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.
Testing Strategies arid Methods Through the analysis of a series of published and unpublished studies, the committee found that two key steps must be taken for a problem with complex mixtures to be manageable. The steps are related to what questions are to be answered and what is known about the characterization of potential exposures. Another factor, related to those two steps, that is of crucial importance in selecting the appropriate testing strategy is whether the mixture under study is a known entity of expected uniformity and formulation or an unknown mixture of varied origins, such as leachates and runoff. This chapter examines problem definition and choice of testing strategies from the perspectives of potential biologic effects of complex mixtures, the agents present in the mixtures, and the predictability of the data on one mixture with respect to other, related mixtures. Throughout, there is a concern for exposure potential. Without exposure, there is obviously no risk and therefore no reason for testing. The nature, extent, duration, and frequency of actual or anticipated exposure can all influence the selection of testing strategies and protocols. In particular, to the extent that any test protocol attempts to model the real world, extensive exposure data must be available in the study-design phase. (Chapter 2 deals specifically with such exposure-dose considerations.) The definition of the problem will dictate the overall strategy to be followed. Reasonably standardized techniques that were developed for the testing of sin- gle chemicals can usually be adapted to study complex mixtures; only rarely must new approaches to toxicity testing be sought. The possibility of failing to observe adverse health affects or toxicity end points must be kept in mind, but this will probably occur no more often with complex mixtures than in the testing of single chemicals. Predictability is an important factor in determining the approach to be used 39
40 COMPLEX MIXTURES in testing mixtures. Stated most simply, it is of considerable value if, in the case of a series of mixtures, the testing strategy might provide information not only on the specific sample tested, but also on other, related mixtures. This potential use of data from one sample as a basis for prediction in other samples is far more difficult to resolve in the case of complex mixtures than for single agents, because of the possibility that interactions will influence physicochemical properties, bioavailability, biotransformation, and related phenomena. Successful development of strategies for testing complex mixtures requires careful structuring of the questions to be asked, decisions regarding the orderin which they are asked with respect to effects and composition of the mixture, and identification of the intended use of the inflation to be obtained. If the questions asked are not defined properly, the testing strategy required cannot be developed, nor can one expect to obtain results that are applicable to poten- tial adverse health effects in humans. The major difference between mixtures and single chemicals is not in the testing strategies chosen, but in the impor- tance of problem definition before testing. PROBLEM AND QUESTION DEFINITION The evaluation of the toxicity of a complex mixture should be preceded by understanding of the problem and definition of the questions to be answered. The extent and nature of testing should be guided closely by knowledge of what is known and what is to be learned. This section outlines the types of complex- mixture problems encountered and some questions that may be asked before selection of a strategy. QUESTIONS RELATED~TO EFFECTS The type of exposure can also influence the selection of a testing strategy and specifically will effect the experimental design. (See Chapter2 for a discussion ofthe relation between administered dose, or exposure, and biologically effec- tive dose.) The nature and magnitude of toxic effects dictate a number of questions: · Under the conditions of expected human exposure, is the mixture poten- tially hazardous? Experimental designs should consider the routes, extents, and durations of exposure. Attention needs to focus on the extrapolation from experimental exposure to expected human exposure. · If a toxic effect is observed epidemiologically, can a particular complex mixture be identified as responsible for the effect? · Given an observable toxic effect in humans, can animal models be used appropriately to elicit it? · How toxic is the mixture, compared with others under consideration? This question is not peculiar to complex mixtures, of course, but it is routine in
TESTING STRATEGIES AND METHODS 41 the assessment of the comparative toxicity of single chemicals in relation to related agents. · What are the yardsticks of concern in preventing adverse health effects in the exposed general population? Should a spectrum of toxicity tests in a stand- ardized format be done, or should a specific end point for example, muta- genesis, teratogenesis, immunotoxicity, or behavioral aberrations be the fo- cus of testing? In retrospective epidemiologic studies, these responses tend to be subtle and inconclusive, the relationship between effect and causative agent being tenuous at best. Whatever end points are chosen must be considered in relation to human exposure to a given complex mixture and a set of plausible circumstances of exposure. QUESTIONS RELATED TO CAUSATIVE AGENTS Questions related to causative agents are peculiar to situations in which a toxic effect has been observed but the causative agents are unknown. They arise for such reasons as a desire to understand mechanisms of toxicity and ways to reduce toxicity. The questions include the following: · What is the source of the toxicity of the complex mixture? The effects might be associated with a component that is active because it constitutes a major proportion of the mixture, is a minor constituent with high intrinsic toxicity, or is the component with the greatest bioavailability (due to physico- chemical properties, etc.~; or the effect might be a property of a combination of components. · Is the chemical composition of the mixture known? Knowledge of the toxic potential of mixture constituents might permit the development of strate- gies to examine toxic end points of special interest. · Can the toxicity of a product be reduced by altering its composition? This question generally requires an identification of the causative agents, if one is to determine whether removal of such components is both possible and likely to reduce toxicity. QUESTIONS RELATED TO PREDICTABILITY Questions related to predictability arise under two circumstances: when chemical constituents of a mixture are known and effects of the mixture need to be predicted, or when both agents and effects are known for one situation but need to be predicted for another. The following are examples of predictability questions: · How might interactions of constituents contribute to the toxicity of a mix- ture at different extents and durations of exposure? · Are the measured biologic effects reproducible with different samples of the mixture? A positive answer would suggest that one sample would suffice to
42 COMPLEX MIXTURES characterize the predominant toxicity attributable to the mixture. A negative answer should lead one to query the stability or heterogeneity of the mixture and address problems of sampling. · Are the toxicity results useful for predicting potential toxicity of similar mixtures? Many complex mixtures are heterogeneous, their compositions varying both qualitatively and quantitatively from source to source and being influenced by the conditions governing their formation. For example, is each municipal waste incinerator unique in its emissions, or can conclusions be drawn about incinerators in general by sampling only a few sources? If so, how many samples are needed regarding the range of burn conditions? STRATEGIES Once the appropriate questions have been formulated, the strategies to an- swer them must be considered. The strategies for evaluating complex mixtures have been organized into three categories, according to the kinds of questions to be answered those related to effects, to causative agents, and to predictability. Although the strat- egies are often applied in combination to address a series of questions, they are described separately here to exemplify the specific approaches that can be used. The integration of strategies is discussed later in this chapter. STRATEGIES RELATED TO EFFECTS Biologic effects of complex mixtures can be the same as those of single agents. There is no compelling reason to expect their effects to be different, and the investigator will normally follow the same patterns of observation and recording used for single agents. If the mixture contains representatives of a wide range of chemical types and a test involves intact animals, the mixture might be expected to affect many organ systems. As Chapter 2 notes, documented dose-response relationships in humans for exposures to most mixtures are relatively rare. Strategies for testing mixtures should avoid overcomplex programs and take into account the normal dose rate and frequencies of exposure. However, testing strategies must always be sensi- tive to unpredicted outcomes. Effect-related strategies generally involve well-established stepwise toxico- logic evaluations of a test substance in this case a mixture (NRC, 1977~. They are concerned initially not with causative agents, but with observable outcomes (effects). This approach uses the complete material, and increas- ingly elaborate studies are based on the outcome of early, simpler studies. Causation may be inferred from thoughtful evaluation of experimental obser- vations. Such factors as time of onset of a toxic response, duration of effect, reversibility of effect, physiologic system involved, organ pathology, and pos-
TESTING STRATEGIES AND METHODS 43 sible links between affected organs can all be parts of a pattern attributable to specific agents or classes of agents. Examples would include cholinesterase inhibitors, heavy metals, and hydrocarbon-based solvents. In the case of some mixtures, identification of causative agents might not even be the information desired. The main interest might be to confirm or eliminate a causal relation- ship between an observed effect and an uncharactenzed matenal. In such in- stances, it is proper to test the sample mixture without conducting charactenza- tion studies. An investigator has a number of design options that avoid premature commitment to an extensive slate of protocols. Several options are described below. Tier Testing Tier testing may be likened to a series of sieves with smaller and smaller pores (NRC, 19841. At each stage, some findings require followup and others are considered adequate or conclusive. Figure 3-1 illustrates a form of tier ~ >11~ us C C IU z ~ ~ m ~ U) O G UJ ~ . Z ._~ Exposure, Physicochemical Data, Production Volume ~,~- Mutagenicity, Accute Toxicity, Irritation Reproduction, Teratology, Subchronic Toxicity Chronic Toxicity, Oncogenesis FIGURE 3-1 Schematic diagram of tier testing where the first tiers consist of less-expensive and less-time-consuming screening tests and progress at higher times and more-expensive and time-consuming tests.
