Risk Assessment

6. Risk Assessment Methodologies for Biotechnology Impact Assessment

JAMES W. GILLETT

No clear basis presently exists for assessing the potential impacts of the application of biotechnology products in the environment. However, there are several possible approaches, employing presently available technology and technology which might reason- ably be developed. This chapter therefore focuses on methods of risk assessment and those portions of current methodologies that might be applicable to biotechnology.

Registration of pest-controlling organisms under the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) provides some general features and data requirements equivalent to the comprehensive assessment usually afforded a new chemical pesticide and pertinent to biotechnology products (EPA 1984). About a dozen such "pesticides" have been registered on an ad hoc basis, under rules relieving the registrant of responsibility for massive data development (see chapter 3). In spite of parallels to issues raised else- where in this report, such procedures for pesticide registration have not been particularly instructive. Al- though the release of pest-controlling organisms is deliberate and intended to affect populations in natural ecosystems, the registered organisms used to date already exist in nature. Monsanto Chemical Company recently announced (press conference, 10 December 1984) plans for submission of extensive data in support of a petition to EPA for an insecticidal soil organism generated by genetic engineering of Bacillus thuringensis x Pseudomonas fluorescens, constituting the initial experience in this area (Sun 1985).

Current toxic substance regulation also provides little insight--though considerable concern--about the potential for predicting adverse impacts if the Toxic Substances Control Act (TSCA) were to be used as the major regulatory mechanism for controlling deliberate releases of biotechnology products. The vast majority of premanufacturing notifications (PMNs) on new chemical substances, submitted under section 5 of TSCA, lack any substantive data about either chemistry (providing details for exposure assessment) or toxicity (for effects assessment) (Auer 1983). However, as described in chapter 3, it is possible to acquire the requisite data by taking a section 5(e) action. No genetically engineered organisms are presently considered as "existing chemicals in commerce" (TSCA section 4), so there is no particular data base against which data

on novel organisms might be compared, further un- derscoring the difficulty of the problem being addressed. What data might EPA's Office of Toxic Sub- stances (OTS) need in order to assess a new genetically engineered product and prevent untoward exposure and adverse effects? A wide range of data suggested as pertinent has been reviewed by expert panels and is currently awaiting public comment (OSTP 1984). On the basis of material presented in chapters 4 and 5 of this issue, it is clear that new methods and improved basic knowledge in several fields will be required to provide answers to these questions.

There is an apparent dilemma regarding the development of quantitative risk assessments in the absence of adequate technology to conduct them. Risk assessments, even for conventional chemicals, do not lead to conclusions that are necessarily absolute or ultimate truths. They are conducted within the confines of what is technically possible at the time. Risk assessment methodologies are dynamic processes which must change in accordance with advances in the state of the art in associated technologies. Therefore, it is not in- consistent to expect regulatory agencies to evaluate the potential risks posed by products of genetic engineering while the tools to do so are still under development or improvement. This means that initial assessments which may be accompanied by fairly large areas of uncertainty should be performed on a case-by-case basis. As technologies are developed which improve the quality of risk assessments, these zones of uncertainty should narrow accordingly, as has been demonstrated historically for radiation, pesticides, and so forth, and assessments may eventually take on a more generic character.

The process of developing an effective and efficient assessment methodology involves more than pre- scribing tests and data-reporting requirements. The assessment paradigm to be used, the tests to meet information needs of such a paradigm, the criteria that might be applied to the results of such tests, and the specific manner in which risk management may in- teract with testing and risk assessment all are important and are therefore considered here.

T h e Risk A s s e s s m e n t P a r a d i g m

Risk assessment of environmental impacts of new chemicals regulated under TSCA has been divided into two major components: (a) hazard assessment, in which the potential adverse impacts of the substance

Environmental Management Vol. 10, No. 4, pp. 515-532 �9 1986 Springer-Verfag New York Inc.

516 J.W. Giliett

product on organisms and processes are first identified, then quantified in relation to experimental exposure; and (b) exposure assessment, in which the distribution of the substance/product is described in relation to the activities of the affected species. For a new product lacking data, hazard assessment often involves comparison of known activities of related or similar structures to those of the candidate. Exposure assessment emphasizes the fate of the chemical in the environment and the cumulative amounts which might reach people or other organisms.

These assessments must them be combined with other information to characterize the risk (NAS 1983) on at least a judgmental basis. When data on sources, rates, and so on, are adequate, a more quantitative analysis may be performed through a series of realistic scenarios anticipating situations in the environment. Risk assessments are therefore likely to differ so suffi- dently f rom each other that they must be evaluated on a case-by-case basis.

