Toxicological Evaluation and Risk Assessment of ...

Critical Reviews in Toxicology, 28(1):73–101 (1998)

Toxicological Evaluation and Risk Assessment of Chemical Mixtures Flemming R. Cassee*, John P. Groten, Peter J. van Bladeren, and Victor J. Feron

Critical Reviews in Toxicology Downloaded from informahealthcare.com by RIVM on 09/13/10 For personal use only.

TNO Nutrition and Food Research Institute, Toxicology Division, P.O. Box 360, NL3700AY, Zeist, The Netherlands Present address and corresponding author: Dr. F. R. Cassee, National Institute for Public Health and the Environment, Laboratory for Health Effect Research, P.O. Box 1, NL3720BA Bilthoven, The Netherlands. Tel: #31 30 274 2406; Fax: #31 30 274 4448; e-mail: [email protected].

ABSTRACT: A major objective of combination toxicology is to establish whether a mixture of chemicals will result in an effect similar to that expected on the basis of additivity. This requires understanding of the basic concepts of the combined toxicological action of the compounds of the mixture: simple similar action (dose addition), simple dissimilar action (effect or response addition), and interaction (synergism, potentiation, antagonism). The number of possible combinations of chemicals is innumerable, and in vivo testing of these mixtures is unattainable from an ethical, economical, or pragmatic perspective. Prediction of the effect of a mixture based on the knowledge of each of the constituents requires detailed information on the composition of the mixture, exposure level, mechanism of action, and receptor of the individual compounds. Often, such information is not or is only partially available and additional studies are needed. Research strategies and methods to assess joint action or interaction of chemicals in mixtures such as whole mixture testing, physiologically based toxicokinetic modeling and isobologram and dose response surface analyses are discussed. Guidance is given for risk assessment of both simple and complex mixtures. We hypothesize that, as a rule, exposure to mixtures of chemicals at (low) non-toxic doses of the individual constituents is of no health concern. To verify the hypothesis is a challenge; to timely detect exceptions to the rule is the real challenge of major practical importance. KEY WORDS: isoboles, response surface analysis, physiologically based toxicokinetic modeling, weight of evidence approach, complex mixtures, simple mixtures, combination toxicology.

I. INTRODUCTION For most chemical mixtures data on exposure and toxicity are fragmentary, and roughly over 95% of the resources in toxicology is still devoted to studies of single chemicals.1,2 Concurrent or sequential human exposure to a multitude of chemicals (e.g., food additives, pesticides, indoor and outdoor air pollutants, and occupational environments) dictates the necessity of exposure assessment, hazard identification, and risk assessment of chemical

mixtures. Moreover, public concern regarding exposure to chemical mixtures and the interest of scientists and regulators in this challenging vanguard of toxicology have increased. Topics of continuous debate are concepts of similar joint action and independent joint action vs. interaction of chemicals, design of experimental studies, and analysis of the results, including the use of statistical methods, dose response relationships and lowdose extrapolation, mechanisms of toxicokinetic and toxicodynamic interactive effects,

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material without the consent of the publisher is prohibited.

73


and, last but not least, the lessons learned from risk assessment of real-life examples of exposure to chemical mixtures.3–9 It is important that the right parameters are used for summation of the effects of the individual compounds in assessing joined effects. For example, the addition of 1 l water to 1 l ethanol will result in a volume that is less than 2 l. In this case the parameter weight rather than the parameter volume should have been used. Another example, given by Steel and Peckham,10 is the amalgamation of two spherical balls of clay. The sum of the radii of each single sphere is dissimilar to the radius of the newly formed spherical ball. In this case the volume or the weight should be used to predict the result of the amalgamation. These examples illustrate the paramount importance of the use of the appropriate parameters and evaluation models to predict the hazard of a mixture relative to its constituents. Complex chemical mixtures consist of tens, hundreds, or even thousands of compounds. Their composition is qualitatively and quantitatively not fully known and may change. Examples are wood smoke, diesel exhaust, welding fumes, fly ash, food, and occupational and environmental settings. Often adequate testing of such mixtures is impossible because they are virtually unavailable for testing and a sufficient number of dose levels cannot be applied. Moreover, highdose levels of the mixture may have different types of effects than low-dose levels, and low-dose extrapolation may be meaningless.7,11 In some cases (diesel exhaust, coal dust), testing of complex mixtures as a whole has been shown to produce relevant toxicological information. For all of the aforementioned reasons, one of the main interests of scientists in the field of combination toxicology is to find out whether the toxicity of a mixture is different from the sum of the toxicities of the single compounds; in other words, will the toxic effect of a mixture be determined by additivity of dose or effect or by supra-additivity

(an effect stronger than expected on the basis of additivity) or by infra-additivity (an effect less than expected on the basis of additivity)? The toxic effect of a mixture appears to be highly dependent on the dose (exposure level), the mechanism of action, and the target (receptor) of each of the mixture constituents. Thus, information on these aspects is a prerequisite for predicting the toxic effect of a mixture. In this article the basic concepts of similar and dissimilar action of chemicals and of toxicological interactions between chemicals in a mixture are briefly discussed. Furthermore, strategies are described for studying both simple and complex mixtures. An important aspect of toxicity studies of mixtures is the impracticability of ‘complete’ testing. If all combinations are to be studied at different dose levels, an increasing number of chemicals in a mixture results in an exponential increase in number of test groups: to test all possible combinations (in a complete experimental design) at only one dose level of each chemical in a mixture consisting of 4, 5, or 6 chemicals, 16 (24–1), 32 (25–1) or 64 (26–1) test groups, respectively, would be required. Such in vivo studies are virtually impossible from a practical, economical, and ethical point of view. To reduce the number of test groups without losing too much information about possible interactions between chemicals, several statistical designs have been proposed. Some of these designs are discussed in this paper. Guidelines from national and international organizations involved in setting exposure limits often suggest the use of simple “dose addition” models for assessing the hazard of a chemical mixture without taking into account the mode of action of the individual chemicals. Clearly, such an approach may greatly overestimate the risk in case of chemicals that act by mechanisms for which the additivity assumptions are invalid. For example, essential nutrients (vitamins, trace elements, essential amino, and fatty acids) have relatively small margins between the


74


dose we need and the lowest toxic dose;12 simultaneous consumption of these chemicals at their recommended intake levels would prove to be a rather unhealthy habit if the toxicity of this mixture is assessed on the basis of the “dose addition” concept (similar joint action). In this respect it is fair to stress that the US EPA guidelines for risk assessment of chemical mixtures13,14 allow for the use of different additivity models and different interaction models, including effect addition. If chemicals in a mixture are known to have simple dissimilar modes of action, studies in our institute have shown that they do not constitute an evidently increased hazard compared with that of exposure to the individual chemicals, provided the exposure level of the chemicals in the mixture is at most equal to or lower than their own “No-ObservedAdverse-Effect level (NOAEL).” 15 The health risk of such mixtures is entirely determined by the health risk associated with the “most risky chemical” in the mixture, provided the risk quotients of the other chemicals in the mixture do not exceed unity. Using as examples rat studies with mixtures of different “classes”of chemicals, the importance of applying the right basic concept to the data obtained in assessing the potential health risk of the mixtures is illustrated in this article.

II. BASIC CONCEPTS OF CHEMICAL MIXTURES TOXICOLOGY Three basic concepts for the description of the toxicological action of constituents of a mixture have been defined by Bliss16 and are still valid half a century later.

A. Simple Similar Action Simple similar action is also known as ‘simple joint action’,16 or ‘concentration/dose

addition’.13 This is a noninteractive process, which means that the chemicals in the mixture do not affect the toxicity of one another.13,17 Each of the chemicals in the mixture contributes to the toxicity of the mixture in proportion to its dose, expressed as the percentage of the dose of that chemical alone that would be required to obtain the given effect of the mixture. All chemicals of concern in a mixture act in the same way, by the same mechanisms, and differ only in their potencies. This form of action also serves as the basis for the use of “toxic equivalency factors”,17 and is used to describe the combined toxicity of isomeres or structural analoges (e.g., PCBs and dioxines). Similar joint action allows us to describe the additive effect mathematically using the summation of the doses of the individual compounds in a mixture after adjustment for the differences in potencies. This method together assumed to be only valid for linear dose response/effect curves.10,18,19

B. Simple Dissimilar Action Simple dissimilar action, also referred to as ‘simple independent action’, ‘independent joint action’,17 or ‘response addition’.13 In this case the agents of a mixture do not affect each other’s toxic effect. The mode of action and possibly, but not necessarily, the nature and site of action differ among the constituents of the mixture. Response addition is referred to if each individual of a population has a certain tolerance to each of the chemicals of a mixture and will only exhibit a response to a toxicant if the concentration exceeds the tolerance dose. In this case the number of responders will be given rather then the average effect of a mixture on a group of individuals. By definition, reponse addition is determined by summing the responses of the animals to each toxicant in a mixture.20 Three different concepts have been developed for response additivity depending


75

on the correlation of susceptibility of subjects to toxic agents 1.


2.

3.

Complete negative correlation: Individuals most susceptible to one toxicant are least susceptible to the other. The percentage P of the individuals responding to the mixture is equal to the sum of the response to the each of the components of the mixture. In a formula: Pmixture A,B = PA + PB ≤ 1 Complete positive correlation: Individuals most susceptible to one toxicant are also most susceptible to the other. The proportion of individuals responding to the mixture will be equal to the response to the most toxic agent in a mixture. In a formula: Pmixture A,B = PA if toxicity A ≥ B No correlation: The proportion of individuals responding to the mixture is equal to the sum of proportion of individuals responding to each of the toxicants taking into account that those individuals that respond to constituant A cannot react to B as well: In a formula: Pmixture A,B = PA + PB · (1 –P A) = PA + PB – PAPB Note that although this type of correlation seems similar to complete negative correlation, the difference is that in this case an individual can respond to both agent A and B but not to both at the same time.