44 COMPLEX MIXTURES testing used in the study of commercial mixtures; predetermined triggers (or end points) dictate whether the next stage is required. A comparable system is a part of the premarketing notification system of the European Economic Com- munity; in this instance, the triggers are related to the amounts of material intended to be produced annually. Not unexpectedly, the cost and time required for each tier are greater than those for the preceding one. A tier-testing approach has been applied to a number of complex mixtures (e.g., diesel emissions, synthetic fuels, and concentrated organic mixtures from drinking water) to identify potential toxic effects resulting from expo- sures to these mixtures. Screening Studies Screening studies are less integrated than tier-testing programs. The investi- gator first determines which biologic end points are of interest and then screens or gains preliminary information regarding those effects, rather than conduct- ing definitive work for each effect. The simplest test that reliably points to an effect is chosen. The criterion should be sensitivity (few false negatives), rather than selectivity (few false positives). Assays are done sequentially until the investigator is able to rank the test material with respect to effects of inter- est. Where necessary, more definitive studies are done on agents about which there is uncertainty, whose exposure potential is greatest, about which there is substantial social concern, or whose potency or other biologic considerations warrant further study. The carcinogenicity studies initially performed by the National Cancer Institute's Bioassay Program relied on a screening program designed to identify potential carcinogenic agents (NCI, 19761. Agents were selected generally because of widespread use or because of suspicion of poten- tial carcinogenicity. Doses were high, and the result was a measure of carcino- genic potential (not risk or hazard). Current National Toxicology Program long-term carcinogenicity bioassays have been modified to provide more in- formation (e.g., by using more than two doses and collecting pharmacokinetic data) (Huff, 19821. Screening studies have been successfully applied to mixture problems, such as the hexacarbon-neuropathy case. These studies revealed that an organized, scientific approach to problem-solving has a high probability of success, even with complex mixtures, where there is a combination of focused questions, assay methods, and good luck. The focused questions (referred to previously) resulted in the isolation of the problem to specific work areas that used materi- als (mixtures) of known or knowable composition. The same focused ques- tions went beyond a general description of the disease to its specific character- istics and permitted the establishment of links to other case reports. The availability of suitable assay systems allowed the screening of hundreds of workers for necrologic deficits and permitted the concerns about the workplace to be concentrated in the print department. The assays also allowed some
TESTING STRATEGIES AND METHODS 45 agents to be eliminated as causes, because the necrologic findings were not of the expected kind. An animal assay allowed mixtures or individual chemicals to be screened for the necrologic disorder and for structure-activity studies to be conducted. (For additional information on this case, see Appendix C.) Matrix Testing Matrix testing involves the systematic manipulation of several variables to define the "universe" of materials and outcomes. Once critical variables have been identified, the agents can be arranged as loci in a matrix field whose dimensions are defined by those variables, which are presented normally in a two-dimensional plot. Additional dimensions can be considered, although the possibilities for graphic representation diminish as dimensions are added. An example of this approach may be found in a series of animal toxicology studies on hydrocarbon solvents published by Carpenter et al. (1975a-h, 1976a-e, 1977a-c, 19781. The underlying premise was that the toxicity of this class of materials was related to boiling range and aromaticity. The matrix is shown in Figure 3-2. Boiling range was a routine specification available for all such materials and is related to number of carbon atoms (and molecular weight). The aromaticity dimension reflected the existing literature, which shows that, for a comparable number of carbon atoms, toxicity decreases from aromatic hydrocarbons to naphthenic hydrocarbons to paraff~nic hydrocarbons (Scala, in press). Sample Matrix of Critical Variables FIGURE 3-2 Example of matrix testing in which critical variables for series of mixtures are manipulated to define aromaticity ~ ~ boundary conditions such that toxicologic testing of selected Miniature I 41; variables wou d result in pre ictability of effects within o, ~ _ 7~ 7t 9 ~ ~7 12 13 14 15 ~ ~ Ah 16 Selected Toxicologic Test(s) Interpretation
46 COMPLEX MIXTURES The individual mixtures of commercial importance were placed on a matrix with these dimensions. Selected mixtures were tested in varied toxicologic screening paradigms. The results of each test dictated where in the matrix the next sample would be chosen. It was hoped that defining "boundary" condi- tions would make it possible to position effects within the matrix. Battery Approach In the battery approach, an array of bioassays is used. This strategy (Figure 3-3) has been used in the field of genetic toxicity, in which the emphasis is on determining whether agents are genotoxic, whether metabolic conversion is a prerequisite for activity, and what sort of mechanisms might be involved (Heussneret al., 19851. Just as a diner might construct a meal by going through several courses on a complex menu, so an investigator can provide answers to the specific questions posed by selecting tests from an available battery. Comparative-Potency Approach Comparative approaches in toxicology generally involve studies of the ef- fects of one substance in different species or bioassays (as in the battery ap- proach) or studies of Me effects of a series of substances in one or a few bioas- Select end points Select battery of bioassays Bioassay Battery Test all mixtures in the entire battery gene- I DNA- I chromosomal- mutation i damage ~ effects assay ~ assay ~ assay 1 1 ~ I aneuploidy assay - - ~1~ - - FIGURE 3-3 Example of battery approach to toxicologic testing of mixtures in which genetic- toxicity test batted is constructed to test for different genetic effects.