Risk management seeks to reduce the risk to "reasonable" levels through controls on manufacture, use, and disposal, and through mitigation, monitoring, and other activities. This risk assessment process can be taken a step further, to compare risks and costs to benefits and other values to determine if the risk is or is not "unreasonable." All of these aspects of risk assessment and risk management, as illustrated for a genetically engineered organism in Figure 1, are interactive and adaptive, evolving with technical experience and shifts in societal values. Whatever paradigm might be applied to biotechnology products deliberately released to the environment, it likely will contain most of these elements. However, the interaction of responsi- bilities of several federal agencies operating under different legislative mandates and executive orders may require additional considerations.

An alternative model of regulation and assessment might include some absolute notion of safety, such as the Delaney Clause of the Federal Food, Drug, and Cosmetic Act (FFDCA), which prohibits the addition of any carcinogen to feed or food. Quarantines and outright prohibitions are commonly used regulatory tools for exogenous and exotic species, whether pathogenic or not. Essentially, such models ignore any benefits and permit only "zero risk" to be the "reasonable" level of risk. Testing is relatively simple and may be limited to only the single prohibited property of the organism/agent.

FIFRA requires consideration of the "unreason- ableness" of risk, and establishes "safety" by defining recommended use patterns, application rates, and methods of use of the agent based on hazard and ex-

. . . . . . . . . . . . . . . . . . . . -4 TEST METHODS I - t

SURVIVAL. I)SAGJ EFFECTS PATHOGENICITY ~ROWTH, AND SITES AND PROCESS

....... \ 7 - " , , / I E x P o S u R E - - - - H .Z .no IDI~NTIFICATION ~ S S E S S M E N T I [ . . . . . . . u ~ . . . . . . . . [

c . . . . . . . . . . . . "K---- - ~ l ~HAZARO ASSESSMENT t

IRISK ASSESSMENT I

1 CO.T.OL TEc..oLoqv--.IRISK MANAGEMENTI '~coET/~HEP'T A.ACVS,E

m~GULATIOM M O H I T O n l N G E M E R G E N C Y R E S P O N S E

Figure 1. Risk assessment and risk management of genetically engineered organisms. In this model of assessment, pat- terened after that for toxic substances, hazard assessment is combined with exposure assessment in realistic scenarios to yield a risk assessment, which is employed in risk management. These latter activities feed back to alter methods and criteria for decisions as experience is gained in the overall assessment process.

posure assessments. To accomplish these and develop risk management, the EPA under FIFRA and its amendments asks for very detailed data to assist in making such judgments, periodically reviews the registration, and collects monitoring data over time to determine the effectiveness of risk management with respect to both the assessment system and the candidate pesticide. Risk is rarely held to zero levels, but many candidates may be rejected because the net outcome (benefits minus risks) is negative.

Other assessment procedures consider either exposure or hazard (but usually not both), which results in establishment of standards for either specific responses to an agent or levels of that agent in specific environments. The Clean Water Act and the Clean Air Act thus largely concentrate on control technologies and waste treatment methods (retrospective and ameliorative approaches) for particular materials with demonstrated hazard. Little predictive testing is employed for exposure or hazard assessments per se, once criteria are developed for the qualifying control technologies. Concern then focuses on whether a given technology is "feasible," "practicable," or the "best available" to meet the criteria within economic limits.

By contrast, regulations formulated on a prospec- tive bas is - - for example, developed under TSCA, FIFRA, FFDCA, RCRA (Resource Conservation and Recovery Act), and the Ocean Dumping Act - -combine experience and test data to ascertain risk before a

Biotechnology Impact Assessment 517

candidate material is released to the environment. These typically consider not only the nominal material and its composition, but also impurities, congeners, metabolites (that is, biological or environmental products derived from or associated with the candidate substance), and even supposedly inert ingredients associated with use of the candidate material.

All assessments are based on incomplete data; we never know enough to prevent all problems or foresee every difficulty. Risk assessment and risk management form an iterative process. Management may have to operate at higher levels of risk than desirable, because information needs often remain after available tests are performed. Large areas of uncertainty are nar- rowed by experience and attention to monitoring of allowed uses. Often, however, an environmental crisis seems to be required before adequate attention is given to the assessment process. For biotechnology products, such a crisis has not yet occurred and may never develop if the risk assessment and management processes are well formulated. We thus may not have watersheds of discovery as was the case of raptor eggshell thinning with DDT or the birth defects caused by thalidomide.

For the most part, traditional toxicity testing is based on the substance as an entity, with structural identity and associated physicochemical characteristics and activities in biological systems. A certain predictability of the toxicological and chemical behavior of that structure in different media and organisms is therefore expected and found. A biotechnology product is both an entity and an idea, in the Platonic sense of a concept embodying knowledge beyond that attained in reality. The potential threat may be posed not only by expression, in one specific context, of the biological information encoded into genetic material, but also by the possibilities of the information being multiplied and/or transferred to another context en- tirely, where the outcome is unforeseen. The inability to predict the outcomes of connotative (phenotypic) expression of the capabilities encoded in the entity constitute a major uncertainty in both exposure and hazard assessments of these products.

