Click for next page ( 122


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 121
Risk Assessment of M ixtures of Systemic Toxicants in Drinking Water Even if the resources now devoted to studies of chemical interactions could be multiplied by 1,000 or more, those studies would provide only a small portion of the information required to determine precisely the toxicity of complex mixtures prevailing in the environment. It would still fall to toxi- cologists and other scientists to integrate the separate pieces of information into a form useful for risk assessment. With the goal of providing adequate but not excessively conservative exposure standards for health protection, this chapter presents various approaches to using incomplete empirical in- formation and scientific judgment to assess the health risks associated with exposure to mixtures in drinking water. The dominant concern with mixtures, as noted earlier, is unexpected am- plification of toxicity arising from combinations of mixture components. The extent of the problem increases with the complexity of the mixture. For example, in a combination of 10 toxicants there are ('), or 45, possible 2- factor interactions; (~3), or 120, possible 3-factor interactions; (I''), or 210, possible 4-factor interactions; ('I), or 252, possible 5-factor interactions; (a), or 210, possible 6-factor interactions; (a), or 120, possible 7-factor interactions; ('), or 45, possible 8-factor interactions; (a), or 10, possible 9-factor interactions; and 1 possible 10-factor interaction. The total number of interactions that must be considered in a study of combinations of 10 substances is: 'The notation ('I) refers to the number of two-substance combinations possible from a group of 10 substances; the general notation, (m), can be rewritten as m!lc!(m - c)!, where ! is the symbol forfactorial.m!equalsm(m - l)(m-2)...(2)(1)i.e., 10! = (10)(9)(8)(7)...(2)(1). 121

OCR for page 121
122 DRINKING WATER AND HEALTH ~0 (',"y = 2"'- ll = 1~013. =2 (1) It is difficult to imagine a procedure developed from consideration of the effects of 10 separate agents that adequately accounts for each of the 1,013 possible interactions. To model interaction requires precise definition of the term "interaction." As described in the 1988 publication by the Committee on Methods for the In Vivo Toxicity Testing of Complex Mixtures (NRC, 1988, p. 1001: Int~r~rtinn char Hill rPfP~rPA to tic clPvi~tinn from Alla Lo V__~lJ 4~ V ~ A- I---- . . . additive behavior expected on the basis of dose-response curves obtained with individual agents. [But] to a biostatistician, the definition includes information on the underlying dose-response model and on units of measure. For ordinary linear models, "interaction" refers to a departure from response additivity. Different measures of response can lead to different conclusions concerning departure from additivity. For nonlinear models, such as log-logistic, log-linear, log-probit, and multistage models, . . . there is additivity for the logarithm of response. Thus, when the word "interaction" is used, one must make certain of the units in which toxicity or dose is expressed, as well as the assumptions about the nature of joint action that was predicted by the model used. THE PROBLEM OF SYNERGISM Of greatest concern is the possibility that exposure to mixtures could result in a much more severe toxic response than that expected on the basis of the potencies of the individual components. "Synergism" is here defined as a response greater than additivity. The mixture problem for drinking water is mainly a problem of relatively low concentrations of individual agents on which human observational data are lacking. Demonstrations of synergism come largely from the laboratory and are usually based on high doses. A limited review of the literature uncovered no cases in which the potency of a particular agent at high doses had been multiplied by more than 100 by combination with another agent. Furthermore, in the extensive literature on drug interactions, novel effects are rarely observed. Instead, if binary combinations act synergistically, they tend to do so through an increase in the effects of one of the agents. Assume, then, that 100 represents the maximal enhancement of toxicity due to the addition of a second agent. That is also the number by which no-observed- adverse-effect levels (NOAELs) or their advocated replacement, bench- mark doses are divided to yield acceptable daily intakes (ADIs), or "reference doses" (the EPA term), although the uncertainty factor of 100 is usually designed to protect against uncertainties other than those arising from inter- actions. The number 100 has been said to incorporate a species variation of