C. Interactions Compounds may interact with one another, modifying the magnitude and sometimes the nature of the toxic effect. This modification may make the composite effect stronger or weaker. An interaction might occur in the toxicokinetic phase (processes of uptake, distribution, metabolism, and excretion) or in the toxicodynamic phase (effects of chemicals on the receptor, cellular target, or organ).

For the latter case numerous examples from the field of pharmacology are available. These include terms such as synergism and potentiation (i.e., resulting in a more than additive effect), or antagonism (i.e., resulting in a less than additive effect). The most obvious cases for interactions in the toxicokinetic phase are enzyme induction and/or inhibition, because metabolism is an important determinant of toxicity (either reactive intermediates are formed — activation — or the toxic agent is removed — detoxification). Compounds that influence the amount of biotransformation enzymes can have large effects on the toxicity of other chemicals. Uptake and excretion may also be active processes. Interaction between substrates for the same membrane receptors or pumps as well as for the biotransformation enzymes themselves follow the rules for receptors described above. In order to be able to predict the effect of a mixture of compounds with the same target receptor, but with (different) nonlinear doseeffect relationships, mathematical or physiological modeling can be applied. For example, the interaction between chemicals and a target receptor or enzyme, the mechanism of joint action can be described by MichealisMenten kinetics.21 In general, this results in a phenomenon called “competitive agonism” and the ultimate combined effect is less than expected on the basis of effect-addition because of the competition for the same receptor. In fact, this type of interaction can also be considered as a special case of similar joint action (dose addition). In reality, however, probably we will often have to deal with partial dose and/or effect addition. For example, different chemicals can act both on the same and on different target sites or receptors at the same time.

D. Bliss Independence These three basic principles of joint action are theoretical. In reality, however, one


76

often has to deal with these concepts at the same time, especially when mixtures consist of more than two compounds and when the targets (individuals rather than cells) are more complex.


III. ASSESSMENT OF JOINT ACTION AND INTERACTION OF CHEMICALS IN MIXTURES A. Strategy for Experimental Studies In the last decade a number of approaches has been presented to evaluate the effect of a mixture compared withthe effect of its components.4, 22–24 Ideally, one should identify all chemicals in a mixture and determine the toxicity of each of the constituents or obtain toxicological information from existing literature. Several test scenario’s can be followed to obtain toxicological information on mixtures with a limited number of test groups. The study design will largely depend on the number of compounds of a mixture and on the question whether it is desirable to assess possible existing interactions between chemicals in a mixture:

• One pragmatic approach is to test the toxicity of the mixture without assessing the type of interactions. Whole mixture effects can be assessed by testing only the whole mixture. However, is extrapolation to lower doses allowed? Thus, preferably mixtures should be tested both at high (effective) concentrations and at low (realistic) concentrations. • Interactive effects between two or three compounds can be identified by physiologically based toxicokinetic modeling, isobolographic or dose-effect surface analysis, or comparison of dose-effect curves. Physiologically based toxicokinetic modeling would be especially useful for inter-

actions in the toxicokinetic phase — note that despite similar biotransformation pathways mechanisms of activated toxicants might be quite different or toxicants might be influencing each others biotransformation through induction or inhibition. Isobolographic and dose-effect surface analyses are most suitable for situations where the target is the same for a limited number of compounds, thus where possible interaction takes place in the toxicodynamic phase. • Interactive effects of compounds in mixtures with more than three compounds can be best ascertained with the help of statistical designs such as (fractionated) factorial designs, ray designs or dose-effect surface analysis.

B. Whole Mixtures The simplest way to study effects of mixtures is to compare the effect of a mixture with the effects of all its constituents at comparable concentrations and duration of exposure at one dose level without testing all possible combinations of two or more chemicals. Although this approach requires a minimum number of experimental groups (n + 1, the number of compounds in a mixture plus the mixture itself), it is impossible to describe the effect of the mixture in terms of synergism, potentiation, antagonism, etc. if there are no dose-effect curves of each of the single compounds. It is even doubtful if deviation from additivity can be concluded from such studies when the effect of the individual compounds is not properly assessed. This strategy has been recommended for mixtures that are not well characterized,4 but it has also been applied for assessing the combined toxicity of defined chemical mixtures consisting of nephrotoxicants, pesticides, carcinogens, and/or fertilizers.25–30 This approach would be of interest for a first screening of adverse effects of a mixture. The designs of these studies were chosen to reflect the net effect


77


of all compounds in the mixture. In order to limit the number of test groups, possible interactive effects of the chemicals in relation to the effects of individual chemicals were not and cannot be taken into account. The question of whether high-to-low-dose extrapolation is allowed remains to be resolved. This can be illustrated by the following example. If two chemicals have dose-effect relationships with different slopes (A > B) and agent A inhibits the action of agent B, one can understand that the effect of agent B at low dose levels cannot be predicted from the information on the effects at high dose levels.

C. Physiologically Based Toxicokinetic Modeling In the risk assessment of xenobiotics, adequate human data are often not available. The current practice is to extrapolate the risk for man, using animal studies combined with arbitrary safety factors. However, for many compounds, their metabolism is the major determinant of the risk and for several hazardous substances a considerable insight in the role of metabolism in toxic reactions has become available from laboratory experiments. In vitro toxicological studies with cell or tissue cultures, subcellular fractions, and pure enzymes have provided information on the nature of several reactive intermediates. Moreover, the occurrence and the adverse effects of these intermediates have been demonstrated in several species. Recently, several new approaches have been developed making use of this knowledge. In principle, the in vivo human metabolism can be predicted using in vitro enzyme kinetics, and thus may be compared with in vitro and in vivo data from animal studies.31–33 In several cases, experiments with microsomal fractions and with hepatocytes have been shown to predict accurately the in vivo velocity of metabolism, for a single metabolic

pathway. Such data may be applied in physiologically based toxicokinetic modeling. This technique has been reviewed extensively.34–37 To date, several physiologically based toxicokinetic models have become available (for list see Ref. 35), including, for example, methylene chloride 38 and ethylene dibromide.39 The enzymes involved in biotransformation processes occur often as gene families with several (iso)enzymes with overlapping substrate specificities.40 As a rule, the description of the rate constants (Vmax, Km, etc.) obtained from individual (iso)enzymes obeys the Michaelis-Menten description of the kinetics even when several (iso)enzymes metabolize a drug. Thus, the consequences of interindividual differences in expression levels, and genetic polymorphisms may easily be modeled, using a model built on individual (iso)enzymes. Of course, this approach benefits heavily from the increased availability of the human enzymes produced by recombinant DNA technology. Recently, a strategy was presented by Ploemen et al.39 to combine physiologically based toxicokinetic modeling with human in vitro metabolic data, to explore the relative and overall contribution of critical metabolic pathways in vivo in man. Four distinct steps can be recognized in this strategy: first, the principal isoenzymes involved in the metabolism, and their rate of catalysis is determined in vitro. Second, the rate is scaled up to the rate per milligram subcellular fractions (cytosol and microsomes). After validation of the prediction at this second level, the rate is expressed per gram whole liver for the relevant metabolic routes (third phase). These values are then used to calculate the ratio of these pathways at “fixed” liver concentrations. In the fourth step, the metabolic parameters are integrated in a physiologically based toxicokinetic model to predict the relative contribution of the metabolic pathways in vivo. This shows exactly to what extent physiologically based toxicokinetic models may be


78


used in combination toxicology. First, one of the components of the mixture is regarded as the “prime toxicant”, the effect of which is modified by the other compounds. Effects of induction or inhibition of specific biotransformatic isozymes may then be incorporated in the model and its effects determined and compared with the differences that occur in any event because of already existing individual differences. Effects of competition between chemicals in a mixture for the same biotransformation enzymes can also be incorporated by translating the effects into effects on the Michaelis-Menten parameters that are then introduced into the model. In principle, it is possible to extend such a model to include the toxicodynamic phase, if a direct relationship exists between the concentration of reactive intermediates and the toxic effect. For instance, recently it has been shown that both glutathione depletion as well as covalent binding to cellular macromolecules can easily be predicted in the human situation for the model compound 1,2-dichlorobenzene.41 D. Isobole Methods An isobole, originally developed by Loewe and Muischnek,23 is a contour line that represents equi-effective quantities of two agents or their mixtures. The theoretical line of additivity is the straight line connecting the individual doses of each of the single agents that produce the fixed effect alone (Figure 1). The method requires a number of mixtures to be tested and is used for a graphical representation to find out if mixtures of two compounds behave in a dose-additive manner and subsequently can be regarded as chemicals with a similar mode of action. When all equi-effect concentrations are connected by a downward concave line, the effect of the combinations is antagonistic, and a concave upward curve indicates synergism. The use of the isobole procedure to evaluate the effects of binary mixtures is widely used,3,42 but is very labo-

rious and requires large data sets in order to produce sufficiently reliable results. Moreover, the departure of the observed data from the theoretical isobole strongly depends on the accuracy of estimated intercepts of the theoretical isobole with the axis (representing the dose of one of the compounds of the mixture inducing the desired effect). In many cases large standard deviations of these intercepts do not allow a reliable conclusion concerning the deviation from additivity. Berenbaum43 has presented a mathematical equation for calculating an interaction index that enables the effects on noninteractive combinations to be calculated directly from the dose effect relationship of the individual compounds, regardless of the particular types of dose effect relations involved. CI =

d1 d 2 d + +… n D1 D 2 Dn

(1) (1)

In this equation d1, d2, … dn are the doses of agents in a mixture and D1, D2, … Dn are the doses of the individual agents producing the same effects as the mixture. For binary mixtures this equation describes a straight line (isobole) joining D1 and D2 and passing trough (d1,d2). CI is an interaction index that is 1, 1 when the combination shows zero interaction, synergy, or antagonism using dose addition, respectively. Even if a heterogenous combination is considered, in which agent A produces an effect and B does not, Eq. 1 is valid. In this case D2 is considered to be infinite and the equation is reduced to d =1 D1

(2)

This procedure requires a number of calculations and, in case of departure from additivity, the magnitude of CI will depend on the ratio of the concentrations of the constituents of the mixture (Figure 1) and thus results in one CI for a mixture of chemicals with specific concentrations rather than for


79


FIGURE 1.

