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 103
Animal Sentinels 6 in Risk Assessment The formalization of risk assessment has resulted in part because of the po- tential threats that toxic chemicals in the environment pose to human and environmental health (NRC, 1983~. A risk assessment is a tool for rational risk estimation and a guide for regulating exposures in cases where risk is judged to be excessive. Risk assessments can be conducted in a wide variety of circumstances, but often are focused on single chemical agents, limited geographic areas or modes of exposure, and defined populations. The assessment of risk due to environmental contaminants depends, to a large extent, on scientific data. When such data are incomplete, as is often the case, assumptions based on scientific judgments are made to calculate potential exposures and effects. Specifically, when direct observations of the effects of environmental contaminants on human or environmental health are incomplete or missing, assumptions must be made to estimate the risks (Coth- ern, 1989~. Those assumptions often are imprecise or speculative, so esti- mates of risks are uncertain. In some cases, the use of animal sentinels can reduce uncertainties by providing data on animals exposed in parallel to the humans whose risks are to be determined. The animal data can help risk assessors to make more accurate estimates. Animal sentinel data include data obtained from epidemiologic studies (descriptive and analytic) and from animal and food-chain monitoring pro- grams, as described in Chapter 3 of this report. Data from animal sentinel studies can often be obtained more quickly than data from human epidemio- logic studies, because the ideal sentinel responds to toxic insults more rapidly than humans (long before clinical manifestations of disease) and at environ- mentally relevant doses, i.e., doses similar to those at which humans are ex- posed. In addition, animal sentinels, like humans, are exposed to complex and variable mixtures of chemicals and other environmental agents. Those charac- teristics of animal sentinel studies offer important advantages over laboratory animal studies, in which animals are usually exposed to high, constant doses of a single chemical substance that is under investigation. Thus, the use of animal sentinels constitutes an approach to identifying hazards and estimating 103

OCR for page 103
104 ANIMALS AS SENTINELS risks in circumstances similar to those in which actual human exposures occur, and it is at least a complementary or alternative to traditional chemical toxicity testing through standardized laboratory studies. Data obtained in studies of animal sentinels also can lead to insights into human health by stimulating epidemiologic studies of humans exposed to agents that might not have been previously identified as potentially hazardous. They can be used to identify diseases related to chemicals in the environment (Schaeffer and Novak, 1988~. Systematic collection of such data in disease registries can help to identify unusual clusters of deaths, cases of disease, or cancers in defined populations and geographic areas. Collection of compara- ble information (i.e., exposures, toxicoses, and environmentally caused diseas- es) for humans and animals likely will improve understanding of diseases in humans, provide clues to etiology that cannot be evaluated in laboratory ani- mals, and provide a basis for evaluating the validity of sentinel data. Although risk assessment might not be the end use to which those data are applied, data collected through animal sentinel programs can provide some of the informa- tion necessary for risk assessment. Data from animal sentinels have been used In each of the four steps of risk assessment an in some aspects of risk management. Risk assessment and risk management were distinguished and defined as follows by a previous NRC committee (NRC, 1983~: · Risk assessment: "The characterization of the potential adverse health effects of human exposures to environmental hazards." · Risk management: "The process of evaluating alternative regulatory actions and selecting among them." The same committee divided risk assessment into four components and defined them as follows: · Hazard identification: "The process of determining whether exposure to an agent can cause an increase in the incidence of a health condition.. · Dose-response assessment: "The process of characterizing the relation between the dose of an agent administered or received and the incidence of an adverse health effect in exposed populations and estimating the incidence of the effect as a function of human exposure to the agent." · Exposure assessment: "The process of measuring or estimating the inten- si~, frequency, and duration of human exposures to an agent currently present in the environment or of estimating hypothetical exposures that might arise from the release of new chemicals into the environment." · Risk characterization: "The process of estimating the incidence of a