OCR for page 121
Risk Assessment of Mixtures of Systemic Toxicants 123 10 and a range of individual human susceptibility of 10, though formal documentation of this view is lacking. THE CONCEPT OF COMMONALITY One concept that might be useful in combining numerical estimates of toxicity of different substances is that of commonality, formulated by Weiss (1986) in relation to food additives. (Food additives are regulated on the basis of individual ADIs, no explicit account being taken of the possibility that their basic actions might not be mutually independent.) Commonality reflects the extent to which several agents are likely to act on the same target organ and elicit the same toxic response. For instance, several organophos- phorus and carbamate chemicals have the commonality of inhibiting acetyl- cholinesterase. Compounds of the heavy metals mercury and cadmium disrupt renal cellular function and produce renal necrosis. "Commonality" does not now translate readily into any simple arithmetic manipulations, but Figure 3-1 is an attempt to illustrate the concept. The ADI, or acceptable level (AL), is transformed on the Taxis into total toxic burden or exposure. The x-axis corresponds to the number of chemicals in the mixture to which there is exposure. The y-axis (commonality) represents the overlap in biologic effects of various chemicals. It represents the degree to which the individual constituents of a mixture contribute to a particular process. It recognizes that most agents have broad biologic actions, so their toxic contribution depends on the choice of a specific end point. Fewer agents or a lower commonality would lead to a lower total burden. A higher com- monality, assuming a fixed intake as a proportion of the ADI for each agent, would raise the total burden. A commonality factor of less than l implies less than full identity of action. MODIFYING THE HAZARD INDEX EPA (1986b) has published guidelines for assessing the health risks as- sociated with chemical mixtures. If data on the toxicity of a specific mixture as a whole are unavailable, the risks may be estimated on the basis of what is known about the individual constituents. A hazard index (HI) associated with a mixture of k toxicants may be defined as: Ek ,2 ALi (2) where Ei is the exposure level of the ith toxicant in the mixture and ALi is the maximum acceptable concentration of the ith toxicant. When the HI exceeds 1, it evokes the same concerns about the mixture

OCR for page 121
124 DRINKING WATER AND HEALTH /~CR=~_ ~ FIGURE 3-1 Three-dimensional idealized representation illustrating increasing toxic burden with the addition of several agents. Lower commonality would reduce total burden; higher commonality would increase it. This illustration applies to only one system or organ at a time. Total burden would be measured in percent of ADI. as those evoked when any individual AL is exceeded, which is, in fact, a special case, with k = 1. This computation of the hazard index is based on the assumption of adding toxically equivalent doses i.e., the absence of interactions among the components of the mixture. However, such dose- additive models as are implied by the hazard index might not provide the most biologically plausible approach to describing the effects of a complex mixture of toxicants if the compounds do not have the same mode of toxic action. EPA (1986b) has suggested that separate hazard indexes should be developed for separate end points of concern. It further notes that "dose addition for dissimilar effects does not have strong scientific support, and,

OCR for page 121
Risk Assessment of Mixtures of Systemic Toxicants 125 if done, should be justified." In addition, EPA recognizes that the biologic meaning of the index given above is unknown and that much additional research is required before its utility as a predictor of toxic severity can be validated. Other Risk-Oriented Approaches Identification of risk associated with the consumption of chemical mixtures in drinking water will come from many sources: laboratory data, such as those obtained from bioassay-directed fractionation; epidemiologic data, such as those based on similar exposures; and models based on knowledge of the toxicity of individual constituents. When models are the primary source, the assumptions described above and the already available empirical data point to an empirical policy: adoption of the additivity assumption for low-response exposures. Three strategies for protection against the risks associated with exposure to drinking water contaminants might be defined in this argument. The choice of strategy is a policy issue based on public concern; the data are clearly inadequate to dictate a firm choice. 1. Least adequate. Each agent's ADI is determined independently; hence, regulations would consider each agent individually. This alternative is un- satisfactory, because of the low likelihood that each agent can truly act in biologic isolation of every other agent. Furthermore, individually acceptable risks might lose their acceptability when there are 50 or 100 of them. 2. Usually adequate. A water quality or hazard index similar to the EPA index would be defined under some assumptions of additivity. Given that ADIs already incorporate uncertainty factors, usually of 100, and given that only one study (Murphy et al., 1959) has reported a 100-fold increase in toxicity through a binary combination (of organophosphates), the simple dose-additive model should be satisfactory under most circumstances. For materials that are assumed to have thresholds for response, special concern must be given to the biologic mechanisms leading to a toxic response. If the mechanisms of toxicity of two or more toxicants are the same, combining below-threshold doses (i.e., doses with zero response) could lead to an above- threshold dose and produce a nonzero response. This concept is inherent in the consideration of toxic equivalents and the idea of relative potency, which implies that one material is (in effect) a dilution of the other material. For these reasons, dose-additivity must be considered for materials that affect the same toxic end point. 3. Most adequate. Although most attempts to examine the joint actions of binary combinations find that toxic responses are well described by simple additivity (e.g., S myth et al., 1969) or that additivity is violated to only a