Isobologram of two agents A and B that act synergistically.

the combinations of chemicals in general. Actually, Eq. 2 is valid in the case of simple similar joint action (agents with similar-shaped concentration-effect curves but different potencies), but the theoretical basis for a more general use of this equation has been questioned.44,45 In addition, the statistical basis for testing whether a specific CI deviates from 1 (additivity) is very complex. Pöch and Reiffenstein46 have argued that large differences in the slopes of dose response curves may result in the wrong conclusion derived from isobolograms. A solution to overcome this problem has been proposed by Loewe47 and is based on the construction of an additivity envelope for nonlinear dose response curves (Figure 2), which was modified and extended by Steel and Peckham.10 The envelope of additivity is an area in which those combinations A and B are lying that have a specific effect and may reasonably be

considered as showing no interaction.10,18 It reflects the inherent uncertainty of how to perform summation given the nature of the dose response curves.3 The envelope is constructed from the dose response curves of both compounds of a binary mixture. The boundaries of the envelope are found by using mode I and II addition. Mode I addition, also referred to as heteroadditivity,18 is the simplest form of effect addition and is describe by the formula: E( d1 , d 2 ) = E( d1 ) + E( d 2 )

in which d1 and d2 represent the doses of two compounds each resulting in an effect E that (3) can be summed to give E(d1,d2), the fixed effect due to exposure to the mixture. Mode II addition or isoadditivity18 represents the general form of the concentration additivity concept. In this case a dose d1 will result in an effect E1 and the second agent will


80

add up with a complementary dose of d1 to result in E(d1,d2) (Figure 2). Kodell and Pound18 have provided a mathematical equation for mode I:p E(d1 , [d 2 − p 2 ⋅ d1 ]) = E( p 2 ⋅ d1 ) = E(d 2 )


(4) in which p2 is defined as the relative potency factor of agent 2. However, when the dose response curves are nonlinear and cannot be linearized by biologically sensible transformation, there is no obvious way to determine a potency factor. The envelope of additivity is useful when two chemicals are believed to act in some additive way but with an unknown type of additivity18,48 and also reflects those cases in which two chemicals show some intermediate form of dose and effect addition. Berenbaum43 has argued that for combinations of two agents that are actually the same chemicals and have an exponential survival curve (and as a consequence will result in straight-line isoboles showing no interaction based on using dose addition), mode I calculations will result in a curved additive isobole. Berenbaum43 concluded that in this particular case calculations according to mode II will best describe the joint action of the agents. The great majority of biologically active agents has nonexponential types of dose response curves (sigmoid, log-linear, log-log, etc.), because their mechanisms of action are different and they do not behave in a dose-additive manner. Despite the fact that isoboles are very illustrative representations of combined action or interactions, the use of isobolograms has a number of drawbacks. For instance, equieffective doses of each of the constituents of the mixtures have to be known very precisely: large standard deviations make the observed effects plotted in the isobologram difficult to interpret. As a consequence, extensive studies must be carried out with the

single compounds. Moreover, the huge number of valuable data are not fully exploited in this method and there is no easy statistical basis for a conclusion drawn from the isobologram. Although a graphical presentation is illustrative, it is obvious that an isobole analysis can only be done at clear effect levels and their graphical nature limits the number of agents to be evaluated in a mixture to three. Recently, a system of parallel coordinate axes representing hyperdimensional figures are proposed to overcome this problem to some extent and permitting the ready visualization and characterization of interaction effects with a polynomial model.49

E. Comparison of Individual Dose Response Curves Comparison of dose response curves (DRCs) of an agent in the absence and presence of a second agent has been proposed.22,51–52 For this purpose, the corresponding theoretical DRCs for dose-additive and independent combinations have to be constructed in order to compare the observed combined response with the expected response. An independent effect can be calculated with Eq. 5. E AB = E A + E B − ( E A ⋅ E B )

Independent DRCs are upward-shifted (Figure 3). For calculation and construction of the (5) additive DRC the equi-effective dose dA,equi of compound A resulting in the same effect as dB is estimated. Under the assumption that additive combinations of compound B must behave like dA,equi, the same effect is expected in additive combinations at the applied doses of dA – dA,equi (Figure 3). All additive DRCs of A + B reach the same maximum as the maximum of DRCA, provided the effect of B is smaller than Amax. However, in case of competitive agonism, the effect of B does not af-


81


FIGURE 2. Isobolic diagram of the effect of joint effect of two chemicals A and B (1B) with the corresponding concentration effect curves (1A,1C) illustrating the construction of an additivity envelope. In an isobologram (1B) combinations of two compounds that result in the same effect (e.g., 50% reduction of the viability), and also the concentrations of each compound that, given alone, have the same effect as the combinations, are plotted. All points of an isobole reveal the mixing values of both components at which a specific, quantitative effect is observed (——). The envelope of additivity is an area in which those combinations AB are lying that have a specific effect and may reasonably be considered as showing no interaction (Modified from 3.)


82


FIGURE 3. Illustration of how dose-additive (a,b) and independent (c,d) dose response curves (DRCs) of compound A in the presence of a fixed dose of B are constructed with the doses on a linear (a,c) or a log (b,d) scale. Dose-additive DRCs (a,b) of A in the presence of B (arrowheads) are shifted to the left by the dose x, representing the dose of A with which the fixed dose of B is equieffective. Independence is exemplified in (c,d) for B causing 45% of the maximum increase in the absence and presence of various doses of A. As B acts independently of A, the fraction of the maximal response to B is 0.45 from 0 dose of A to the highest dose of A (adapted with permission from Pöch.50)


83


fect the effect of A + B at higher doses of A. Considerably less data are then needed to describe the combined effects using this evaluation method compared with isobolographic evaluations. Although this method is easier and more straightforward than the isobole method, it still requires complete DRCs of compound A in the presence and absence of a fixed effective concentration of compound B. The method, including a statistical interpretation, is restricted to binary mixtures and has been presented for dose frequency as well as “quantitative” curves.22,51

F. Response Surface Analysis (ESA) response or effect surface methodology will yield a statistically based mathematical relationship between the doses of each of the agents of a mixture and the effect parameter.24,44,45,52,53 This mathematical equation is obtained by multiple linear regression. An example of an equation for a 3-compound mixture is given below: E = α + β1 ⋅ d1 + β 2 ⋅ d 2 + β3 ⋅ d 3 + γ 1 ⋅ d1 ⋅ d 2 + γ 2 ⋅ d 1 ⋅ d 3 + γ 3 ⋅ d 2 ⋅ d 3 + δ1 ⋅ d ⋅ d 2 ⋅ d 3 (6)

in which dn represents the dose of a chemical in the mixture. By means of a t test and the standard error of a coefficient, the p value for the regression coefficient can be estimated. The p value measures the probability of observing the value of the coefficient or a more extreme value given the null hypothesis (the coefficient is zero) is true. In the case of significant interaction terms, these coefficients should always be interpreted with the corresponding main effects. Coefficient α represents the control situation (e.g., 100% viability), and the constants β are associated with the main effect of each of the compounds. The coefficients γ and δ are indications for two- and three-factor interactions,

respectively. In this example, for viability a positive value of γ or δ (in association with a negative value for β) indicates a less than additive effect due to an interaction between two (or three) compounds of the mixture. Zero values of γ or δ suggest the absence of a particular interaction. The advantage of this method is that it includes all data points obtained. However, knowing that at relatively high exposure levels saturation and competition will play a major role in the establishment of the combined effect, it should be avoided to work at high-effect levels in the ESA. Instead, we recommend choosing a concentration range not exceeding an (arbitrary) 50% effect level, or even less, because at present there is an increasing awareness that in the field of (combination) toxicology it would be more interesting to investigate the effects of mixtures at no-effect levels or minimal-effect levels of the individual constituents. G. Statistical Designs Whenever a mixture consists of more than two chemicals, all kinds of two- or threefactor interactions are possible. If all these factors are taken into account in an experiment, the number of possible test combinations increases exponentially with increasing numbers of compounds in a mixture. Moreover, the number of experimental groups will also increase with the number of doses of each compound. Both ethical and practical reasons force investigators to minimize the number of test groups. As it is of major interest to find out whether two or more chemicals can give rise to a synergistic or antagonistic effect, more experimental groups are needed. For this purpose one should carefully explore all possible and readily available study designs.54,55 A number of statistical designs are available to evaluate the effects of mixtures compared with their constituents,56 for instance, ray, central composite, and factorial designs (Figure 4). One way to detect interactive ef-


84

study has been used to identify nonadditive effects of three compounds on developmental toxicity at five dose levels.59 Full factorial designs, however, lead to very costly experiments, and even if only two dose levels are used, it is already virtually impossible to perform complete, conventional toxicity tests using 2n – 1 test groups to identify interactions between all chemicals of interest. One way to deal with this problem is the use of fractionated factorial designs.60 Fractionated designs still identify most of the interactions between the


fects, that is, nonadditive effects, between more than two chemicals in a chemical mixture is to use factorial designs. The use of factorial designs, in which n chemicals are studied at x dose levels (xn treatment groups) has been suggested by the U.S. Environmental Protection Agency as one of the valuable statistical approaches for risk assessment of chemical mixtures.57 Very recently, a 25 study was presented to describe interactions between the carcinogenic activity of 5 polycyclic aromatic hydrocarbons at two dose levels58 and a 53

FIGURE 4. Designs used to study the joint effect of mixtures of two chemicals. Each design is plotted on an isobologram. (Modified from 55.)


