Animals as Sentinels of Environmental Health Hazards (1991)

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.

7 Selection and Application of Animal Sen tine! Systems in Risk Assessment This chapter discusses the principles that should guide the selection and appli- cation of animal sentinel systems in risk assessment. It also discusses develop- ments and improvements in methods that will be required if the full potential of animal sentinel systems is to be realized. Chapter 6 described how animal sentinel systems have been used in the various phases of risk assessment and pointed out their advantages and disad- vantages for each phase. Animal sentinel systems have been most useful in general environmental monitoring, in assessing exposure of monitored organ- isms and their consumers, in assessing bioavailability of contaminants, and in assessing ecologic risk; in these applications, data from animal sentinels can often be used quantitatively. Animal sentinel systems have potential value as early warning systems for new hazards, as indicators of potential human eypo- sure to complex mixtures or in complex environments, and as monitors of the effectiveness of remediation measures or other environmental management actions; in these applications, data from animal sentinels are usually used qualitatively, but there is at least a potential for semiquantitative assessments. Animal sentinels have more limited value as surrogates for exposed human populations in hazard identification, dose-response assessment, and risk char- acterization; they have rarely been used for these purposes, and to date such applications have been entirely qualitative. The traditional method for conducting risk assessments for humans exposed to environmental contaminants is to measure (or calculate) the concentrations of the contaminant in various environmental media, calculate human exposure on the basis of rates of contact with the contaminated media, and combine the estimates of exposure with dose-response data derived from animal studies (NRC, 1983~. Each of those steps is subject to error and uncertainty, so that risk estimates are often extremely rough. As discussed in Chapter 6, animal sentinels can be used to complement the traditional approaches; if appropri- ately used, they offer the potential for reducing some of the errors and uncer- tainties inherent in the traditional methods. An investigator planning an environmental assessment should always con- 121

1= ANIMALS AS SENTINELS sider using an animal sentinel system, when it is practicable, as an adjunct to conventional assessment procedures. Animal sentinel data are likely to be especially useful in circumstances where the conventional procedures are most prone to uncertainty. Those circumstances include the following: · Accumulated chemicals—cases in which the agent under investigation is persistent and stored in animal tissues, so a sentinel animal could serve as a sampling and averaging device. · Complex mixtures—cases in which humans are exposed to complex or poorly characterized mixtures of chemicals. · Complex exposures—cases in which humans are exposed to contaminants at concentrations that vary widely in time or space or are exposed via multiple routes, so total exposure is difficult to characterize. · Uncertain biouvailabili~cases in which contaminants are present in media from which their availability for uptake into biologic systems is difficult to predict. · Poorly characterized agents—cases in which humans are exposed to agents of uncertain toxicity, pharmacokinetics, or other characteristics. Whether an appropriate sentinel can be selected in a specific case depends on the circumstances. Factors to consider in determining whether to use an animal sentinel system include the following: · Media. Humans can be exposed to a contaminant via several media, including indoor air; animals can frequent different media and be exposed in different ways. · Scale of averaging. Animal sentinels average their exposure over spatial and temporal scales that are determined by the animals' behavior (home range) and physiology (pharmacokinetic characteristics); it might or might not be feasible to find an animal species that averages over scales appropriate for human risk assessment. · Sensitivity. The animal species selected should be appropriately sensitive to the biologic effects that are being investigated. · Specificity. The species selected should respond reasonably specifically to the contaminant under investigation. · Availability. If a wild-animal species is selected, it should be reasonably abundant and readily trapped for sampling; if no wild species is suitable, in situ sampling should be considered. Consideration of those factors requires communication among specialists in several disciplines: risk assessment, environmental chemistry, toxicology,