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 105 health effect under the various conditions of human exposure described in exposure assessment.. Although the division of the risk-assessment process and the definitions of the four components have been widely accepted and used for a variety of purposes, they were formulated specifically to assess human health risks, especially to estimate the risk of human cancer associated with exposure to chemical carcinogens. For wider application, including the uses discussed in this report, the definitions should be broadened to refer to risks to animals other than humans and to refer to agents other than chemicals. In the context of ecologic risk assessment, the distinctions between hazard identification and dose-response assessment and between hazard identification and risk charac- terization often are not clear. Nevertheless, the N~C definitions are useful in clarifying the steps involved in risk assessment. This chapter discusses the role of animal sentinels and animal sentinel data in the process of risk assessment related to human health. Risk assessment for nonhuman species (including domestic and wild arumals) also is discussed briefly. The chapter reviews how data from the systems described in Chapters 2-5 have been or could be used in the various steps of risk assessment and risk management and points out the value and limitations of animal sentinel systems for each purpose. Most animal sentinel systems provide some data on exposure, even when they are designed primarily to help in other steps of risk assessment. Therefore, exposure assessment is considered first in this chapter, and some studies are used as examples of exposure assessment and other steps of risk assessment. USE OFANIAiAL SENTINEL SYSTEMS IN EXPOSURE ASSESSMENT Animus as Gent Era Alias Animal sentinel systems have been used widely as components of general environmental monitoring schemes, many of which were discussed in earlier chapters. Although those monitoring systems provide information about the exposure of the animals that are sampled, their primary purpose is to provide information about contamination of the environment. The extent to which they do so reliably and quantitatively depends on the species selected for sampling, the sampling design, and other features of the monitoring program. For example, animal sentinel systems are useful for monitoring contaminants that are persistent in animal tissues, such as halogenated organic compounds

OCR for page 103
106 ANIAL4LSAS SENTINELS and metals. Some animals yield good samples for those contaminants, be- cause their tissues integrate exposures over appropriate temporal scales (such as the retention time for the contaminant in their tissues) and spatial scales (such as the foraging range of the animals during the same period of interest). However, the relationships between concentrations of contaminants in animal tissues and those in the environment are not known a priori and usual- ly must be determined by calibration or by pharmacokinetic modeling. Except in the Mussel Watch program, the constancy and stability of the relationships have not been investigated systematically (Farrington et al., 1983~. Thus, in many programs, the precision with which spatial and temporal patterns of contamination in the sentinel animals reflect those in the environment is open to question. More investigation is needed to improve the quantitative reliabili- ty of the animal sentinel systems. The relationships between ambient and tissue concentrations of contami- nants are difficult to establish and verify in free-ranging animals; for some purposes, sessile animals, such as mussels, offer important advantages (Far- rington et al., 1983~. Some of the best monitoring systems are in situ systems, in which the sentinel animals are placed and controlled so that the location and duration of their exposure are known precisely. To date, in situ systems have been used mainly for investigating very small-scale patterns of contami- nation or for real-time monitoring of effluents. In situ systems used for those purposes suffer the same drawbacks as do the systems based on wild animals- the precision and reliability with which they track spatial and temporal varia- tions in ambient concentrations are unknown. If such limitations could be overcome by better calibration, the systems would be very promising, at least for smald-scale applications. The utility and limitations of animals as monitors of environmental contam- ination have been discussed extensively elsewhere (e.g., NRC, 1979~. It should be emphasized here that the monitoring schemes generally provide informa- tion on patterns of environmental contamination i.e., information on the context of exposure, rather than on exposure itself. Al as Mo~ of ~ 0~ E - sum An exception to the generalization just stated is the sampling of tissues of animals that are of the same species as those whose exposure is to be as- sessed. Some examples of such studies have been reviewed In Chapters 3, 4, and 5. Those examples include the monitoring of human tissues or body fluids for pesticides, metals, and volatile organic compounds; the monitoring of predatory birds and mammals to assess their exposure to organochlorine

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 107 compounds; and several programs involving analysis of domestic or wild ani- mals that are thought to have suffered lethal poisoning or reproductive impair- ment as a result of localized contamination. As in the general environmental monitoring systems discussed in the previ- ous section, these studies involve measurement of blood or tissue concentra- tions of contaminants, rather than direct measurement of exposure. In some cases (e.g., lead in human blood, DDE in bird eggs), blood or tissue concen- trations provide immediately useful measures of exposure, because relation- ships between these concentrations and measures of effect are known from observation or experiment (Blus et al., 1972; Fyfe et al., 1988; Nisbet, 1988~. In other cases, however, blood or tissue concentrations cannot be used directly as measures of exposure- it might be necessary to derive a conversion factor, either empirically or by pharmacokinetic modeling, to obtain estimates of exposure or dose from measurements of tissue concentration. Aninais as Maims of Exposure of Their (I Animals can be used to monitor exposure most directly when the animal species sampled is used as food by the species whose exposure is to be deter- m~ned. Several monitoring systems of that type have been reviewed in Chap- ters 3 and 5. In particular, federal, state, and local agencies monitor contami- nants in human foods, including marketed foods of animal origin and wild fish and game species. Calculation of human intakes from measured contamination of food com- modities is not always straightforward. Contaminant concentrations in wild animals vary widely, and concentrations in animals sampled in the field can differ substantially from concentrations in those consumed after preparation and cooking (Humphrey, 1976~. Human consumption of fish and game ani- mals is extremely variable and poorly documented. Careful attention to sam- ple design and population characterization is needed, if studies of this kind are to identify the most highly exposed persons and population groups and to provide reliable estimates of their exposure. Mass Monitors of Humm Exposure Another way in which animals have been used in exposure assessments is as surrogates for exposed humans. When humans are exposed to contami-