TESTING STRATEGIES AND METHODS - o c E I / o / Coke Oven `~' Roofing Tar / ./ / / / / Cigarette Smoke E A: Animal Tumorigenesis Potency 47 / D / ·/ / C Bacterial Mutation Potency FIGURE 3~ Example of comparative-potency approach to evaluating diesel emissions. says. The latter is described here as the comparative-potency approach; it is particularly useful for the evaluation of complex mixtures (Figure 3-41. Many mixture problems involve the grouping of mixtures, such as fuels derived from different sources (e.g., petroleum, shale, and coal), soots emitted from different combustion sources, hazardous wastes from different sites, ef- fluents from different industries, emissions subject to different control technol- ogies, and emissions from the burning of different products. The problem posed to the toxicologist might be whether one mixture is more or less toxic than another. All soots from incomplete combustion are expected to contain carcinogenic polycyclic organic compounds, so the toxicologist might have to address the question of whether a new combustion source will produce soot that is more potent in a carcinogenicity bioassay than soot from current sources. A comparative-potency method for cancer risk assessment has been devel- oped and tested with estimates of human lung-cancer risk (Albert et al., 19831. The data base used was associated with several mixtures, including emissions from coke ovens, roofing-tar pots, cigarette-smoking, and automotive engines (Lewtas, 1983; Nesnow et al., 1982a,b; Lewtas et al., 19831. The method was based on the hypothesis that relative potency of different carcinogens in differ- ent bioassay systems is constant (Lewtas et al., 19831. The mathematical ex- pression for the constant-relative-potency model is as follows: (relative po- tency in bioassay 1/relative potency in bioassay 2) = k. This assumption of constancy is implicit in any comparison that uses the relative toxicity of two substances in animals to predict which one would be less toxic in humans. The assumption is testable if the relative potency of two mixtures or components can be determined in one bioassay (e.g., humans) and compared with the rela- tive potency in a second bioassay. The hypothesis was tested for three complex organic emissions from a coke oven, a roofing-tar pot, and cigarettes by using human lung-cancer data from epidemiologic studies of humans exposed to
48 COMPLEX MIXTURES these emissions (Albert et al., 1983) and testing the emissions in a series of short-term mutagenicity bioassays and animal tumorigenicity bioassays. The comparative approach has been widely used in the petroleum industry. In one such effort, the comparative acute toxicity was determined for a series of 19 petroleum hydrocarbon products ranging from light oils and gasoline to heavy fuel oils. The testing included standard oral and dermal acute and sub- chronic toxicity tests, as well as a series of sensitization and irritation tests (Beck et al., 19821. Another example involved shale-derived fuels and other synfuels whose toxicity was determined primarily in a wide variety of compar- ative studies, including general toxicity, target-organ, behavioral, mutagenic- ity, carcinogenicity, teratogenicity, and neurotoxicity tests (see MacFarland et al., 1984, and Mehlman et al., 19841. These studies were not conducted spe- cifically to provide a quantitative estimate of human risk associated with the mixtures, but rather to gauge the comparative toxicity potential of this series of mixtures. The design considerations for comparative-potency studies include factors that might not always be appropriate in other strategies. Some examples are the following: · Simultaneous evaluation of all comparisons in one experiment. This might tee desirable if bioassay variability between experiments is large. It is not possible in many chronic bioassays involving large numbers of animals, but it is often the best approach in some in vitro bioassays. This approach has been recommended specifically for comparative studies with the Salmonella typhi- murium reverse-mutation plate incorporation assay (Lewtas, 19831. · Exposure doses (or concentrations) needed for statistical analysis and potency measurement. The use of identical measures of dose across all the mixtures being studied usually facilitates the statistical analysis, including esti- mation of potency. If the potency range of the mixtures being tested is very large, the dose ranges for the assay might not overlap. Range-finding studies are usually needed to determine doses. Comparative potency (e.g., effect per unit dose or dose causing an effect) must be based on an exposure dose or concentration and can be expressed in any terms normally used to describe or characterize the toxic effect. Various statistical methods are available for deter- mining linear and nonlinear slopes, as well as the relative dose required to cause a specific effect (Lewtas et al., 19831. · Study objectives and integrated use of the data. The objectives of a spe- cific comparative-potency study and the expected use of the data are important in study design. Feder et al. (1984) have proposed a strategy for evaluating the toxicity of mixtures of gasoline blends that uses stagewise organization, frac- tional factorial designs, multiple bioassays, and a standardized reference fuel as a center point. (This "global" approach is discussed further in Chapter 5.) The purposes are to minimize the testing required to indicate the variability in
TESTING STRATEGIES AND METHODS 49 toxicity among members of a class of similar mixtures (e.g., blends of gaso- line) and to provide a data base that will allow future predictions based on composition and response-surface modeling. STRATEGIES RELATED TO CAUSATIVE AGENTS Bioassay-Directed Fractionation The objective of bioassay-directed fractionation is to identify the biologi- cally active (bioactive) components of mixtures. The choice of bioassays depends on the mixture being tested and on what is known about its chemical composition or the toxic effects. Bioassays that have been used include assays of tumor initiation, tumor promotion, mutagenicity, cytotoxicity, target-organ effects, and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) cellular receptor binding. Bioassay-directed fractionation has been extensively used to identify biolog- ically active components of cigarette smoke, particularly with cigarette-smoke condensate (CSC). Fractionation of CSC followed by mouse-skin carcinoge- nicity bioassays led to the identification of specific fractions of tobacco smoke that contain tumor-initiating activity and other fractions that contain tumor- promoting or carcinogenic activity (Bock et al., 1969; Hoffmann and Wynder, 1971~. Later, short-term mutagenicity and cellular-transformation bioassays were used to identify the most biologically active fractions and in some cases to identify specific chemicals in those fractions that could account for a portion of the mutagenic or tumorigenic activity (DeMarini, 19831. These studies con- cluded, for example, that polycyclic aromatic hydrocarbons in the neutral frac- tion contribute significantly to the tumor-initiating activity of the neutral frac- tion of CSC, whereas amines, aza-arenes, and other nitrogen-containing compounds account for much of the mutagenic activity of the basic fraction of CSC (DeMarini, 1983; IARC, 19861. (For additional information on this case, see Appendix C.) Nearly all the methods that have been described or are available for chemical analysis have been used to fractionate mixtures for bioassay. Often, several methods are used sequentially. Many of these approaches are described in Chapter 4, and the overall approach is diagrammed in Figure 3-5. From the perspective of the toxicologist, there are several important points in selecting fractionation methods and conducting these studies, including the following: · Recovery of mass and bioactivity. At each point in the fractionation scheme where bioactivity is measured, recovery of mass and bioactivity should be accounted for. If the sum of the bioactivities of the mass-weighted fractions equals that of the unfractionated mixture and that of a reconstituted mixture,