SELECTION AND APPLICA TION OF SYSTEMS 123 ecology, and veterinary science. If a suitable sentinel species or system can be found, several tests of practicality should be applied. Is there any previous experience with the same animal species that would indicate that it is likely to provide the information sought? Can the system be put into operation within a reasonable time and with the professional resources that are available? Will the system provide information that will supplement or complement informa- tion provided by traditional methods of assessment? Will the system be cost- effective in comparison with traditional methods? Will the system meet legal requirements and animal-welfare codes? Will it provide information that win be defensible as a basis for regulation or other risk-management activities? As a guide to those who wish to use animal sentinel systems in risk assess- ment, some of the advantages and disadvantages of different approaches are summarized in Table 7-1. SYSTEM DESIGN Once an animal species is found that meets the initial tests of availability, efficacy, and practicality, the design of an appropriate system to use it should be considered. The following design issues are important (Steele, 1975~: · Nature of the problem. The nature and magnitude of the public-health, veterinary, or wildlife problem for which the system is to be designed should be clearly defined. The merits of the animal sentinel program, including its expected sensitivity and specificity and its complementarily to other programs, should be explicitly stated to justify implementing the proposed program. · Objectives. Animal sentinel programs can serve a variety of purposes. They are generally most likely to be useful if chronic low-level exposure is suspected and human data are absent, incomplete, or inconclusive. Animal- monitoring programs can be used in hazard identification, exposure assess- ment, or risk characterization (see Chapter 6~. In situ monitoring programs can enable researchers to assess bioavailability of contaminants, can provide surrogate measures of potential human exposures, or can enable researchers and risk managers to determine the efficacy of remediation measures taken at sites determined to be hazardous. Animal epidemiologic studies can pro- vide answers to specific questions about health effects of chronic low-level exposure or long-term trends of a particular disease. The objectives of an animal sentinel program must be formulated realistically at the outset, because they will be bases of all other major decisions in the design and operation of the program. · Event and unit of observation. The exposure or event that will be under

124 ANlhL4LS AS SENTINELS TABLE 7-1 Advantages and Disadvantages of Animal Sentinel Systems for Risk Assessment Characteris- tic of System Epidemiolog- ic Studies in Wildlife and Fish Epidemiolog- ic Studies in Domestic Animals In Situ Field Studies r Availability Plentiful; Plentiful; Can select Can select of animals diversity of limited diver- desirable desirable species sity of spe- species species cues Existence of baseline data for disease occurence Limited Yes, from existing dis- ease surveil- lance systems plus available medical rec- ords Yes for labo- ratory ani- mals; no for many others Yes Existence of Limited to a Very limited Not. NA baseline data few existing exposure for exposures monitoring data programs Knowledge Unknown Usually un- Well defined Well defined of total pop- known, but ulation at defined in risk some circum- stances Ability to Usually none Partial; can Good; mim- Good control for be done by ics laboratory potentially study design setting confounding or during factors data analysis Exposure route; com- parison with humans Usually dif- ferent Often very . . slm1 ar Similar or different, depending on location of study site Can be ma- nipulated

SELECTION AND APPLICATION OF SYSTEMS 125 Characterize tic of System Epidemiolog- ic Studies in Wildlife and Fish Epidemiolog- ic Studies in Domestic Animals In Situ Field Studies Laboratory Animal Studies Complexity Usually com- Usually cam- UsuaLly com- Can be man- of exposures plex plex plex, but can ipulated, but manipulate might not doses measure hu- man expm sure Latency peri- Usually Usually Usually Usually od shorter than shorter than shorter than shorter than for humans; for humans; for humans; for humans; varies with varies with can select a can select species species species for species for short latency short latency Interspecies extrapolation of results Necessary; physiology and metabo- lism of many species often ill-defined or very different from humans Animal-wel- Usually none fare concerns Necessary; physiology and metabo- lism of many species often well defined and similar to human (e.g., beagle dog) Necessary; can select well-defined species for study, but similarity to humans still questionable Minimal to Great Great none Necessary, can select well-dehmed species for studier, but similarity to humans still questionable N^. = Not applicable. surveillance must be defined precisely. The unit of observation such as the individual, flock, herd, or population—must be specified. · Sources of data. The most effective animal sentinel systems are those in which data are collected by the researchers themselves according to a de- signed protocol. In other cases, specimens or observations might be available from farmers, veterinarians, veterinary schools, diagnostic laboratories, disease registries, food-monitoring programs, hunters, or state and federal agencies responsible for environmental quality, fish and wildlife monitoring, or manage-