85


compounds and determine which compounds are important in causing effects and have the advantage that the number of test groups is manageable.56,61 Fractional factorial designs have been shown to be an efficient, that is, cost effective, approach to identify interactive effects between seven trace elements and the cadmium accumulation in the body62 and to determine structure-activity relationships for 10 halogenated aliphatic hydrocarbons.63 Groten et al.64,65 used a fractionated twolevel factorial design for a combination of nine chemicals in a subacute rat study. Main and interactive effects of the agents could be detected using 16 different test groups, whereas a full design would have required 29 – 1 (511) experimental groups. The design is based on a general balance of groups with and without one of the compounds. By subtracting the mean of the groups not containing the agent from the mean of the other groups, the main effects of each agent can be found based on effect addition. A measure of non-additivity (interaction) is the difference between the effect of an agent in the presence of another one and the effect of the agent in the other one’s absence. This study applied a 1/32 fraction of a complete design of all possible combinations of nine chemicals. Undoubtedly, the confounding pattern would have been less complex and interpretation of the data would have been much easier when a 1/16 fraction or a 1/8 fraction had been applied. However, one has to realize that the number of test groups in such cases will increase to unmanageable numbers, viz. 32 or 64, respectively. For the same reason the analysis was restricted to one dose level, although a three-level factorial analysis might have produced additional information about dose response relationships.59 This also implies that we have to be careful to extrapolate the cases of non-additivity found to lower dose regions. If the risk assessor attempts to model the complete dose response surface of the mixture, another approach might be followed in which the interest is focused on departure from additivity of the mixture as

whole, rather than on the identification of specific interactions. This approach, suggested by Berenbaum,66 utilizes single chemical dose response information in addition to the observed responses induced by the particular combinations of interest. Clearly, the advantage offered by such an approach becomes apparent after comparing the number of experimental groups employed in its application to that required by the use of factorial designs.67 Apart from this limitation, the strength of the fractionated factorial approach is the possibility to formulate a more accurate hypothesis as to most of the relevant or critical interactions, and to show which compounds are important in a particular mixture in terms of hazard. The major drawback of fractionated factorial designs is the possible occurrence of inferences that result from a high level of aliasing, and of prior knowledge input needed to be able to interpret experimental results. For instance, if in the study by Groten et al.65 the treatment combinations were selected in such a way that first-order formaldehyde interactions were disregarded that greatly helped the interpretation of the alias pattern. Thus, to some extent the use of prior knowledge is essential for properly designing this type of study. Other useful designs are available, for example, a ray design that can be used for calculating the combination index (Eq. 1) or isoboles of a mixture. At least 10 points (3 points on each ray and the control point) are needed for studying a 2-compound mixture. This design can be used as a follow-up after a first screening study with, for example, a factorial design.57 Also, a central composite design can be useful to study deviations from additivity comparing the central point with all other points. It is also used to determine an optimum effect and it is very useful in combination with response-surface analysis. All these designs are based on the assumption that the dose response curves are monotonic. Statistical designs that account for nonlinear dose response relationships are lacking. It may be emphasized that the use of statisti-


86


cal designs may lead to misinterpretations of results when information on individual compounds of a chemical mixture is missing.

an answer to this question, the paragraphs that follow deal with a number of major aspects of risk assessment of simple and complex chemical mixtures.

IV. RISK ASSESSMENT OF CHEMICAL MIXTURES

A. Simple (Defined) Mixtures

The reason for concern regarding risks of mixtures is obvious. Man is always exposed to more than one chemical at a time: environmental contaminants, voluntary exposures (e.g., food, medicines, tobacco smoke), or exposure at the workplace (e.g., coke ovens, forges, pesticide spraying). Moreover, a wealth of experimental evidence exists that various chemicals interact toxicologically.3 Whenever this information is available, it should be noticed in risk assessment. However, there are all sorts of problems with risk assessment of chemical mixtures and, as has been described in the previous sections, with hazard identification as well. Current approaches and new developments in risk assessment have been elegantly discussed and summarized for non-carcinogens17,52,57,68 and for carcinogens.68 Testing of all kinds of (complex) mixtures of chemicals existing in the real world or of all possible combinations of chemicals of a simple (defined) mixture at different dose levels is virtually impossible. Moreover, even if toxicity data on individual compounds are available, we are still facing the immense problem of extrapolation of findings obtained at relatively high exposure concentration in laboratory animals to man being exposed to (much) lower concentrations. This problem has been recognized recently as one of the key issues in the assessment of possible health risks from disinfection byproducts in drinking water.69 Indeed, what is the predictive value of combined or interactive adverse effects occurring (in laboratory animals) at relatively high exposure concentrations for man being exposed to (much) lower concentrations of the chemical mixture? Focusing on

A simple, defined mixture consists of a relatively small number of chemicals, say at most 10, and its composition is qualitatively and quantitatively known.70 In the last 7 years, short-term repeated-dose toxicity studies in rats with a variety of defined mixtures (combinations) of chemicals have been carried out at our institute. The combinations consisted of chemicals (a) with different target organs and/or different modes of action, (b) with the same target organ but with different modes of action, (c) with the same target organ but with different target sites, or (d) with the same target organ and the same mode of action. The basic concepts of simple dissimilar action or simple similar action were applied to the toxicity data obtained in a number of these studies in order to assess the potential health risks of these combinations in a descriptive way.6,7 This assessment is summarized briefly hereafter. Currently, the potential health risks of these combinations are being assessed in a numerical way.

1. Combinations of Chemicals with Different Target Organs and/or Different Modes of Action Chemicals and dose levels used in two different 4-week toxicity studies in rats with combinations of eight or nine compounds with different target organs and/or different modes of action are presented in Tables 1 and 2.65,71,72 The main purpose of these studies was to determine whether simultaneous administration of the compounds at a dose level equal to the NOAEL for each of the chemicals would result in a NOAEL or an adverse effect level


87

TABLE 1 Test Compounds and Their Organ(s) as Well as the Dose Levels (ppm in Diet or Drinking Water) Used in a 4-Week Oral Toxicity Study of a Combination of Eight Chemicals in Rats71,a Compound

NOAEL/10b

NOAEL/3

NOAEL

LOAELc

10 100 500 20 0.5 0.5 3 0.6

33 330 1670 70 1.7 1.7 10 2

100 1000 5000 200 5 5 30 6

300 3000 20000 1000 25 80 100 30


KNO2d SnCl2 ⋅ 2H2O NasS2O5 Metaldehyde Loperamide Mirex Lysinoalanine DOTCe a b c d e

Target organ Adrenals Body wt, blood, liver Blood, stomach, liver Liver Body wt. Body wt., liver, blood Kidneys Thymus, liver

The study also included an appropriate control group. NOAEL = No observed adverse effect level. LOAEL = Lowest observed adverse effect level. Only KNO2 was administered in the drinking water. DTOC = di-n-octyltin dichloride.

for the combination. It may be emphasized that the NOAEL and the LOAEL of the individual chemicals had been established previously in similar 4-week toxicity studies in the same strain of rats kept under comparable experimental conditions. At the LOAEL of the

combinations both more and less severe effects were observed than at the LOAEL of each of the individual compounds present in the respective combinations, indicating interaction of toxic effects of the combinations at this dose level. Except for some minor chang-

TABLE 2 Test Compounds and Their Target Organ(s) as Well as the Dose Levels (ppm in Diet or Air) Used in a 4-Week Toxicity Study of a Combination of Nine Chemicals in Male Rats67,a Compound Aspirin Cadmium chloride Stannous chloride Loperamide Spermine BHAd DEHPe Dichloromethane Formaldehyde a b c d e

NOAEL/3b

NOAEL

LOAELc

Target

330 3 260 2 130 330 65 30 0.3

1000 10 800 6 400 1000 200 100 1

5000 50 3000 30 2000 3000 1000 500 3

Liver, stomach Red blood cell, liver Red blood cell Liver Heart, liver Stomach Liver Blood Nose

The study also included an appropriate control group. NOAEL = No observed adverse effect level. LOAEL = Lowest observed adverse effect level. BHA = Butyl hydroxyanisol. DEHP = Di(2-ethylhexyl)phthalate.