so Bioassay Activity O Mass Recovery 10% COMPLEX MIXTURES ~ Total Mixture l Bioassay Fractionate Mixture Is separated on chemical or physical properties Bioassay Activity Mass Recovery 1 1 1 l 3 I Fractions I+ (+) O 10% 30% 60% r I I ~ Subfractions I++ ~ 20% 70% Identify Components FIGURE 3-5 Diagrammatic representation of bioassay-directed fractionation where mixture is sequentially separated into fractions and fractions are bioassayed. Effort is concentrated on further fractionation and chemical characterization of biologically active fractions. then it at least appears that all the bioactivity is accounted for. One should be cautious, however: simultaneous chemical changes as a result of the fractiona- tion procedure could increase the bioactivity of some components and decrease that of others and thus result in the appearance of no change. · Identification of minor components. In several reported cases, the bioac- tive components of mixtures were minor, according to their mass, but-ex- tremely bioactive. Examples of such components are the strongly mutagenic dinitropyrenes, found in xerographic toners (Rosenkranz et al., 1980) and die- sel exhaust (Salmeen et al., 19841; the potent toxicant aflatoxin, found in pea- nuts (Dichter, 1984; Linsell, 19791; and the potent contaminants of poly- chlorinated biphenyls, such as polychlorinated dibenzofurans, found in emissions from burning transformers (Milby et al., 1985), and the bicy-
TESTING STRATEGIES AND METHODS S1 clophosphate ester found in the thermal decomposition products of a fire-re- tardant polyurethane foam (Petajan et al., 19751. · Quantitation of contribution of bioactive components. Once a highly bioactive fraction is obtained and chemical analysis provides a list of Me con- firrned or suspected components, additional information and sample availabil- i~ requirements must be anticipated. Bioactivity data on each of the com- pounds identified must be available. Pure standards must be obtained to assay the identified compounds and determine their contribution to the total activity observed, or fractionation and biologic testing must continue. Those issues are descnbed from the perspective of the analytic chemist in Chapter 4. Pairing Painug (depicted in Figure 3-6) combines information obtained indepen- dently from We literature or computer data bases with information on chemi- cals known to be bioactive and chemicals identified analytically as present in the mixture of interest. This approach is used in literature reviews (Bndbord and French, 1978; NRC, 1983; gingham et al., 19801. Chemical Search In a chemical search (as illustrated in Figure 3-7), bioassay data on pure chemicals and analytic data on the bioactive components of a mixture are used. Data Base 1 Chemicals In a Mixture - - Toxic Chemicals in a Mixture \ Data Base 2 \ Chemical Toxicity Data FIGURE 3-6 Two data bases can be paired or combined; overlap represents toxic chemicals identified in mixture. This method is described as pairing
52 (; Determine Toxicity of the Chemicals Chemical Toxicity Mutagenicity Carcinogenicity 2 3 o a) in - ._ n 1 ++ + + +++ · ~ ++ ++ +++ Mixture Chemical Analysis Chemical Presence Concentration o 2 3 n COMPLEX MIXTURES 10 ppm + o 0 s On l a) LU FIGURE 3-7 Illustration of chem~cal-search and effect-search methods for determining causative agents in refixture. This method has been used with nitroparaffins (Hite and Skeggs, 1979) and with polycyclic aromatic hydrocarbons (Bos et al., 19841. It can be used to evaluate a series of similar mixtures rapidly after bioassay-directed fractiona- tion has identified the class of compounds of interest, such as nitrated polycy- clic aromatic hydrocarbons (Pederson and Siak, 19811. Later, it can be used to search chemically for the most mutagenic compounds in other related mixtures (Nishioka et al., 19851. Effect Search Effect search (Figure 3-7) is the reverse of chemical search. Initially, chemi- cal methods are used to identify as many chemicals as possible in the mixture. That step is followed by bioassay studies or literature searches on the chemi- cals, in an attempt to identify those related to the effects of interest. The ap- proach has been widely used in dealing with hazardous wastes (Houk and
TESTING STRATEGIES AND METHODS 53 Claxton, 1986; Andon et al., 1984) and was used by Florin et al. (1980) to identify hundreds of compounds found in tobacco smoke. A combination of chemical-search and effect-search strategies has been used to identify acutely toxic agents in fires (Alarie and Anderson, 1979; Crane et al., 1977; Hilado, 1976; Klimisch et al., 1980; Levin et al., 1982; and Yusa, 19851. Concern about the toxicologically responsible single or multiple com- ponents in fire atmospheres has been approached through extensive chemical analyses (Levin, 1986~. The prediction of acute toxic hazard, which includes many other components in addition to toxicity, has used a computer model that integrates many of the necessary pieces of the puzzle. The strategy to deter- mine the toxicity factors needed for input into the hazard assessment has been to determine the individual and combined toxicities of a small select number of fire gases, predict the toxicity of the thermal decomposition products from materials based on the chemical analysis of the selected gases, and verify the prediction with a few animal exposures (Levin et al., in press). (For additional information on this case, see Appendix C.) Bioassay Identification Bioassay identification relies on the development of specialized bioassays that can detect chemicals of a particular class or structure. Examples of such assays are the TCDD receptor assay, monoclonal antibody assays, and geneti- cally engineered bacterial strains that can metabolize, and thus mutate, only after exposure to particular compounds. STRATEGIES RELATED TO PREDICTABI~TY AND MODELS The goal of strategies related to predictability is to predict the toxicity of materials related to a sample of known toxicity. In many instances, inadequacy of resources (such as time and sample availability) prevents the design and conduct of laboratory tests on a series of complex mixtures. Hazardous-waste sites, for example, present problems associated with temporal and spatial van- ability of leachate composition. It might be necessary to predict the risk con- nected to a time- and space-integrated sample that is representative of the prob- able human exposure. These strategies are best described as model-driven, because they are based on empirical, theoretical, statistical, or mechanistic models (described in greater detail in Chapter 51. Testing Mixture Components When the composition of a mixture is known and no biologic data are avail- able on the whole mixture, its toxicity can be approximated by summing the known toxicities of the components (see Chapter 5 and Appendix D). A weak-
54 COMPLEX MIXTURES ness of this approach is the likely lack of complete chemical composition data. Also, the strategy assumes that no interactions occur. That assumption can be tested biologically at several doses with interaction studies. Interaction studies have been performed with many chemicals. Usually, however, only two agents were examined; in a few, three were examined. The fire-toxicology case study illustrates the use of matrix testing to examine the interaction of mixture com- ponents. (For more information, see Appendix C.) When two substances are involved in toxicity testing, their individual char- acteristics are investigated completely, as well as those of the combination. Usually, dose-response relationships are worked out in detail for one agent and then essentially repeated in the presence of the second agent (Klaasen, 19861. The purpose is to establish whether the presence of the second agent alters the dose-response profile of the first. The two agents can be used as a mixture or sequentially. The protocols involved in this type of strategy are very effect- dependent. Essentially, the purpose of this strategy is to test the hypothesis that the effect seen is other then additive. Mitchell (1976) pointed out that, if both substances are active, the first step is to determine the potency of one relative to the po- tency of the other in effecting the response. The second step is to combine fractional doses of the substances and compare the results with the results of using standard doses of the substances individually. The null hypothesis is that the two substances will behave as though they are different forms of the same substance and hence produce additive effects. The modification of the dose-response curve of a toxicant given in combina- tion (concurrently or sequentially) with another substance can have important