126 ANIMALS AS SENTINELS meet. Most of those sources of data are potentially subject to collection bias, sample contamination, error in documentation, incomplete followup, or other types of errors. Any program that is based on voluntary or discretionary cooperation of sources of those kinds should take into account the likelihood of bias and error and should seek to investigate them and minimize or assess their effects. Even when a program is designed and conducted by research scientists, the potential for biases and errors needs to be considered. Sam- pling design, sample collection, analytic methods, validation, quality assurance, and quality control are critical in animal sentinel studies (as in anv other tone of field study). .__, ~ ~r~ · Characterization of the system. It is important in system design to su~m- marize the characteristics of the species or system that is to be used. If a wild-animal species is to be used as a sentinel, its behavior, ecology, physiolo- gy, and population should be characterized, to determine what the system is likely to measure (e.g., to establish spatial and temporal scales of averaging). If important species-specific or site-specific information is missing, the system should be designed to obtain this information. If a domestic-animal species is to be used, characteristics of its population (e.g., breeds, age structure, diet, nutrition, and morbidity patterns) need to be established, either at the outset or as part of the investigation. · Selection of controls. Many types of program (e.g., case-control epidemi- ologic studies) require selection of appropriate control populations. General- ly, control populations will need to be established in uncontaminated (or less contaminated) areas. Selection of appropriate control locations and popula- tions requires careful and often complex, multidimensional comparisons of sample locations and populations, to minimize the potential for confounding and bias. · Characteristics of the program. The program can be either active or passive; the first requires a deliberate effort to collect new information, and the second uses data generated without solicitation or intervention. Data collection can be continuous or intermittent and for long or short periods. A decision must be made whether an entire population will be tested (as for some livestock diseases) or samples will be selected—and if so, how. IMPr CEMENTATION Once a system is designed, its implementation and operation raise addition- al issues: · Professional and institutional issues. Effective operation of an animal

SELECTION AND APPLICATION OF SYSTEMS ID sentinel system requires effective cooperation by professionals in several disci- plines; in this regard, animal sentinel systems are more complex than other systems used in risk assessment. Provision should be made from the outset for regular communication, program integration, and peer review. Many prod grams will require cooperation among institutions in unusual combina- tions—e.g., military installations with veterinary schools and hazardous-waste facilities with wildlife research institutions. Special efforts might be needed to overcome institutional barriers and problems in communication. · Long-term continuity. Some programs are designed to be conducted over a long term (e.g., the National Contaminants Biomonitoring Program (NCBP) and other programs designed to monitor long-term consequences of remedial activities). Such programs require long-term institutional commitment, includ- ing stable funding and provision for storage of data and archived specimens. They also require long-term stability of methods or measures to ensure long- term comparability of data (e.g., intercalibration of results when analytic methods are changed). In practice, long-term continuity has been very diffi- cult to achieve in the programs reviewed for this report. · Mechanisms of recording, coding' and storing data. Forms used for re- cording information should be easy to use. To be useful for analysis and interpretation, chemical and biologic data must be codable with widely accept- ed and standardized nomenclature. Temporal and geographic data should be integrated, where possible, into geographic information systems (GISs). A wide variety of applications can benefit from the use of GIS technology. One of the more common applications of GIS technology relevant to the use of animals as sentinels is in resource management, specifically in defining wildlife habitat. Placement of food, water sources, and terrestrial components in prescribed forms and relationships establishes the habitat for a particular wildlife species. For example, a habitat change in the use of a specific area in Georgia by a colony of wood storks was determined (Hodgson et al., 1988~. A more focused potential application would result in the juxtaposition of data on animals (individuals or populations), distribution of toxicant sources and concentrations, occurrence of adverse health effects, and a multitude of relat- ed geographic and nongeographic data. Thematic maps can be produced that outline, for example, the relationship between high concentrations of a partic- ular pollutant and the incidence of a specific effect in a sentinel animal. The distribution and density of the human population at risk, the infrastructure, health facilities, water supply, etc., could be overlaid on the graphic displays. Data should be stored in data bases that are accessible to users who wish to link them with other data bases or to use data for other purposes. Many of the advantages of GISs for the study of environmental conditions are obvious. The storage of large amounts of data in a logically retrievable

