Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Concepts and Definitions CHARAC1~:RISTICS OF ANNEAL SENTINEL SYSTEMS The biologic effects of suspected toxic substances in animals can be evaluated while the animals remain in their natural habitat, such as a field, farm, body of water, or human home. Such settings offer an opportunity to assess the intensity of exposures, measure the effects of chemical mixtures, and deter- mine results of low-level exposures over a long period. Animals also can be placed deliberately in an area of special interest to permit collection of data for health assessments, determine the extent of contamination, or determine temporal changes in contamination; for example, animals might be placed at a site that had been contaminated to determine the efficacy of remedial ef- forts. Animal sentinel systems can include data collection through epidemio- logic studies, in situ studies, or food monitoring programs. Before the type of program is chosen, several characteristics must be selected, including the species to be used as a data source, the kind of exposure to which the species mill be subjected, the length of exposure, and the way in which effects of exposure will be measured. Specms Various species of animals~omestic, wild, and exoticl- are potentially useful as animal sentinels. Several attributes of an animal contribute to its suitability as a sentinel. 1A domestic species is one that traditionally has been tamed, bred, and kept for service to humans or as pets. A wild species is one that is free-ranging in its native habitat. Exotic species are those not native to an area; they can be domestic or wild. 33
34 ANIA~1LS AS SENTINELS A sentinel should have a measurable response (including accumulation of tissue residues) to the agent or class of agents in question. The animal might be exquisitely sensitive to an agent, or it might be resistant, accumulate the agent to a high concentration in its tissues, or undergo physiologic or behavioral changes in response to the agent (Lower and Kendall, 1990~. In the case of extreme sensitivity, the species serves as a sentinel by rapidly decreasing in numbers after exposure; the presence of carcasses or the lack of animals seen or captured alerts investigators to a problem. In the case of species resistance to change, animals can be captured at intervals and tested for tissue residues and body burdens of the agent or monitored for long-term nonlethal effects (e.g., reproductive, necrologic, and immunologic responses); the animal must have a life span long enough for substantial accumulation of the agent or for adverse reactions to occur. In the case of some chemicals, an animal itself can act as a dosimeter if the relationship between dose and response is known; an example is DNA-adJuct formation after exposure to polycyclic aromatic hydrocarbons. A sentinel should have a territory or home range that overlaps the area to be monitored. If a small and discrete location, such as a hazardous-waste site, is to be monitored, it would not be appropriate to use an animal that ranges over many square kilometers and visits the site only occasionally or an animal that visits several contaminated sites. Animals with small home rang- es such as some birds, rodents, and reptileswould be appropriate. Migrato- ry animals, therefore, are not good sentinels for environmental contaminants with point sources. Nonmigratory animals could be good indicators of point- source pollution in a stream, because they generally are subject only to con- taminants whose source is upstream from their location. A sentinel species should be easily enumerated and captured. For exam- ple, small mammals, such as mice and voles, are easier to capture than large mammals, and their population characteristics and dynamics are easier to assess over a short period. The size of an animal can be important in itself If an animal is large enough, various types of monitoring devices can be at- tached to transmit radio signals to indicate location, allow determination of whether the animal is alive, or permit collection of data on physiologic charac- teristics and exposure. A sentinel species must have sufficient population size and density to permit enumeration. Rare or endangered species might not be good candi- dates for sentinels, because they often are difficult to locate, are under popula- tion stresses that could obscure pollution effects, and are protected by statute from the types of collection and manipulation that might be associated with sentinel studies. That does not preclude carefully defined, nondestructive studies such as analysis of unhatched eggs or shell membranes for contami-
CONCEPTS AND DEFINITIONS 35 nants. But, in general, the population of a sentinel species should be large enough to sustain the harvesting required by a monitoring study without major adverse impact. In some situations, the most desirable species (with respect to sensitivity, longevity, etc.) might not be present in the study area. Deliberate placement of a sentinel species in the area might then be appropriate. In appropriate circumstances, animals might have to be caged or penned and special atten- tion paid to prevent dispersal and to facilitate relocation. Stray domestic animals or other commercial species sometimes can be considered as sentinels. In urban areas, stray dogs and cats are often abun- dant and easy to study, as are rats, mice, and birds, such as pigeons, starlings, sparrows, and gulls. E - sum Sources Sources that can be monitored with sentinel animals include soil, air, plants, water, and human habitats. A sentinel species should have a close association with the source of interest. For example, animals that could be considered as soil monitors include small digging animals, such as earthworms, soil insects, gophers, moles, mice, and voles. The National Contaminants Biomonitoring Program uses starlings to monitor soil contaminants; starlings feed on soil invertebrates and range over wide areas, so they are exposed to contaminants over areas as wide as 10 km. Any above-ground animal can be suitable for monitoring air pollution, especially if it is large or