implications for prediction. These implications have been described for the potentiation of chloroform-induced liver injury by the pesticide chlordecone (Plea and Hewitt, 1982a) and the potentiation of haloalkane-induced liver and kidney injury by ketones or ketogenic agents (Plea and Hewitt, 1982b). If the dose-response curve is displaced to the left but is parallel to the curve obtained with animals that are not pretreated, the potentiating agent decreases the effec- tive dose of the toxicant under study. If the curve does not shift laterally (effec- tive dose is unchanged) but the slope ofthe curve increases, the response to the toxicant is exaggerated by the potentiating agent. If the curve is displaced to the left and the slope is increased, the potentiating agent lowers the effective dose of the toxicant and exaggerates the response to it. The implications of each situation are different. With a parallel lateral dis- placement, the effective dose is smaller, but the effect threshold can be esti- mated rather accurately. If the curve has rotated, the effective dose is unal- tered; the established no-observed-effect dose remains the same. If the curve has shifted laterally and the slope has changed, neither that dose nor the sever- ity of response can be predicted from the responses observed in the absence of the second chemical. Chlordecone pretreatment in mice results in a leftward
TESTING STRATEGIES AND METHODS 55 shift of the chloroform dose-response curve; in rats, however, the curve ap- pears both to shift laterally and to rotate (Plea and Hewitt, 1982a). In selecting doses for interaction studies when both agents are active, it is necessary to decide the fraction of the effective dose of each agent that should be administered. If the dose-response curves for the two agents are parallel, equivalent fractions of the median effective doses or equipotent doses of the agents can be selected. However, if the curves are not parallel, one must use equipotent doses; equivalent fractions of the median effective doses can result in misinterpretation of the action of the combination (Clausing and Bieleke, 1980; Mitchell, 19661. Dose-dependent interactions are not always monophasic linear relation- ships. In ketone potentiation of haloalkane-induced liver or kidney injury, bi- phasic phenomena can occur, with low doses of ketone potentiating and larger doses protecting (MacDonald et al., 1982; Hewitt and Plaa, 1983; Brown and Hewitt, 19841. In the case of simple mixtures, knowledge about the mechanisms of action facilitates the designing of experiments. The more one knows of how the toxic response is initiated and propagated, the better the study. It is essential to un- derstand the pharmacodynamic and pharmacokinetic characteristics of the agents under study. Mechanistic Studies and Models for Interactions An analysis of the mechanistic principles underlying the development of interactions has shown that only a few biologic phenomena can be mitigated or amplified through multichemical exposure (Witschi, 19821. Acute effects am- plified by interaction most often include cell death in particular organs or tis- sues, and such an event can often be traced back to the increased formation of toxic compounds. Potential mechanisms of chronic effects of interactions in- clude the formation of DNA adducts, prolongation of the life of free radicals, enhancement or impairment of transport across epithelial barriers, and induc- tion of mixed-function oxidases. Components of the disease process and its secondary effects (hepatotoxicity, nephrotoxicity, and tumor promotion) can influence the development of a primary disease. It might be appropriate to see whether mixtures or selected components af- fect the potential mechanisms for interaction. In many cases, that can be done in simple in vitro systems. Empirical Models The simplest experimental approach to estimating the effects of an unknown mixture is to derive an empirical model. The requirements for such a model are a biologic effect that can be quantified and one or more characteristics of the
56 COMPLEX MIXTURES mixture that can be quantified (e.g., boiling point and solubility). It is not necessary to know the chemical composition of the mixture or the causative agents, but knowledge of the concentration of one or more chemical constitu- ents could be useful. The extent to which an observed effect is correlated with measured charac- teristics indicates the predictive utility of the model, although it does not imply causation or mechanism. Aquatic Test Systems Although this report focuses on mammalian toxicity testing, the committee recognizes that aquatic toxicity test systems are being used for hazard evalua- tion and risk assessment in a number of fields, particularly in chemical-mixture toxicity tests. The standard acute toxicity tests are still performed in a number of laboratories and for the most part yield the evaluations required by regula- toIy agencies. In recent years, however, interest has shifted away from the standard tests, and more and more information concerning the biochemical aspects of aquatic species is becoming available. Along with the development of this data base, toxicologists are gaining new insight into how chemicals interact with various biochemical and physiologic pathways. In the last few years, a number of field studies have attempted to correlate chemical exposure to the concentrations of xenobiotic metabolizing enzymes (the mixed-ffinction oxidase system) in feral fish populations (Stegeman et al., 19861. In the laboratory, it has been shown that fish will respond to exposure to a number of chemicals by an increase in or induction of specific isozymes of the metabolizing system (tech et al., 1982; Buhler and Rasmusson, 1968~. Some of the enzymes of the rainbow trout and soup have been purified and cloned (Klotz et al., 1984; Williams and Buhler, 19831. A number of fish species have been shown to be sensitive to various chemi- cal carcinogens, both direct and indirect (e.g., diethylnitrosamine or methyl- azoxy methanol acetate). Exposure to these chemicals has resulted in rapid tumor formation (Stanton, 1965; Ishikawa et al., 1975) and unscheduled DNA synthesis (Ishikawa et al., 19781. Because of these types of responses, there is a great deal of interest in their use as tier II test species in the evaluation of chemicals and chemical mixtures as potential carcinogens. INTEGRATION OF STRATEGIES The strategy for toxicity testing of mixtures depends on at least three factors: complexity of the mixture, knowledge of the composition (constituents) and effects of the mixture, and the problem or question. The influence of each of these factors, individually and in combination, on testing strategy is examined below.
TESTING STRATEGIES AND METHODS Sr? COMPLEXITY OF MIXTURE As the complexity of a mixture increases, the likelihood that its toxic compo- nents can be completely characterized decreases. The optimal testing strategy forge simplest two-component mixture is to test the components in such a way as to determine whether combining them causes interactions. Toxicologic test- ing of components with a design that elucidates interactions and their mecha- nisms will provide data that can be useful in models for predicting the toxicity of other simple mixtures that contain the same components. For combinations of more than two constituents, Me task of assessing the interactions of all components becomes overwhelming and impractical, be- cause the number of binary combinations equals (n2 _ nyl2. Environmental mixtures are therefore evaluated by adm~n~stenng either the total mixture or a few fractions or ingredients. If We latter approach is selected, efficient expen- mental designs can be extracted from the statistical literature on what are called fractional factorial designs. These are discussed in Chapter 5. Figure 3-8 de- scnbes a design for the preliminary assessment of 16 substances with 24 mice. Note Me sequences within each group and the fact that the number of animals A GROUP B GROUP C GROUP S1S2S3S4 S5S6S7S8 SgS10S11S12 s1 a S5 GROUP Sg S13 MICE ~ ~ ~ ~ ~ ·E ~M~E Z Me 2 MICE 2 MICE 2 MICE 2 MICE 1 MICE 2 MICE s2 GRgUP s1o S14 GROUP s3 s7 S11 S15 s4 GROUP s8 S12 S16 FIGURE 3-8 A fractional factorial design for the preliminary assessment of 16 substances with 24 mice. The intersection of the ~ group and the C group is cross-hatched to indicate a toxic response in one of the animals in that cell.