OCR for page 103
lag ANI~lLSAS SENTINELS nants in complex environments (e.g., in the home or in the workplace), it might be difficult to estimate their exposure by the conventional procedure of measuring ambient concentrations of the contaminants and calculating their intakes from the contaminated media. One approach to solving that problem is to use animals as surrogate monitors; blood or tissues of animals exposed in the same environments as the people in question are taken for analysis and yield a measure of their exposure. If the animals' contact with the contaminat- ed media is sufficiently similar to that of the humans, the animals' exposure can be translated to a reasonable measure of the humans' exposure. Most examples of such animal sentinel systems involve the use of domestic or com- panion animals; several examples of the use of pets in such systems were reviewed in Chapter 4. The principal advantage of using animals as surrogate monitors is that their blood and tissues are often sampled. As discussed in Chapter 4, using animal sentinel data as quantitative measures of human exposure is difficult for sever- al reasons, including different types and concentrations of exposure in the same environment for animals than for humans and differences between animals and humans in metabolism and pharmacokinetics. In all the examples cited in Chapter 4, animal data have been used as qualitative or relative mea- sures of human exposure. One specific application of this type of monitoring is the use of animals to investigate the bioavailability of contaminants in the environment. Many inorganic and hydrophobic organic compounds are strongly adsorbed onto soil particles and airborne or waterborne particles. Even when human exposure to the particles (e.g., by inhalation or ingestion) can be estimated, the extent to which contaminants are desorbed from the particles and absorbed into the human body often is uncertain (Hawley 1985; Paustenbach et al., 1988~. Animals exposed by the same routes as humans can sometimes be used to determine bioavailability. For example, rats, guinea pigs, hamsters, and fish have been used In the laboratory to investigate the bioavailability of chior~nat- ed dioxins and dibenzofurans from contaminated fly ash and sediments (Van den Berg et al., 1983; Kuehl et al., 1987~. Similarly, rats and guinea pigs have been used in the laboratory to determine the bioavailability of tetrachlorodi- benzop~ioxin (TCDD) from contaminated soil after ingestion (McConnell et al., 1984; Umbreit et al., 1986~. Such studies have shown marked differences in bioavailability of TCDD in soil collected from different sites and thus illus- trate a role for animal sentinels in assessing differential bioavailability. In less-controlled conditions, wild animals have been used as indicators of the bioavailability of TODD in contaminated terrestrial environments (Fanelli et al., 1980; Young and Shepard, 1982; Bonaccorsi et al., 1984; Heida et al., 1986; Lower et al., 1989~. Studies have revealed broadly similar patterns of

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 109 uptake in animals from soil and food (Lower et al., 1989~. However, it is not clear that the results could be used other than qualitatively to infer the poten- tial for human exposure. Wild animals have been used widely to assess the bioavailability of metals, such as lead from soil, and hence to determine patterns of contamination and potential exposure (Williamson and Evans, 1972; Gish and Christensen, 1973; Goldsmith and Scanlon, 1977; Clark, 1979; Ash and Lee, 1980; Hutton and Goodman, 1980; Ohi et al., 1981~. Again, the results have not been used to develop quantitative estimates of human exposure, and it is not clear how they could be so used without extensive calibration exercises. USE OF ANIMAL SENTINEL SYSTEMS IN HAZARD IDEN71FICATION The primary method of identifying hazards posed by toxic chemicals—toxi- cology studies using laboratory animals- is not usually regarded as an animal sentinel system and does not fall within the definition of such systems in Chapter 1. Animal sentinel systems for identifying hazards in the environment can be categorized as early warnings and systems for screening mixtures of chemicals. Warty Wanun.gs: Initial Ideraifica~orl of Hn7~dals A,gf7'ts Table 1-1 lists a number of examples in which incidents of poisoning in wild or domestic animals provided the first indications of hazards posed by env~- ronmental contaminants or other agents. In many other cases that could be cited, the species first noted as affected were probably the species most at risk c.g., arsenic and selenium in domestic herbivores; pesticides in crusta- ceans, fish, and birds; and acidic pollutants in fish. Table 1-1 focuses on agents whose observation in animals was thought to have provided early warn- ing of potential effects in humans. In some of the cases listed in Table 1-1, it is still uncertain whether the agents pose important hazards to humans at concentrations commonly en- countered in the environment; the animals might have been more heavily exposed (e.g., to dioxins) or more susceptible (e.g., to agene). In other cases, it is now known that the agents pose similar hazards to exposed humans, and the animal data have been important in the stepwise procedure of human risk assessment. In most of the latter cases, however, it took a long time after the