mobile enough to be free of filtering vegetation. It is generally difficult to monitor air for contamination with sentinel animals, because many routes of exposure including respiratory, dermal, or oral e~o- sures must be taken into account. Honey bees are excellent monitors of air pollution (Bromenshenk et al., 1985), and other flying insects might be equally suitable. Little work has been done to examine the potential of birds as moni- tors of air pollution, although birds have a unique respiratory system that consists of a network of air sacs in the body cavity and some long bones that allows rapid, whole-body distribution of airborne pollutants. Many birds have a relatively high respiratory rate, and some have an apneustic (inhalation) period in each inhalation-exhalation cycle that lasts up to 60% of the entire cycle duration (Brackenbury, 1981~. Those characteristics increase the contact time of inhaled chemical with pulmonary tissue and might increase uptake and sensitivity. However, the use of birds for monitoring air pollution would be confounded by uptake of the same pollutant from prey species, unless captive
36 ANIM'4LSAS SENTINELS birds used as sentinels were fed on clean food. Swallows have been used monitor lead contamination (Grue et al., 1984), and bats have been used to monitor organochlorine compounds (D.R. Clark et al., 1985~; in both cases, the predominant exposure probably was ingestion of contaminated insects. Caged rabbits have been used to monitor airborne pesticides (Arthur et al., 1975~. In the determination of plant contamination, herbivorous animals are espe- cially useful as sentinels. If a particular species or type of plant (for example, shrubs or trees) is of interest, an animal for which that plant constitutes a major portion of the diet should be selected. If all plants in a given area are of equal interest, an animal with broad and varied eating habits should be used. For example, deer are primarily browsers and prefer to eat woody plants, whereas sheep are primarily grazers and prefer to eat grasses; general- ists, such as rabbits and goats, eat both. Water contamination is best monitored with wholly aquatic organisms. Populations of fish and other aquatic species sometimes are absent from an otherwise suitable habitat when toxic chemicals are present. Fish living In contaminated environments might develop tumors, most commonly in the liver, and so have been used as indicators of contamination. Some fish exhibit fin rot when subjected to pollutant stress, and an index of the condition has been devised to signal the degree of interest warranted in relation to fin rot (O'Connor et al., 1987~. Fish can respond not only to acutely toxic conditions in the water, but also to the presence of chemical carcinogens. Fish develop neoplasms in response to many known mammalian carcinogens (NCI, 1984; Couch and Harshbarger, 1985~. Studies have been conducted with freshwater and brackish-water species to evaluate their sensitivity to carcinogens. The freshwater Japanese medaka and guppy (king cobra strain), and to a lesser extent the brackish-water sheepshead minnow, proved to be the most suscepti- ble. When exposed for 1-3 months and held for another 3-9 months, those species expressed tumors in many organs and tissues (Cameron, 1988~. In situ bioassays with caged fish have been used effectively for many years to detect the presence of toxic chemicals in lakes and streams. Fish held in tanks have been used for continuous monitoring of the quality of wastewater discharges from industrial plants. Caged-fish toxicity bioassays have included investigations of fish mortality related to field applications of pesticides (Jack- son, 1960), effluent discharges from pulp and paper mills (Ziebell et al., 1970) and chemical-manufacturing plants (Kimerle et al., 1986), and metal releases from mining-waste sites (Davies and Woodling, 1980~. Because caged-fish bioassays have become so well accepted in the monitoring of water pollution, the U.S. Department of the Interior has adopted them as one method to evaluate the effects of hazardous waste on wild fish populations (DOI, 1987~.
CONCEPTS AND DEFINITIONS 37 Bivalves, such as mussels and oysters, accumulate many chemicals to con- centrations much higher than those in the ambient water; bioconcentration factors range up to 104 or even 105 for some chemicals. Bivalves have been used in the Mussel Watch monitoring program sponsored originally by the U.S. Environmental Protection Agency (EPA) (Butler, 1973) and currently by the National Oceanic and Atmospheric Administration (NOAA) (Farrington et al., 1983~. Terrestrial animals that use water as a source of food or as habitat can be good indicators of aquatic pollution. Piscivorous animals are high on the food chain and often are exposed to chemicals that are much more concentrated in the fish they consume than in the water. Detection of residues is easier in those animals because of the concentration of chemicals in their food and might also lead to discovery of adverse effects earlier than they would be seen in organisms that are lower on the food chain. Ospreys, gulls, otters, seals, and various reptiles and amphibians are some of the animals that can be used for this purpose, as can many others, such as waterfowl and moose that eat marsh vegetation, such as canary grass and duckweed, and the invertebrates isopods, mayflies, and snails, which accumulate pollutant chemicals. Contamination in human homes can be monitored with domestic animals, such as cats and dogs. Other domestic species such as rabbits, gerbils, ham- sters, and caged birdscould also be used, although the committee is unaware of actual examples where studies have been done following intentional expo- sure. Cats and dogs use living spaces in much the same way as their owners, and many share their owners' food. But cats and dogs are more exposed than their owners to soil, house dust, and airborne particles. Cats are exposed differently to airborne contaminants, such as lead, because they lick their coats regularly. Felines in urban zoos have proved to be good indicators of lead contamination. Duradon of Exposwe A monitoring study can last minutes (as in the case of fish) to a few months or even many years, depending on the questions asked and the end points measured. The likely duration will influence the choice of sentinel species. If a study is to be a short-term effort looking for acute toxicity, the sentinel should have a high sensitivity to the chemical of interest and either die quickly on exposure or show some obvious physiologic, pathologic, or behavioral response. A study designed to look at the long-term health of an ecosystem or at the effects of continual exposure to small amounts of a chemi- cal needs to use a species that manifests nonlethal toxic responses or accumu-
