toxic chemical species in humans exposed to DCM by inhalation or by drinking water. Target tissue doses in humans exposed to low concentrations of DCM are 140- to 170-fold lower (inhalation) or 50- to 210-fold lower (drinking water) than would be expected from the linear extrapolation and body surface area factors which have been used in conventional risk assessment methods (D. V. Singh, H. L. Spitzer, and P. D. White (1985). Addendum to the Health Assessment Document for Dichloromethane (Methylene Chloride). EPA/600/8-82/004F). The PB-BK analysis thus suggests that conventional risk analyses greatly overestimate the risk in humans exposed to low concentrations of DCM. PB-PK considerations provide a scientific basis for risk assessment, improve experimental design in chronic studies, and structure collection of quantitative metabolic constants required for risk assessment.

Risk assessment extrapolations and physiological modeling.

The process of assessing the risk associated with human exposure to environmental chemicals inevitably relies on a number of assumptions, estimates and rationalizations. One of the more challenging aspects of risk assessment involves the need to extrapolate beyond the range of conditions used in experimental animal studies to predict anticipated human risks. The most obvious extrapolation required is that from the tested animal species to humans; but others are also generally required, including extrapolating from high dose to low dose, from one route of exposure to another and from one exposure timeframe to another. Several avenues are available for attempting these extrapolations, ranging from the assumption of strict correspondence of dose to the use of statistical correlations. One promising alternative for conducting more scientifically sound extrapolations is that of using physiologically based pharmacokinetic models that contain sufficient biological detail to allow pharmacokinetic behavior to be predicted for widely different exposure scenarios. In recent years, successful physiological models have been developed for a variety of volatile and nonvolatile chemicals, and their ability to perform the extrapolations needed in risk assessment has been demonstrated. Techniques for determining the necessary biochemical parameters are readily available, and the computational requirements are now within the scope of even a personal computer. In addition to providing a sound framework for extrapolation, the predictive power of a physiologically based pharmacokinetic model makes it a useful tool for more reliable dose selection before beginning large-scale studies, as well as for the retrospective analysis of experimental results.