Criteria Development

In assessment methodolgy, decisions are based on criteria accepted by the regulators, the regulated community, and those in the broader society who may be affected. The ultimate criteria are regulatory endpoints (Clements 1984) that define unacceptability of outcomes in the environment. These might be human fatalities, economic loss in a commercial fishery, aes-

SOURCE (INPUT)

~ MOTILITY SOR PTIOH/ADHESION

TO PARTICLES MEDIA MOVEMENT

�9 BIOLOGICAL VECTORS

;TORS } lANCING =OSURE ~CEPTABLE CONDITIONS DAPTATION )MPETITIVEHESS JOIDANCE ZSISTANCE 3MPBNSATION IDUCTION

REVITALIZATION S I T E OF I M P A C T

ORGANISM, COMMUNITY, OR ECOSYSTEM PROCESS

Figure 2. Critical pathway for exposure to a genetically engineered microorganism. Factors mitigating or enhancing exposure (that is, relative magnitude of the population of candidate organisms, indicated by the relative width of arrows) affect the potential for adverse effects of an organism released from a source and following a critical pathway (indicated by solid arrow) of transport and dispersal in the environment. Not shown is the potential interaction of the genetic material of the released organism with species already in the environment, potentially acting as a revitalizing mechanism.

thetic loss of a recreational resource, energy costs in restoring a damaged landscape, or the loss of an endangered species or its habitat.

To the extent that these endpoints are connected to the effects of a chemical or other environmental stress, criteria can be developed concerning adverse impact and the safety margins within which management decisions can be made. Discovery of an adverse effect in the laboratory or field does not necessarily result in development of a regulatory endpoint. However, every regulatory endpoint requires development of appropriate tests. Because results of these tests may provide limited predictive power, evaluative criteria for responses are heavily emphasized in environmental testing and monitoring.

The connection between a test result and an environmental impact depends on at least three important aspects for predictions: (a) the dose-response relationship within the test, (b) statistical tools for evalu- ating responses, and (c) the identification of a critical path from a source to a target showing the effect. A critical path is the description of the steps in the fate of an agent yielding the highest meaningful exposure of the target (Figure 2).

518 J, W. Gillett

Consequently, criteria for regulation of biotechnology products are needed in the following areas:

1) Regulatory endpoints and definitions of unreasonable and unacceptable risk

2) Exposure scenarios with critical pathways 3) Temporal and spatial distributions of the released

organism with respect to the affected systems, processes, or organisms

4) Exposure-response relationships for adverse effects

5) Quality control measures to assure the validity and applicability of data used in the assessment

6) Statistical tools for assessing testing and monitoring data

7) Evaluation of whether immediate control, further testing, or waiver of testing is most appropriate before environmental use

Each criterion may require substantial modification from criteria developed for conventional chemicals, if those criteria are to be adapted to regulation of genetically engineered organisms.

Evaluative Precepts

Techniques in biotechnology assessment are being explored at a level comparable to the origins of modern environmental chemistry in the 1950s and 1960s. Means of identification of the specific genetic makeup of candidate organisms, detection of plasmids in the environment, and correlation of genetic structure with function are in development, but not widely available. By analogy to pesticides, drugs, and other toxic or hazardous materials, a biotechnology product must have an analytical method available for detection of its functional character, in whatever form that functional character might subsequently appear, before it is released into the environment. As a corollary, methods are needed to detect functions and identify contaminants and transformation products associated with release of genetically engineered organisms to the environment.

DDT is frequently cited as an example of oversight in assessment, not simply because the reasons for its subsequent restriction were not considered in early evaluations, but also because the tests used falsely im- plied predictability and safety. Inadequate criteria were applied to these test results. Two analytical methods (gravimetric and colorimetric) were incapable of detecting differences between DDT and some of its metabolites and were insensitive to environmentally active levels in biota. Functional assays (such as acute insect mortality) also failed to find DDE, a derivative of DDT, since it is less acutely toxic than DDT.

When an analytical system (chromatography plus sensitive detectors) was developed and linked to the appropriate measures of effect, much of the problem attributed to DDT was found to be due to the perva- siveness and persistence of DDE. New functional tests for chronic effects in various organisms (for example, eggshell thinning in raptorial birds, mixed-function oxidase induction in mammals) redefined hazard, directing attention to longer, lower-level exposures. These exposures, in turn, were tracked with new and more adequate analytical methods. Examination of chronic effects and of transformation products formed either in the environment or in manufacturing processes is now required for pesticides and drugs and is desirable for other chemicals.