38 ANIMALS AS SENTINELS tales the chemical in tissues or other products (e.g., bird eggs or beeswax). Short-term studies can use species with relatively short life spans, but long- term studies do not necessarily need to use long-lived animals; for example, a species with a short life span and high reproductive capacity (e.g., mice and voles) would be suitable for monitoring effects on reproductive and dispersal behavior. In selection of a species, the duration of the monitoring effort must be considered with other characteristics, including the end points being mea- sured. The possibility of sampling biologic fluids or products that can be collected without killing the animals (e.g., blood, hair, or eggs) is important in selecting a species for study. Measures of Effect An animal-sentinel system can be used to monitor concentrations of pollut- ants and their distribution in the environment much as strategically placed mechanical devices can. However, the advantage of using a biologic system is that it can couple measures of exposure Lath a variety of subclinical or clinical effects. Biologic systems therefore can yield a better evaluation of hazard to humans or to the animal population itself than can be obtained with inanimate sampling devices. Once an animal (or a human) has been exposed to a toxic chemical, a series or set of biologic events often can be detected. If an animal is to func- tion as a sentinel, biologic responses must be observed soon after exposure. Therefore, changes in ordinarily measured biologic characteristics, such as the hematologic profile and serum chemical values, probably are more generally useful end points than are reproductive characteristics, mutagenesis, teratogen- esis, or neoplasia. Structural changes generally are easier to measure than functional changes, but both can provide important information after exposure. For example, germ cells in female mammals are extremely sensitive to the polycyclic aromatic hydrocarbons that destroy primordial oocytes and decrease the functional life span in the ovary (Dobson and Felton, 1983~. If the ovary of an exposed animal is examined soon after exposure, the degree of oocyte atresia can be assessed; however, it would take much longer to relate the degree of oocyte atresia to a deficit in reproductive capacity. The advantage of an animal sentinel system is that it can be used to detect acute structural changes and compare them with functional sequelae. Such an approach cer- tainly is more sensitive than, for instance, using age at menopause as an indi- cator of reproductive toxicity of smoking or of materials thought to be toxic to ovarian tissue (Mattison, 1985~. Animals can respond to pollutant effects in many ways, with several meas-
CONCEPTS AND DEFINITIONS 39 urable end points. They can be monitored for subcellular changes (e.g., ad- duct formation on DNA and hemoglobin molecules), cellular changes that result in tumorigenesis, physiologic changes, organ-system malfunctions, and the presence of chemical residues in tissues. These indices can be useful for assessing relatively short-term toxic effects or for extrapolation to human health. Population dynamics of fish and other wildlife species can be monitored to obtain measures of effects of environmental pollution. In addition to the information suggested earlier for species selection, it is necessary to have some knowledge of the natural history of a species (e.g., the 10-year cycles of snowshoe hares) and of biologic disease agents that could affect its population dynamics. Population studies of this kind are often prolonged, expensive, and difficult to conduct. Moreover, populations of wild animals are influenced by many natural factors that are difficult to control, as well as by the contami- nants that are under investigation, and regulation of animal populations in- volves complex interactions among the various controlling factors. Although changes in animal populations might be an end point of primary interest, it is usually easier to measure physiologic or behavioral effects in individual ani- mals than to determine their population consequences. Reference Populations Census data on livestock and poultry are collected in the Agriculture Cen- sus (U.S. Department of Commerce, 1988), and census data are available on some species of fish and other wildlife. Numbers of game fish and other wildlife are estimated annually by state conservation agencies and the U.S. Fish and Wildlife Service. The Christmas bird count, breeding-bird census, and winter-bird population study are long-standing wildlife censuses. Their results are available to the public and to researchers in various publications. But the pet-animal population has not been clearly defined. Estimates often are achieved through the marketing surveys of dog and cat food sales, but those data are suspect, in part because it is believed that some animal-food products are consumed by humans and vice versa. To calculate incidence and rates, epidemiologic research and disease sur- veillance require knowledge of the population at risk and of the number of cases of disease. In human populations, those are generally determined through a census or a special survey in a defined geographic area. Effective use of pet animals as sentinels of environmental health hazards requires simi- lar information (although it usually is lacking). Once the population at risk is defined, it can provide the basis for calculating incidence and risk. Accurate