OCR for page 103
110 AN1~4LS AS SENTINELS initial observation of animal poisonings to identify the causative agents and to confirm their toxicity. During that time, significant human exposure or injury had taken place (e.g., as a result of exposure to aflato~nn, dioxins, ergot, mer- cury, PBBs, and PCBs). In only a few cases were the early warnings provided by sentinel animals distinctive and decisive enough to trigger control measures before human exposure was recognized. One such case might be that of the pesticide isobenzan, whose manufacture was discontinued after substantial wildlife damage was reported in association with effluents from early produc- tion (Koeman, 1972~. However, human poisonings were also a factor (eager, 1970), and the decision to discontinue manufacture appears to have been influenced by the combination of occupational poisonings with environmental persistence and wildlife toxicity. Hn7~d Identification for Musical Mores Among the best examples of the use of animal sentinels to identify environ- mental hazards are the investigations of the prevalence of cancers in fish and shellfish living in polluted environments (reviewed in Chapter 5~. If the preva- lence of cancers in such species is high, it is reasonable to presume that they were exposed to carcinogenic combinations of pollutants, even if the specific agents primarily responsible cannot be identified (Matins et al., 1988~. It is then reasonable to infer that humans exposed to similar mixtures of pollutants would also be at risk. For example, it is reasonable to infer that human con- sumers of shellfish from the same waters would be at risk, because shellfish accumulate most of the pollutants without changing them. It is less clear that human consumers of fish would be at risk, because fish metabolize many pollutants and often do not accumulate them in their tissues. A general prin- ciple is that the extent to which an animal sentinel animal species can serve as an indicator of human hazard depends on the degree of similarity in eypo- sure. That in turn depends on many factors that influence the exposure of the humans and the sentinels, which must be assessed case by case. A more-controlled animal-based system for screening complex environmen- tal mixtures is the mobile laboratory developed by Legator et al. (1986) for investigating hazards posed by polluted air. Mobile-laboratory systems are still in the experimental phase of development and have not been validated or used to any substantial extent in risk assessment. If they can be validated, they will be useful for assessing hazards posed by environmental contaminants in real-world conditions. Any system in which animals are exposed to air, water, or other media in the natural environment involves exposure to many chemical agents in poorly

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 111 controlled combinations. Some environmental mixtures have been shown to be more toxic than would be predicted on the basis of their principal chemical constituents (e.g., Hornshaw et al., 1983~. The responses of animals exposed to those mixtures in situ are expected to provide more reliable and more timely information on the degree of hazard than could be derived from labo- ratory studies of their individual constituents. USE OF ANIMAL SENTINEL SYSTE1~S mr DOSE-RESPONSE ASSESSMENT Animal sentinel systems have not often been used to elucidate dose-re- sponse relationships, mainly because quantitative information on sentinel animals doses or exposures is rarely available. The major exception is a series of studies of effects of DDE on eggshell-thinning and reproductive success in wild birds. Those studies not only have yielded the effective ranges of dose (measured as DDE concentration in eggs) in various species, but also have provided information on the shape of the dose-response curves (Blus et al., 1972; Fyfe et al., 1988; Nisbet 1988~. Other studies have provided information on responses of animals to contamination gradients. With the exception of one, the examples in this report yielded exposure-response relationships that are directly applicable only to the species that were studied (rather than for use in assessing human risks). The only example of which the committee is aware that showed a dose-response relationship directly useful in assessing human risks is that of bladder cancer in dogs (see Chapter 4~. Extensive dose-response data are available on the effects of PCB-contaminated fish on reproduction in mink (Aulerich and Ringer, 1977; Hornshaw et al., 1983; Ringer, 1983~; but risk characterization for humans exposed to the same fish has been conducted by considering data on more conventional laboratory species (Swain, 1988; National Wildlife Federation, 1989; Tilson et al., 1990~. USE OF ANIMAL SENTINEL SYSTEMS IN RISK CHARAC1~;RIZAT7ON Many of the studies referred to in preceding sections of this chapter as examples of the use of animal sentinel systems in exposure assessment, hazard identification, or dose-response assessment also included applications to risk characterization. In addition, many other studies with animal sentinel systems have provided information directly for the characterization of risk, either to the animals that were studied or, by extension, to humans. This section pre-