OCR for page 121
126 DRINKING WATER AND HEALTH small degree, maximal protection might require an additional margin of safety beyond that provided by the additivity model. Because of the 100-fold in- crease observed by Murphy et al. (1959), adjustments of an index based on additivity by an equivalent factor should provide a rather robust safety margin. Such a margin, however, would be excessive in most cases. That is, just as it seems unreasonable to expect totally independent actions of the agents identified in a drinking water source, it seems equally unreasonable to expect complete overlap in toxic targets, so even the additivity alternative already embodies this additional but unrecognized safety factor. Later in this chapter, it is suggested that an uncertainty factor can be considered to compensate for any synergism. In addition to the EPA hazard index, there are other precedents for com- bining the toxic manifestations of individual components of a mixture. Threshold limit values (TLVs), which are workplace standards, are combined for si- multaneous exposure to more than one agent of the same class of compounds (ACGIH, 1983; NRC, 19801. For example, the joint action of several volatile organic solvents, which induce central nervous system impairment, is esti- mated by adding the individual ratios of exposure concentration to TLV. For some combinations, such as n-hexane and methylethyl ketone, the assump- tion of additivity might underestimate toxicity, even though both compounds are organic solvents. The additive model might seem reasonable only for conditions in which the mode of action of the constituents is well understood and greater than additive effects can therefore be excluded on the basis of mechanistic processes. It might be adapted, however, for situations in which drinking water standards, such as MCLs, are based on ADIs derived from chronic studies devoid of mechanistic information. One way to apply the additive model to extract an overall assessment of drinking water quality is to use the dose-additive procedure adopted by the State of New York to regulate aldicarb and carbofuran. The prescribed limits are 7 and 15 ~g/liter, respectively. The actual concentrations at any site are used in the following expression: (actual aldicarb) (actual carbofuran) T. 15 (3) If T ' l, no action is taken. If T > 1, filters are installed, and the cost is charged to the manufacturer whose product is calculated to give the higher ratio. The procedure is based on the assumption that both agents act on identical systems and are totally interchangeable. A corresponding expression can be based on many more than two agents, as for workplace TLVs. If dose additivity is assumed, as in the case in New York, then the sum of the ratios identified for the hazardous constituents can

OCR for page 121
Risk Assessment of Mixtures of Systemic Toxicants 127 be required to be no more than 1.0, on the basis of the sum of quotients determined as described above in Equation 3. The hazard index defined by Equation 2 can be modified to take into account the sensitivity of each toxic end point. As noted earlier, applying the uncertainty factor incorporated in ALs or ADIs is intended to provide a safe (nontoxic) exposure of the most sensitive persons for the toxic mani- festation or end point that occurs at the lowest dose if a given material has several manifestations of toxicity. The operating assumption is that an ex- posure that is safe for the most sensitive end point will be safe for any other toxic manifestation. If sensitivity is defined by the no-observed-adverse-effect level (NOAEL) the most sensitive end point being defined as the one with the lowest NOAEL- then the relative sensitivities of various end points can be estimated by comparing NOAELs or ratios of NOAELs. That is, sensitivity is taken to be the inverse of the NOAEL, and relative sensitivity is defined as the ratio of the NOAEL for the most sensitive end point to the NOAEL for a specific subsidiary end point. If this relative sensitivity, or ratio of NOAELs, is used as a weighting factor, called Wij (which will always be less than or equal to 1), then the equation for HI can be modified to read: j ,2 AL`, (4) where i is the toxic material and j is some specific end point. This procedure assumes that the AL for each substance is not known for each end point of concern; if it were, 1/ALij could be substituted for Wij/ALij in the above equation. The HIj would then be computed by summing HIj across all ma- terials i. If any of the HIj values exceeded 1, the sum also must exceed 1, and the actions indicated above for the unmodified HIj would be taken. This method is sensitive to fractionation by the number of toxicity categories used; thus, it might be most reasonable to sum across all toxic end points: HI= 2, HIj, i=] (5) where I is the number of toxic end points considered. Modifying HI by using the weighting factor Wij takes into account the possibly different toxic spectra of different materials, avoids the lumping together of unrelated toxicities, and still incorporates all reported toxicities into a unified measure. Grouping Agents with Common Toxicities With a small number of agents, it might be possible to invoke expert opinion to separate components of toxic mixtures into clusters and to assign