40 ANIMALS AS SENTINELS counts require accurate inventories of household pets. The Tufts Center for Animals and Public Policy recently has undertaken surveys of pet ownership in New England to determine who owns pets, the types of pets owned, and the pet population. A potential way to count household pets nationally would be through inclu- sion of a few questions on animal ownership in the decennial census conduct- ed by the U.S. Bureau of the Census. To be included in the census question- naire, questions must meet several criteria; for example, only questions that are deemed necessary to obtain essential information with demonstrated broad societal relevance are included. The subject matter for the census in the year 2000 will have to be submitted to Congress by April 1997, and specific ques- tions, including new and modified questions, by April 1998. Pet census data would be useful in the establishment of a large national pet population data base, which would represent the population at risk for calcula- tions of disease incidence or prevalence; the data would potentially enable correlations of disease or exposure patterns between pet and owner popula- tions (through retrospective veterinary epidemiology of pets counted by the census) and allow for prospective prediction of human risk. Thus, questions related to pet ownership, specifically dogs and cats, by household, collected in concert with human population data, would provide the opportunity for sys- tematic investigation of incidence, risk, and the relationships to human and pet animal populations. Pet population data compared with veterinary and other sentinel data may identify geographic areas where human populations are at risk from exposure to unknown or suspected hazards. For example, the Toxics Release Inventory (EPA, 1987) is an example of a recently instituted national information collec- tion program that could be used in conjunction with animal sentinel data or coupled with exposure modeling and veterinary epidemiologic studies of pet populations near industrial facilities. OBJECTIVES OF MONITORING ANIMAL SENTINELS The objectives of monitoring animal sentinels include data collection to aid in the estimation of human health risks, identify contamination of the food chain, determine environmental contamination, and identify adverse effects on animals.
CONCEPTS AND DEFINITIONS 41 Hurrah Health Elects Behavioral changes in animals warned our ancestors of environmental hazards: early in history, abnormal behavior in animals undoubtedly alerted people to environmental dangers, such as the presence of predators and ene- mies and the imminence of weather changes. Animals and humans breathe the same air, drink the same water, eat from the same food supply, are ex- posed to the same environmental chemicals, and are subject to many of the same disease organisms and stresses of daily life. No one can know when humans first noticed the ingestion of natural emetics by dogs, the rejection of particular plants and water sources by grazing ungulates, and other forms of animal behavior that protect against environmental hazards. Undoubtedly, mysterious animal deaths or unusual numbers of animal illnesses also raised human sensibilities. By Roman times, people knew that birds were especially sensitive to coal gas and other pollutants in coal mines (Schwabe, 1984b). Caged birds, more sensitive than humans to the invisible, odorless toxic gas carbon monoxide, were taken into coal mines; their death from exposure to it gave miners time to race to the surface. Several major environmental toxicants were discovered because of their effects on domesticated or wild animals (see Table 1-1~. In some cases, those effects enabled investigators to predict adverse health effects in humans. Observational studies and followup research on numerous environmental and feed-contamination problems in domestic animals enabled investigators to predict probable hazards to human health. For example, aflatoxin Be, a po- tent carcinogenic mycoto~nn produced by the genus Asper~llus, was first dis- covered to cause hepatitis X, a severe hepatic degenerative disease, in dogs, cattle, swine, and turkeys fed moldy peanut meal and grains. When hatchery- raised trout developed primary hepatic tumors while being fed grain that was overgrown with A. paves mold (Halver, 1965), the potential hazard to human health associated with foods made from moldy grains was recognized. In many other cases, however, the effects on animals were not appreciated, understood, or made known to public-health officials until humans had been exposed and severely affected. For example, in 1966, a Swedish researcher reported finding PCBs in fish and an eagle taken from the Baltic Sea area (Jensen, 1966~. The structural resemblance of PCBs to persistent organo- chlorine insecticides raised questions as to their hazard to humans, but the possibility of human toxicity was not fully appreciated. In 1968, a strange Edema disease" epidemic in chickens occurred in western Japan. The disease stemmed from feed to which rice oil that was contaminated with a commercial blend of PCBs had been accidentally added. The PCB contamination was traced to leaking coils of a heat-transfer system that was used to deodorize the
