Complex Mixtures: Methods for In Vivo Toxicity Testing (1988)

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.

- Concepts for Analyzing Human Exposure to Complex Chemical Mixtures People are seldom, if ever, exposed to single chemicals. Instead, most natu- ral and artificially produced substances are mixtures of chemicals, such as those in acid deposition, fire products, hazardous wastes, pesticides, surface drinking water, and the products of Mel distillation or combustion. Appendix A discusses the origins of major categories of the complex mixtures that people encounter today. A few studies of human populations exposed to particular mixtures have demonstrated toxic effects (see Appendix B). Analyses of ex- perimental studies, mostly toxicologic studies in animals, have later verified the toxic nature of a mixture or even identified its specific toxic constituents. However, the number of mixtures identified through this process of epidemio- logic hypothesis generation followed by toxicologic verification is small. This chapter discusses many of the problems inherent in the process, not the least of which is that such a system is premised on confirming past human hazards, rather than seeking to predict, and hence prevent, such hazards. The impetus, therefore, for studying complex mixtures is the need to gener- ate reliable predictions on which to base estimates of potential health risks for humans and ultimately to guide exposure controls. Except for clinical trials of therapeutic agents, deliberate high-dose exposure of large populations to mate- rials of unknown potential for harm is unacceptable in modern society. The prediction of human health risks uses a "weight-of-evidence" approach that considers data from various sources—epidemiology, analyses of accidental or intentional human exposures, animal tests, cell culture studies, metabolic and mutagenic assays, and so on and weights the data from all those sources in relation to their relevance to human populations (WHO, 1981; Ballantyne, 19851. For any chemical exposure, an effect of concern must be identified and a dose-response relationship must be estimated. The dimensions of the dose- 10

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 11 response relationship usually prove particularly troublesome, primarily be- cause estimates of dose (biologically effective dose, as described in Chapter 1) are poor in most collections of epidemiologic evidence. Even the quantifica- tion of exposure can be difficult. This chapter begins with a discussion of factors that affect dose and dose surrogates (using the case of inhalation exposure to demonstrate the complex- ity of the process), including biologic, analytic, and statistical problems that typically arise in dose characterization. With that discussion as a base, the second and third sections discuss the qualitative and quantitative problems associated with attempting to characterize human exposures. Those discus- sions include not only the complexity of the materials to which people are exposed and the complexity of the exposed populations, but also the interactive effects that can ensue from exposure to mixtures of biologically active sub- stances.. In theory, the most relevant effects data are those generated from human exposures. In practice, however, such data are usually unavailable or so poorly characterized (by confounding exposures, inadequate dose quantification, etc.) that data from experimental animal studies must be relied on. The fourth section of this chapter discusses various types of data that are commonly col- lected on both human and animal exposures, in an attempt to show how they might be compared, and suggests reasons for potential divergence. The final section of this chapter presents eight examples of human experi- ence with complex mixtures that involve a number of failures to anticipate human harm and includes discussion of types of data that could have been helpful in predicting problems. (For readers interested in more complete ac- counts, Appendix B contains more detailed chronologies of those examples.) GENERAL CONCEPTS Exposure represents a contact between a chemical agent and an object. This report is concerned with exposure of living organisms, particularly humans. Exposure, which may be considered an independent variable, is different from the dose that reaches a target organ or tissue (the biologically effective dose or, simply, dose), which may be considered a dependent variable. The magnitude of a dose is a reflection of the magnitude of an exposure modified by a series of intervening processes, including inhalation or ingestion; transfer of inhaled or ingested material across epithelial membranes of the skin, respiratory tract, and gastrointestinal tract; transport via circulating fluids to target tissues; and uptake by the target tissues. Those processes in turn vary with a subject's activity level, age, sex, health status, and inherent constitutional makeup with respect to race or species, size, and so on. When people are subjected to chronic or repetitive exposures to chemical agents, other important factors affect the magnitude of the dose. If the agent or

12 COMPLEX MIXTURES its effect on the target tissue is cumulative and clearance of the agent or recov- e~y from its effect is slow, the dose is a function of the total cumulative uptake. But if the agent is rapidly eliminated or its effect is rapidly and completely reversible on removal from exposure, the dose is a function of the rate of uptake. (See the 1985 NRC report Epidemiology and Air Pollution for other discussion of these concepts.) DOSE AND DOSE SURROGATES We can seldom measure dose directly. The two most commonly used surro- gates for dose are exposure and biologic fluids from exposed persons, which can provide evidence that absorption has taken place or that biologically effec- tive doses have been delivered. Dose surrogates can include the presence ofthe chemical of interest itself in blood, urine, or exhaled air or the presence of metabolites or enzymes induced by uptake of the chemical. Therefore, assess- ing exposure and its resulting dose is a complex process involving a set of multidisciplinary activities (NRC, 1985~. The generation of data useful for dose estimation through sampling is a function of two processes. The first is the design of sampling protocols to obtain biologically relevant data; the second is the actual collection of samples themselves. Techniques for sampling air, water, food, and biologic fluids are relatively well developed, as are techniques for sample separation from copol- lutants, media, and interferences and for quantitative analysis. The design of appropriate sampling protocols requires an understanding of the critical vari- ables that determine the magnitude of dose. We need to know when, where, how long, and at what rate and frequency to collect samples to obtain data relevant to the exposures of interest; obtaining such data requires knowledge of the temporal and spatial variabilities of exposures and concentrations and knowledge of metabolic pathways and kinetics. We seldom have enough infor- mation of these kinds to guide our collection of samples. The use of exposure as a dose surrogate introduces the further problem of inadequate quantitative data, particularly in the case of retrospective studies. Assignments to the various exposure groups, either "exposed" versus "nonex- posed" or graded exposures, are usually based on work histories, question- naires, and some knowledge of process conditions that justify judgments about relative magnitudes of exposure. All these methods of assessing exposure are indirect and lead to a degree of misclassification (NRC, 1986a). However, when they are used in conjunction with measurements of biologic fluids or of exposure, the rate of misclassification can be estimated. The issues just described are related to quantifying dose on the basis of exposure. These issues are more complicated when the exposure is to complex mixtures, rather than single entities.

CONCEPTS FOR ANALYZING HUMAN EXPOSURE INDICATORS OF EXPOSURE TO COMPLEX MIXTURES AS DOSE SURROGATES 13 Because it is impractical or impossible to measure the concentrations of all the constituents of a complex environmental mixture in an exposure environ- ment or in body fluids, exposures or retained amounts are sometimes estimated from data on one or several constituents or their metabolites. The use of indica- tors can be rationalized on the basis either that the retention of the constituent is representative of the retention of one constituent of interest or that the toxicity of the marker is representative of the toxicity of the mixture. An example of a representative constituent is benzoLa~pyrene (BaP), which is one of the most frequently used indicator compounds for complex organic mixtures. However, BaP alone does not always accurately reflect the carcino- genic potency of a complex mixture. Its concentration in air has been corre- lated with relative risks of lung cancer for coal-tar workers, coke-oven work- ers, and others (Speizer, 19831. It is a potent animal carcinogen, and it can be routinely analyzed by analytic techniques of proven reliability (IARC, 19831. Although BaP is likely to be present to some degree in most environmental mixtures, its mass or biologic activity can be highly variable (NRC, 19851. An epidemiologic study of roofers exposed to very high concentrations of BaP found some excess lung cancer among those exposed for more than 20 years (Hammond et al., 19761. The standardized mortality ratios (SMRs) for those exposed for 9-19, 20-29, 30-39, and 40 or more years were 92, 152, 150, and 247, respectively. (The smoking histories of the workers were not known.) Airborne BaP concentrations measured in roofing operations range from 14 ,ug/m3 in the roof-tarring area to 6,000 ~g/m3 in the coal-tar roofing kettle area (Sawicki, 1967) . The amounts of BaP recovered from masks, which were worn by a minority of roofers studied, indicated an average of 16.7,ug of BaP inhaled per day (Hammond et al., 19761. Another group exposed to high concentrations of airborne BaP were British gasworkers, who inhaled about 30 ,ug/day (Lawther et al., 19651. The mainstream smoke of a cigarette contained about 3.5 x 10-2,ug of BaP in 1960 (Kotin and Falk, 19601. Thus, a 2-pack/day smoker inhaled about 1.4 ,ug/day. Ambient urban air in 1958 contained about 6 ng/m3, which could account for an inhaled mass of about 120 ng/day (i.e., 0. 12 ,ug/day). Ambient rural air has about 10% as much BaP (Faoro, 1975~. The research cited above on roofers exposed to BaP indicates that their ambi- ent exposures are considerably higher than those of cigarette smokers. How- ever, the rate of lung cancer is much higher for cigarette smokers than for roofers. This indicates that BaP at best is a crude indicator of the carcinogenic potential of a complex mixture. This example is pertinent because it. suggests that additional compounds in a complex mixture can be important in the ex-