88


es in a few parameters (over 80 different parameters were measured), no treatment-related adverse effects were found at the NOAEL of the two combinations. Both studies allowed the overall conclusion that exposure to the combinations of chemicals did not constitute an evidently increased hazard, provided the dose level of each individual chemical is a NOAEL. The results of both studies also suggested, although in an indirect manner because they were designed for a different purpose, that in the case of a quantitatively arbitrary composition of these combinations the NOAEL of the combination would largely, if not entirely, be determined by the NOAEL of the chemical with the smallest margin between its actual level in the combination and its true no adverse effect level, assuming that the actual level of each of the chemicals in the combination is lower than its true no adverse effect level, or expressed in terms of health risk, the risk of the combinations studied is largely, if not entirely, determined by the risk associated with the chemical in the combination with the highest risk quotient, provided the risk quotient of each of the other chemicals in the combination is at most equal to unity. We suggest designating the chemical in a combination or mixture with the highest risk quotient as the “most risky chemical” of the combination or mixture. What factors determine the nature and extent of the health hazard of a mixture or combination of chemicals administered at adverse effect levels for several or most of the chemicals may be of interest from an academic point of view but is nearly always practically irrelevant. Clearly, the combined action of the chemicals in the aforementioned combinations very closely, if not fully, meets the criteria of the basic concept “simple dissimilar action”: at levels around or lower than the NOAEL the chemicals in the combinations did not clearly affect each other’s toxicity. Nevertheless, as stated before, at the NOAEL of each of both combinations some minor effects were observed. If indeed these minor changes were

treatment related, they are indicative of some form of additivity or interaction. However, in each case the degree and nature of the additive or interactive effects were so minor that in our view they do not need to be taken into account in assessing the health risks of the combinations. Such minor additional uncertainties would undoubtedly be adequately covered by the safety factor applied to the NOAEL of the combination that, as we have seen, is identical to the NOAEL of the most hazardous chemical in the combination.7,15

2. Mixture of Chemicals with the Same Target Organ but with Different Modes of Action Chemicals and dose levels used in a 4-week oral toxicity study in rats with a mixture of four nephrotoxicants with different modes of action are presented in Table 3.27 The aim of the study was to find out whether simultaneous administration of the compounds at a dose level equal to the no observed nephrotoxic effect level (NONEL) for each of the chemicals would result in a NONEL or a nephrotoxic effect level for the mixture. At the low-

TABLE 3 Test Compounds and Dose Levels (ppm in Diet) Used in a 4-Week Oral Toxicity Study of Four Nephrotoxicants in Rats27,a Nephrotoxicant Lysinoalanine Mercuric chloride Hexachloro-1,3butadiene d-Limonene a

b c

NONEL/4b

NONEL

LONELc

7.5 3.75

30 15

240 120

5

20

100

125

500

4000

The study also included an appropriate control group. NONEL = No observed nephrotoxic effect level. LONEL = Lowest observed nephrotoxic effect level.


89


est observed nephrotoxic effect level (LONEL) of the mixture, both greater and lesser effects were seen than at the LONEL of each of the individual compounds, indicating additive or interactive nephrotoxic effects at this dose level. No nephrotoxic effects unambiguously related to treatment were found at the NONEL of the mixture, allowing the conclusion that simultaneous administration of these chemicals with the same target organ (the kidneys) but with different working mechanisms did not constitute a clearly increased hazard, provided the administered dose of each of the nephrotoxicants is a NONEL. Similar to what appeared from the results of the studies described in the previous section, also in this case the NONEL of the mixture is determined by the NONEL of the chemical with the smallest margin between its actual level in the mixture and its true nephrotoxic effect level, assuming that the actual level of each of the compounds in the mixture is lower than its true no nephrotoxic effect level. In terms of health risks, the risk of this mixture of nephrotoxic chemicals is determined by the risk associated with the risk of the most risky chemical

(= the chemical with the highest risk quotient) of the mixture.7,15

3. Mixtures of Chemicals with the Same Target Organ but with Different Target Sites Chemicals and exposure levels used in a number of 3-d inhalation toxicity studies in male rats with mixtures of two or three toxicants and with the individual toxicants as well are presented in Table 4.73,74 Formaldehyde induced clear nasal changes in rats exposed to 3.2 ppm; no formaldehyde-related changes were seen at 1.0 ppm. In olfactory epithelium minor changes of doubtful toxicological significance were found in some rats exposed to 750 or 1500 ppm acetaldehyde. Acrolein induced marked nasal changes at 0.67 ppm that were similar in degree and incidence to those observed in rats exposed to 3.2 ppm formaldehyde. Similar but less distinct nasal changes were seen in rats exposed to 0.25 ppm acrolein. Nasal changes in rats exposed to 1.0 ppm formaldehyde + 0.25 ppm acrolein (Mix 1) or to 1.0 ppm formalde-

TABLE 4 Exposure Concentrations (ppm) of Formaldehyde, Acetaldehyde, and Acrolein Used in a Series of 3-D (6 h/d) Inhalation Toxicity Studies in Male Rats73 Groupsa

Formaldehyde

Form/low Form/high Acet/low Acet/high Acro/low Acro/high Mix 1 Mix 2 Mix 3

1.0 3.2

a

Acetaldehyde

Acrolein

750 1500

1.0 1.0 3.2

750 1500

0.25 0.67 0.25 0.25 0.67

Each separate study included an appropriate control group exposed to clean air only.


90


hyde + 750 ppm acetaldehyde + 0.25 ppm acrolein (Mix 2) were very similar in site, type, degree, and incidence to those induced by 0.25 ppm acrolein alone. Consequently, the nasal lesions seen in rats exposed to Mix 1 or Mix 2 are considered acrolein-induced effects uninfluenced by co-exposure to 1.0 ppm formaldehyde or to 1.0 ppm formaldehyde + 750 ppm acetaldehyde. Pronounced nasal lesions were observed in rats exposed to 3.2 ppm formaldehyde + 1500 ppm acetaldehyde + 0.67 ppm acrolein (Mix 3). The lesions were always much more marked and more extensive than those found after exposure to the individual aldehydes at comparable concentrations. This latter finding indicates at least summation of the acrolein and formaldehyde effects on the nasal respiratory epithelium and probably potentiation of the acetaldehyde effect on the olfactory epithelium by formaldehyde and/or acrolein. Whatever, the mechanism underlying the more severe effects of Mix 3 may be, no clear dose (concentration) addition occurred following exposure to mixtures of these aldehydes at the NOAEL or LOAEL (of one of the aldehydes) because the severity of the nasal changes seen after exposure to Mix 1, Mix 2, or 0.25 ppm acrolein alone was similar. Clearly, only when dose effect curves of each of the single compounds of a mixture are available, can one actually calculate a combined effect, which should be compared with the observed effect of the mixture to find out whether there is interaction between two or more chemicals in the mixture. In this example, neither effect addition nor potentiating interactions occur, provided the exposure concentrations of the aldehydes are NOAELs. Even if one of the aldehydes (acrolein) is administered at the LOAEL, this conclusion seems to be valid. In summary, these findings suggest that combined exposure to these chemicals (aldehydes) with the same target organ (nose) and exerting the same type of adverse effect (nasal cytotoxicity), but with different target sites (different regions of the nasal mucosa), is

not associated with a greater hazard than exposure to the individual chemicals, provided the exposure concentrations are NOAELs. This may be true also in the case of the exposure concentration of one of the compounds in the mixture is a LOAEL. These studies also indicate that the type of combined action or interaction found at clearly toxic effect levels (e.g., 3.2 ppm formaldehyde plus 0.67 ppm acrolein) is not very helpful in predicting what will happen at no toxic effect levels, including levels only slightly lower than the LOAEL. Even if one of the chemicals occurs at a slightly toxic effect level (in the example 0.25 ppm acrolein being the LOAEL) the type of combined action of the mixture may be different from that occurring at clearly toxic effect levels. However, precisely what happens at no toxic effect levels (including exposure levels only slightly lower than the LOAEL) is what counts in assessing the potential health risk of humans exposed to mixtures of these chemicals. Obviously, as soon as the exposure levels of the chemicals in the mixture are in the range of NOAELs, no additivity and no potentiating interaction were found, indicating the applicability of the basic concept “simple dissimilar action” (or “independent joint action”). The implication is that, as with the previous examples, the NOAEL of the “most risky chemical” in the mixture (acrolein) would also be the NOAEL of the mixture. A safety factor of “normal” size applied to the NOAEL would do to establish an acceptable exposure limit for such a mixture.

4. Mixture of Chemicals with the Same Target Organ and the Same Mode of Action Chemicals and dose levels used in a 4-week oral (gavage) toxicity study in female rats with mixtures of three or four nephrotoxicants with the same mode of action are presented in Table 5.75


91

TABLE 5 Test Compounds and Dose Levels Used in a 4-Week Oral Toxicity Study in Female Rats with Mixtures of Nephrotoxicants75 Treatment (doses in mg/kg body weight)

Total dose in toxicity units


Control: corn oil 10 ml/kg 0 Individual compounds at NONELe TETRAa 600 mg/kg TRIb 500 mg/kg 1.5 mg/kg TCTFPc 1.0 mg/kg HCBDd

1/4 1/4 1/4 1/4

Individual compounds at LONELf TETRA 2400 mg/kg TRI 2000 mg/kg TCTFP 6.0 mg/kg HCBD 4.0 mg/kg

1 1 1 1

Combinations of all 4 compounds at NONEL at LONEL/2

1 2

Combinations of 3 compounds TETRA + TRI + TCTFP at LONEL/3 TETRA + TRI + HCBD at LONEL/3 TETRA + TCTFP + HCBD at LONEL/3 TRI + TCTFP + HCBD at LONEL/3

1 1 1 1

a b c d e f

TETRA = TRI = TCTFP = HCBD = NONEL = LONEL =

tetrachloroethylene. trichloroethylene. 1,1,2-trichloro-3,3,3-fluorpropene. hexachloro-1,3-butadiene. No observed nephrotoxic effect level (= LONEL/4). Lowest observed nephrotoxic effect level.

The nephrotoxicity of these (chlorinated) chemicals results from initial conjugation to glutathione in the liver, and cysteine conjugate (lyase)-mediated formation of reactive metabolites in the proximal tubular epithelial cells. Relative kidney weight was increased following exposure to the individual compounds at their LONEL and, to about the same extent, following combined exposure at the NONEL (= LONEL/4) or LONEL/3. The other endpoints used to assess renal toxicity (viz. histopathology, concentrating ability, urinary excretion of glucose, protein, and marker enzymes,

and plasma creatinine and urea) were not or only scarcely affected after combined exposure at the NONEL or LONEL/3. Interpretation therefore is complicated because these endpoints, unlike kidney weight, were not affected at the LONEL of each of the individual compounds. Co-administration at the LONEL/2 resulted in clear nephrotoxicity, as indicated by the effects on most of the above endpoints. It is concluded that the renal toxicity, as assessed by the effect on kidney weight, of a mixture of similarly acting nephrotoxicants, at levels slightly below the LONEL of the individual compounds corresponded to the effect


92


expected on the basis of the additivity assumption.75 It is self-evident that for this example the “dose addition” model (simple similar action or similar joint action)17 represents the basic concept to be used for risk assessment. This model is applicable over the whole range of exposure levels from low non-toxic levels to NOAELs (= the highest level of each chemical at which no adverse effect was observed). The “dose addition” concept is the most common approach to risk assessment of mixtures.17 However, its use is only justifiable from a scientific point of view when all chemicals in the mixture act in the same way, by the same mechanism, and thus differ only in their toxic potencies.17 Of course, in reality, the existence of mixtures that meet such conditions are the exception rather than the rule. As evidenced by the other examples, described application of the “dose addition” model to mixtures of chemicals that act by mechanisms for which the additivity assumptions are invalid would greatly overestimate the risk.