42 ANIA~1LS AS SENTINELS rice oil. Six months after the cause of the chicken disease was discovered, an epidemic of a peculiar chloracne-like condition occurred in humans in western Japan. The chronic, debilitating disease, known as ~yusho,. was traced to PCBs and polychlorinated dibenzofurans in rice oil from the same source as that which had contaminated the chickens (Kuratsune et al., 1972~. Although the contamination of the human food and the chicken feed occurred at the same time, the chickens displayed toxic signs within a few weeks, whereas the hu- man disease took several months to develop. Con~n~nafiior' of He Food Cam Many monitoring programs and observational epidemiologic studies of livestock, poultry, fish, and other wildlife have identified potential contami- nants of the food chain. Prominent examples are the monitoring of milk, eggs, and red meat for residues of natural contaminants of crops and various chemical contaminants (e.g., drugs, feed additives, pesticides, and agricultural and industrial chemicals). Agencies and programs involved in this type of monitoring include the Food Safety Inspection Service (FSIS) of the U.S. Department of Agriculture (USDA), which inspects poultry and livestock before and after slaughter at establishments whose marketing area are in more than one state; departments of health in various states, which inspect milk that is sold commercially; animal-disease diagnostic laboratories, which have been established in almost every state and assist practicing veterinarians and livestock and poultry producers in diagnosing diseases; the Food Animal Residue Avoidance Database (FARAD), which is maintained by the USDA Cooperative Extension Service in cooperation with several colleges of veteri- nary medicine throughout the United States; and the U.S. Food and Drug Administration (FDA) Chemical Contaminants Monitoring Program, which monitors foods for pesticide and other chemical residues. In numerous instances, fish and other wildlife have been found to have accumulated environmental pollutants to the point where their tissues are hazardous to predators of these species and to humans who consume them (see Chapter 5~. Such exposures have often occurred where agricultural or industrial wastes have been discharged into lakes, rivers, or the atmosphere (Bergman et al., 1985~. ~ _ ~ _ _ _ ,, ~ Er~v~ronmentol Cor~tamir~n Domestic animals~ompanion animals and livestock- have long served as
CONCEPTS AND DEFINITIONS 43 sentinels of environmental contamination. Until recently, the role of sentinel has been limited to human observation of animals under husbandry conditions. The use of livestock for monitoring lead- and fluoride-containing effluents near lead mines, aluminum factories, steel mills, fertilizer plants, and smelters has been routine since the early 1950s (Shupe and Alther, 1966; NRC, 1974; Osweiler et al., 1985a). Livestock have also been used to monitor the environ- ments of mining and processing operations that emit arsenic, molybdenum, cadmium, copper, and other elements (Lloyd et al., 1976; Osweiler et al., 1985b), and livestock and poultry have flagged environmental contamination with numerous organic industrial and agricultural compounds, including halo- genated hydrocarbons, polybrominated biphenyls, PCBs, hexachlorobenzene, dibenzodioxins, and organochIorine pesticides (Mercer et al., 1976; Osweiler et al., 1985b). More recently, wildlife have come into use as monitors of contaminant exposures in many different environments. Starlings, mallards, and various fish species, for example, have been used since 1965 as indicators of pesticide contamination patterns across the United States. The National Contaminant Biomonitoring Program (formerly called the National Pesticides Monitoring Program) of the U.S. Fish and Wildlife Service uses free-ranging wildlife to detect trends and magnitudes of contamination with some persistent pesticides and heavy metals. The wildlife are chosen on the basis of their wide distribu- tion, abundance, ease of collection, exposure to the chemicals of interest in specific environmental settings, and tendency to accumulate the chemicals in their tissues. In addition to revealing trends in contaminant concentrations, the data collected have been used by EPA in identifying exposures to some hazardous substances, and hence in regulating the release of some of these substances into the environment. Adverse Elects on Animal Most uses of domestic animals to monitor environmental pollutants have been unplanned byproducts of veterinary services directed at alleviating health problems in the animals involved, rather than organized monitoring programs. Many domestic animals and wildlife are routinely presented to veterinary clinics and diagnostic facilities for clinical examination or necropsy. Few programs for using healthy domestic animals as biologic monitors have been proposed (Schwabe et al., 1971; Buck, 1979~. One example of widespread surveillance of a domestic species is Market Cattle Identification (MCI), a cooperative state-federal program established in 1959 primarily to facilitate eradication of brucellosis (Schwabe, 1984b). The
4' ANIMflLS AS SENTINELS program collects blood samples when cattle that have identifying tags that can be traced to a farm or ranch of origin are slaughtered. Another major sur- veillance program is the National Animal Health Monitoring System (NAHMS), a population-based, cooperative state-federal program designed to obtain information on the occurrence of domestic-animal diseases and conditions and their associated costs. It is now well established that environmental pollutants have had substantial effects on fish and other wildlife populations. Few historical examples are as well known as the rise and fall of the use of persistent organochlorine pesti- cides and industrial chemicals (e.g., DDT and PCBs) its adverse effects on wildlife populations. By the time Rachel Carson's Silent Spring was published in 1962, a large body of information already existed showing detrimental effects of organochlorine pesticides and industrial chemicals (e.g., on wildlife populations). At that time, most documented effects were related to acute poisonings that resulted in large-scale dicoffs of fish, birds, and mammals (Rudd and Genelly, 1956; Carson, 1962; Turtle et al., 1963~. In the flurry of public attention and additional research in the aftermath of Silent Spring, scientists were able to demonstrate further that persistent organochlorine compounds, even when used judiciously, had the potential to cause bird popu- lations to decline and even vanish through the chemical induction of reproduc- tive dysfunctions. In North America, DDT and its metabolite DDE contribut- ed to the endangerment and regional extinction of some species (e.g., bald ~ · r ~ ~ ~ 1~ ~ _ ~ ~ _1 ~__1~ eagle, peregrine falcon, ana grown pelican) ana severe regional ut;clmes others (e.g., osprey, Cooper's hawk, and various other fish-eating birds). Thus, the bulk of the evidence that initially supported the need for a