14 COMPLEX MIXTURES pression of the response and that the physical phase of the indicator compound can play a role. There are also indicators for pollutant mixtures with other hearth effects. The U.S. Environmental Protection Agency (U.S. EPA, 1986) uses ozone (03) as an indicator for the presence of photochemical oxidants, nitrogen dioxide (NO2) as anindicator for the various nitrogen oxides, sulfur dioxide (SO2) as an indicator for the various sulfur oxides, and the mass concentration of sus- pended particulate matter (PM) for all pollutants present as droplets or solid particles. In the committee's judgment, the validity of these indicator chemi- cals as hazard indexes varies from very good (03) to very questionable (PM). EFFECTS OF MATRIX AND PHYSICAL STATE ON DOSE Risk of health effects is generally considered to be a function of both toxicity and exposure, so it is important to consider bioavailability to determine the extent of potential exposure once a potentially toxic agent has been identified by chemical analysis. Bioavailability defines the biologically effective dose, a variable fraction of the administered dose, and should be considered in estimat- ing the body burden of a chemical. Bioavailability can be affected (either in- creased or decreased) by agents or conditions that alter release, uptake, metab- olism, distribution, or biologic effects. With knowledge of bioavailability, one can better understand the biologic end points and characterize the toxicity of a complex mixture with regard to additivity, antagonism, and synergism. If bioavailability assays are conducted immediately after chemical characteriza- tion, the risk assessor will have important end points with which to develop a hazard evaluation by minimizing the uncertainty and clarifying exposure infor- mation (U.S. EPA, 19851. An example of how bioavailability can differ is the case of 2,3,7,8-tetra- chlorodibenzo-p-dioxin (TCDD), whose bioavailability from environmental samples can vary from less than 0.1 % up to 85 %, depending on the matrix to which it is bound, the media from which it entered the environment, the dura- tion of the binding to environmental substrates, and the presence of other com- pounds in the mixture. This can be best illustrated by studies conducted with TCDD-contaminated soil samples from Times Beach, Missouri, and Newark, New Jersey (De Caprio et al., 1986; McConnell et al., 1984; Umbreit et al., 1986b; Van der Berg et al., 19851. Both soils were contaminated by a number of agents in addition to TCDD. The biologic assay of the contaminated soils involved administering TCDD at equivalent concentrations to rats and guinea pigs and monitoring for appearance of TCDD-related symptoms. A compari- son revealed that both soils induced cytochrome P-450 enzyme systems, but only the Times Beach soil induced the TCDD syndrome and death in guinea pigs. These differences in response were associated with the ease of extraction

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 15 of TODD and other agents (shaking in solvent followed by column chromatog- raphy) from the Times Beach samples, in contrast with the rigorous methods (48-72 hours of exhaustive Soxhlet extraction) required to extract similar chemicals from Newark soils (Umbreit et al., 1986b). The bioavailability of components was the decisive factor in the evaluation of the toxicologic hazard of these environmental samples. An underlying concern of researchers and regulators is whether the risk assessor should consider only the presence of xenobiotics in a medium and the toxicity of these xenobiotics (assuming expo- sure) or should attempt to use bioavailability data to complete the exposure assessment. The organic content appeared to be higher and binding sites appeared to be more abundant in Newark soil than in Times Beach soil. Whether it can be generalized that matrices with higher organic content have greater binding affinity than similar matrices with lower organic content is not clear. Poiger and Schlatter (1980) and Rappe et al. (1986) showed that fly ash, sediments, and carbonaceous materials bind several organic compounds, but have a high affinity for chlorinated dibenzodioxins and dibenzofurans. The presence of solvents or the continued release of solvents at a site might aid in the percola- tion of compounds through the soil and increase binding to soil particles. This phenomenon has been hypothesized as an explanation for some of the soil binding of polychlorinated biphenyls (PCBs) in Japan. Some of the constituents of a mixture might alter the absorption of others, or the mixture might alter gastrointestinal transit time and thus affect absorption of several compounds, including nutrients (Hollander, 19811. Analytic meth- ods are available to determine differential uptake; indeed, Bandiera et al. (1984) have recently demonstrated selective hepatic and adipose retention of specific chlorinated dibenzofurans from complex mixtures of them and PCBs found at Yusho, Japan. Bioavailability varies with route of exposure. The oral, dermal, and respira- tory routes differin selectivity of uptake, rejection, and storage. Thebioavaila- bility of a constituent of a complex mixture will depend on that constituent's solubility, volatility, charge, and concentration and on those of other com- pounds in the mixture. Research efforts and resources should be aimed at dif- ferential bioavailability of constituents of complex mixtures in different physi- cal states. The case of fuels and fuel exhausts or effluents might be slightly simpler, because the compounds (those not particle-bound) are in a nonpolar liquid medium (highly lipid-soluble) or in a vapor phase; bioavailability is a function of membrane effects and differential uptake in the absence of a matrix (Klaasen, 19861. Bioavailability also depends on the physical state of the mixture. Several investigators have studied the bioavailability of dioxins and PCBs from liquid or semisolid media and have found reasonable agreement between theoretical

16 COMPLEX MIXTURES and actual biologic values. However, in the case of mixtures bound to solid substrates, there are marked differences from site to site (Umbreit et al., 1986a). FACTORS INFLUENCING RELATIONSHIPS BETWEEN EXPOSURE AND DOSE Estimating human pulmonary damage involves making a number of as- sumptions about effective dose. The problems of estimating the effective dose of inhaled materials are described here to illustrate the complexity of the dose estimation process. The surface and systemic uptake of chemicals from inhaled air depends both on the physical and chemical properties of the chemicals and on the anatomy and pattern of respiration within the airways. Gases and vapors rapidly contact airway surfaces by molecular diffusion. For water-soluble gases, dissolution or reaction with surface fluids on the airways facilitates removal from the air- stream. Highly water-soluble vapors, such as SO2, are almost completely re- moved in the airways of the head, and very little penetrates into lung airways. For water-insoluble compounds, surface uptake is limited, and thus the decline in concentration with depth in the lung can be low. For such compounds, the greatest uptake is often deep in the lung, where the residence time and surface areas are the greatest. For airborne particles, a critical characteristic affecting surface deposition patterns and efficiencies is particle size. A model to predict the percentage deposition of particles in various regions of the respiratory tract was developed by the Task Group on Lung Dynamics (ICRP, 1966) of the International Com- mission on Radiological Protection. Almost all the mass of airborne particulate matter is found in particles with diameters greater than 0.1 ,um. The penetration of airborne particles into the lung airways is determined primarily by convective flow that is, the motion of the air in which the particles are suspended because these particles are small in relation to the airways in which they are suspended. Some deposition by diffusion does occur for particles smaller than 0.5 ,um in small airways, whereas for particles larger then 0.5 ~m, deposition by sedimentation occurs in small to midsized airways. For particles with aerodynamic diameters greater than 2 ,um, particle inertia is sufficient to cause particle motion to deviate from the flow streamlines; that results in deposition by impaction on surfaces down- stream of changes in flow direction, primarily in midsized to large airways that have the highest flow velocities. The extent of deposition on limited surface areas within the large airways is of special interest with respect to dosimet~y and the pathogenesis of chronic lung diseases, such as bronchial cancer and bronchitis. Quantitative aspects of particle deposition are summarized in a recent air

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 17 quality criteria document (U.S. EPA, 19821. The deposition efficiencies in different regions of the human respiratory tract are highly variable among healthy subjects. Additional variability results from structural changes in the airways associated with disease processes. Generally, these changes involve airway narrowing or localized constrictions that act to increase deposition and concentrate it on limited surface areas. The preceding discussion of particle deposition is based on the assumption that each particle has a specific size. However, hydroscopic particles enlarge considerably as they take up water vapor in the airways (NRC, 19781. Materials that dissolve in the mucus of the conductive airways or the surfac- tant layer of the alveolar region can rapidly diffuse into the underlying epithelia and the circulating blood, thereby gaining access to tissues throughout the body. Chemical reactions and metabolic processes can occur within the lung fluids and cells, limiting access of the inhaled material to the bloodstream and creating reaction products with either greater or less solubility and biologic activity. Few generalizations about absorption rates are possible for individual chemicals, let alone for complex chemical mixtures. Particles that do not dissolve at deposition sites can be translocated to remote retention sites by passive and active clearance processes. Passive transport depends on movement on or in surface fluids lining the airways. There is a continuous proximal flow of surfactant to and onto the mucociliary escalator, which begins at the terminal bronchioles, where it mixes with secretions from Clara and goblet cells. Within midsized and large airways, there are additional secretions from goblet cells and mucous glands, producing a thicker mucous layer having a serous subphase and a more viscous gel layer. The gel layer, lying above the tips of the synchronously beating cilia, is found in discrete plaques in smaller airways and becomes more nearly continuous in the larger airways. The mucus reaching the larynx and the particles carried by the mucus are swallowed and enter the gastrointestinal tract. The total transit time for particles deposited on ciliated conductive airways extending to the terminal bronchioles varies from about 2 to 24 hours in healthy nonsmoking humans. Macrophage-mediated particle clearance via the bron- chial tree takes place over a period of several weeks. The particles deposited in alveolar zone airways are ingested by alveolar macrophages within about 6 hours; but turnover of the macrophages normally takes several weeks. At the end of several weeks, the particles not cleared to the bronchial tree via macro- phages have been incorporated into epithelial and interstitial cells, from which they are slowly cleared by dissolution or as particles passing through pleural and eventually hilar and tracheal lymph nodes via lymphatic drainage path- ways. Clearance rates for these later phases depend strongly on the chemical nature of the particles and their sizes, with half-times ranging from about 30 days to 1,000 days or more. All the characteristic clearance times cited above refer to inert, nontoxic