We have become familiar with the problems generated by congeners, impurities and by-products in pesticides and commercial chemicals. Notable attention has been devoted to polychlorinated dioxins in certain herbicides and chlorinated solvents, for example. It is relatively safe and simple to collect Douglas fir tussock moth larvae infected with nuclear polyhedrosis virus, which can be extracted and applied to a few hundred hectares as an experiment. The problem becomes much more complex, however, when one must certify that the manufactured quantities of virus sufficient to combat heavy infestations on 500,000 ha contain only the nominal infective material, devoid of similar contaminants with a different spectrum of pathogenicity. Rohrmann and others (1978) found, for example, that DNA hybridization and restriction endonuclease treatment, presumably the most specific techniques for determining purity, could identify as the nominal virus DNA 97% of that grown in tissue culture, with the re- maining 3% probably being provirus DNA and related precursors. However, a similar baculovirus with a different pattern of pathogenicity for the same strains of Douglas fir tussock moth (Orygia pseudotsugata) had less than 1% homology with the other virus (Schafer and others 1979).

Similar problems arising from scale-up of products of genetically engineered organisms can be anticipated. Hence, at least part of the evaluation must consider the technical use product, not just the nominal material in its purer, laboratory-scale form.

Considering the high degree of uncertainty in assessment methodology for genetically engineered organisms, peer-review not only of research and development of methods, but also of the judgmental aspects characterizing risk should be extended as a necessity. Although this would be complicated by proprietary interests and the rules regarding confidential business information, it would be critical to achieving both professional and


public trust in decisions. Current assessment develop- ments (OSTP 1984) undoubtedly have been strength- ened by the extent of peer review to date and the present general openness of the assessment process.

Regulatory Endpoints Examples of generic, ecologically significant regula-

tory endpoints include:

�9 Direct loss of economic values associated with natural or managed systems

�9 Reduced options in land or water use �9 Degradation of aestb.etic values and recreational

activities of sensitive ecosystems �9 Adverse impacts on endangered species or their

habitats �9 Climatic effects from atmospheric alterations

Note that these categories, being value related, include the perception of loss as well as losses actually measured.

Chapter 5 reviews examples of the numerous opportunities for genetically engineered organisms to affect ecosystem processes and community structure. We have many examples of how accidental or deliberate releases of exotic species have caused serious problems, often worsened by attempts at mitigation. Adverse environmental impacts of the pesticides used in various attempts at gypsy moth control--effects on indigenous birds, fish, and other wildlife--constitute one of the watersheds of awareness in chemical assessment.

Ecological concerns will probably be more important in regulation of environmentally released biotechnology products than in any previous regulatory effort. This importance stems from two major facets of the introduction of a product with new biological information into natural ecosystems, especially when that product is designed to change ecosystem performance. First, the nature of ecological processes is hidden from common observation. Monitoring of acute or massive bird kills from a pesticide application di}'fers from a situation in which trained observers must make particular measurements over a period long enough to ascertain the impacted status, and these must be compared against a "normal" or unim- pacted baseline of data on the process.

The acute effects of pesticides stimulated research that resulted in ecological effects being uncovered. Eventually these ecological effects were taken into ac- count in regulation, so that we can no longer discover in the United States widespread, irreversible ecological impacts due to pesticides. Acute effects, such as patho-

genicity, may or may not be apparent for genetically engineered organisms introduced in the environment. Chronic or lasting effects may stem from pathogenicity to a vital (but unknown) participant in a process about which we know or observe very little. Indeed, even now many are concerned about such undetected effects for pesticides and toxic substances, given that so little direct effort has been devoted to the ecotoxicology of such chemicals (Levin and Kimball 1984).

The second aspect is the interactive nature of species in determining ecological outcomes of anthropogenic actions. Introduction of genetically engineered organisms per se changes the constellation of those interactions, in ways that we cannot predict or sometimes even detect. These interactions are at the heart of ecology of natural ecosystems. Although not every set of interactions will be significant, our inability to predict which will assume key roles in ecosystem development, stability, and function, especially at the microbial level, makes even apparently benign introductions suspect until we have data to the contrary.

Both of these major aspects are embodied in concern over introduction of the ice(-) bacterium to im- part frost tolerance in crops (Foundation on Economic Trends vs. Heckler) by application of a species occupying sites on plant surfaces which otherwise might be occu- pied by bacteria acting as frost nucleation sites. Such applications might also affect pests (weed or insect species), changing the interactions in crop-pes t coevo- lution. Moreover, since the role(s) of the normal geno- type ice (+) may be poorly understood, unobserved processes may also be altered significantly before mitigation can occur.

Hence, establishment of cause-and-effect relationships and direct connections between the biotechnology products and regulatory endpoints are critical, even though deficiencies exist in current understanding of these processes. Neuhold and Ruggiero (1975) listed ecosystem dysfunctions known to have adverse consequences and which therefore can con- tribute to the ecological endpoints categorized above:

1) Reduction of primary productivity (energy assimilation)

2) Reduction of material processing rates (especially carbon compounds) and energy assimilation in secondary production

3) Reduction of rates of, or capacity for, cycling of nutrients (especially N, P, S, and trace elements)

4) Extreme fluctuations in population size or age- class structure

5) Altered species diversity, especially simplification of community and ecosystem structure

520 J.W. Gillett

6) Loss of critical species (that is, a species of pivotal ecological importance, as in any of the above relationships)

On the basis of details in chapter 5, we can judge that ecosystems may be vulnerable to many of these dysfunctions as a result of genetically engineered organisms specifically being released to affect ecological processes. Accidental releases of organisms, of course, may also be a problem. Concerns about ecological responses to pesticides, toxic chemicals, and other anthropogenic inputs have resulted in considerable effort to develop tests for these responses. The problems of extrapolating laboratory test results to the field make direct connections from tests to regulatory endpoints difficult. However, as a regulatory framework for biotechnology products develops, there will be increased attention devoted to validity of lab-to-field and field-to-field extrapolations.