OCR for page 103
112 ANIMALS AS SENTINELS sents some examples of the use of animal sentinel systems in risk charactenza- tion and discusses their advantages and disadvantages. Risk to Andnal Species Ruler Study Except for studies whose purpose was limited to monitoring of patterns of contamination, most of the studies that have been cited in this chapter provid- ed information that could be used in characterizing risks to the animal species that were studied. Among the most complete risk assessments are those for birds of prey. In the peregrine falcon, for example, epidemiologic studies revealed patterns of reproductive impairment and population decline and suggested an association with pesticide use; ecotoxicity studies identified the causative agents (DDE and dieldrin) and yielded dose-response information on the wild birds; controlled toxicity studies in related (surrogate) species confirmed the hazard identification and provided additional dose-response data; monitoring of peregrine falcons and their prey in the wild revealed spatial and temporal patterns of exposure; all this information was combined into a risk assessment that explained the results of the epidemiological studies and predicted the success of reintroduction efforts. Details of the assessments are published in various chapters of the symposium volume edited by Cade et al. (1988~. Similar information is available on other birds of prey, including the osprey (Poole, 1989) and the European sparrowhawk (Bogan and Newton, 1979~. The information developed in those extensive risk assessments can be extended to other bird species exposed to DDE or dieldrin, although species differences in exposure and susceptibility must be taken into account. Howev- er, one of the principal mechanisms of toxicity inhibition by DDE of ATPase, which is responsible for membrane transport of calcium in the shell gland is known only in birds, so the information cannot be used in risk assessment for other groups of animal species. Other risk characterizations for wild animals exposed to environmental contaminants have generally been based on less-detailed information, especial- ly on dose-response relationships. However, risks have been characterized for a number of species, including seals (Anas and Wilson, 1970a,b), bats (D.R. Clark et al., 1988), benthic assemblages (Varanasi et al., 1989), and lacustrine faunas (Smies et al., 1971~. The common feature of these risk characteriza- tions is that they have been based on observation of actual adverse effects, which could be extended or generalized to other species or other locations. A different approach to environmental risk characterization is exemplified by the EPA program of establishing ambient water quality criteria (AWQCs) to protect aquatic life (EPA, 1972~. Dose-response data on the effects of each

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 113 contaminant on aquatic species are reviewed, and the AWQC is established on the basis of the lowest chronic-effect concentration measured in or calcu- lated for an aquatic species, sometimes incorporating a safety factor or other modification. Aquatic species are considered at risk of adverse effects if the ambient concentration of a contaminant exceeds the AWQC for a specified period, although the degree of risk or the likely magnitude of adverse effects is not calculated. A similar program is under way to develop quality criteria for contaminated sediments. Those programs are analogous to safety assess- ments for human health, in which Acceptable daily intakes. or Risk reference doses" are calculated in the basis of toxicity data on laboratory mammals. Testing procedures for aquatic species are less standardized than for laborato- ry mammals, interspecies scaling is less well validated for aquatic species than in laboratory animals, and safety factors are generally smaller for aquatic species than for humans. Thus, risks to aquatic animals exposed to concen- trations at or near the AWQC are relatively poorly characterized by this procedure. Risk to Corm of Anir7tal Species trader S - y When animal sentinel systems are used to assess exposure of direct con- sumers of the animals that are monitored, the exposure assessments can be used directly in risk characterizations. For example, risks to human consum- ers associated with food contaminated with pesticide residues have been calcu- lated on the basis of measurements of residue concentrations (Foran et al., 1989) or calculations of potential residue concentrations (NRC, 1987~. Similar assessments have been made for other contaminants, such as PCBs and poly- cyclic aromatic hydrocarbons (Swain, 1988; National Wildlife Federation, 1989~. Risks posed to nonhuman consumers, such as seals (Anas and Wilson, 1970a,b; Anas, 1974a,b) and birds (Bunck et al., 1987) have been characterized in the same way, although the inferred risks have not been put into numerical terms. An extension of this kind of risk characterization is to model the transfer of contaminants in the food chain and thus characterize risks to con- sumers posed by various concentrations of contaminants in soil, water, or forage (e.g., Fries and Paustenbach, 1990~. Travis and Arms (1988) have developed a general model for exposure assessment (and hence risk character- ization) based on generalized transfer coefficients. T. Clark et al. (1988) have developed a general model for toxicokinetics in food webs based on the con- cept of fugacity. These generalized models can be used to derive estimates of human exposure under average or typical conditions. For use in risk char-