B. Complex Mixtures A complex mixture consists of tens, hundreds, or thousands of chemicals, and its composition is qualitatively and quantitatively not (fully) known and may change.70 Examples are cigarette smoke, welding fumes, food products, workplace atmospheres, and total volatile organic chemicals in indoor or outdoor environments. There are all kinds of problems with hazard identification and risk assessment of complex mixtures. Often adequate toxicity testing of the entire mixture is impossible because a sufficiently high dose cannot be applied, for example, in the case of a new food product a high dose may lead to an imbalanced diet, or cigarette smoke that contains highly acutely toxic compounds such as carbon monoxide, rendering the use of a high dose of cigarette smoke in long-term studies impossible. Low-dose extrapolation may not be allowed; at high concentrations, the contri-

bution of the different chemicals to the toxicological effect may be proportionally different from that at low concentrations.7 On the other hand, testing of complex mixtures occasionally has produced relevant toxicological information, for example, on diesel exhaust carcinogenicity, coal dust toxicity, and finally also on cigarette smoke carcinogenicity. Testing of all individual chemicals is virtually impossible, and even if we were able to generate the toxicity data of all individual chemicals, we would still be facing the problem of a complex mixture with its numerous possible combined actions or interactions. Sometimes fractionation is appropriate, for example, for identification of toxic or mutagenic compounds or groups of compounds, and for substitution of compounds to create a less toxic mixture. However, fractionation leads to other artifactual complex mixtures, the toxicity of which may be very different from that of both the compounded chemicals and the mixtures as a whole.6 A way to tackle the safety evaluation of large or complex mixtures might be the socalled two-step procedure recently described by Feron et al.7 (Figure 5). The first step consists of selection of chemicals with the highest risk potential, for example, the top ten chemicals. The second step comprises assessment of the health hazards and potential health risks associated with exposure to the mixture of the selected (top ten) chemicals. Selection of the top ten chemicals should be based on the level of exposure and the level of toxicity of the individual chemicals. The ratio between the two is called the risk quotient. Hazard identification and risk assessment of the mixture of the top ten chemicals should be based on toxicity data and on data on mechanism of action of the individual chemicals and on prediction of presence or absence of additive or interactive effects. Where data are absent, expert judgement should be used. In brief, identify the, say 10, most risky chemicals in the complex mixture and approach these 10 chemicals as a simple mix-


93


FIGURE 5. Two-step procedure for safety evaluation of complex chemical mixtures. (Modified from Feron et al.)

ture, assuming that hazard and risk of this simple mixture are representative for hazard and risk of the entire complex mixture. For more details of this two-step procedure the reader is referred to Feron et al.6,7

C. Hazard Index (HI) and Weight-of-Evidence (WOE) Approach The hazard index as put forward in the U.S. EPA mixture guidelines should be regarded as an approximation of the risk posed


94

by exposure to the mixture. In this approach the hazard quotients are calculated for individual compounds and the quotients for each compound in the mixture are then added. The ‘hazard index’ (HI)13 can be calculated from equation HI =

E1 E E + 2 +… n DL1 DL 2 DL n


(7) in which E is the level of exposure and DL is some defined limit exposure value. If the equation exceeds unity, the concern is the same as if an individual chemical exposure exceeded its acceptable level by the same proportion.13 In the case of independent joint action, the HI is that of the agent with the highest quotient. Thus, in a three-compound mixture ABC with exposure levels of A = 1, B = 0.5, and C = 75 and a DLs of A = 10, B = 1, and C = 100, the HIs are 0.1, 0.3, and 0.75 for the individual chemicals, respectively. In the case of dissimilar joint action, the HI of the mixture ABC will be 0.75, whereas with similar joint action the HI of the mixture is 1.25. If the hazard index is greater than 1, there might be an increased risk of exposure to this mixture. The HI method is based on the assumption of dose additivity (similar joint action) and thus is only valid under the condition that the compounds in the mixture induce similar toxic effects through the same mode of action. Often it will be impossible to get sufficient and adequate toxicological information on each of the compounds of a mixture to make these calculations. Thus, data on interaction are not included in this approach. Modifications of the Standard HI Approach are being developed to take account of the data on interactions. For instance, the U.S. Subcommittee on Mixtures of the National Research Councils Safe Drinking Water Committee has put forward the idea of an additional uncertainty factor (between 1 and 100) to protect against potential synergistic interactions.76 In this approach an uncertainty factor of 10

would indicate that relatively high concentrations of the mixture compounds are present, but without clear knowledge about the synergistic interactions to occur. A different approach, originally published by Mumtaz and Durkin,77 takes into account both synergistic and antagonistic iterations in the derivation of the HI. In this approach an weight-of-evidence (WOE) classification is followed to estimate the joint actions (additivity, antagonism, and synergism) for binary mixtures of chemicals based on information about the individual compounds. Several “weighing factors” are taken into account in the final classification, such as the mechanistic understanding of the binary interactions, the demonstration of toxicity, and additional uncertainty factors, that is, modifiers of interactions, such as route of exposure, in vitro data, etc. The better the data set on the individual compounds is, the more precise the joint action can be predicted. Therefore, in support of the comments of Seed et al.,78 it is very plausible that no interaction can be estimated for several of the possible binary pairs of chemicals. In the WOE method the chance of interactions is anticipated on worst case conditions and is mostly built on data at high dose levels. One major drawback in this method is that the extrapolation of interactions from high doses to low doses is not incorporated in the evaluation. One of the challenges in future risk evaluation using this so-called binary weight of evidence (BINWOE) approach is to estimate with more precise interaction factors at realistic, environmentally relevant dose levels. Nevertheless, at this moment the WOE approach should be regarded as one of the most universal approaches in estimating the interaction between compounds in a given mixture. Although the method is based on classic concepts of joint action (i.e., dose addition and relatively simple deviations from dose addition), the method has not been evaluated using experimental or epidemiological data. Therefore, the Agency for Toxic Sub-


95


stances and Disease Registry (ATSDR) and the Netherlands Organization for Applied Research (TNO) in a joint effort are currently in the process of determining whether the general WOE approach can be used to account for observed interactions in experimental mammals. In this evaluation, ATSDR prepared BINWOEs on combinations of chemicals included in mixture studies by TNO. For instance, the analysis of a four-component mixture of a similarly acting nephrotoxicants (trichloroethylene, tetrachloroethalene, hexachloro-1,3,butadiene, and 1,1,2,-trichloro3,3,3,fluoropropene) revealed that the WOE approach correctly predicted the observed the toxicity in the experimental studies.85 Preliminary results indicate that for the mixtures with similar-acting toxicants, the WOE method quantitatively predicts the observed interactions. For the mixtures with dissimilaracting toxicants, the quantitative evaluation seems to be less accurate.79 In coming years the WOE method will be under continuing review by ATSDR, TNO, US EPA and the Netherlands Ministry of Public Health, Spatial Planning and Environment to adjust the quantitative evaluation of dissimilar-acting compounds in a mixture. V. SUMMARIZING CONCLUSIONS From a public health point of view, it is most relevant to answer the question whether chemicals in a mixture interact in a way that results in a reduced or increased overall response when compared with the sum of the responses to the individual chemicals in the mixture, or indeed in an effect that is simply a summation of the expected effects. By and large, regulatory actions and industrial practices are based on the use of dose-addition as the default assumption addition has been stated most explicitly for assessing carcinogenic risks.80 Although, in general, this is a prudent practice, this assumption may usually overestimate risk and sometimes underestimate risk. Approaches for dealing with risk assess-

ment of mixtures rely heavily on some form of additivity model unless data are adequate for a direct risk assessment of the mixture of concern. Although the previously described additivity models are mathematically simple, they require assumptions about the mechanisms and the mode of action. The number of mixtures to which direct risk assessment or risk assessment using the relative potency method have been devoted is limited. The theoretical considerations in risk assessment of chemical mixtures should be verified by simple case studies as published recently.21,26,27,64,65,72–75,79,81–84 The usefulness of a method for hazard identification and risk assessment of a complex mixture very much depends on the context in which one is confronted with a complex mixture and also on the information that is available on the chemistry and toxicity of the mixture. The way to approach a complex mixture of which the components are largely known but whose hazard is almost unknown is different from the way to approach a complex mixture whose hazard has been established but the chemical composition of which is largely unknown. The development of decision trees to tackle complex mixtures has been suggested70 and should be pursued. The weight-of-evidence approach77 might prove to be very useful in evaluating simple defined mixtures. In conclusion, to study the toxicology of chemical mixtures and to assess their potential health risks it is essential to be familiar with the basic concepts of combined action and interaction of chemicals in a mixture and to understand the available methods for designing studies and for analyzing the data obtained. The introduction of concepts of combination toxicology in procedures for hazard identification and risk assessment of complex mixtures seems a prerequisite. The use of effect addition to predict the risk of a mixture will frequently result in an overestimation of the risk and from time to time in an underestimation. Overall, the potential adverse health effects of realistic exposure lev-


96


els to (mixtures of) chemicals should be a primary research topic of toxicologists for the near future, focusing on mechanisms of action and development of approaches to assess interaction data.11 It is a challenge for toxicologists to verify the hypothesis that as a rule exposure to chemical mixtures at nontoxic (realistic) exposure levels of each of the individual constituents of the mixture is of no health concern. To timely detect the exceptions to this rule is the real challenge. Issues (or non-issues) to be taken up by the mixture toxicologist are endocrine disrupters, organophosphates, and other cholinesterase-inhibiting chemicals, outdoor air respiratory tract irritants, and indoor air volatile organic chemicals.