8 COMPLEX MIXTURES particles in healthy lungs. The presence of toxicants can drastically alter clear- ance times. Inhaled materials that affect mucocilia~y clearance rates include cigarette smoke (Albert et al., 1974, 1975), sulfuric acid (Lippmann et al., 1982; Schlesinger et al., 1983), O3 (Phalen et al., 1980), SO2 (Wolff et al., 1977), and formaldehyde (Morgan et al., 1984~. Macrophage-mediated alveo- lar clearance is affected by SO2 (Ferin and Leach, 1973), NO2 (Schlesinger, 1986), sulfuric acid (Schlesinger, 1986), O3 (Phalen et al., 1980; Schlesinger, 1986), and silica dust (Jammet et al., 19701. Cigarette smoke is known to affect the later phases of alveolar zone clearance in a dose-dependent manner (Bohning et al., 19821. Both clearance pathways and rates can be affected by these toxicants, which thus influence the distribution of retained particles and their dosimet~y. Given the sparseness of current knowledge and dose-related clearance rate data, pulmonary retention of complex chemical mixtures cannot be predicted. That might pose greater difficulties for interpreting exposures to complex mix- tures in occupational settings, where the exposures could be high enough to alter clearance pathways and rates. For general population exposures at lower concentrations, it might be reasonable to assume, as a first approximation, that pulmonary clearance rates are nodal. The above discussion illustrates the importance of determining characteris- tics of biologic systems directly affected by exposure. The physiology of nor- mal and abnormal or diseased organs should be well characterized. EFFECT OF POPULATION AND EXPOSURE COMPLEXITY ON QUALITATIVE ASSESSMENT Although the choice of appropriate indicators for environmental measure- ments of exposure and the nature of bioavailability are complicated issues, perfect knowledge of these would still leave many other problems in assessing exposure of human populations unresolved. Epidemiologic studies necessarily assess the effects of both exposures and other factors that occur in the complex and variable social and physical environment in which people live and work. Few exposures, even to single chemicals, occur in the absence of other envi- ronmental factors. COMPLEXITY OF EXPOSURE In occupational or general environmental settings, it is important to know the temporal and concentration characteristics of exposures. Short-term high- level exposures might occur in some industrial settings, for example, during recharge or cleaning of normally sealed batch reaction chambers. Long-term low-level exposures occur in the general environment; for example, a broad

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 19 spectrum of the population, some of whom might be in compromised health, receive relatively low-level exposures to air pollutants continuously. Expo- sures to mixtures of chlorination byproducts in drinking water are usually at low levels, but occur frequently (NRC, in press). Because exposure can start in infancy and continue throughout a lifetime, the cumulative dose can be substantial. Similar types of exposure can occur among different groups of exposed per- sons at different levels and frequencies. For example, x irradiation fordiagnos- tic purposes occurs at relatively low levels and infrequently in the general population and usually results in a cumulative lifetime dose of under 10 reds (Redford, 19861. But x rays are also used frequently at much higher levels to control tumor growth in cancer patients. In the past, x irradiation was also used to treat many other medical conditions that were not life-threatening, such as severe acne, tinea capitis, and anlylosing spondylitis. As with x rays, exposure to industrial and environmental chemicals can vary in intensity and frequency across the population. Important considerations in designing epidemiologic and toxicologic studies are the origin and degree of inherent complexity of mixtures, as well as the ways in which exposure to them can be modified or controlled to decrease the risk of disease. Adverse effects of exposure to proprietary blends of well- characterized substances often can be ameliorated simply by removing or substituting for toxic constituents, but this strategy is obviously ineffective for mixtures whose toxic factors are integrally related to their intended use. Prod- ucts of combustion from a variety of sources and chlorination byproducts re- sulting from the disinfection of drinking water are examples of this type of mixture (NRC, 1986a,b, in press). Identifying all the toxic components in these cases is usually not possible. Even if the toxic components could be identified, removing individual constituents is seldom technologically feasible (two notable exceptions are reduction of the nicotine content of tobacco and the caffeine content of coffee and cola drinks); control consists of reducing expo- sure to all constituents, toxic and nontoxic alike. Cigarette smokers, for exam- ple, who switch to filtered brands reduce exposure to the total mixture, but not necessarily to the key toxic ingredients; the obvious way to decrease exposure to the toxic ingredients is to decrease or cease smoking. Decreases in the expo- sure of workers and the general public to coke-oven emissions are accom- plished by controlling all the emissions, not only the most toxic fractions. Concentrations of organic substances in drinking water can be controlled by activated-carbon filtration. Activated carbon removes the relatively small frac- tion of chemicals that are toxic, as well as the bulk of the ones that are not. Thus, in designing toxicologic testing approaches, one must consider whether the toxic components are intrinsic to the mixture and difficult to control as individual substances or whether they are easily identifiable and potentially separable from the mixture.

20 COMPLEX MIXTURES COMPLEXITY OF POPULATIONS In populations, the exposure setting and nature of the affected groups can profoundly influence the type and severity of an adverse effect. Are exposures limited primarily to the occupational setting, or is the general population at risk? If the latter, special attention is warranted for high-risk groups, such as infants, the elderly, pregnant and lactating women, the poor, and others whose sensitivity might be increased by previous exposure, chronic medical condi- tions, or inherited characteristics. Even if groups are not identified as being at especially high risk, age, sex, and normal variation of metabolic characteris- tics can have important consequences for the reaction of exposed persons. Nutritional status and preexisting medical conditions in the host organism might have a profound influence on the response to environmental exposures. For example, a variety of micronutrients appear to confer protection against chemical carcinogens (NRC, 1983a). That can occur through direct interaction of the micronutrients with a potential carcinogen or indirectly through enzy- matically mediated detoxification, whereby the micronutrient is a critical en- zyme cofactor. Selenium, for example, is a necessary cofactor of glutathione peroxidase, which serves as a reducing agent for potentially carcinogenic per- oxides (Ip and Sinha, 1981~. Some micronutrients that might confer protection from carcinogens are vitamin A, many of the retinoids, carotene, vitamin C, and indoles (NRC, 1983a). It is important, then, to consider nutritional status of test animals in evaluating complex mixtures. Some human groups might be especially sensitive to particular environmen- tal exposures. People might be sensitized to the dermal effects of irritating chemicals by extensive previous exposures. Asthmatics could suffer severe bronchoconstriction on exposure to sulfur oxides at concentrations that would have little or no impact on nonasthmatics. INTERACTIVE EFFECTS OF MULTIPLE EXPOSURES Toxicologic testing of complex mixtures can answer some questions raised by observations among humans. As discussed below, experimental designs must carefully consider exposure conditions, host characteristics, appropriate health end points, and potential interactions between constituents of mixtures and multiple types of exposure. Interactive effects of concurrent exposure to two or more known risk factors have been evaluated in a number of epidemiologic studies. Some examples are lung cancer resulting from exposure to cigarettes and radon daughters (Little et al., 1965; Rajewsky and Stahlhofen, 1966; Pershagen et al., 1987), lung can- cer due to cigarette-smoking and asbestos exposures (Selikoff et al., 1968; Berm et al., 1972; Hammond and Selikoff, 1973), and oral cancer from alco- holic-beverage consumption and smoking (McCoy et al., 19801. Statistical

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 21 modeling of the available epidemiologic data in most cases has not resolved and is unlikely to resolve whether these exposures interact in an additive or a synergistic manner. Even less clear is the most appropriate biologic or physio- logic model that can explain the action of multiple environmental insults. Toxi- cologic testing of joint effects of two or more chemical exposures, under care- fully controlled conditions, would help answer many of the outstanding questions. Toxicologic evaluation of complex mixtures can address many of the ques- tions raised specifically by epidemiologic studies of workplace and other expo- sures to mixtures. Large-scale analyses of extensive data bases with occupa- tional and mortality data have identified groups of workers in several industries at high risk of cancer and other diseases. Some of these surveys have used mortality records kept by vital-statistics bureaus in several states that now in- clude the usual data on occupation and industry, in addition to underlying cause of death and other items abstracted from death certificates (Milham, 1983, 1985~. Several surveys of this type have been conducted in the United States and Britain (see, for instance, Office of Population Censuses and Surveys, 1978; Logan, 19821. Typically in these studies, the proportion of deaths by cause has been analyzed within each of many occupations or industries to indicate the causes of death that might be linked to specific job classifications. The exposure in these studies is limited to inferences based on the occupation or industry entered on the death certificate. Many occupational groups that show excess risk of dying of particular causes have been identified by these surveys and warrant more detailed evaluation with both epidemiologic and toxicologic approaches. (Further discussion of the use and limitations of these surveys can be found in Kazantzis and McDonald, 1986, and Hernberg, 1986.) For many occupational groups at high risk, specific exposure information is often not available or is too complex to interpret. That is true for several indus- tries with occupations at high risk of cancer that are the subject of several recent monographs from the International Agency for Research on Cancer (IARC) in the series IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans. The first 24 volumes of the IARC monograph series restricted attention to the carcinogenic properties of individual chemicals. But starting in 1981, with Volume 25 (Wood, Leather, and Some Associated Indus- trzes), and continuing with Volume 28 (The Rubber Industry, 1982) and Vol- ume 34, Part 3 (Polynuclear Aromatic Compounds: Industrial Exposures in Aluminium Production, Coal Gasification, Coke Production, and Iron and Steel Founding, 1984), the monographs have reviewed the carcinogenic risk of employment in selected industrial occupations. The changed focus in this se- ries of reports reflects a growing recognition that exposure to complex mixtures or multiple insults in some occupational settings is not amenable to simple analysis or interpretation on the level of individual chemicals. In these indus- tries, workers who held specific jobs with exposure to complex mixtures are