Exposure Scenarios

The connection between regulatory endpoints and use of biotechnology products in the environment requires identification of critical pathways of exposure to the agent (product). Detailed knowledge of releases, transport, transformation, and storage of bioactive forms is used to construct mathematical models of exposure for chemicals. Similarly, detailed toxicologic knowledge of sensitive life stages of the "target" organisms or systems and their behavior in response to the agent is needed to evaluate the connection to particular exposures. The critical pathways for exposure and toxicology combine to yield quantitative risk assessment for selected scenarios of use.

Earlier we noted differences and similarities between conventional chemicals and biotechnology products regarding features important to risk assessment. For genetically engineered organisms intention- ally released in the environment to affect ecosystem processes, exposure scenarios take on a much different and more complex form than those for chemicals. Assessment scenarios must include information o n "

�9 Locale, rate, and form of release �9 Survival of the introduced organism (immediate

and long-term), including determination of survival form(s) under adverse conditions, the minimum propagule size, if such exists, and the range of physiological and nutritional stresses tolerated

�9 Growth and propagation of the organism and its genetic material, including physiological and nutritional requirements and interaction with other organisms

�9 Transfer of genetic information to other organisms

�9 Dispersal of organisms with the engineered genetic information beyond the site of release

�9 Interaction of the characteristics and activities of the organism having the introduced genetic material with components of the natural system

Chapter 4 has considered some of these and related factors in examining the survival, growth, transport, and dispersion of microbial populations. Under- standing these requires many facets of traditional microbiology (systematics, physiology, ecology). Recent teaching and research efforts have largely emphasized applied microbiology, molecular biology, and genetics, but there is a need for direct support of method and data base development for exposure assessment.

Fortunately, biotechnology product development has spawned theoretical and applied research which is yielding data on survival of some genetically engineered species and on manipulative techniques characterizing genetic material. Mechanisms of interorgan- ismic transfer of information have been explored. In- troduction of specific genetic markers to identify released organisms may facilitate study of mechanisms of survival, propagation, dispersal, and information transfer.

These mechanistic studies will be helpful in assessments, but the basic function and interrelationships of organisms involved must be known. Review of survival and growth (chapter 4) shows that there are both con- flicting information and a lack of detailed knowledge in areas vital to creation of useful exposure scenarios. This will pose a severe handicap in exposure assessments until procedures can be agreed upon and applied widely enough to afford a useful data base. How- ever, there appear to be useful techniques in the literature for labeling candidate organisms or their surrogates which, in combination with other analytical techniques, enables the organism to be tracked in the environment. Systematic method development and in- terlaboratory testing should therefore be expected to provide means of obtaining the necessary information, once appropriate research is initiated.

Exposure-Response Relationships Traditional testing of chemicals and other toxic

substances relies heavily on the measured response of defined populations of organisms to controlled exposures of the agent. Subsequently the measured exposure-response relationships thus developed are employed to estimate effects from various exposure scenarios. The most useful expressions of the exposure-response relationship are: (a) the exposure con-


centration affecting hal f of the test population, for example, LDs0 or ECs0; (b) the slope of the log dosage vs probit (of effect) plot, or the 95% confidence limits of the ECs0; and (c) the "no-effect" level, which may be expressed as either greater than the highest level tested without effect or less than the lowest level tested showing a statistically significant effect. The first cate- gorizes the response by placing it within a range (that is, super toxic, moderately toxic, and the like), while the second characterizes the distribution of response within the population. The third estimates (experi- mentally or by extrapolation) the ability to detect an effect and has the greatest implications for assessment.

No such simple relationships appear to be applicable across the board for genetically engineered organisms and anticipated effects. Certainly some effects will be proportional to the numbers or density of added organisms. When an organism is used as a pathogen, as in pest control, the target-species density affects the effectiveness of the applied dose, usually expressed as initial numbers of viable organisms per unit surface area. High densities of target organisms result in rapid spread of the pathogen, whereas low densities require heavier or more widespread treatment to achieve the same result. In other instances, the inital inoculum size may have a threshold for effect, but further increases do not affect the outcome when generation times are short. Finally, the effect may be delayed or may involve other forms of the organism, so that responses are not directly connected with initial conditions. This lack of knowledge of the relationship of exposure to outcome adds greatly to the uncertainty of assessments.