ACKNOWLEDGMENT The authors gratefully acknowledge the advice of Prof. Dr. Gerald Pöch.

REFERENCES

and the Environment, The Toxicology Forum, Berlin, 1994, 390–421. 6. Feron, V. J., Groten, J. P., Jonker, D., Cassee, F. R., and van Bladeren, P. J., Toxicology of chemical mixtures: challenges for today and the future, Toxicology, 105, 415– 427, 1995. 7. Feron, V. J., Woutersen, R. A., Arts, J. H. E., Cassee, F. R., de Vrijer, Fl., and van Bladeren, P. J., Safety evaluation of the mixture of chemicals at a specific workplace: theoretical considerations and a suggested twostep procedure, Toxicol. Lett., 76, 47–55, 1995. 8. Feron, V. J. and Bolt, H. M., Combination toxicology, Food Chem. Toxicol., 34, 1025– 1185, 1996. 9. Yang, R. S. H., Toxicology of Chemical Mixtures, Academic Press, San Diego, 1994, 720. 10. Steel, G. G. and Peckham, M. J., Exploitable mechanisms in combined radiotherapychemotherapy: the concept of additivity, Int. J. Radiation Oncol. Biol. Phys., 5, 85–91, 1979. 11. Henschler, D., Bolt, H. M., Jonker, D., Pieters, M. N., and Groten, J. P., Experimental designs and risk assessment in combination toxicology: panel discussion, Food Chem. Toxicol., 34, 1183–1185, 1996.

1. Krishnan, K. and Brodeur, J., Toxicological consequences of combined exposure to environmental pollutants, Arch. Complex Environ. Studies, 3, 1–104, 1991.

12. Feron, V. J., van Bladeren, P. J., and Hermus, R. J. J., A view point on the extrapolation of toxicological data from animals to man, Food Chem. Toxicol., 28, 783–788, 1990.

2. Yang, R. S. H., Introduction to the toxicology of chemical mixtures, in: Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994, 1–10.

13. US EPA, Guidelines for the Health Risk Assessment of Chemical Mixtures, US Environmental Protection Agency, Fed. Regist., 51(185), 34014–34025, 1986.

3. Calabrese, E. J., Multiple Chemical Interactions, Lewis Publishers, Chelsea, 1991, 704. 4. Mumtaz, M. M., Sipes, I. G., Clewell, H. J., and Yang, R. S. H., Risk assessment of chemical mixtures: biologic and toxicologic issues, Fund. Appl. Toxicol., 21, 258–269, 1993. 5. Feron, V. J., Jonker, D., van den Berg, M., Bogaards, J. J. P., Groten, J. P., and van Bladeren, P. J., Some principles of combination toxicology applied to defined mixtures of (chlorinated organic) chemicals, in: Proceedings of the Conference on Chlorinated Organic Chemicals: Their Effect on Human Health

14. US EPA, Technical Support Document on Health Risk Assessment of Chemical Mixtures, EPA/600/8-90/064, Office of Health and Environmental Assessment, Environmental Assessment and Criteria Office, Cincinnati, Ohio, 1990. 15. Feron, V. J., Groten, J. P., Jonker, D., Cassee, F. R., and van Bladeren, P. J., Risk assessment of simple (defined) mixtures of chemicals, in Toxicology and Air Pollution: Risk Assessment, Burgat-Sacaze, V., Descotes, J., Gaussens, P., and Leslie, G. B., Eds. Faculté de Médicine et de Pharmacie, Université de Bourgogne, Dijon, 1995, 31–42.


97

16. Bliss, C. I., The toxicity of poisons applied jointly, Ann. Appl. Biol., 26, 585–615, 1939. 17. Mumtaz, M. M., DeRosa, C. T., and Durkin, P. R., Approaches and challenges in risk assessments of chemical mixtures, in Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994, 565–598.


18. Kodell, R. L. and Pound, J. G., Assessing the toxicity of mixtures of chemicals, in Statistical Methods in Toxicological Research, Krewski, D. and Franklin, C., Eds., Gordon and Breach Science Publ., New York, 1988. 19. Reif, A. E., Synergism in carcinogenesis, J. Nat. Cancer Inst., 73, 25–39, 1984. 20. Weber, L. J., Anderson, P., Muska, C., Yinger, E. K., and She-Hon, D., Concentration and response addition of mixtures of toxicants using lethality, growth and organ system studies. Presented at the ASA/EPA Conference on Joint Toxicity, Washington, DC, 1987. 21. Cassee, F. R., Arts, J. H. E., Groten, J. P., and Feron, V. J., Sensory irritation to mixtures of formaldehyde, acrolein, and acetaldehyde in rats, Arch. Toxicol., 70, 329–337, 1996. 22. Pöch, G., Dittrich, P., Reiffenstein, R. J., Lenk, W., and Schuster, A., Evaluation of experimental combined toxicity by use of dose frequency curves: comparison with theoretical additivity as well as independence, Can. J. Physiol. Pharmacol., 68, 1338–1345, 1990. 23. Loewe, S., and Muischnek, H., Über Kombinationswirkungen, Arch. Exp. Pathol. Pharmak., 114, 313–326, 1926. 24. Carter, W. H., Gennings, C., Staniswalis, J. G., Campbell, E. D., and White, K. L., A statistical approach to the construction and analysis of isobolograms, J. Am. Coll. Toxicol., 963– 973, 1988. 25. Yang, R. S. H. and Rauckman, E. J., Toxicological studies of chemical mixtures of environmental concern at the National Toxicology Program: Health effects of groundwater contaminants, Toxicology, 7, 15–34, 1987. 26. Jonker, D., Jones, M. A., van Bladeren, P. J., Woutersen, R. A., Til, H. P., and Feron, V. J., Acute (24 h) toxicity of a combination of four nephrotoxicants in rats compared with the toxicity of the individual compounds, Food Chem. Toxicol., 31, 45–52, 1993.

27. Jonker, D., Woutersen, R. A., van Bladeren, P. J., Til, H. P., and Feron, V. J., Subacute (4-week) oral toxicity of a combination of four nephrotoxins in rats: comparison with the toxicity of the individual compounds, Food Chem. Toxicol., 31, 125–136, 1993. 28. Chaturvedi, A. K., Toxicological evaluation of mixtures of ten widely used pesticides, J. Appl. Toxicol., 13, 183–188, 1993. 29. Heindel, J. J., Chapin, R. E., Gulati, D. K., George, J. D., Price, C. J., Marr, M. C., Myers, C. B., Barnes, L. H., Fail, P.A., Grizzle, T. B., Schwetz, B. A., and Yang, R. S. H., Assessment of the reproductive and developmental toxicity of pesticide/fertilizer mixtures based on confirmed pesticide contamination in California and Iowa groundwater, Fund. Appl. Toxicol., 22, 605–621, 1994. 30. Ito, N., Hasegawa, K., Imaida, K., Kurat, Y., Hagiwara, A., and Shirai, T., Effect of ingestion of 20 pesticides in combination at acceptable daily intake levels on rat liver carcinogenesis, Food Chem. Toxicol., 33, 159–163, 1995. 31. Miners, J. O., Veronese, M. E. and Bricket, D. J., In vitro approaches for the prediction of human drug metabolism. Chap. 31. Annual Reports in Medical Chemistry, 29, 307–316, 1994. 32. Houston, J. B., Relevance of in vitro kinetic parameters to in vivo metabolism of xenobiotics, Toxic. In Vitro, 8, 507–512, 1994. 33. Houston, J. B., Relevance of in vitro drug metabolism data in predicting in vivo metabolic clearance, Biochem. Pharmacol., 47, 1469–1479, 1994. 34. Andersen, M. E., Clewell, III, H. J., and Frederick, C. G., Applying simulation modeling to problems in toxicology and risk assessment—a short perspective. Toxicol. Appl. Pharmacol., 133, 181–187, 1995. 35. Leung, W.-W and Paustenbach, D. J., Physiologically based pharmacokinetic and pharmacodynamic modeling in health risk assessment and characterization of hazardous substances, Toxicol. Lett., 79, 55–65, 1995. 36. Fiserova-Bergerova, V., Extrapolation of physiological parameters for physiologically based simulation models, Toxicol. Lett., 79, 77–86, 1995.


98

37. Yang, R. S. H., El-Masri, H.A., Thomas, R. S., Constan, A. A. and Tessari, J. D., The application of physiologically based pharmacokinetc/pharmacodynamic (PBPK/PD) modeling for exploring risk assessment approaches of chemical mixtures, Toxicol. Lett., 79, 193–200, 1995.