22 COMPLEX MIXTURES shown to be at increased risk of cancer, although specific exposures cannot be identified. Further chemical characterization of mixtures, followed by their toxicologic testing, would help to clarify excess cancer risk among workers in a number of workplace settings described in these monographs. The rubber industry (IARC Monographs, Volume 28, 1982) provides an excellent illustration of the potential value of further chemical characteriza- tions and toxicologic testing. Through a collective-bargaining agreement, the United Rubber Workers and several U.S. rubber end tire companies funded an extensive series of occupational studies on the risk of cancer and other dis- eases. These studies revealed excess risk of several cancers within specific job categories. For example, they revealed excess mortality due to lymphocytic leukemia among employees whose jobs might have exposed them to benzene and other solvents. But the epidemiologic data were unable to discriminate further among the solvents. Toxicologic testing of solvent mixtures thus could help to pinpoint the sources of excess risk in this industry and could serve as a basis for removing or replacing offending substances. COMPLEXITY OF HEALTH EFFECTS Cancer is but one end point that should be considered in evaluating health effects of complex mixtures. Similar types of exposure can induce other ef- fects, depending on the conditions of exposure, exposure setting, and host characteristics. Some examples that might be appropriate for toxicologic eval- uation include mutagenesis, teratogenesis, neurotoxicity, bronchoconstric- tion, hepatotoxicity, renal insufficiency, and a variety of dermatologic effects. Systematic review of the epidemiologic literature can help to focus toxicologic testing of particular mixtures on questions directly applicable to the human experience. CONSIDERATIONS IN QUANTITATION OF HUMAN EXPOSURES IN EPIDEMIOLOGIC STUDIES Documented effects of environmental chemicals on humans seldom contain quantitative exposure data and only occasionally include more than crude ex- posure rankings based on known contact with or proximity to the materials believed to have caused the effects. Interpretation ofthe available human expe- rience requires some appreciation of the uses and limitations of the data used to estimate the exposure side of the exposure-response relation. The discussion that follows is an attempt to provide some relevant background for interpreting the available, retrospective data and for specifying the kinds of data that might be collected in prospective studies. There are both direct and indirect sources of exposure data and data that can

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 23 be used to rank exposed subjects into exposure-intensity groups. Among the quantitative direct approaches are direct measurements of external exposures and measurements of exposure via analysis of biologic fluids. The indirect measures generally rely on work or residential histories and data on exposure intensity at each exposure site or on enumeration of the frequency of process upsets or effluent discharges that result in high-intensity short-term exposures. AMBIENT EXPOSURE The problems of exposure characterization differ with the nature of the expo- sure. Many epidemiologic studies of complex mixtures are based on occupa- tional exposures. These involve repetitive daily exposures lasting up to about 8 hours, with uptake predominantly by inhalation, but sometimes also by skin absorption. Most studies of the health effects of community air pollution in- volve complex mixtures with highly variable temporal exposure patterns (NRC, 19851. The biologically relevant exposure times can also be highly variable. The temporal patterns of ambient concentration can depend on photo- chemical reaction sequences and strong gradients between indoor and outdoor concentrations. Ingestion exposures from food and water depend strongly on methods of food preparation, dietary preferences and variety, and other nutri- tional factors or deficiencies. CONCENTRATIONS IN AIR, WATER, AND FOOD Historical data are occasionally available on the concentrations of materials of interest in environmental media. But the data might or might not relate to the exposures of interest. We must ask some important questions before attempt- ing to use such data: · How accurate and reliable were the sampling and analytic techniques used in the data collection? Were they subjected to any quality assurance protocols? Were standardized or reliable techniques used? · When and where were the samples collected, and how did they relate to exposures at other sites? Air concentrations measured at fixed (area) sites in industry might be much lower than those occurring in the breathing zone of workers close to the contaminant sources. Air concentrations at fixed (gener- ally high) community air sampling sites can be either much higher or much lower than those at street level and indoors, owing to gradients in source, strengths, and rates of pollutant decay in indoor and outdoor air. · What is known or assumed about the ingestion of food or water containing the measured concentrations of the contaminants of interest? Time at home and dietary patterns are highly variable among at-risk populations.

24 COMPLEX MIXTURES BIOLOGIC SAMPLING DATA Many of the questions that apply to interpretation of environmental media concentration data also apply to biologic samples, especially quality assur- ance. The time of sampling is especially critical in relation to the times of the exposures and to the metabolic rates and pathways. In most cases, it is difficult to separate the contributions to circulating concentrations from recent expo- sures and those from long-term reservoirs. EXPOSURE HISTORIES Exposure histories themselves are generally unavailable. However, work histories and residential histories can yield information on exposure, as can reliable data on norms with respect to hygiene programs in specific industries. Job histories, as discussed above, are often available in company or union records and can be converted into relative rankings of exposure groups with the aid of long-term employees and managers familiar with the work processes and the history of process changes. Knowledge about materials handled and about tasks performed is critical. So is knowledge about the installation and effec- tiveness of engineering modification for exposure control. Routine, steady- state exposures might be the most important and dominant exposures of inter- est in many cases. But for some health effects, occasional or intermittent peak exposures could be of greater or primary importance. In assessing or accumu- lating exposure histories or estimates, it is important to collect evidence on the frequency and magnitude of the occasional or intermittent releases associated win process upsets. SOME STATISTICAL ISSUES IN QUANTITATION OF EXPOSURE The preceding discussions have addressed some complications in the assess- ment of exposure for complex mixtures, including the parameters of exposure- response relations and measurement of exposure. The section below addresses the role of statistical methodology in these issues and highlights problems for which additional methodologic research is desirable. A number of descriptive and mathematical models have been developed to permit estimation of exposure from knowledge of exposure and one or more of the following factors: translocation, metabolism, and effects et the site of toxic action. The use of these models for airborne particulate matter generally requires a knowledge of the concentration within specific particle-size intervals or of the particle-size distribution of the chemicals of interest. The most widely used of these models was developed by the International Commission on Radiological

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 25 Protection and Measurement (ICRP, 19661. The model describes, in probabi- listic terms, the sites of deposition of particles of various sizes in the lungs. A key analysis issue is the consequence of mismodeling exposure history in estimating or testing an exposure-response relation or association. For exam- ple, in a study of the association between one or more air pollutants and the risk of chronic respiratory impairment, we must determine whether cumulative, peak, recent, or some other measure of exposure history is used. In many if not most situations, biologic knowledge is not sufficiently detailed to indicate which measure is most appropriate. We know from general principles that using an incorrect measure can result in biased estimates of the exposure- response relation and in reduced sensitivity (i.e., statistical power) in the detec- tion of an exposure-response association (see, for instance, Inskip et al., 19871. Another analysis issue is the consequence of inexact measurements of expo- sure. In most applications, the measured value of exposure is only an approxi- mation of the actual that is, absorbed exposure. Two key issues are the effects of inexact exposure on the estimation of the exposure-response relation and the loss of statistical power for detecting an association between exposure and response. Exposure assessment issues also have important implications for the design of exposure-response studies. Although it often is clear that obtaining exact exposure information would be prohibitively expensive, relatively little is known about how to balance the added value of more precise exposure data with the corresponding increase in cost. This problem can be regarded as one of experimental design, and statistical literature on methods of determining de- signs that are "optimal" for some prespecified criteria, which account for both cost and accuracy, could play an important role in the design of future expo- sure-response studies. PREDICTIVE VALUE OF LABORATORY STUDIES FOR HUMAN EFFECTS OF ENVIRONMENTAL EXPOSURES The rationale for using animal studies to evaluate the human risks associated with environmental pollutants is that evolution has endowed mammalian spe- cies with similar genetic, biochemical, and physiologic makeups. These simi- larities extend to toxification and detoxification mechanisms and to target sites for the adverse effects of pollutants. Toxicity testing is based on principles discovered in the study of comparative toxicology, which is founded on the basic disciplines of comparative biology, comparative physiology, and comparative biochemistry. Those disciplines have demonstrated that successful extrapolation of animal toxicity studies to