OCR for page 121
128 DRINKING WATER AND HEALTH an average commonality to each cluster. Or, as an upper limit in the dose- additive model, a commonality of 1.0 could be assumed within each cluster; this would lead to dose additivity. The concept could also be extended by estimating commonalities between clusters defined as chemical classes. For example, clusters as diverse as the heavy metals, organic solvents, and insecticides might all have a given end point, such as peripheral nerve dam- age. Clustering could be on the basis of mechanistic, toxic, or structural properties. Structural classification, in fact, is the first step by which EPA assesses potential hazards associated with new chemicals, so the process is not unfamiliar for estimating commonalities within a group. With larger numbers or a lack of sufficient data to describe a toxicity profile, such a procedure might prove unwieldy; but commonalities among clusters could still be considered after commonality within a cluster is established. The concept of commonality clearly could be helpful, but more work is needed to develop its application. Incorporating an Uncertainty Factor for Synergism The issue of toxic interactions synergistic or antagonistic is central in the development of a risk assessment strategy for chemical mixtures in drink- ing water. Even though the concentrations of contaminants in most sources of drinking water for the general public are likely to be very low, there is insufficient evidence about the toxicity of chemical mixtures after long-term, low-dose exposure to support a definite conclusion that toxic interactions are absent under these conditions. For instance, a combination of chemicals, even at low concentrations, could conceivably act to modify the immune system, thereby compromising natural defense systems. Evidence supporting the existence or absence of such a response process is clearly not available for all relevant mixtures. The argument for the consideration of greater than response additive effects is strengthened by the possibility that water sources in heavily polluted areas (e.g., hazardous-waste sites or point-source acci- dental spills) contain much higher concentrations of contaminants. Thus, at least a small fraction of the population is sometimes exposed to relatively high (parts per million) concentrations of mixtures of chemicals in drinking water. These types of concerns could lead to approaches that are more pro- tective than those given above. It is important to address the issue of toxic interactions in the risk assess- ment of chemical mixtures, but flexibility must be provided to avoid inap- propriate regulation. Therefore, the additive approach represented by Equation 5 might be modified by incorporating an additional uncertainty factor (UF), which could be applied to the hazard index defined above to yield: HI = (UF) ~ HI j=, (6)