128 ANIMALS AS SENTINELS form that retains their geographic integrity is the central theme underlying the current surge of interest in this rapidly evolving technology. The use of com- puter hardware and software to process widely disparate data interactively almost in real time yields insight into issues and processes that were hard to study because of lack of understanding, time, or trained staff. Decisions can be made on the basis of a hierarchy of deliberately structured algorithms, and the data sets can readily be updated to allow temporal and geographic rela- tionships to be examined more closely than is possible with conventional techniques. Models, statistical algorithms, and graphics can be incorporated into this analytic scheme. The trend toward greater use of GIS technology will continue as computer hardware and software improve, GIS technology develops, and our understanding of the complexity of environmental issues increases. For it to be effectively integrated into programs that use animals as sentinels, a substantial commitment must be made to identify uses of the data generated, to complement consistent systems and data management, and to incorporate existing data bases and networks into the system. · Characteristics of intended report. Reporting of methods, validation, and early results of current animal sentinel systems is important for the further development of such systems. As programs are implemented and begin to yield useful results, the results should be reported regularly to interested parties to ensure the maintenance of individual and institutional commitments to long-term programs. Important considerations are how often and to whom reports will be distributed. Typically, enthusiasm for monitoring diminishes as the interval between reports increases. Some reports should contain inter- pretative summaries of key findings; some need not. · Quality assurance. Periodic assessment of a monitoring program is important to determine whether it is achieving its stated objectives and at what cost. A method for program evaluation and the frequency of evaluation should be established when a program is instituted. VALIDATION Many of the animal sentinel systems discussed in this report are exploratory or experimental, in the sense that each type of system has been used only once or a few times. Before any of these systems can be used on a wide scale as an element in exposure assessment, hazard assessment, or risk characteriza- tion (see Chapter 6), it will require an extensive process of validation. In this context, validation of a system includes the following elements: · Charactenzation of the systen'. What are the characteristics of the spe-

SELECTIONAND APPLICATION OF SYSTEMS 129 cies selected for the system? What are the characteristics of the end point selected for measurement? How variable is the end point, and how do the variations depend on such factors as age, sex, strain, environmental conditions, and other contaminants? · Replicabili~. Does the system yield replicable results? Are the results similar in replicate groups within a study? Between studies? Between labora- tories? Are the results stable over time? · Sensitivity. Is the system responsive to the contaminant under investiga- tion? Does it respond at environmentally relevant concentrations? What is the form of the dose-response relationship? · Specipci~. Does the system respond only to one or a few agents, or does it respond similarly to a variety of contaminants or environmental stress- es? · Predictive value. Are the results useful in predicting human exposure or human effects? Can the results be correlated with direct observations in humans? If a study yields information on the sentinels' exposure, are the sentinels exposed in ways similar to those of humans? If not, are the differ- ences predictable? Is the bioavailability of contaminants from environmental media similar in the animals and humans? If the study yields information on biologic effects, is there reason to expect that humans will respond in ways similar to those of the sentinels? Except for some systems that have been designed to monitor exposure (e.g., the NCBP), no animal sentinel systems have been fully validated, in the sense that all five elements has been considered. If the potential value of any other animal sentinel system is to be realized, it must be subjected to a thorough program of validation. The lack of a systematic program of validation is probably the most important obstacle to the wider use of animal sentinel systems in risk assessment and risk management. PROGRAlLf INTEGRATION Many existing programs have been designed for specific purposes, and the resulting data have been used sparingly. In some cases, different programs collect data on the same contaminants and in the same areas but are poorly integrated. For example, programs in several federal agencies measure or- ganochlorines in fish. If those programs could be better integrated, each could tap a larger data base and could become more cost-effective. State departments of natural resources, federal laboratories, and universities already field questions from the public bringing in material for inspection. State

130 ANIMALS AS SENTINELS departments and law enforcement agencies obtain a wide range of material year-round from road kills, poaching, and similar events. In addition, tag- return programs in fisheries have been well supported. A second way in which program integration could yield more efficient use of resources is the use of specimens for multiple purposes. Many animal specimens are collected for single purposes and discarded after single analy- ses. For example, the National Animal Health Monitoring System and Market Cattle Identification programs collect and analyze tissue and blood, respective- ly, from market cattle. The programs handle large numbers of documented specimens that could be useful for other purposes; analysis of subsets of the specimens for organochlorines or other contaminants could provide a new and efficient way to assess human exposure to these contaminants. Another way to extend the value of existing programs is by archiving or banking specimen material from monitoring programs for analysis when new contaminants are discovered or new analytic methods developed. A third desirable form of program integration is the integration of data from animal sentinel programs with data from traditional environmental sam- pling. Animal sentinel data could become more valuable if they could be correlated with environmental data, such as measurements of ambient concen- trations of the same contaminants. Such correlations could improve not only the utility of each type of data, but also the basis for modeling of environmen- tal transport and exposure assessment. The committee is aware of the technical and institutional obstacles to program integration of the types mentioned here. However, much of the information now collected in animal sentinel programs is underused. Even modest efforts to extend and integrate existing programs could lead to a sum stantial improvement in applications.