26 COMPLEX MIXTURES humans requires knowledge not only of general similarities among species, but also of differences between species. Interspecies comparisons of toxic effects of pollutants can be carried out at a number of levels of varied complexity. These levels include gross end points (such as lethality, tumorigenesis, and mutation induction in bacteria), more descriptive toxicokinetic events (such as absorption, distribution, metabolism, and elimination), and changes at the cellular level. Although molecular toxi- cology holds great promise, preliminary findings from this emerging field indi- cate that basic biochemical pathways are very similar for all forms of life and nearly identical for related species. That implies that molecular mechanisms of toxicity are qualitatively the same for related species and that interspecies dif- ferences in toxicity are due primarily to quantitative (toxicokinetic) differences in biochemical pathways. At present, the most productive approach to toxicity testing continues to be bioassays for gross toxic end points in species related to humans (mammals). Agents that have adverse effects in test animals are assumed to be dangerous to humans, although the degree and specific manifestation of toxicity in humans might differ from those in test animals, because of underlying toxicokinetic variables. For that reason, successful extrapolation of animal toxicity data to humans requires comparing animal and human toxicity data and elucidation of toxicokinetic events that are relevant to interspecies variability in toxicity. The process is aided by knowledge gained from surveys of occupationally exposed persons and from both unintentional and carefully controlled experimental hu- man exposures. In rare cases, such as when therapeutic agents are given to critically ill persons, animal and human toxicity data can be compared directly. EXPOSURE AS A VARIABLE Acute responses to short-term exposures in animal experiments are often difficult to compare with analogous events in human populations. Short-term human exposures to toxic amounts of substances usually result from accidental spills or workplace accidents that often are difficult to characterize. Imagine the difficulty in accurately determining the human LDso for methylisocyanate from the Bhopal tragedy. In addition, most toxicologic research is oriented toward establishing safe concentrations for long-term exposures, which are generally much lower than concentrations that produce acute toxicity. Human and animal acute responses to short-term exposures have been com- pared directly for some anticancer drugs (Freireich et al., 1966; for a review, see Calabrese, 19831. In these studies, maximal tolerable doses (MTDs) of anticancer drugs that physicians gave to patients as a normal part of cancer chemotherapy were compared with MTDs for the same drugs in a number of different species. When the doses were expressed per unit area of body surface, all species, including humans, had comparable MTDs.

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 27 Most of the agents tested in the above studies were electrophilic alkylating agents or antimetabolites for which there is apparently little interspecies differ- ence in critical toxicokinetic variables. However, many chemicals exhibit sub- stantial interspecies toxicokinetic variability that accounts for differences in toxic effects between humans and test animals. (This topic is discussed in more detail later in this chapter.) In contrast with the short-term testing mentioned above, most toxicity test- ing is oriented toward protecting human populations from chronic disease that results from long-term exposures. The most widely studied chronic disease associated with long-term exposures to environmental pollutants is cancer. It is therefore not surprising that studies of carcinogenesis provide the best example of the utility of animal toxicity testing to predict human diseases that result from exposure to environmental pollutants. The concentration and chronicity of the exposure can influence responses to single agents or to mixtures of agents. Nitrogen dioxide in concentrations present in polluted ambient air (0.05-1.0 ppm) can in the presence of sunlight oxidize organic compounds and produce peroxides and peroxyacetyl nitrate, a strong conjunctival irritant. High concentrations of NO2, such as those in ciga- rette smoke (50-100 ppm), can produce ciliastasis and inflammatory reaction in the respiratory tract. When cigarette smoke is inhaled chronically in amounts generated from one to two packs per day for 20 or more years, chronic lesions develop in several organs, including the lung and cardiovascular system, and neoplasms develop that are not predictable from any of the acute lesions. Em- physema and chronic bronchitis develop as a result of chronic prolonged ciga- rette-smoke inhalation; however, their development in humans cannot be pre- dicted from the acute pulmonary effects associated with cigarette-smoke inhalation in animals (U.S. Surgeon General, 19841. Coronary arterial disease and hypertension both chronic lesions of the cardiovascular system cannot be predicted by laboratory studies of cigarette smoke, even with chronic and prolonged exposures (U.S. Surgeon General, 19831. Neoplasms of the lung, larynx, mouth, and esophagus likewise are not produced in animals by inhala- tion exposure to cigarette smoke (U.S. Surgeon General, 19821. COMPLEXITY OF THE AGENT AS A VARIABLE In general, the simpler the agent, the more likely that the laboratory data will demonstrate changes that are useful as a guide to its actions in humans. The acute effects of NO2 are an example. NO2 at concentrations over 40 ppm for 12 hours will produce profound, and even fatal, pulmonary edema in humans, as well as in experimental animals (NRC, 1977, 19811. As the agent-becomes increasingly complex, so too does the effect on the animal. Beagles exposed to SO2 and diluted auto exhaust showed increased airway resistance and de-

28 COMPLEX MIXTURES creased lung volume and compliance (Hyde et al., 1978; Bloch et al., 1972), and they developed mild emphysema. Cigarette smoke is a mixture of great complexity with medical manifesta- tions (NRC, 1986a). Cigarette smoke contains over 3,800 compounds, each in a different concentration (Dube and Green, 19821. It has a particulate phase and a gaseous phase. The nature and amounts of particles generated depend on the volume of the inhaled puff, the temperature of the burning tobacco, its dilution with air, the nature of tobacco, and the distance the smoke must travel from ignited tobacco to target organ. New polycyclic aromatic hydrocarbons are generated by pyrolysis; in addition, new compounds are formed by oxida- tion, and free-radical formation initiates peroxide formation in unsaturated fatty acid and alkoxyl intermediates. The biologic effects of this galaxy of compounds are virtually impossible to predict from a knowledge of the individ- ual constituents, not only because of their huge number, but also because tran- sient chemicals generated during the processes of pyrolysis, oxidation, and free-radical formation might dissipate or change with time and temperature. LIMITATIONS OF ANIMAL MODELS In several ways, laboratory animals respond in a manner different from that of humans when exposed to some environmental agents. Allergy and hyper- sensitivity, common in humans, appear not to affect most laboratory animals in a similar manner. Atmospheric pollutants can irritate and aggravate respiratory abnormalities in people with allergic diatheses or bronchospastic airways. Hy- persensitivities to ingested agents such as aspirin, antibiotics, and other drugs can commonly produce organic diseases, including pulmonary fibrosis and drug-induced hepatitis, that are not commonly observed in animals. Fur- thermore, laboratory animals do not commonly develop chronic disease states such as arteriosclerosis, emphysema, and malignant neoplasms- in re- sponse to small exposures to environmental toxicants. For instance, neither emphysema nor arteriosclerosis has been adequately created experimentally as a result of cigarette-smoke inhalation or other agents. Finally, animal studies are not reflective of human disease when the agent or mixture tested causes a recurrence or exacerbation of an existing disease. Peo- ple with allergic asthma, chronic nonspecific reactive bronchitis, or acute in- fections might respond vigorously to inhaled irritant mixtures, whereas healthy animals exposed to similar mixtures will not. In spite of the previously described difficulties associated with the identifica- tion of human carcinogens, 30 chemicals or chemical mixtures and 9 industrial processes have been identified as definitive human carcinogens studies on humans have confirmed that the agents cause cancer in those who are exposed (IARC, 1982; Vainio et al., 1985; Rall et al., 19871. A review of IARC mono- graphs found 288 chemicals, industrial processes, and complex mixtures for

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 29 which some data on human carcinogenicity or sufficient evidence of carcinoge- nicity in experimental animals existed (Vainio et al., 19851. For the purpose of classifying carcinogenic risk in humans, IARC concludes that, in the absence of epidemiologic studies, sufficient evidence of carcinogenicity in animals is reasonable grounds for regarding a chemical as carcinogenic in humans (IARC, 1986~. In a study of the adequacy of toxicologic testing of chemicals in commerce, a National Research Council committee reported that most such chemicals are not adequately tested for potential toxicity (NRC, 1983b). Fur- thermore, many substances cannot be evaluated for carcinogenicity in humans, because adequate data on human exposure are not available (Karstadt and Bobal, 1982~. Several chemicals have been shown to be carcinogenic in hu- mans, but cannot be adequately tested in a laboratory setting. A 1975 report (NRC, 1975) reviewed much of the information and stated: It may be noted that the organs affected in man are not always the same as those that are found to develop tumors in laboratory animals, nor is the the organ specificity the same in different species of rodents. Nevertheless, the evidence, based on a limited number of carcinogens, suggests that most agents that pose a carcinogenic threat to man will be carcin- ogenic in laboratory tests on animals. However, this leaves open the possibility that such tests may also identify chemicals carcinogenic to rodents that do not pose such a threat to man. That report went on to describe a number of examples in which the organ specificity, a qualitative effect, is the same in humans and animals but the magnitude of the effect is decidedly different (aflatoxin Be and diethylstilbes- trol) or in which the effect in the most sensitive animal species is qualitatively different from that in humans (vinyl chloride) or in any other animal species (benzidine). Of course, these comparisons have been made in the absence of information on molecular dosimetry. However, when study results are com- pared on the basis of milligrams of substance per kilogram of body weight, both qualitative and quantitative differences between humans and animals are observed (NRC, 19751. In some circumstances because of methodologic shortcomings, species differences, or particular end points of interest animal studies might not be predictive of human carcinogenicity. In studies of SO2 and suspended particu- late matter, animal studies have not adequately reflected the effects of polluted atmospheric air in humans (see Appendix B for details). The use of laboratory animals has been of limited value in the study of combined exposures, such as exposure to radon daughters and cigarette smoke and to asbestos and cigarette smoke (see Appendix B for details). The effects of radon daughters and of asbestos alone are adequately reflected in animal stud- ies. But in experimental studies of combined exposures in which cigarette smoke is one of the agents, it is not reliably delivered to the target organ in adequate quantities. Adequate investigation of the effects of inhalation expo-