OCR for page 121
Risk Assessment of Mixtures of Systemic Toxicants 129 The UF could vary from 1 to 100, depending on the amount of information available and the concentrations of the contaminants. If a great deal of toxicologic information is available on the individual contaminants, if toxic interactions are not likely (on the basis of the knowledge available), or if the concentrations of the contaminants are "low," the UF might be set at l (thus assuming simple additivity). If less is known about the toxicity of individual components and the concentrations of the contaminants are higher, the UF might be set at 10. The greater the uncertainty involved (because of the lack of information) and the higher the concentrations of the contaminants, the higher the UF would be set. Other Considerations In keeping with the NRC report on complex mixtures (1988, p. 8), we have here defined "low" dose for carcinogens as one associated with a true relative risk of less than or equal to 1.01. Such a small relative risk is likely to be unmeasurable, and the "low" dose considered would be one estimated to produce a relative risk of 1.01 or less. The present subcommittee first considered tying the previous report's definition of "low" dose to detection limits, but decided not to do so, because this concept is related more to the physicochemical properties of an agent than to its biologic or toxic properties. Moreover, detection limits are likely to change as measurement technology improves. Further investigation into the magnitude of synergism as a function of dose might yield guidance here. Such investigation would involve the analysis of empirical data, alternative dose-response models, and additional laboratory experiments. "Amount of information available" remains unavoidably ambiguous. Ob- viously, if perfect information were available about the existence and nature of any synergism, it could be included in a risk assessment. In the absence of complete knowledge, information about synergism among similar agents or about the mechanisms of toxicity of agents would be helpful. In considering uncertainty factors, we recognize that a different type of uncertainty factor is already incorporated into the estimation of ADIs and ALs. Traditionally, this safety margin (usually a factor of 100) was provided to allow for possible interspecies and interindividual variations. Potential synergistic effects were not used in this development of safety margins. From a different perspective, however, the incorporation of a UF is not intended to trigger regulatory action every time contaminants are detected in drinking water sources. The hazard index in the new approach must be assessed in actual situations. The above approaches apply to mixtures of systemic toxicants. For car- cinogens, models that assume no threshold for response are recommended?

OCR for page 121
130 DRINKING WATER AND HEALTH and indeed they are used; hence, the constructs associated with ALs or ADIs do not apply. Mixtures of carcinogens are considered in Chapter 7. For noncancer end points, dose-response models are not widely available (or accepted), because most toxicity evaluations have described effects in terms of single numerical values, such as the EDso or NOAEL. That is reasonable when a single compound is under consideration, but it is not necessarily realistic for a combination. When it is known that the compounds interact to produce dose-additive effects, it might be useful to use the hazard index. RESPONSE-SURFACE MODELING Just as knowledge of the concentration-response curve is necessary to characterize the toxicity of a single agent, knowledge of the concentration- response surface is required to characterize the toxicity of a combination. Response-surface methods are mathematical and statistical techniques that have been developed to aid in the solution of particular types of problems in scientific and engineering processes (Box and Draper, 1987; Carter et al., 1983; Cornell? 1981; Khuri and Cornell, 1987; Myers, 19761. The report ASAIEPA Conferences on the Interpretation of Environmental Data. 1. Cur- rent Assessment of the Combined Toxicant Effects (EPA' 1986a) referred briefly to the potential applicability of response-surface methods for describ- ing the effects of combinations of toxicants The methods include experi- mental design, statistical inference? and mathematical techniques that, when combined, enable an experimenter to explore empirically the process of interest. One important aspect of this approach is experimental design. It can be used to guide the generation of data suitable for estimating the parameters of a statistical model of the dose response relationship (NRC 19881. Ex- perimental designs have been developed to estimate the effects of each com- ponent of a mixture and the effects of interactions on the basis of a small number of experimental points. Such designs could be useful for indicating the presence or absence of low-order interactions, once an appropriate model has been determined. One design that permits the estimation of the inter- actions of interest and requires only a small number of experimental test groups is the 2-factorial experiment. Each of k factors is present at two levels, and each level of each factor is combined with each level of every other factor. Such a protocol requires 2t observations. The data from such experiments allow the estimation of the effect of each of k toxicants and each of the interactions of 2,3, . . . ? k toxicants (factors). If some interactions are of less interest, more compact experimental designs (fractional factorial designs), which require fewer groups of exposed subjects, are available (NRC, 19881.