ban on DDT was related to effects on wildlife, and only later were potential hazards to human health identified; both were cited as primary bases for the cancellation of DDT uses in the United States in 1972 (Federal Register, June 14, 1972~. Over the past quarter-century, wildlife toxicology has played a major role in highlighting and reversing the general deterioration of natural environments caused by chemical pollutants (Kendall, 1988; Hoffman et al., 1990~. One legacy of the DDT era is that negative effects on wildlife will continue to be an important aspect of the environmental risk-assessment process. Demon- strations of adverse effects on wildlife populations are now sufficient grounds for restricting or banning the use of a toxic substance, regardless of human health considerations. However, it took some 25 years for the public to react to the threats that persistent organochlorine compounds posed to wildlife, despite ample evidence of food-chain-related phenomena (bioconcentration and bioaccumulation) and severe disruption of the dynamics of wildlife popu- lations (through effects on reproduction and survival). The ultimate evidence that organochlorine pesticides were responsible for the environmental catas-
CONCEPTS AND DEFINITIONS 45 trophes was the spontaneous recovery of many affected wildlife populations in the years after the curtailment of widespread use of the chemicals (Ander- son et al., 1975; Grier, 1982; Cacle et al., 1988~. In the same manner, in situ studies might now be user! to determine the efficacy of cleanup regulations for hazardous-waste sites. ANIMAL SEN71NEL SYSTEMS IN OBSERVATIONAL EPIDEMIOLOGIC STUDIES Epidemiology2 may be defined as the study of the patterns of disease that exist under field conditions and of the specific determinants of health and disease in populations (Martin et al., 1987~. Animal sentinel studies can be designed as descriptive epidemiologic studies or as analytic epidemiologic studies. Descriptive epidemiologic studies are conducted to estimate the frequencies and patterns of diseases in a population, so that unusual increases in frequency (e.g., epidemics) can be identified and hypotheses regarding possible underlying causes or risk factors or determinants can be generated. Analytic epidemiologic studies analyze or test causal hypotheses in a con- trolled manner and can yield quantitative measures of effects (e.g., relative risks or attributable risks). Descriptive and analytic epidemiologic methods are considered observational, in that they are nonexperimental and do not entail intervention in a natural sequence of events. Observational studies can be contrasted with experimental epidemiologic studies, such as preventive or therapeutic clinical trials, which are tightly controlled and designed to deter- mine whether a change in an independent variable (e.g., treatment) has an effect on the dependent variable (e.g., disease). In situ studies with animal sentinels often are experimental epidemiologic studies, provided that they incorporate appropriate controls. Most other animal sentinel studies are observational in nature. Epidemiologic studies in animals have been important in recognizing the causes of diseasesusually infectious diseases. Epidemiology has more recent- ly addressed the health effects of toxic chemicals in the environment from production facilities, accidental spills, and toxic-waste sites (Anderson, 1985~. When public-health epidemiologists need information on relationships between diseases and specific chemical exposures, they often begin retrospective case- 2The committee chose to use the term epidemiology rather than epizootiology, because the basic approaches and methodology are the same; it also chose to use the term epidemics rather than epizootics.
46 ANIlt~ALS AS SENTINELS control studies~ases are identified and compared with controls, and an asso- ciation with prior exposure to an etiologic agent is sought. If the time be- tween a suspected exposure and a case of disease has been long, it might be difficult to assess exposure status for cases and controls accurately and there- fore difficult to estimate the risks associated with the exposures. In prospec- tive studies, groups of persons or animals with different degrees of exposure are followed for changes in health status. A long latency period or a small effect, however, could make adequate followup impossible. Despite their limitations, human epidemiologic studies are the most reliable basis for esti- mating the risk of toxicity associated with exposure of humans to environmen- tal agents (Woods, 1979~. Descriptive and analytic epidemiologic methods that can be used to moni- tor animal populations for environmental health hazards are discussed below. Desenp~ve Epidemiology Descriptive epidemiology is used to characterize the distribution of an event (exposure) or disease in a population by subject, place, and time. The goal is to identify nonrandom variations in distribution; from these variations, hypotheses can be generated regarding etiology and risk and can be tested with more rigorous, controlled epidemiologic study designs. The number of new cases of a disease or new exposures in a population over a specified interval is referred to as incidence; the number of cases or of exposed subjects in a population at a given time is referred to as prevalence. Prevalence and incidence can be measured over time to determine trends and can be used to compare populations in different geographic areas or ecologic settings. The descriptive epidemiologic approach in animals can be particularly useful if exposure information and disease information are collected simulta- neously and if comparable data on humans in the same area are available. An outbreak or epidemic is the occurrence of a group of similar cases in a population or region that exceed normal expectation. Outbreaks can involve few or many animals in a population; they can be confined to a small area or occur in a large geographic region; they can encompass any period, from a few hours to many years. An endemic disease or exposure is one that is constantly present at low levels in a given geographic area; an exposure or disease that remains epidemic over many years might eventually be considered endemic. The objective of an investigation of an outbreak among animals is to determine its causes, sources, and extent. That information is used to take immediate corrective action and to make recommendations aimed at prevent-