OCR for page 103
114 ANIMALS AS SENTINELS acterization, the estimates of exposure must be combined with estimates of the exposure-response relationship; the reliability of the risk characterizations is limited by uncertainties arising from both parts of the assessment (McKone and Ryan, 1989~. Risk to Humus: Al 5~b ~ 5~ The third way in which animal sentinel systems can be used in risk char- acterization is as surrogates for exposed humans. This is the most challenging use of animal sentinel systems, and its evaluation is at the heart of the task assigned to the committee. The exposure-assessment section of this chapter pointed out that animal sentinel systems have been used in three ways to assist in exposure assess- ment: (1) charting temporal and spatial patterns of contamination, (2) moni- toring environments as they are used by humans (usually by using pet or other domestic animals), and (3) measuring bioavailability of contaminants from environmental media. In each case, the derived information on the potential for human exposure can be used in risk characterization. Some examples of such uses of each type of system follow. · Measurements of the bioavailability of TODD from soil, fly ash and other media have been used widely in risk assessments of sites and facilities contaminated with this compound (e.g., Paustenbach et al., 1986~. The main limitation of risk characterizations generated in this way is uncertainty about whether bioavailability is the same in different species. · Data derived from companion animals have been used primarily in exploratory epidemiologic studies (e.g., Schneider, 1972, 1977; Reif and Cohen, 1979; Glickman et al., 1983, 1989; Sonnenschein et al., in press). Although the studies have suggested that companion or other sentinel animals could be useful in screening for hazardous human exposures, the committee is unaware of any cases in which such systems have been used in formal risk characteriza- tions. If they were to be so used, their quantitative reliability would be limited by behavioral and pharmacokinetic differences between the animals and hu- mans. However, they have potential value in identifying relative risks, e.g., in identifying households in which residents are at high risk. · As mentioned earlier, monitoring studies of sentinel animals provide information primarily on the context of human exposure, rather than on e~o- sure itself Nevertheless, they are useful in identifying Hot spots. of contamina- tion, i.e., locations where humans are or could be at high risk. For example, the National Pesticide Monitoring Program has identified specific rivers where

OCR for page 103
ANIMAL SENTINELS IN ~SKASSESSMENT 115 fish are highly contaminated with pesticide and PCB residues (Schmitt et al., 1985~; this can provide the basis for more focused risk characterizations of fish consumers. Some monitoring programs have provided information that de- clines in contaminant concentrations (EPA, 1983; Schmitt et al., 1985; Prouty and Bunck, 1986; Bunck et al., 1987~; this information provides the basis for broad inferences about decreases in risk. Some animal sentinel systems are designed to provide information on exposure, but others provide information directly on effects and so can be used directly to support inferences of risk to humans. The classical and proto- typical example of such a system is the miner's canary. More recently, com- panion or other animal sentinels have been used as biologic monitors to screen human environments for carcinogens (Wang et al., 1984; Glickman et al., 1989; Schuckel, 1990) or lung irritants (Donham and Leininger, 1984~. A specific advantage of animal sentinels for that purpose is that animals usually develop cancer in response to carcinogens more rapidly than do hu- mans. Hence, in principle, identification of an excess cancer rate in sentinel animals could be used to identify risks to humans that have the same environ- ments and to trigger remedial measures, even though the magnitude of the risks to humans could not be estimated quantitatively from the animal data. In practice, however, the potential value of such systems has not been realized. Other than the in situ use of rodents to study the relationship between N- phenyl-2-napthylamine (PBNA) and cancers for the Shanghai rubber industry (Wang et al., 1984), the Committee is not aware of any animal sentinel sys- tems that have been used in this way for "real-time" characterization of human cancer risks. Even in the Quincy Bay study (see Chapter 5), the high preva- lence of liver cancer in fish used for human consumption was not used directly to characterize risks to the consumers; instead, human risks were assessed by the conventional procedure of calculating exposure to identified carcinogens and multiplying by a cancer potency factor inferred from studies in laboratory mammals. Likewise, the fact that Great Lakes fish impair reproduction in domestic mink (Hornshaw et al., 1983) has not been used by any public-health agency as the basis for characterization of human risks. Animal sentinel systems require much more development and validation before they can be used as more than qualitative underpinning for conventional procedures in risk characterization.