OCR for page 121
Risk Assessment of Mixtures of Systemic Toxicants 131 The response-surface approach has been used to evaluate combinations of toxic substances (Kinne, 1972; Schnute and McKinnell, 1984; Voyer and Heltshe, 1984; Voyer et al., 19821. Its use seems appropriate in this setting and should be encouraged. The availability of computer graphics has per- mitted plotting of observed and fitted data as either the surface or contours of constant response. Such plots are often informative in assessing interac- tions among components. However, their use is limited to combinations of only a few toxicants. There is some expectation that graphic procedures can be developed to display the associated surfaces in higher dimensions; these efforts should be encouraged and supported. CONCLUSIONS Given the current availability of information about the presence and mag- nitude of synergism among contaminants in drinking water, no single risk- assessment approach can be justified scientifically for systemic toxicants. The possibility of synergism cannot be ignored, so approaches that consider only individual toxic agents are likely to be inadequate. Approaches based on additivity of response or additivity of doses that produce a given response are likely to be useful in many situations. In risk assessments of mixtures, these approaches should be limited to groups of agents that have similar mechanisms of action and act at the same biologic site. Nevertheless, the possibility remains that some unexpected and important synergism (or antagonism) exists and that either additive model could seri- ously underestimate risk. Societal and policy concerns about the existence of such interactions could lead to the introduction of further uncertainty factors in risk assessment. Uncertainty factors should not be uniform, but should increase with increasing exposure and decrease with increasing knowl- edge about the agents in mixtures. REFERENCES ACGIH (American Conference of Governmental and Industrial Hygienists). 1983. TLVs: Threshold Limit Values for Chemical Substances and Physical Agents in the Work Envi- ronment, with Intended Changes for 1983-1984. Cincinnati Ohio: American Conference of Governmental and Industrial Hygienists. Box, G. E. P., and N. R. Draper. 1987. Empirical Model-Buildin~ and Response Surface. New York: John Wiley & Sons. 669 pp. Carter, W. H., G. L. Wampler, and D. M. Stablein. 1983. Regression Analysis of Survival Data in Cancer Chemotherapy. New York: Marcel Dekker. 209 pp. Cornell, J. A. 1981. Experiments with Mixtures: Designs' Models, and the Analysis of Mixture Data. New York: John Wiley & Sons. 305 pp. EPA (U.S. Environmental Protection Agency). 1986a. ASA/EPA Conferences on the Inter

OCR for page 121
132 DRlNKlNG WATER AND HEALTH pretation of Environmental Data. 1. Current Assessment of the Combined Toxicant Effects, May 5-6, 1986. EPA (U.S. Environmental Protection Agency). 1986b. Guidelines for the health risk assessment of chemical mixtures. Fed. Regist. 51(185):34014-34025. Khuri, A. I., and J. A. Cornell. 1987. Response Surfaces: Designs and Analyses. New York: Marcel Dekker. 405 pp. Kinne, O., ed. 1972. Environmental factors. Part 3 in Marine Ecology. A Comprehensive Integrated Treatise on Life in Oceans and Coastal Waters, Vol. 1. New York: Wiley In- terscience. Murphy, S. D., R. L. Anderson, and K. P. du Bois. 1959. Potentiation of the toxicity of malathion by triorthotolyl phosphate. Proc. Soc. Exp. Biol. Med. 100:483-487. Myers, R. H. 1976. Response Surface Methodology. 246 pp. [Distributed by R. H. Myers. Department of Statistics, Virginia Polytechnic Institute and State University, Blacksburg. Virginia. ] NRC (National Research Council). 1980. Principles of Toxicological Interaction Associated with Multichemical Exposures. Washington, D.C.: National Academy Press. 213 pp. NRC (National Research Council). 1988. Complex Mixtures: Methods for In Vivo Toxicity Testing. Washington, D.C.: National Academy Press. 227 pp. Schnute, J., and S. McKinnell. 1984. A biologically meaningful approach to response surface analysis. Can. J. Fish. Aquat. Sci. 41:936-953. Smyth. H. F., C. S. Well, J. S. West. and C. P. Carpenter. 1969. An exploration of joint toxic action: I. Twenty-seven industrial chemicals intubated in rats in all possible pairs. Toxicol. Appl. Pharmacol. 14:340-347. Voyer, R. A. ~ and J. F. Heltshe. 1984. Factor interactions and aquatic toxicity testing. Water Res . 18:441 -447. Voyer, R. A., J. A. Cardin. J. F. Heltshe, and G. L. Hoffman. 1982. Viability of embryos of the winter flounder Pseudopleuronectes americanus exposed to mixtures of cadmium and silver in combination with selected fixed sabinities. Aquat. Toxicol. 2:223-233. Weiss, B. 1986. Emerging challenges to behavioral toxicology. Pp. 1-2 in Neurobehavioral Toxicology. Z. Annau, ed. Baltimore: Johns Hopkins University Press.