58 COMPLEX MIXTURES exposed to each pair of substances is at least double the number in each cell. The intersection of the ~ group and the C group is cross-hatched to indicate a toxic response in one of the animals in that cell. To identify the particular substances responsible for that response, the seven substances assigned to the cell could be divided into another fractional factorial design or tested in pairs individually. KNOWLEDGE OF CONSTITUENTS AND EFFECTS The selection of strategies and their application to toxicologic testing are often to a great extent either agent-driven or effect-driven. For example, if a very complex mixture is known to be carcinogenic and mutagenic, the driving force will often be the identification of agents (or components) that are causing the effect. That will lead to a bioassay-directed fractionation strategy. The research published earlier on unleaded-gasoline vapors and diesel particle emissions has been driven by the observed carcinogenic effect (MacFarland et al., 1984). When an agent or class of agents, such as TCDD or polychlorinated bi- phenyls (PCBs), is known to be present in a mixture, the testing will often be driven to elucidate the effects or the influence of other components on an iden- tified effect. Knowledge of components or effects has direct influence on the analytic and toxicologic testing strategy. When neither the effects nor the con- stituents are known, there is a driving force to elucidate both, and the strategy will be influenced more by the specific problems or questions being asked about the mixture. PROBLEMS OR QUESTIONS IN RELATION TO STRATEGIES Problem definition and the formulation of specific questions are important steps that must be taken before selection of a testing strategy. Examples of the types of complex-mixture questions and proposed strategies were discussed early in this chapter. They can be grouped into questions regarding effects and risk, causative agents, and predictive value. The questions usually arise sequentially, so questions about the effects of a mixture would logically be raised before questions on causative agents. Ques- tions on predictive value, however, have been raised when both a mixture's effects and agents were known and when very little was known about it other than what some of its constituents were. It is more productive to address the question of predictive value with interaction studies when effects and agents are known. When neither the effects nor all the agents are known, the strategy should be to elucidate the agents and effects before attempting to address how
TESTING STRATEGIES AND METHODS TABLE 3-1 Relationship Between Questions/Problems and Associated Strategies Questions Strategies Effects and risks Toxicologic evaluation of mixtures Tier/screening Matrix/comparative Causative agents Bioassay-directed fractionation Predictive value Toxicologic evaluation of components 59 and whether predictions could be made from knowledge of a limited number of constituents. The strategies can clearly be grouped according to the general type of questions they address, as shown in Table 3-1. FORMULATION OF OVERALL STRATEGY Each of the three factors above needs to be considered in selecting the final strategy for toxicologic studies (Figure 3-9~. Most research on mixtures in- volves a logical progression of strategies as knowledge is gained about effects and agents. As new concerns are raised over exposures to mixtures, the factors discussed above should be considered in selecting strategies to provide data useful in answering questions about potential human risk and in selecting steps to protect public health. Figure 3-10 diagrams a hypothetical flow of a process that would consider each of the above factors and use the strategies presented here to address specific problems and questions. EFFECTS En l Known Unknown ~ ._ ],_ FIGURE 3-9 The selection of a strategy for toxicologic studies to increase knowledge of both agents and effects.
60 COMPLEXITY AGENTS PROBLEM DEFINITION SELECTION OF STRATEGY COMPLEX MIXTURES Mixture INFORMATION determine complexity ......... ...... Simple 111 Known Unknown C_ Complex fir Chemical Composition Known I Effects I .o 0 ° c .O 3 E x ~ 0 0 _ ._ _ ~ 0 ~ _ 0 Agents s Unknown ~$~.~.~$.~Y. . z ~ ~ mm A) z a: Cal Is _ ~ c' _' c Is Q In O O O ~ ~ O t~ t ~ ~ O X If O ~ _ All O Bioassay-Directed Fractionation or other Causative Agent Strategy 1 1 FIGURE 3-10 Illustration of effect of knowledge on selection of testing strategy and logical sequence from first determining effects of mixture and then determining causative or toxic agents. REFERENCES Alarie, Y. C., and R. C. Anderson. 1979. Toxicologic and acute lethal hazard evaluation of thermal decomposition products of synthetic and natural polymers. Toxicol. Appl. Pharmacol. 51:341-362. Albert, R. E., J. Lewtas, S. Nesnow, T. W. Thorslund, and E. Anderson. 1983. Comparative potency method for cancer risk assessment application to diesel particulate emissions. Risk Anal. 3:101-117. Andon, B., M. Jackson, V. Houk, and L. Claxton. 1984. Evaluation of Chemical and Biological Methods for the Identification of Mutagenic and Cytotoxic Hazardous Waste Samples. Health Ef- fects Research Lab., Research Triangle Park, N.C. (Available from NTIS as PB84-201615.) (27 pp.) Beck, L. S., D. I. Hepler, and K. L. Hansen. 1982. The acute toxicology of selected petroleum hydrocarbons, pp. 1-12. In H. N. MacFarland, C. E. Holdsworth, J. A. MacGregor, R. W. Call,
TESTING STRATEGIES AND METHODS 61 and M. L. Kane (eds.). The Toxicology of Petroleum Hydrocarbons. American Petroleum Institute, Washington, D.C. gingham, E., R. P. Trosset, and D. Warshawsky. 1980. Carcinogenic potential of petroleum hydrocar- bons. A critical review of the literature. J. Environ. Pathol. Toxicol. 3:483-563. Bock, P. G., A. P. Swain, and R. L. Stedman. 1969. Bioassays of major fractions of cigarette smoke condensate by an accelerated technic. Cancer Res. 29:584-587. Bos, R. P., J. L. G. Theuws, C. M. Leijdekkers, and P.T. Henderson. 1984. The presence of the mutagenic polycyclic aromatic hydrocarbons benzo[a]pyrene and benz[a]anthracene in creosote Pi. Mutat. Res. 130:153-158. Bridbord, K., and J. G. French. 1978. Carcinogenic and mutagenic risks associated with fossil fuels, pp. 451-463. In P. W. Jones and R. I. Preudenthal (eds.). Carcinogenesis. Vol. 3: Polynuclear Aromatic Hydrocarbons. Raven Press, New York. Brown, E. M., and W. R. Hewitt. 1984. Dose-response relationships in ketone-induced potentiation of chloroform hepato- and nephrotoxicity. Toxicol. Appl. Pharmacol. 76:437-453. Buhler, D. R., and M. E. Rasmusson. 1968. The oxidation of drugs by fishes. Comp. Biochem. Physiol. 25:223-239. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975a. Petroleum hydrocarbon toxicity studies. I. Methodology. Toxicol. Appl. Pharmacol. 32:246-262. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975b. Petroleum hydrocarbon toxicity studies. II. Animal and human response to vapors of varnish markers' and painters' naphtha. Toxicol. Appl. Pharmacol. 32:263-281. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975c. Petroleum hydrocarbon toxicity studies. III. Animal and human response to vapors of Stoddard Solvent. Tox- icol. Appl. Pharmacol. 32:282-297. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975d. Petroleum hydrocarbon toxicity studies. IV. Animal and human response to vapors of rubber solvent. Toxicol. Appl. Pharmacol. 33:526-542. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975e. Petroleum hydrocarbon toxicity studies. V. Animal and human response to vapors of mixed xylenes. Toxicol. Appl. Pharmacol. 33:543-558. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975f. Petroleum hydrocarbon toxicity studies. VI. Animal and human responses to vapors of "60 Solvent." Toxicol. Appl. Pharmacol. 34:374-394. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975g. Petroleum hydrocarbon toxicity studies. VII. Animal and human response to vapors of "70 Solvent." Toxicol. Appl. Pharrnacol. 34:395-412. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., L. J. Sullivan, and J. M. King. 1975h. Petroleum hydrocarbon toxicity studies. VIII. Animal and human response to vapors of " 140° Flash Aliphatic Solvent." Toxicol. Appl. Pharmacol. 34:413-429. Carpenter, C. P., E. R. Kinkead, D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1976a. Petroleum hydrocarbon toxicity studies. IX. Animal and human response to vapors of "80 Thinner." Toxicol. Appl. Pharmacol. 36:409-425. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1976b. Petroleum hydrocarbon toxicity studies. X. Animal and human response to vapors of "50 Thinner." Toxicol. Appl. Pharmacol. 36:427-442. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1976c. Petroleum hydrocarbon toxicity studies. XI. Animal and human response to vapors of de- odorized kerosene. Toxicol. Appl. Pharmacol. 36:443-456. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and I. M. King. 1976d. Petroleum hydrocarbon toxicity studies. XII. Animal and human response to vapors of "40 Thinner." Toxicol. Appl. Pharmacol. 36:457-472.