The biological equivalent of chemical concentration is population density, which is difficult to measure for some species. Microbial enumeration for soils and sediments presents special problems, whether for total biomass or individual species. The more heteroge- neous the medium, the greater the difficulty in determining which organisms are actually participating in the process. Distinguishing between live and dead organisms, between active and inactive individuals, and between competent and noncompetent strains becomes critical to quantitative expression of "exposure." Lacking an unequivocal expression of exposure-response, heavier reliance will be placed on statistical in- ference as opposed to deterministic analyses.

Irrespective of the mode of expression of the exposure-response relationship, genetically engineered organisms must be tested regarding functional processes likely to be affected. Chapter 5 details the processes or systems of concern in natural ecosystems, but does not provide specific guidance for testing. This is not an oversight; rather, it reflects the lack of standardized,

useful test protocols for ecotoxicological effects beyond single species responses. Although a variety of multispecies test systems have been investigated (see Hammonds 1981) and are viewed as necessary for chemical testing (Cairns and others 1981, Cairns 1983), agreement on structure, operation, and interpretation of such tests has eluded the scientific and regulatory communities.

It might be argued that, since we are unsure of the thresholds for effect and cannot describe the organ- ismic and community relationships determining processes (nutrient cycling, respiration, and so on), these tests might be required for any organism released. In instances where the application explicitly affects an ecological process, relevant fate and effects of the candidate product may be pursued as a part of product development, so that this information might be readily available. For others, such needs will not be self-evi- dent. In either case, testing for effects in microcosms or in the field (once sufficient assurances of safety can be provided) will be needed.

Test Evaluation

Experience in the enforcement of FFDCA, FIFRA, and TSCA has shown the need for developing the widest possible data base in support of evaluative testing criteria. In tiered testing schemes, such as for chemical testing of toxic substances (EPA 1982 and 1983a), evaluative criteria provide answers to questions concerning the order, extent, and direction of testing. When information supplied is sparse (as for most PMNs), generic data bases from standard tests form the basis for constructing quantitative structure-activity relationships (QSAR) and thus permit appropriate estimates of hazard, exposure, and risk in a timely, if imprecise, mam,er.

Numerous laboratories have had at least two decades of experience in single-species toxicity testing and physicochemical characterizations, creating a data base for chemical QSARs. Laboratory tests for ecosystem processes are much more recent and limited in terms of the numbers and types of chemicals exam- ined; no QSARs for these tests have been developed. The absence of published studies of novel organisms in model ecosystems that deal with ecological process measurement means that no QSARs for biotechnology hazard and exposure assessments, equivalent to those for chemicals, exist now and for the foreseeable fu- ture. Again, this emphasizes the need for case-by-case data evaluation from tests of fate and effects for genetically engineered organisms.

Two sets of criteria for test evaluation appear in need of development in support of biotechnology impact assessments: those for microcosm and test system

522 J. W, Gillett

operations (to establish the quality and validity of the data generated), and those for test results (to evaluate the fate and effects of the product). The former are significant for new systems extracted from untested portions of the environment and will require substantial field comparison for purposes of extrapolation and validation. Microcosm and mesocosm systems have not been studied so extensively as to be included in the regulatory process, but a number of standard protocols with supporting documentation (for example, Perez 1983, Van Voris and others 1984) are now under review for toxic substances regulation by the Office of Toxic Substances in EPA. Evaluative criteria for ecological process tests also need further development for regulatory use, although measurements employed in such tests (for example, CO 2 release or uptake, 02 uptake, nutrient loss, net biomass change, and the like) have been used widely for some time.

Statistical Procedures

One of the needed efforts in biotechnology impact assessment will be amplification and application of statistical techniques in experimental design and interpretation to determine distribution, fate, and activity of organisms under a variety of physical conditions and in diverse ecological structures. With a large number of variables affecting fate and effects of biotechnology products, partial factorial designs and related techniques can be used economically to explore primary factors, but cannot test the significance of interactions. Microcosm studies and other prefield tests may present operational and statistical difficulties regarding extrapolation to the field and from one site to another.

The role of statistical approaches appears to be important in stochastic phenomena including ecosystem processes potentially affected by biotechnology products and processes involved in organism dispersal and propagation. Because these products may be self- perpetuating, questions have been raised concerning the applicable safety margins and statistical confidence intervals to be used. Finally, even if survival of introduced organisms is viewed a priori as of low proba- bility, how are risk assessments to be performed if potentially devastating phenomena are evidenced in tests?

Quality Assurance

In cooperation with other federal and international agencies, the EPA (1983b) has promulgated requirements of Good Laboratory Practice (GLP) regarding data submissions in support of regulatory actions

under FIFRA and some parts of TSCA. Agencies potentially involved in regulation of biotechnology probably will have little practical ability (due to staffing limitations or legal authority) to generate data on each product. The data submitted should meet acceptable professional and regulatory standards, including ac- ceptability for evidentiary use in hearings and trials.