38. Andersen, M. E., Clewell, III, H. J., Gargas, M. L., Smith, F. A., and Reitz, R. H., Physiologically based pharmacokinetics and the risk assessment for methylene chloride, Toxicol. Appl. Pharmacol., 87, 185–205, 1987. 39. Ploemen, J. H .T. M., Wormhoudt, L. W., Haenen, G. R. M. M., Oudshoorn, M. J., Commandeur, J. N. M., Vermeulen, N. P. E., de Waziers, I., Beaune, P. H., Watabe, T., and van Bladeren P. J., The use of human in vitro metabolic parameters to explore the risk of hazardous compounds: the case of ethylene dibromide, Toxicol. Appl. Pharmacol., 193, 56– 64, 1997. 40. Wolf, C. R., Metabolic factors in cancer susceptibility, Cancer Surv., 9, 437–473, 1990. 41. Hissink, A. M., Oudshoorn, M. J., Van Ommen, B., Haenen, G. R. M. M., and Van Bladeren, P. J., Differences in cytochrome P450-mediated biotransformation of 1,2-dichlorobenzene by rat and man: implications for human risk assessment, Chem. Res. Tox., 9(8), 1249–1256, 1996. 42. Gessner, P. K., A straightforward method for the study of drug interactions: an isobolographic analysis primer, J. Am. Col. Toxicol., 7, 987– 1012, 1988. 43. Berenbaum, M. C., Criteria for analyzing interactions between biologically active agents, Adv. Cancer. Res., 35, 269–335, 1981. 44. Greco, W. R., Park, H. S., and Rustum, Y. M., Application of a new approach for the quantitation of drug synergism to the combination of cis-diamminedichloroplatinum and 1β-arabinofuranosylcytosine, Cancer Res., 50, 5318–5327, 1990.

47. Loewe, S., The problem of synergism and antagonism of combined drugs, Arzneim. Forsch., 3, 285–290, 1953. 48. Smuziel, I. and Nias, A. H. W., Isobologram analysis of the combined effects of anti-tumor platinum complexes and ionizing radiation on mammalian cells, Br. J. Cancer, 42, 292–296, 1980. 49. Gennings, C., Dawson, K. S., Carter, W. H., and Myers, R. H., Interpreting plots of a multidimensional dose response surface in a parallel coordinate system, Biometrics, 46, 719– 735, 1990. 50. Pöch, G., Evaluation of combined effects with respect to independent action, Arch. Compl. Environ. Studies, 3, 65–74, 1991. 51. Pöch, G., Combined Effects of Drugs and Toxic Agents: Modern Evaluation in Theory and Practice, Springer Verlag, New York, 1993. 52. Carter, W .H. and Gennings, C., Analysis of chemical combinations: relating isobolograms to response surfaces, in: Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994, 643–664. 53. Plummer, J. L. and Short, T. G., Statistical modeling of the effects of drug combinations. J. Pharmacol. Methods, 23, 297–309, 1990. 54. Schoen, E. D., Statistical designs in combination toxicology: a matter of choice, Food Chem. Toxicol., 34, 1059–1065, 1996. 55. Box, G. E. P., Hunter, W. G., and Hunter, J. S., Statistics for Experimenters, John Wiley & Sons, New York, 1978, 653. 56. Svengaard, D. J. and Hertzberg, R. C., Statistical methods for toxicological evaluation, in: Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994 599-642.

45. Greco, W. R., Bravo, G., and Parsons, J. C., The search for synergy: a critical review from a response surface perspective, Pharmacol. Rev., 47, 332–385, 1995.

57. Nesnow, S., Mechanistic linkage between DNA adducts, mutations in oncogenes, and tumorgenesis of carcinogenic polycyclic aromatic hydrocarbons in strain A/J mice, in Chemical Mixtures and Quantitative Risk Assessment, Abstract of the Second Annual HERL Symposium, November 7 to 10, Raleigh, NC, 1994.

46. Pöch, G. and Reiffenstein, R. J., Inappropriate and wrong conclusions derived from isobolograms, Arch. Complex Environ. Studies, 6, 33–36, 1994.

58. Narotsky, M. G., Weller, E. A., Chinchilli, V. M., and Kevlock, R. J., Nonadditive developmental toxicity of mixtures of trichloroethylene (di(2-ethyl-hexyl)phtalate, and hep-


99

tachlor in a 5x5x5 design, Fund. Appl. Toxicol., 27, 203–216, 1995. 59. Eide, I. and Zahlsew, K., Inhalation experiments with mixtures of hydrocarbons. Experimental design, statistics and interpretation of kinetics and possible interactions. Arch. Toxicol., 70, 397–404, 1996.


60. Plackett, R. L. and Burman, J. P., The design of optimum multifactorial experiments, Biomatrika, 33, 305–339, 1946. 61. Groten, J. P., Sinkeldam, E. J., Muys, T., Luten, J. B., and van Bladeren, P. J., Interaction of dietary Ca, P, Mg, Mn, Cu, Fe, Zn, and Se with the accumulation and oral toxicity of cadmium in rats, Food Chem. Toxicol., 29, 249–258, 1991. 62. Eriksson, L., Jonsson, Hellberg, E. J., Lindgren, F., Skagerberg, B., Sjöström, M., Wold, S., and Berglind, R., A strategy for ranking environmentally occurring chemicals. III. Multivariate quantitative structure-activity relationships for halogenated aliphatics, Environ. Toxicol. Chem., 9, 1339–1351, 1991. 63. Groten, J. P., Schoen, E. D., van Bladeren, P. J., and Feron, V. J., Subacute toxicity of a combination of nine chemicals in rats: a twolevel factorial design to predict interactive effects, Toxicologist, Abstract no. 1153, 1994. 64. Groten, J. P., Schoen, E. D., van Bladeren, P. J., Kuper, F. C. F., van Zorge, J. A., and Feron, V. J., Subacute toxicity of a combination of nine chemicals in rats: detecting interactive effects with a two-level factorial design, Fund. Appl. Toxicol., 36, 15–29, 1997. 65. Berenbaum, M. C., The expected effect of a combination of agents: the general solution, J. Theor. Biol., 114, 413–431, 1985. 66. Gennings, C., An efficient experimental design for detecting departure from additivity in mixtures of many chemicals, in Chemical Mixtures and Quantitative Risk Assessment, Abstract of the Second Annual HERl symposium, November 7 to 10, 1994, Raleigh, NC. 67. Portier, J. C. and Sherman, C. D., Potential effects of chemical mixtures on the carcinogenic process within the context of the mathematical model, in: Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994, 665–686.

68. ILSI, Disinfection byproducts in drinking water: critical issues in health effects research, International Life Sciences Institute, Health and Environmental Sciences Institute, Washington, D.C., 1996, 168. 69. Fay, R. M. and Feron, V. J., Complex mixtures: hazard identification and risk assessment, Food Chem. Toxicol., 34, 1175–1176, 1996 70. Jonker, D., Woutersen, R. A., van Bladeren, P. J., Til, H. P., and Feron, V. J., 4-week oral toxicity study of a combination of eight chemicals in rats: comparison with the toxicity of the individual compounds, Food Chem. Toxicol., 28, 623–631, 1990. 71. Groten, J. P., Schoen, E. D., and Feron, V. J., Use of factorial designs in combination toxicity studies, Food Chem. Toxicol., 34, 1083– 1089, 1996. 72. Cassee, F. R., Groten, J. P., and Feron, V. J., Changes in the nasal epithelium of rats exposed by inhalation to mixtures of formaldehyde, acetaldehyde and acrolein, Fund. Appl. Toxicol., 29, 208–218, 1996. 73. Cassee, F. R., Groten, J. P., and Feron, V. J., Combined exposure of rat upper respiratory tract epithelium to aldehydes, Hum. Exp. Toxicol., 13, 726, 1994. 73a. Cassee, F. R., Stenhuis, W. H., Groten, J. P., and Feron, V. J., Toxicity of formaldehyde and acrolein mixtures: in vitro studies using nasal epethelial cells. Exp. Toxic. Pathol., 48, 481–483, 1996. 74. Jonker, D., Woutersen, R. A., and Feron, V. J., Toxicity of mixtures of nephrotoxicants with similar or dissimilar mode of action, Food Chem. Toxicol., 34, 1075–1082, 1996. 75. National Research Council, Selected Issues in Risk assessment. II. Mixtures (Drinking Water and Health). Vol. 9. Natl. Acad. Press, Washington, DC, 1989, 95–184. 76. Mumtaz, M. M. and Durkin, P. D., A weight of evidence approach for assessing interactions in chemical mixtures, Toxicol. Ind. Health, 8, 377–406, 1992. 77. Seed, J., Brown, R. P., Olin, S. S., and Foran, J. A., Chemical mixtures: current risk assessment methodologies and future directions, Reg. Toxicol. Pharmacol., 22, 76–94, 1995.


100

78. DeRosa, C. R., Johnson, B. L., Fay, M., Hansen, H., and Mumtaz, Public health implications of hazardous waste sites: findings, assessment and research, Food Chem. Toxicol., 34, 1131–1138, 1996. 79. McClellan, R. O., Foreword, in Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994, xxi–xxiv.


80. Cassee, F. R. and V. J. Feron, Biochemical and histopathological changes in nasal epithelium of rats after 3-day intermittent exposure to formaldehyde and ozone alone or in combination, Toxicol. Lett., 72, 257–268, 1994. 81. Reuzel, P. G. J., Wilmer, J. W. G. M., Woutersen, R. A., Zwart, A., Rombout, P. J. A., and Feron, V. J., Interactive effects of ozone and formaldehyde on the nasal respira-

tory lining epithelium in rats, J. Toxicol. Environm. Health, 29, 279–292, 1990. 82. Simmons, J. E., Nephrotoxicity resulting from multiple chemical exposures and chemical interactions, in Toxicology of Chemical Mixtures, Yang, R. S. H., Ed., Academic Press, San Diego, 1994, 335–360. 83. Mautz, W. J., Kleinman, M. T., Phalen, R. F., and Crocker, T. T., Effects of exercise exposure on toxic interactions between inhaled oxidant and aldehyde air pollutants, J. Toxicol. Environm. Health, 25, 165–177, 1988. 84. Mumtaz, M. M., DeRosa, C. R., Groten, J. P., Feron, V. J., Hansen, H., and Durkin, P. R., Estimation of toxicity of chemical mixtures through modeling of chemical interactions. Environ. Health. Perspect., in press.


101