62 COMPLEX MIXTURES Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1976e. Petroleum hydrocarbon toxicity studies. XIII. Animal and human response to vapors of toluene concentrate. Toxicol. Appl. Pharmacol. 36:473-490. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1977a. Petroleum hydrocarbon toxicity studies. XIV. Animal and human response to vapors of "High Aromatic Solvent." Toxicol. Appl. Pharmacol. 41 :235-249. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1977b. Petroleum hydrocarbon toxicity studies. XV. Animal response to vapors of "High Naphthe- nic Solvent." Toxicol. Appl. Pharmacol. 41:251-260. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, and J. M. King. 1977c. Petroleum hydrocarbon toxicity studies. XVI. Animal response to vapors of "Naphthenic Aromatic Solvent." Toxicol. Appl. Pharmacol. 41:261-270. Carpenter, C. P., D. L. Geary, Jr., R. C. Myers, D. J. Nachreiner, L. J. Sullivan, andJ. M. King. 1978. Petroleum hydrocarbon toxicity studies. XVII. Animal response to n-nonane vapor. Toxicol. Appl. Pharmacol. 44:53-61. Clausing, P., and R. Bieleke. 1980. Aspects of methodology employed in the investigation of com- bined chemical effects on acute oral toxicity. Arch. Toxicol. Suppl. 4:394-395. Crane, C. R., D. C. Sanders, B. R. Endecott, J. K. Abbott, and P. W. Smith. 1977. Inhalation Toxicol- ogy: I. Design of a Small Animal Test System. II. Determination of the Relative Toxic Hazards of 75 Aircraft Cabin Materials. FAA-AM-77-9. U.S. Federal Aviation Administration, Oklahoma City, Okla. DeMarini, D. M. 1983. Genotoxicity of tobacco smoke and tobacco smoke condensate. Mutat. Res. 114:59-89. Dichter, C. R. 1984. Risk estimates of liver cancer due to aflatoxin exposure from peanuts and peanut products. Food Chem. Toxicol. 22:431-437. Feder, P. I., E. Margosches, and J. Bailar. 1984. A Strategy for Evaluating the Toxicity of Chemical Mixtures. Draft Report. EPA Contract No. 68-01-6721, Task 76. U.S. Environmental Protection Agency, Washington, D.C. Florin, I., L. Rutberg, M. Curvall, and C. R. Enzell. 1980. Screening of tobacco smoke constituents for mutagenicity using the Ames test. Toxicology 15:219-232. Heussner, J. C., J. B. Ward, Jr., and M. S. Legator. 1985. Genetic monitoring of aluminum workers exposed to coal tar pitch volatiles. Mutat. Res. 155: 143-156. Hewitt, W. R., and G. L. Plaa. 1983. Dose-dependent modification of 1,1-dichloroethylene toxicity by acetone. Toxicol. Lett. 16: 145-152. Hilado, C. J. 1976. Relative toxicity of pyrolysis products of some foams and fabrics. J. Combust. Toxicol. 3:32-60. Hite, M., and H. Skeggs. 1979. Mutagenic evaluation of nitroparaff~ns in the Salmonella typhimu- rium/mammalianmicrosome test and the micronucleus test. Environ. Mutagen. 1 :383-389. Hoffmann, D., and E. L. Wynder. 1971. A study of tobacco carcinogenesis. XI. Tumor initiators, tumor accelerators, and tumor promoting activity of condensate fractions. Cancer 27:848-864. Houk, V. S., and L. C. Claxton. 1986. Screening complex hazardous wastes for mutagenic activity using a modified version of the thin-layer chromatography Salmonella assay. Mutat. Res. 169: 81 -92. Huff, J. 1982. Carcinogenesis bioassay results from the National Toxicology Program. Environ. Health Perspect. 45: 185-198. International Agency for Research on Cancer (IARC). 1986. IARC Monographs on the Evaluation of Carcinogenic Risk of Chemicals to Humans. Vol. 38: Tobacco Smoking. International Agency for Research on Cancer, Lyon, France. Ishikawa, T., T. Shimamine, and S. Takayama. 1975. Histologic and electron~microscopy observa- tions on diethylnitrosamine-induced hepatomas in small aquarium fish (Oryzias latipes). J. Natl. Cancer Inst. 55:909-916.