30 COMPLEX MIXTURES sure to cigarette smoke requires long-term chronic studies, which are expen- sive and difficult. An effective alternative has been to apply the concentrated extract of tobacco tars to a surrogate organ during exposure to the other agent, such as radon daughters or asbestos. However, the predictive validity of this altered method for the human conditions can be questioned. This strategy could be applied to studies of noncancer effects of other complex mixtures to provide cost efficiency and time-saving. Studies in laboratory animals provide the best means available to anticipate and hence prevent human harm. TOXICOMNETICS Differences in toxicokinetics are often the explanation for interspecies vari- ability in susceptibility to toxicants. Much of the research in this subject con- sists of studies of efficacy and side effects of drugs. In many cases, molecular events have been elucidated that are critical control points for the toxico- kinetics of chemicals. One of the control points is the activity of toxification and detoxification processes. For example, the duration of activity of hexobar- bital varies greatly from species to species and is directly correlated with spe- cies ability to metabolize the parent compound to an inactive form (Weiss, 19781. Variations in intestinal and dermal absorption can also influence the activity of an agent; sheep and cattle absorb smaller percentages of ingested lead than humans and thus have greater tolerances for dietary lead (Scharding and Oehme, 19731. After absorption, toxicants often must cross other biologic barriers to reach their targets of action. Species differ in the ability of toxicants to pene- trate barriers, such as the placenta (Koppanyi and Avery, 1966) and the blood- brain barrier (Way, 1967; Rall, 1965, 19711. The differences often explain interspecies variability in sensitivities to teratogens and neurotoxins. Species susceptibilities to toxic effects can vale with differences in plasma-protein binding, bilia~y excretion, and enterohepatic circulation (Albert, 1979; Smith, 1973), all of which can affect the metabolism, distribution, and elimination of toxicants. All these differences have been extensively reviewed by Calabrese (1983, 19861. SUMMARIES OF EXAMPLES OF HUMAN EXPOSURES TO COMPLEX CHEMICAL MIXTURES The final section of this chapter contains eight summaries of situations in which there has been an apparent nonconcurrence between human and animal health responses. (For readers interested in additional information, more de- tailed descriptions are provided in Appendix B.) In the situations discussed, human responses often have differed substantially from what might have been

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 31 expected on the basis of data from controlled laboratory exposures to pure materials. Furthermore, a review of the abstracts shows that for many cases more careful interpretation of laboratory research presaged these findings of toxic effects in humans. For other cases summarized here, epidemiologic evi- dence first signaled toxicity, and the search for an animal model continues. Thus, experimental studies can play an important role in predicting human effects and reducing disease. The detailed case studies (Appendix B) discuss the experimental data that suggest or have helped to establish causal relationships. They also provide information that might have helped, if it had been used or interpreted differ- ently (a judgment admittedly based on hindsight). The purpose of the latter is not to criticize past decisions, but to identify the types of experimental data that are most useful in extrapolation to human disease potential. SULFUR DIOXIDE AND SUSPENDED PARTICULATE MATTER The sulfur-oxide/particulate-matter complex common to urban areas has been clearly associated with excess daily mortality and respiratory tract mor- bidity (NRC, 1985; D. V. Bates, personal communication). Only in recent years have animal models for relevant functional and morphometric effects proved useful in developing an understanding of the human health effects, and these have involved using animals and exposure protocols not used in conven- tional toxicity testing (beagles and dilute auto-exhaust mixtures, guinea pigs and SO2-particle mixtures passed through a heated furnace, etc.~. Other in- formative studies involved nonconventional functional assays (e.g., particle clearance function in rabbits undergoing brief daily exposures to sulfuric acid aerosols). Collectively, the results of the animal studies and of some studies involving short-term exposures of human volunteers point to one specific com- ponent of the mixture acidic aerosol" as the most likely causal factor for the various demonstrated human health effects. LEAD AND NUTRITIONAL FACTORS—EFFECTS ON BLOOD PRESSURE Recently, lead (Pb) was identified as an independent causal factor for hyper- tension in men among a host of dietary cofactors that were previously known or strongly suspected to affect hypertension. The evidence for the independent role of lead comes primarily from NHANES II, a national population-based study that included a substudy on blood lead, blood pressure, and dietary fac- tors. Supporting evidence for the influence of blood lead on blood pressure comes from a series of recent epidemiologic studies of occupationally exposed populations. In retrospect, published data on blood-lead/blood-pressure asso- ciations in rats could have been used to predict the human association. One

32 COMPLEX MIXTURES reason that the relation was not recognized sooner through toxicologic or epi- demiologic studies was its unusual shape: a plateau of elevation in blood pres- sure at a relatively low concentration of blood lead. Most toxicologic data are generated at relatively large exposures, and effects that occur only at smaller exposures are likely to be overlooked. RADON DAUGHTERS AND CIGARETTE SMOKE Several investigators have speculated that cigarette-smoking and radon daughters have joint actions in the incidence of lung cancer among miners exposed occupationally. Both radon daughters and cigarette-smoking are known to increase lung-cancer incidence in humans independently, and their joint action produces no more than an additive cumulative incidence. Smokers, however, have a shorter latent period for lung cancer than nonsmokers. One study found that nonsmokers chronically exposed to cigarette smoke have an increased risk of lung cancer if radon is also found in the home (Pershagen et al., 19871. Animal models have been of only limited value in studies of joint action of these two agents, at least for the inhalation model of exposure. Ani- mals in groups large enough for cancer studies cannot be exposed to cigarette smoke in a manner analogous to the way humans are exposed. Topical applica- tion of cigarette-smoke condensate to rodent skin provides a means of studying some effects of joint action, but it is of small value as a model for the human effects of concern, because the bioavailability is likely to be different between smoke and smoke condensate. ASBESTOS EXPOSURE AND CIGARETTE_SMOKING Asbestos is known to cause asbestosis, mesothelioma, and lung cancer in humans. Human studies have also demonstrated that cigarette-smoking has no apparent effect on the incidence of asbestosis or mesothelioma among those exposed to asbestos (NRC, 19841. But cigarette-smoking does markedly in- crease the incidence of lung cancer among asbestos-exposed humans, and the increase appears to be multiplicative. There is no suitable animal model for cigarette-smoke-related health effects. Without such a model, animal studies on joint action are not feasible. CIGARETTE_SMOKING AND ALCOHOLIC-BEVERAGE CONSUMPTION AS RISK FACTORS IN THE ETIOLOGY OF ORAL CANCER There is evidence of joint action of cigarette-smoking and alcohol ingestion in the incidence of oral cancer in humans. The evidence provides a rare exam- ple of a demonstrated health effect with available quantification of exposure to both agents. Therefore, it allows a more detailed mathematical analysis of joint

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 33 action than can be performed in most studies on the effects of mixtures. Even here, however, it is not possible to establish definitively whether the effects of the two agents are additive or multiplicative. The absence of a relevant animal model also limits the ability to address the critical issues in controlled experiments. TR]HALOMETHANES AND OTHER BYPRODUCTS OF CHLORINATION IN DRINKING WATER This case concerns the epidemiologic basis for the presumption of human risk of cancers of the gastrointestinal and urinary tracts associated with expo- sure to chlorination byproducts in disinfected drinking water. Of the situations presented here, evidence that exposure to chlorination byproducts increases the risk of cancer in humans is the most difficult to establish. There are several reasons. Increases in human cancer risk are likely to be small, in a relative sense, and are therefore difficult to detect in an epidemiologic setting with many possible confounding exposures. Mixtures of volatile and nonvolatile chlorination byproducts vary geographically and temporally and thus defy un- ambiguous definition (NRC, in press). The chlorinated organics in the nonvol- atile fraction thought to be of greatest toxicologic importance occur in ex- tremely low concentrations and are difficult to detect and measure. Toxicologic evaluations have relied largely on in vitro tests in nonmammalian systems and therefore are of only limited value in suggesting the cancer sites of greatest concern and the magnitude of expected risk. Further evaluation of risk associated with exposure to this complex mixture is a challenge to both epide- miologists and toxicologists. The issue deserves much further attention, be- cause exposures are so widespread and potential absolute risks so high. COKE-OVEN EMISSIONS Epidemiologic studies have shown an increased incidence of lung and geni- tourinary tract cancers in connection with exposure to coke-oven emissions, whereas experimental inhalation studies in animals have not produced compa- rable results. However, technical limitations have prevented exposures of ex- perimental animals to realistic inhalation atmospheres. As a consequence, tox- icologic insights have been obtained from studies involving surrogate exposures and cancers in other tissues in particular, studies in which extracts of particulate emissions were shown to produce skin cancer after topical appli- cations. The animal studies provide opportunities for determining relative po- tencies among the numerous constituents of the coal-tar complex. Despite the substantial differences in the human and animal data, the effects of coal tar in animals are related to the effects of coke-oven emissions in humans. Further- more, the data generated by toxicity testing of complex mixtures in animals can