OCR for page 103
116 ANIMALS AS SENTINELS USE OFANL\IAL SENTINEL SYSTEMS IN RISK ]¢ANAGEAfENT Each of the ways in which animal sentinel systems have been used to sup- port risk assessment has its counterpart in risk management. Although actual examples of the use of animal sentinel systems in risk management are few, the potential is great. Management of Risks to Al Species UP 50 Several EPA programs are designed to manage risks to wild animal species. They include the designation of AWQCs to protect aquatic life, the prospec- tive designation of sediment quality criteria for the same purpose (EPA, 1972), the consideration of risks to wildlife in decisions to register or cancel pesticides, and the characterization of ecological risks as part of the process of evaluating remedial actions at uncontrolled hazardous-waste sites. Some of those programs are still under development, and only the AWQC program has an established body of criteria for use in risk-management decisions. Although risks to wildlife are considered in many risk-management deci- sions taken by EPA, it is difficult to point to any cases in which effects on animals deliberately used as sentinels were the exclusive or primary basis for selection of a specific action. One exception might be the banning of the pesticide TDE (DDD), for which the primary basis was the observation of adverse effects on aquatic birds (EPA, 1976~. In most other cases that might be cited, management of risks to human health was also involved; generally, protection of human health requires more stringent control of magnitudes of environmental contamination than does protection of wildlife. One example of the protection of monitored species as the primary basis for risk-management action is the adoption of regulations by the U.S. De- partment of the Interior that restrict the use of lead shot in waterfowl-hunting areas. The action was taken after much documentation of lead poisoning in waterfowl (Bellrose, 1959) and, more recently, in the endangered bald eagle (Reichel et al., 1984~. Another important example is the adoption of regula- tions by the U.S. Coast Guard that govern the transport and shipment of petroleum products. A major consideration underlying those regulations was the risk posed to wildlife (including sea birds, marine mammals, and sea turtles) by spilled oil. Many of the species at risk are monitored regularly, although detection of effects of spilled oil is only one of several purposes of such monitoring (e.g., Ainley and Boekelheide, 1990~.

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 117 Mana~r~t of Risks to Consume of Animal Species Drum Study Many regulatory actions have been taken to limit human consumption of contaminated animals that are used as sentinels. Perhaps the most frequent such actions are restrictions on the taking of shellfish~ased on magnitude of contamination with fecal coliform bacteria, metals or other pollutants, or paralytic shellfish toxin. Shellfish are continuously monitored for those con- taminants around most coasts of the United States; together, the monitoring programs probably constitute the largest set of animal sentinel systems in current operation and the most direct use of animal sentinel systems in risk management. Local bans on fishing or advisories to limit fish consumption have been promulgated in a number of places (especially around the Great Lakes and on other inland waters) where fish are contaminated with pesti- cides, PCBs, or mercury. Selection of the actions usually was based on data from fish-monitoring programs and on Action levels. developed by the U.S. Food and Drug Administration (Reed et al., 1987~. The action levels them- selves were selected to limit human risks, but in most cases involved consider- ation of other factors as well. Another class of regulatory actions based on consideration of risks to con- sumers of sentinel animals is the regulation of uses of pesticides based on residues in domestic food animals. Those regulations were most frequent in the l950s and 1960s, when some uses of persistent pesticides (e.g., DDT, aldrin, dieldrin, and heptachlor) were found to give rise to residues in milk, beef, poultry, and meat from other farm animals. Residues were detected in monitoring programs conducted by FDA and the U.S. Department of Agricul- ture (Reed et al., 1987~. Successive actions to restrict pesticide uses in ques- tion were based on tolerances or action levels established to protect human health. Those actions were generally effective, although continued monitoring of the same animal products during the 1970s and 1980s has detected several incidents of contamination resulting from misuses. A{anage~r~t of Risks to Hurrah Hem AnDnal Sender ~ 5~ As noted earlier, animal sentinels have not been used often as surrogates for human risk characterization. The few examples cited were limited to qualitative characterization of human risk. There appear to be no recent examples of the use of animal sentinels as the basis for risk-management decisions. Risk management today is based generally on quantitative char- acterization of risks, even though formal risk-benefit balancing is rarely possi-