TESTING STRATEGIES AND METHODS 63 Ishikawa, T., S. Takayama, and T. Kitagawa. 1978. Autoradiographic demonstration of DNA repair synthesis in ganglion cells of aquarium fish at various ages in vivo. Virchows Arch. B. 28:235-242. Klaasen, C. D. 1986. Principles of toxicology, pp. 11-32. In C. D. Klaasen, M. O. Amdur, and J. Doull (eds.). Casarett and Doull's Toxicology: The Basic Science of Poisons. 3rd ed. Macmillan, New York. Klimisch, H. J., H. W. M. Hollander, and J. Thyssen. 1980. Comparative measurements of the toxic- ity to laboratory animals of products of thermal decomposition generated by the method of DIN 53 436. J. Combust. Toxicol. 7:209-230. Klotz, A. V., J. J. Stegeman, and C. Walsh. 1984. Multiple isozymes of hepatic cytochrome P-450 from the marine teleost fish soup (Stenotomus chrysops). Mar. Environ. Res. 14:402-404. Lech, J. J., M. J. Vodicnik, and C. R. Elcombe. 1982. Induction of monooxygenase activity in fish, pp. 107-148. In L. J. Weber (ed.). Aquatic Toxicology. Vol. 1. Raven Press, New York. Levin, B. C. 1986. A Summary of the NBS Literature Reviews on the Chemical Nature and Toxicity of the Pyrolysis and Combustion Products from Seven Plastics: Acrylonitrile-Butadiene-Styrenes (ABS), Nylons, Polyesters, Polyethylenes, Polystyrenes, Poly(Vinyl Chlorides) and Rigid Polyure- thane Foams. NBSIR 85-3267. National Bureau of Standards, Gaithersburg, Md. Levin, B. C., A. J. Fowell, M. M. Birky, M. Paabo, A. Stolte, and D. Malek. 1982. Further Develop- ment of a Test Method for the Assessment of the Acute Inhalation Toxicity of Combustion Products. NBSIR 82-2532. National Bureau of Standards, Gaithersburg, Md. Levin, B. C., M. Paabo, J. L. Gurman, and S. E. Harris. In press. Effects of exposure to single or multiple combinations of the predominant toxic gases and low oxygen atmospheres produced in fires. Fundam. Appl. Toxicol. Lewtas, J. 1983. Evaluation of the mutagenicity and carcinogenicity of motor vehicle emissions in short-term bioassays. Environ. Health Perspect. 47:141-152. Lewtas, J., S. Nesnow, and R. E. Albert. 1983. A comparative potency method for cancer risk assess- ment: Clarification of the rationale, theoretical basis, and application to diesel particulate emissions. Risk Anal. 3:133-137. Linsell, C. A. 1979. Decision on the control of a dietary carcinogenaflatoxin, pp. 111-112. In W. Davis and D. Rosenfeld (eds.). Carcinogenic Risks: Strategies forIntervention. (IARC Scientific Publications No. 25.) IARC, Lyon, France. MacDonald, J. R., A. J. Gandolfi, and I. G. Sipes. 1982. Acetone potentiation of 1,1,2-trichloro- ethane hepatotoxicity. Toxicol. Lett. 13:57-69. MacFarland, H. N., C. E. Holdsworth, J. A. MacGregor, R. W. Call, and M. L. Kane, eds. 1984. Applied Toxicology of Petroleum Hydrocarbons. Advances in Modern Environmental Toxicology. Vol. VI. Princeton Scientific Publishers, Princeton, N.J. Mehlman, M. A., C. P. Hemstreet III, J. J. Thorpe, and N. K. Weaver, eds. 1984. Renal Effects of Petroleum Hydrocarbons. Advances in Modern Environmental Toxicology. Vol. VII. Princeton Scientific Publishers, Princeton, N.J. Milby, T. H., T. L. Miller, and T. L. Forrester. 1985. PCB-containing transformer fires: Decontamina- tion guidelines based on health considerations. J. Occup. Med. 27:351-356. Mitchell, C. L. 1966. Effect of morphine and chlorpromazine alone and in combination on the reaction to noxious stimuli. Arch. Int. Pharmacodyn. Ther. 163:387-392. Mitchell, C. L. 1976. The design and analysis of experiments for the assessment of drug interactions. Ann. N.Y. Acad. Sci. 281:118-135. NCI (National Cancer Institute). 1976. Report of the Subtask Group on Carcinogen Testing to the Interagency Collaborative Group on Environmental Carcinogenesis. NCI, Bethesda, Md. Nesnow, S., L. L. Triplett, and T. J. Slaga. 1982a. Comparative tumor-initiating activity of complex mixtures from environmental particulate emissions on SENCAR mouse skin. JNCI 68:829-834. Nesnow, S., C. Evans, A. Stead, J. Creason, T. J. Slaga, and L. L. Triplett. 1982b. Skin carcinogene- sis studies of emission extracts, pp. 295-320. In J. Lewtas (ed.). Toxicological Effects of Emissions from Diesel Engines. Elsevier, Amsterdam.
64 COMPLEX MIXTURES Nishioka, M. G., C. C. Chuang, B. A. Petersen, A. Austin, and J. Lewtas. 1985. Development and quantitative evaluation of a compound class fractionation scheme for bioassay-directed characteriza- tion of ambient air particulate matter. Environ. Int. 11: 137-146. NRC (National Research Council). 1977. Principles and Procedures for Evaluating the Toxicity of Household Substances. National Academy of Sciences, Washington, D.C. (130 pp.) NRC. 1983. Feasibility of Assessment of Health Risks from Vapor-Phase Organic Chemicals in Gaso- line and Diesel Exhaust. National Academy Press, Washington, D.C. NRC. 1984. Toxicity Testing. National Academy Press, Washington, D.C. Pederson, T. C., and J. S. Siak. 1981. The role of nitroaromatic compounds in the direct-acting mutagenicity of diesel particle extracts. J. Appl. Toxicol. 1(2):54-60. Petaj an, J. H., K. J. Voorhees, S. C. Packham, R. C. Baldwin, I. N. Einhorn, M. L. Grunnet, B. G. Dinger, and M. M. Birky. 1975. Extreme toxicity from combustion products of a fire-retarded polyurethane foam. Science 187:742-744. Plaa, G. L., and W. R. Hewitt. 1982a. Methodological approaches for interaction studies: Potentiation of haloalkane-induced hepatotoxicity, pp. 67-96. In H. F. Stich, H. W. Leung, and J. R. Roberts (eds.). Workshop on the Combined Effects of Xenobiotics, Ottawa, Ontario, Canada, June 22-23, 1981. NRCC No. 18978. National Research Council Canada, Associate Committee on Scientific Criteria for Environmental Quality, Ottawa, Canada. Plaa, G. L., and W. R. Hewitt. 1982b. Potentiation of liver and kidney injury by ketones and ketogenic substances, pp. 65-75. In H. Yoshida, Y. Hagihara, and S. Ebashi (eds.). Advances in Pharmacol- ogy and Therapeutics II. Proceedings of the 8th International Congress of Pharmacology, Tokyo, 1981. Vol. 5: Toxicology and Experimental Models. Pergamon, New York. Rosenkranz, H. S., E. C. McCoy, D. R. Sanders, M. Butler, D. K. Kiriazides, and R. Mermelstein. 1980. Nitropyrenes: Isolation, identification, and reduction of mutagenic impurities in carbon black and toners. Science 209: 1039-1043. Salmeen, I. T., A. M. Pero, R. Zator, D. Schuetzle, and T. L. Riley. 1984. Ames assay chromatograms and the identification of mutagens in diesel particle extracts. Environ. Sci. Technol. 18:375-382. Scala, R. A. In press. Motor gasoline toxicity. Fundam. Appl. Toxicol. Stanton, M. F. 1965. Diethylnitrosamine-induced hepatic degeneration and neoplasia in the aquarium fish, Brachydanio rerio. J. Nat. Cancer Inst. 34:117-130. Stegeman, J. J., P. J. Kloepper-Sams, and J. W. Fa~nngton. 1986. Monooxygenase induction and chlorobiphenyls in the deep-sea fish Coryphaenoides armatus. Science 231: 1287-1289. Williams, D. E., and D. R. Buhler. 1983. Multiple forms of cytochrome P-448 and P450 purified from ,B-naphthoflavone fed rainbow trout. Abstract No. 3618. Fed. Proc. 42:910. Witschi, H. P. 1982. Altered tissue reactivity and interactions between chemicals, pp. 263-288. In Assessment of Multichemical Contamination: Proceedings of an International Workshop, Milan, Italy, April 28-30, 1981. National Academy Press, Washington, D.C. Yusa, S. 1985. Development of laborato~y test apparatus for evaluation of toxicity of combustion products of materials in fire, pp.471-487. In Seventh Joint Panel Meeting of the UJNR Panel on Fire Research and Safety. Proceedings. NBSIR 85-3118. National Bureau of Standards, Gaithersburg, Md.