34 COMPLEX MIXTURES be a reasonable predictor of human disease, if the mixture tested in the labora- toty is representative of mixtures in the human environment, if studies are properly designed to detect diseases, and if we are aware that the specific diseases and target organs found in test animals can vary from those found in humans. COAL-MINE DUST The complex mixture known as coal-mine dust includes particles of coal of varied composition and cytotoxicity, as well as rock dust of varied silica con- tent. Miners have been shown to have an excess of several chronic lung dis- eases, including coal workers' pneumoconiosis (CWP), progressive massive fibrosis (PMF), emphysema, and bronchitis. The largest incidence is that of COOP, as diagnosed by x-ray opacity, but the most serious win respect to the severity of effects and influence on mortality are emphysema, PMF, and bron- chitis. There are two important limitations to our knowledge. First, the role of cigarette-smoking in the development of emphysema and bronchitis is not clear. Second, an acceptable animal model that might be used to provide in- sight into this problem is not available. Thus, our ability to use toxicologic approaches to develop a better understanding of the pathogenesis of ihe dis- eases of coal workers is severely limited. REFERENCES Albert, A. 1979. Selective Toxicity: The Physico-Chemical Basis of Therapy, 6th ed. John Wiley & Sons, New York. (662 pp.) Albert, R. E., J. Berger, K. Sanborn, and M. Lippmann. 1974. Effects of cigarette smoke components on bronchial clearance in the donkey. Arch. Environ. Health 29:96-101. Albert, R. E., H. T. Peterson, Jr., D. E. Bohning, and M. Lippmann. 1975. Short-term effects of cigarette smoking on bronchial clearance in humans. Arch. Environ. Health 30:361-367. Ballantyne, B. 1985. Evaluation of hazards from mixtures of chemicals in the occupational environ- ment. J. Occup. Med. 27:85-94. Bandiera, S., K. Farrell, G. Mason, M. Kelley, M. Romkes, R. Bannister, and S. Safe. 1984. Compar- ative toxicities of the polychlorinated dibenzofuran and polychlorinated biphenyl mixtures which persist in Yusho victims. Chemosphere 13:507-512. Berry, G., M. L. Newhouse, and M. Turok. 1972. Combined effect of asbestos exposure and smoking on mortality from lung cancer in factory workers. Lancet 2:476-478. Bloch, W. N., Jr., T. R. Lewis, K. A. Busch, J. G. Orthoefer, and J. F. Stara. 1972. Cardiovascular status of female beagles exposed to air pollutants. Arch. Environ. Health 24:342-353. Bohning, D. E., H. L. Atkins, and S. H. Cohn. 1982. Long-term particle clearance in man: Normal and impaired. Ann. Occup. Hyg. 26:259-271. Calabrese, E. J. 1983. Principles of Animal Extrapolation. John Wiley & Sons, New York. (603 pp.) Calabrese, E. J. 1986. Animal extrapolation and the challenge of human heterogeneity. J. Pharm. Sci. 75: 1041-1046. DeCaprio, A. P., D. N. McMartin, P. W. O'Keefe, R. Rej, J. B. Silkworth, and L. S. Kaminsky. 1986. Subchronic oral toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the guinea pig: Comparisons with a PCB-containing transformer fluid pyrolysate. Fundam. Appl. Toxicol. 6:454-463.

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 35 Dube, M. F., and C. R. Green. 1982. Methods of collection of smoke for analytical purposes. Rec. Adv. Tobacco Sci. 8:42- 102. Faoro, R. B. 1975. Trends in concentrations of benzene soluble suspended particulate fraction and benzo(a)pyrene. J. Air Pollut. Control Assoc. 25:638-640. Ferin, J., and L. J. Leach. 1973. The effect of SO2 on lung clearance of TiO2 particles in rats. Am. Ind. Hyg. Assoc. J. 34:260-263. Freireich, E. J., E. A. Gehan, D. P. Rall, L. H. Schmidt, and H. E Skipper. 1966. Quantitative comparison of toxicity of anticancer agents in mouse, rat, hamster, dog, monkey, and man. Cancer Chemother. Rep. 50:219-244. Hammond, E. C. and I. J. Selikoff. 1973. Relation of cigarette smoking to risk of death of asbestos- associated disease among insulation workers in the United States, pp. 312-317. In P. Bogovski, J. C. Gilson, V. Timbrell, and J. C. Wagner (eds.). Biological Effects of Asbestos. IARC Scientific Publication No. 8. International Agency for Research on Cancer, Lyon, France. Hammond, E. C., I. J. Selikoff, P. L. Lawther, and H. Seidman. 1976. Inhalation of benzpyrene and cancerin men. Ann. N.Y. Acad. Sci. 271:116-124. Hernberg, S. 1986. Uses of epidemiology in occupational health, pp. 317-340. In M. Karvonen and M. I. Mikheev (eds.). Epidemiology of Occupational Health. (WHO Regional Publications, Euro- pean Series No. 20). WHO Regional Off~ce for Europe, Copenhagen. Hollander, D. 1981. IntestinalabsorptionofvitaminsA, E, D, andK. J. Lab. Clin. Med. 97:449-462. Hyde, D., J. Orthoefer, D. Dungworth, W. Tyler, R. Carter, and H. Lum. 1978. Morphometric and morphologic evaluation of pulmonary lesions in beagle dogs chronically exposed to high ambient levels of air pollutants. Lab. Invest. 38:455-469. IARC (International Agency for Research on Cancer). 1981. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Vol. 25. Wood, Leather and Some Associated Industries. International Agency for Research on Cancer, Lyon, France. LARC. 1982. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Suppl. 4. Chemicals, Industrial Processes and Industries Associated with Cancer in Humans (IARC Monographs, Volumes 1 to 29). International Agency for Research on Cancer, Lyon, Prance. (292 pp.) IARC. 1983. LARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Vol. 28. The Rubber Industry. International Agency for Research on Cancer, Lyon, France. (486 pp.) ~. IARC. 1984. L\RC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Vol. 34. Polynuclear Aromatic Compounds, Part 3, Industrial Exposures in Aluminium Production, Coal Gasification, Coke Production, and Iron and Steel Founding. International Agency for Re- search on Cancer, Lyon, France. IARC. 1986. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Vol. 38. Tobacco Smoking. International Agency for Research on Cancer, Lyon, France. (421 pp.) ICRP (International Commission on Radiological Protection), Task Group on Lung Dynamics. 1966. Deposition and retention models for internal dosimetry of the human respirato~ tract. Health Phys. 12: 173-207. Inskip, H., V. Beral, P. Fraser, M. Booth, D. Coleman, and A. Brown. 1987. Purther assessment of the effects of occupational radiation exposure in the United Kingdom Atomic Energy Authority mortal- ity study. Br. J. Ind. Med. 44: 149-160. Ip, C., and D. K. Sinha. 1981. Enhancement of mammary tumorigenesis by dieta~y selenium defi- ciency in rats with a high polyunsaturated fat intake. Cancer Res. 41 :31-34. Jammet, H., J. Lafuma, J. C. Nenot, M. Chameaud, M. Perreau, M. le Bouffant, M. Lefevre, and M. Martin. 1970. Lung clearance: Silicosis and anthracosis, pp. 435~37. In H. A. Shapiro (ed.). Pneumoconiosis: Proceedings of the International Conference, Johannesburg 1969. Oxford Univer- sity Press, New York. Karstadt, M., and R. Bobal. 1982. Availability of epidemiologic data on humans exposed to animal carcinogens: 2. Chemical uses and production volume. Teratog. Carcinog. Mutagen. 2: 151-168.