CONCEPTS AND DEFINITIONS 47 ing recurrences. The techniques for investigating outbreaks of human disease (Mausner and Kramer, 1985) and animal disease (Kahrs, 1978) have been described. In an outbreak investigation, researchers systematically delineate the characteristics of affected and unaffected animals in the study population and observe, record, and analyze the distribution of cases of the disease with respect to time, place, and a variety of exposure and environmental factors (Kahrs, 1974~. Those characteristics of outbreak investigations ~ animals make such investigations useful in identifying environmental hazards for hu- mans. Ai~c Epi~ntiology Analytic epidemiologic studies usually are initiated when there is sufficient preliminary information from routinely collected data and other sources of data to develop testable hypotheses regarding the etiology or pathogenesis of a disease. Analytic studies tend to be more expensive than descriptive studies; they often entail the collection of new data. If properly designed, they gener- ally allow more definitive conclusions to be reached about disease causation (Kelsey et al., 1986~. For analytic epidemiologic studies in animals to be useful for risk assess- ment in humans, they either should be capable of determining the presence of a known environmental hazard before it produces adverse effects in people (e.g., asbestos as a cause of mesothelioma) or have the power to identify chemicals that cause disease in people but that, owing to methodologic diffi- culties, are less likely to be identified through human epidemiologic studies. For example, older dogs living in a heavily industrialized urban environment were found, in a descriptive study, to have a higher prevalence of nonspecific chronic pulmonary disease based on chest radiographs than were dogs living in a less-industrialized environment (Reif and Cohen, 1970~. The finding led to speculation that the "urban factor" in pulmonary disease was air pollution. In later studies, no significant differences were noted in the urban-rural distri- bution between dogs with cancer of the lungs and bronchi or nose and sinuses, dogs with gastrointestinal neoplasms, and the total hospital population of dogs (Reif and Cohen, 1971~. However, a significant urban association was noted for dogs with tonsillar carcinoma: 73.7% of them resided in heavily industrial- ized areas, compared with 60.8% of the total dog population in hospitals and 47.4% of dogs with gastrointestinal neoplasms. That sequence of studies illustrates that descriptive data can be used to generate hypotheses regarding disease causality that can then be confirmed with analytic epidemiologic stud- ies. Descriptive epidemiologic studies generally are the starting points of or stimuli for analytic epidemiologic investigations.
48 ANIMALS AS SENTINELS ANIMAL SENTINEL SYSTEMS TV EXPERIMENTAL EPIDEMIOLOGY When a known or suspected source of environmental contamination has been identified, animal sentinels can be confined at or near the sourcethat is, in situ. The in situ approach enables the use of animal sentinels with most of the rigors of a laboratory study; laboratory-reared animals may be used, and controls are easily established. All environmental media can be evaluated through the use of in situ studies. An in situ study can involve placement of caged animals or the use of a mobile laboratory that can be taken to a site of interest. Use of animal sentinels in situ can provide information on body burdens or effects that result from small exposures to chemicals in water, air, or soil. They can improve our ability to assess accurately the health risks posed by toxic chemicals, including those at hazardous-waste sites. Data from in situ experiments provide integrated information on exposure to and toxicity of complex mixtures and information for the characterization of multiple end points of environmentally relevant exposures. Thus, in situ experiments are different from laboratory tests, which typically expose animals to high doses of chemicals of interest. Data from in situ testing can be entered into a data base and examined for benefits, weaknesses, correlations with other test meth- ods and data sources, and cost-effectiveness. ADVANTAGES AND ~MITAIIONS OFANIADIL SENTINEL SYSTEMS Multifactorial G`usali~ Disease results from highly complex events involving multiple, heteroge- neous environmental insults occurring over a broad range of individual suscep- tibilities. The impact of these events can be appreciated only by studying population effects under natural conditions over time. Herein lies the strength of epidemiologic methods: If vigorously applied, they can bring us closer to understanding complex interactions and provide a clearer biologic picture. Animals often can be used more effectively than humans as subjects for such investigations. The risk-assessment process can be viewed as a scientific exercise, whose goal is to bring us closer and closer to the truth (Figure 2-1~. Animal sentinel systems can be increasingly important in reducing the uncertainties generated using laboratory-animal experiments. Because of their capacity to integrate natural exposures with biologic effects, they also provide more relevant data