OCR for page 103
118 ANIMALS AS SENTINELS ble (Travis et al., 1988~. Animal sentinel systems appear to have promise for risk assessment and risk management, but there seems to be no current equiv- alent of the canary in the mine. A,~ 502~ S - .ens as Motor of Effectivertess of Risk Mana~r~Acd;or~s The utility of animal sentinel systems in the choice of risk-management actions is highly limited, but they are important in the verification of the effectiveness of such actions. Almost all the environmental-monitoring schemes in this chapter have led to documented local or regional reductions in concentrations of regulated contaminants, such as DDE and dieldrin. Those reductions constitute evidence of the effectiveness of regulatory actions. In several recent cases, animal sentinel systems have been custom-designed to accompany remedial actions, to determine their effectiveness. Notable examples of such systems are those designed by the Department of Defense to accompany remedial actions at sites contaminated with hazardous wastes. The examples include the use of fish as in situ monitors of the toxicity of surface waters and discharged groundwaters (van der Schalie et al., 1988; Gardner et al., in press), the use of shellfish as biomonitors of the toxicity of contaminated sediments (Farrington et al., 1983), and the use of starlings to monitor exposure to and toxicity of contaminants in terrestrial environments (Johnston et al., 1988~. Those systems use the animal sentinels both to moni- tor levels of environmental contamination and to detect potential effects. Although the effects are on the sentinel animals themselves, they are used as surrogate markers of potential effects on other species, including humans. The animal sentinel systems are in relatively early stages of development, and their results will need to be evaluated i.e., it will have to be verified that they can detect changes in magnitude of contamination and in biologic effects that are relevant to human health and other risks that are the primary objects of management. If they can be so validated, the systems appear to be promising. 5~4RY The committee's survey has demonstrated that animal sentinel systems are most useful for persistent environmental contaminants (e.g., halogenated organic chemicals and metals), because these contaminants are retained in the animals' tissues at concentrations that can be measured and can serve as inte- grated measures of the animals' exposure. Animal sentinel systems are most

OCR for page 103
ANIMAL SENTINELS IN RISK ASSESSMENT 119 often used to support risk assessment and risk management in two contexts: when the animal species at risk are the species used as sentinels and when they are the consumers of the species used as sentinels. For those purposes, animal sentinel systems usually provide the most direct and most useful mea- sures of exposure and hence the most useful basis for risk characterization and risk management. In principle, animal sentinels could also serve as surrogate markers of human exposure and effects on humans in circumstances where direct mea- surements on humans are difficult or undesirable. Although animal sentinels are sometimes used as surrogates that way, they have not been used often in formal risk-assessment or risk-management activities. Most such uses have been in experimental epidemiologic studies or in qualitative characterizations of exposure and potential risk. The main reason appears to be that the ani- mal species generally differ from humans in important characteristics: their use of the environment, their contact with contaminated media and uptake of contaminants from these media, their metabolic and pharmacokinetic charac- teristics, and their susceptibility to biologic effects of exposure. Because of those differences, animal sentinels cannot be used quantitatively as surrogate monitors of human exposures and human responses, unless they can be cali- brated against measures of human exposure or response. Calibration would require the measurement of human exposure and response in at least one case. If the animal system can be so calibrated, it could be used in other places or at other times to predict human exposures and risks. Application of calibrated surrogate animal systems could be more convenient and cost- effective than repeated direct studies on humans. At present, such applica- tions are only speculative, because no surrogate animal system has been cali- brated adequately. For the immediate future, the most promising use of surrogate animals is to monitor changes in human exposures and consequent risks, e.g., to monitor the effectiveness of remedial measures. Several opportunities appear to be feasible for animal sentinel systems to decrease the uncertainty of particular regulatory decisions. Future decisions regarding the use of animals sentinel systems in risk management should be made after examination of existing sentinel data on particular issues.

OCR for page 103