36 COMPLEX MIXTURES Kazantzis, G., and J. C. McDonald. 1986. Work, health, and disease, pp. 43-67. In M. Karvonen and M. I. Mikheev (eds.). Epidemiology of Occupational Health. (WHO Regional Publications, Euro- pean Series No. 20) WHO Regional Office for Europe, Copenhagen. Klaasen, C. D. 1986. Distribution, excretion, and absorption of toxicants, pp. 33-63. 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. Koppanyi, T., and M. A. Avery. 1966. Species differences and the clinical trial of new drugs: A review. Clin. Pharmacol. Ther. 7:250-270. Kotin, P., and H. L. Falk. 1960. The role and action of environmental agents in the pathogenesis of lung cancer. II. Cigarette smoke. Cancer 13:250-262. Lawther, P. J., B. T. Commins, and R. E. Waller. 1965. A study of the concentrations of polycyclic aromatic hydrocarbons in gas works retort houses. Br. J. Ind. Med. 22: 13-20. Lippmann, M., R. B. Schlesinger, G. Leikauf, D. Spektor, and R. E. Albert. 1982. Effects of sulphu- ric acid aerosols on respiratory tract airways. Ann. Occup. Hyg. 26:677-690. Little, J. B., E. P. Radford, Jr., H. L. McCombs, and V. R. Hunt. 1965. Distribution of polonium2'0 in pulmonary tissues of cigarette smokers. New Engl. J. Med. 273: 1343-1351. Logan, W. P. D. 1982. Cancer Mortality by Occupation and Social Class, 1851-1971. (IARC Scien- tific Publication No. 36; Studies on Medical and Population Subjects No. 44). H. M. Stationery Off~ce, London, and International Agency for Research on Cancer, Lyon. (253 pp.) McConnell, E. E., G. W. Lucier, R. C. Rumbaugh, P. W. Albro, D. J. Harvan, J. R. Hass, and M. W. Harris. 1984. Dioxin in soil: Bioavailability after ingestion by rats and guinea pigs. Science 223: 1077-1079. McCoy, G. D., S. S. Hecht, and E. L. Wynder. 1980. The roles of tobacco, alcohol, and diet in the etiology of upper alimentary and respiratory tract cancers. Prev. Med. 9:622-629. Milham, S., Jr. 1983. Occupational Mortality in Washington State, 1950-1979. DHHS (NIOSH) Publication No. 83-116. U.S. Government Printing Office, Washington, D.C. Milham, S., Jr. 1985. Improving occupational standardized mortality ratio analysis by social class stratification. Am. J. Epidemiol. 121:472~75. Morgan, K. T., D. L. Patterson, and E. A. Gross. 1984. Frog palate mucocilia~y apparatus: Structure, function, and response to formaldehyde gas. Fund. Appl. Toxicol. 4:58-68. NRC (National Research Council). 1975. Pest Control: An Assessment of Present and Alternative Technologies. Vol. I.: Contemporary Pest Control Practices and Prospects. National Academy of Sciences, Washington, D.C. (506 pp.) NRC. 1978. Sulfur Oxides. National Academy Press, Washington, D.C. (322 pp.) NRC. 1981. Indoor Pollutants. National Academy Press, Washington, D.C. (537 pp.) NRC. 1983a. Diet, Nutrition, and Cancer. National Academy Press, Washington, D.C. (449 pp.) NRC. 1983b. Polycyclic Aromatic Hydrocarbons: Evaluation of Sources and Effects. National Acad- emy Press, Washington, D.C. (460 pp.) NRC. 1984. Asbestiform Fibers: Nonoccupational Health Risks. National Academy Press, Washing- ton,D.C.(334pp.) NRC. 1985. Epidemiology and Air Pollution. National Academy Press, Washington, D.C. (224 pp.) NRC. 1986a. Environmental Tobacco Smoke: Measuring Exposures and Assessing Health Effects. National Academy Press, Washington, D.C. (337 pp.) NRC. 1986b. Fire & Smoke: Understanding the Hazards. National Academy Press, Washington, D.C. (156 pp.) NRC. In press. Disinfectants and Disinfectant By-Products. Drinking Water and Health, Vol. 7. Na- tional Academy Press, Washington, D.C. NRC, Committee on Medical and Biologic Effects of Environmental Pollutants. 1977. Nitrogen Oxides. National Academy of Sciences, Washington, D.C. (333 pp.) Office of Population Censuses and Surveys. 1978. Occupational Mortality. Decennial Supplement, England and Wales, 1970-1972. H. M. Stationery Office, London.

CONCEPTS FOR ANALYZING HUMAN EXPOSURE 37 Pershagen, G., Z. Hrubec, and C. Svensson. 1987. Passive smoking and lung cancer in Swedish women. Am. J. Epidemiol. 125: 17-24. Phalen, R. F., J. L. Kenoyer, T. T. Crocker, and T. R. McClure. 1980. Effects of sulfate aerosols in combination with ozone on elimination of tracer particles inhaled by rats. J. Toxicol. Environ. Health 6:797-810. Poiger, H., and C. Schlatter. 1980. Influence of solvents and absorbents on dermal and intestinal absorption of TCDD. Food Cosmetics Toxicol . 18: 477-481. Radford, E. P. 1986. Ionizing radiation, pp. 726-733. In J. M. Last (ed.). Maxcy-Rosenau Public Health and Preventive Medicine, 12th ed. Appleton-Century-Crofts, Norwalk, Conn. Rajewsky, B., and W. Stahlhofen. 1966. Polonium-210 activity in the lungs of cigarette smokers. Nature 209: 1312-1313. Rall, D. P. 1965. Conference on obstacles to the control of acute leukemia. Experimental studies of the brood-brain barrier. Cancer Res. 25:1572-1577. Rall, D. P. 1971. Drug entry into brain and cerebrospinal fluid, pp. 240-248. In B. B. Brodie and J. R. Gillette (eds.). Handbuch der Experimentellen Pharmakologie, Vol. 28, Pt. 1. Springer-Verlag, Berlin. Rall, D. P., M. D. Hogan, J. E. Huff, B. A. Schwetz, and R. W. Tennant. 1987. Alternatives to using human experience in assessing health risks. Ann. Rev. Public Health 8:355-385. Rappe, C., S. Marklund, L. O. Kjeller, and M. Tysklind. 1986. PCDDs and PCDFs in emissions from various incinerators. Chemosphere 15: 1213 - 1217. Sawicki, E. 1967. Airborne carcinogens and allied compounds. Arch. Environ. Health 14:46-53. Scharding, N. N., and F. W. Oehme. 1973. The use of animal models for comparative studies of lead poisoning. Clin. Toxicol. 6:419~24. Schlesinger, R. B. 1986. The effects of inhaled acid aerosols on lung defenses, pp. 617-635. In S. D. Lee, T. Schneider, L. D. Grant, and P. J. Verkerk (eds.). Aerosols: Research, Risk Assessment and Control Strategies. Proceedings of the Second U.S.—Dutch International Symposium, Williams- burg, Virginia, The United States, May 19-25, 1985. Lewis Publishers, Chelsea, Mich. Schlesinger, R. B., B. D. Naumann, and L. C. Chen. 1983. Physiological and histological alterations in the bronchial mucociliary clearance system of rabbits following intermittent oral or nasal inhala- tion of sulfuric acid mist. J. Toxicol. Environ. Health 12:441-465. Selikoff, I. J., E. C. Hammond, and J. Churg. 1968. Asbestos exposure, smoking, and neoplasia. J. Am. Med. Assoc. 204:106-112. Smith, R. L. 1973. Species variations in biliary excretion, pp. 76-93. In R. L. Smith (ed.). The Excretory Function of Bile. John Wiley & Sons, New York. Speizer, F. E. 1983. Assessment of the epidemiologic data relating lung cancer to air pollution. Envi- ron. Health Perspect. 47:33-42. Umbreit, T. H., E. J. Hesse, and M. A. Gallo. 1986a. Comparative toxicity of TCDD-contaminated soil from Times Beach, Missouri, and Newark, New Jersey. Chemosphere 15:2121-2124. Umbreit, T. H., E. J. Hesse, and M. A. Gallo. 1986b. Bioavailability of dioxin in soil from a 2,4,5-T manufacturing site. Science 232:497-499. U.S. EPA (Environmental Protection Agency). 1982. Air Quality Criteria for Particulate Matter and Sulfur Oxides. EPA-600/8-82-029c. U.S. Environmental Protection Agency, Research Triangle Park, N.C. (3 vole.) U.S. EPA. 1985. Guidelines for Preparing Environmental and Waste Samples for Mutagenicity (Ames) Testing. EPA 600/4-85-058. Environmental Monitoring System Laboratory, Las Vegas, Nevada. (Available from NTIS as PB 85-120144.) (255 pp.) U. S. EPA. 1986. National Primary and Secondary Ambient Air Quality Standards. Code of Federal Regulations, Title 40, Part 50. U.S. Surgeon General. 1982. The Health Consequences of Smoking: Cancer. DHHS (PHS) 82-50179. U.S. Public Health Service, Office on Smoking and Health, Rockville, Md. (322 pp.) U.S. Surgeon General. 1983. The Health Consequences of Smoking: Cardiovascular Disease. DHHS

38 COMPLEX MIXTURES (PHS) 84-50204. U.S. Public Health Service, Office on Smoking and Health, Rockville, Md. (384 pp.) U.S. Surgeon General. 1984. The Health Consequences of Smoking: Chronic Obstructive Lung Dis- ease. DHHS (PHS) 84-50205. U.S. Public Health Service, Office on Smoking and Health, Rock- ville, Md. (545 pp.) Vainio, H., K. Hemminki, and J.Wilbourn. 1985. Data on the carcinogenicity of chemicals in the IARC Monographs programme. Carcinogenesis 6: 1653-1665. Van der Berg, M., E. de Vroom, M. van Greevenbrook, K. Olie, and O. Huzinger. 1985. Bioavailabil- ity of PCDDs and PCDFs adsorbed on fly ash in rat, guinea pig, and Syrian golden hamster. Chemo- sphere 14:865-869. Way, E. L. 1967. Brain uptake of morphine: Pharmacologicimplications. Fed. Proc. 26:1115-1118. Weiss, B. 1978. The behavioral toxicology of metals. Fed. Proc. Fed. Am. Soc. Exp. Biol. 37:22-27. WHO (World Health Organization). 1981. Health Effects of Combined Exposures in the Work Envi- ronment. Technical Report Series 662. WHO, Geneva. (76 pp.) Wolff, R. K., M. Dolovich, G. Obminski, and M. T. Newhouse. 1977. Effect of sulphur dioxide on tracheobronchial clearance at rest and during exercise, pp. 321-332. In W. H. Walton (ed.). Inhaled Particles. IV. Part 1. Pergamon, New York.