These requirements have not been reviewed specifically by the agencies regarding data involved in biotechnology product assessments. Although some changes in quality assurance (QA) and GLP structure may arise from such review, and later from experience in regulation of genetically engineered organisms, the present GLP structure should suffice. The responsibility for QA rests with the submitter where standards are not specifically indicated. The statistical reliability of quality control (QC) procedures and the surety of separation/containment practices are typical issues requiring supporting documentation. Absence of detailed protocols for field studies places the burden for documentation of the particular study on the submitter. The responsibility for requiring and reviewing such documentation still rests with the agencies.

Test Method Development

The foregoing discussion of criteria clearly demon- strates the need for development of standardized tests for fate and effects of biotechnology products. In some cases, the measure used in a test has not been agreed upon; in other cases, the experimental conditions must be made more explicit. The preceding ar- ticles in this issue show there is ample potential for such tests, but these may require further refinement or may be applied by only a few laboratories in a manner appropriate to biotechnology assessments. Some widely standardized tests (such as pathogenicity) may be uninformative for organisms affecting ecosystem processes.

Thus much must be done to elevate many test protocols to the level of guidelines for use in risk assessments. Deficiencies include a lack of a thorough understanding of the connections between tested phenomena and regulatory endpoints and a lack of experience of investigators in use of techniques, such as specific growth tests or microcosm technology. However, the history of risk assessments of conventional chemicals amply illustrates the interactive nature of test development and regulation as iterations of peer-review standardization, application, interpretation, and modification. It is reasonable to expect that this sequence will apply to biotechnology products as well.


Identity

Genetically engineered organisms have both a functional and structural identity. Much attention in the assessment process will be directed at the former, simply because it would be through phenotypic expression of such function(s) that any effects might be manifested. It is important to keep in mind, however, that the technology employed in creating genetically modified organisms may be used to perform structural determinations, such as nucleotide sequencing and in vivo localization (plasmid, chromosome, and so on). The complete sequencing and genetic mapping of a few simple organisms should not be taken as evidence that this can and should be accomplished for all biotechnology products, but it is feasible to begin developing an encyclopedia of relationships between nucleotide sequences, genetic structure, and phenotypic expression in relation to environmental conditions. It is unlikely that higher plants and animals can be included in such compilations, but microorganisms seem appropriate subjects.

Addition of specific, unique nucleotide sequences has been suggested as one means of tracking released organisms. Similarly, addition/deletion of sequences may be a means of "biological contaimnent." That is, genetically engineered organisms could be manipulated further to include factors limiting growth or reproduction under conditions of use of the product, where, for example, nutrients might be provided lo- cally to a deficient mutant. Carrier or promoter sequences have been suggested as markers for genetically engineered R-DNA, particularly for material from prokaryotes transferred to eukaryotes. Of course, there is some question as to whether such markers and controlling genes might not also be transferred in the environment to other organisms with or without the functional sequences for which the biotechnology product was designed. Many extraneous sequences occur in DNAs in nature, some of which may have had a similar, but natural, introduction into the cell.

This problem of identity therefore poses important research questions for both product development and impact assessment. Monitoring of released organisms and any effects attributable to them will require assurances of identity. Eventually, some predictive ecological tools might be developed, if a careful record of the details of chemical structure and outcomes is accumu- lated.

Fate Tests

Survival of introduced genetically engineered organisms must he measured under conditions of poten-

tial use and dispersion. Methods employed must be effective for both viable and debilitated (but potentially viable) cells. Test conditions and initial inocula must be adequate to determine the circumstances where survival is likely and where it is unlikely. As a minimum, these conditions should include environmental factors such as pH, temperature, salinity, and other media properties. It appears important to know the patterns of survival as well as the end result of a particular set of conditions appropriate to a given application. Methods exist for satisfying all of these information needs (see chapter 4), but contradictions and lack of generalizations suggest the desirability of directing efforts at standardization.

Growth (specifically, increase in numbers of a given species) similarly is affected by numerous factors, of which nutrient availability and ecological interactions (predation, parasitism, and so on) seem preeminent. Of particular interest in assessment would be any set of conditions permitting the introduced organism to displace indigenous populations or become pervasive. Because much less is known about the interrelationships surrounding growth of microbial populations, basic research will he required to support test method development and interpretation. Tests measuring survival and identity will clearly be involved in growth assessment, which must similarly define the sets of conditions wherein growth is likely, possible, or unlikely.