CONCEPTS AND DEFINITIONS 49 +4 +3 a) ._ in ~ of +1 ._ ~ a: .m .' a) CD a) ._ ~ C CO C: in - - - "True" Risk 1~ - - - Risk Assessment In Situ Laboratory _ . . Animal Experiments pidemlologic Human Studies Stud es FIGURE 2-1 Moving toward the true risk in risk assessment. The success of risk assessment is reflected in a decrease in the confidence level. than do fixed-station monitors for many environmental pollutants. In animal epidemiologic studies (as in human studies), the ability to approximate the truth depends mainly on rigorous scientific application of accepted epidemio- logic methods and analytic techniques that control for confounding and reduce bias. Complexity is natural in life and should not necessarily be avoided or changed when trying to assess health effects of natural exposures. For exam- ple, when human epidemiologic studies showed that smoking causes lung cancer, it was important not because a single chemical was implicated, but because it identified a hazard that is potentially avoidable. The same can be said for the relationship between asbestos exposure and mesothelioma. It might not be essential to determine whether crocidolite is more or less carci- nogenic than chrysotile asbestos fibers; but it is important to recognize that asbestos fibers are potentially carcinogenic and to develop a strategy to reduce human exposure to them. Many environmentally caused diseases in humans are recognized to be
50 ANIMALS AS SENTINELS multifactorial. Identification of the contribution of each specific factor might be less important than determination of the effect of reducing exposure to all factors simultaneously, in recognition of their usually occurring together. That was the idea behind the NIH-sponsored Multiple Risk Factor Intervention Trial, whose ultimate goal was to reduce the incidence of and death rates from coronary arterial disease in the United States. The primary goal of an animal sentinel system is to identify harmful chemi- cals or chemical mixtures in the environment before they might otherwise be detected through human epidemiologic studies or toxicologic studies in labora- tory animals. Once identified, exposures to them could be minimized until methods can be devised to determine specific Biologic agents. Animal senti- nel systems themselves are not the answer to the latter problem, but might provide additional valuable time in which to search for the answer. Measure of Exposure and E - zpolati;on to Humps Animals have been used in exposure assessments as surrogates for humans. Where humans are exposed to contaminants in complex environments (e.g., In the home or in the work place), it can be difficult to estimate exposures by the conventional procedure of measuring ambient concentrations of the con- taminants and calculating intakes of the contaminated media. One approach to solving the problem is to use surrogate monitors animals exposed in the same environments; blood or tissues of the animals can be taken for analysis and provide an integrated measure of exposure. If the animals' contact with the contaminated media is sufficiently similar to that of humans, the animals' exposure might provide a reasonable indirect measure of the humans' expo- sure. Most examples of such animal sentinel systems involve the use of do- mestic or companion animals. For example, pet dogs have been used as surrogate monitors of human exposure to asbestos (Glickman et al., 1983) and lead (Thomas et al., 1976; Kucera, 1988~. The principal advantage of using animals as surrogate monitors is that their blood or tissues can be sampled at surgery (e.g., when pet animals are rou- tinely surgically neutered). Animals used as surrogate monitors are not sacri- ficed for study purposes; but they have relatively short lives, and their tissues can be sampled at the time of death (in the case of pets) or slaughter (in the case of food animals). Pet animals occupy the same environments as their owners and are expected to be exposed in broadly similar ways. However, their exposures are not exactly parallel to those of their owners; among other differences, animals have greater contact than humans with soil, house dust,
CONCEPTS AND DEFINITIONS 51 and floor surfaces, and they are more likely to ingest contaminants when cleaning or grooming themselves. Animals also differ from humans in metabolism and pharmacokinetics, so animals and humans will differ in the relationships between exposure and tissue concentrations. However, these differences can be adjusted with model- ing techniques (Andersen, 1987~. Using animal sentinel data as quantitative measures of human exposure is challengin~all the examples cited in this chapter were examples of the use of animal data as qualitative or relative measures of human exposure. It is sometimes possible to sample tissues of the animal species whose exposure is to be assessed. The most important examples are the uses of human tissues or body fluids to monitor human exposure to pesticides, metals, and other environmental contaminants, as in the National Human Adipose Tissue Survey, the National Health and Nutrition Examination Surveys, and assessments of lead, arsenic, and mercury. Tissues of predatory birds and mammals are used to monitor their exposure to organochlorine compounds (see e.g., Wiemeyer et al., 1984, 1988~. For example, after the peregrine falcon was reintroduced into the eastern United States, concentrations of DDE and other organochlorine substances in the eggs of the newly established birds were used to assess the residual contamination of their prey and hence the suitability of the regional environment to support a self-sustaining popula- tion (Burnham et al., 1988~. Most of those examples were studies of geo- graphic patterns of exposure. Other studies have involved analysis of animals that were thought to have suffered lethal poisonings (Coon et al., 1970; Auler- ich et al., 1973; Stone and Okoniewski, 1988) or reproductive impairment (Aulerich et al., 1973) from more localized contamination. In some cases, surrogate markers of exposure are used, such as brain cholinesterase.as a marker of exposure to organophosphate insecticides (Grue et al., 1983) and mixed-function oxidases as markers of exposure to inducers of these enzymes (Rattner et al., 1989~. Animal bioassays, whether conducted in the laboratory or in the field, have several recognized disadvantages and limitations for risk assessment. The most notable disadvantage is that quantitative extrapolation of exposure-relat- ed and dose-related effects to humans is at best uncertain. But animal bio- assays might be more predictive of human experience than are short-term in vitro tests, and the use of multiple animal species provide important compara- tive information.