Transport and dispersion of the introduced genetically engineered organism require knowledge of the mechanisms by which it is possible for the species to move (or be moved) in the environment and the environmental conditions enabling the operation of those mechanisms. Information on population density required for growth of survivors at sites other than the release site will be needed to establish a "minimmn e f fective propagule." Current knowledge in this area is derived heavily from plant and animal pathogens (including fungi), all of which may not be particularly applicable to species used in the development of biotechnology products. Therefore, any test methods developed in this area will require substantial basic research to derive conclusions regarding relative exposure hazard or safety. It seems likely, however, that once mechanisms of transport are defined with respect to environmental conditions, knowledge of survival and growth can be used realistically to select more complex test environments (scenarios). Such knowledge must also be applied in design of effective monitoring tools to verify movement in the environment.

Ecosystem Process Tests

Primary productivity, growth, respiration, sec-

524 J.W. Gillett

ondary productivity (growth and reproduction of consumers), and nutrient fluxes are the chief candidates for ecological test method development and application in biotechnology impact assessment. They rely on relatively simple, inexpensive measurements, may be applied in the field and in laboratory model ecosystems, and hence can serve as measures of comparison of microcosms to the field. The advantage of these tests generally is that they integrate function across species, population, and community bound- aries. The disadvantage is that they are not well understood in detail, so that predictive ecological assessment is difficult. Simple tests of function of specific organisms in specific culture have been required and accepted by EPA for pesticide registration, but the results have not been found to be totally pertinent to field experience. Hence, the Office of Toxic Sub- stances (OTS) has not yet developed guidance for effects of toxicants on nitrification, ammonification, sul- fate reduction, and cellulose digestion.

In most microcosm studies of ecological processes, measurements have included net respiration (CO 2 production or 02 uptake of treated systems vs. controls), net photosynthesis or primary productivity (measuring chlorophyll A), or net biomass change over a fixed period (up to several weeks). These net measurements may be adequate for screening potentially adverse novel organisms, but may be too insensitive for defini- tive work. This is because the processes should be evaluated as dynamic, interacting features taking into ac- count the resistance and resilience of the system to the introduction of a new organism.

For example, Van Voris and others (1980) showed how time-series analyses can reveal process complexity, in this case the production/uptake of CO2 in soil core microcosms. The impact, as measured by nutrient leaching, of a cadmium amendment to the soil- plant system was inversely proportional to the number of intense peaks in the time-series analysis. The time- series complexity differed somewhat from gross species diversity, but no measure of microbial species diversity was made. Similar efforts should be directed at other nutrient fluxes (N, P, S, trace minerals) and relative functions [for example, production/respiration (P/R) ratios]. With automation and computer-assisted collection and management of data, such sophisticated studies might be readily adapted to routine use in higher-level (confirmatory) testing of environmentally released agents.

Obviously, support for any ecological testing would benefit by research that would strengthen understanding of underlying mechanisms (to permit simpler tests to have greater predictability) and that would spe-

cifically relate the tests to identifiable regulatory endpoints. Holistic studies, coupling effects and exposure measurements in microcosms, might be then under- taken before field trials and consequently reinforce confidence in initial predictions made on the basis of laboratory studies. With sufficient effort devoted to particular steps in exposure and effects mechanisms, mathematical modeling of these outcomes might be possible.

Microcosm Technology

Various aquatic and terrestrial laboratory microcosms (Figure 3) show considerable promise as means of examining the fate and effects of biotechnology products, even though efforts to date have been based on studies of pesticides, toxic substances, industrial wastes, and other chemical stresses. Ecotoxicological applications of microcosms have been reviewed extensively (Gillett and Witt 1979, Giesy 1980, Hammonds 1981, Cairns and others 1981). Successful hazard identification has occurred at the process level (Van Voris and others 1980 and 1983, Gillett 1983) and in regard to system structure and function (Perez and others 1977, Harte and others 1980). These portend well for application of microcosm technology both to exposure-response studies for hazard assessment and to those of fate and transport for exposure assessment of genetically engineered organisms (Liang and others 1982).

Microcosms have been successfully used to demon- strate ecological processes in the controlled conditions of the laboratory. The most important points regarding test method development are that: (a) experience with a variety of model types and physical ar- rangements provides confidence that practically any environment and accompanying set of conditions may be represented (although costs can be a problem for certain conditions and environments), (b) the systems can be used with environmental safety before release of organisms in the field, and (c) appropriate systems are more cost-effective and accurate than sets of simpler tests performed separately (Van Voris and others 1983).

If, as many feel, environmental safety issues take precedence over feasibility and cost considerations, then application of microcosm technology should be given high priority in both hazard and exposure assessment testing. The problems of containment, whether physical or biological, could be addressed in such studies. Moreover, these systems would provide the variety of microhabitats likely to be experienced in the "real world." Given evidence that results of microcosm studies can be more incisive, accurate, and cost-


Air inlet

Roin inlTt 1

Figure 3. The terrestrial microcosm chamber (TMC). The TMC was developed at the Corvallis Environmental Research Labo- ratory (EPA) for ecotoxicologic assessment of chemical fate and effects. (Left) Schematic diagram of the TMC, a semi-enclosed environment affording the investigator control of inputs and outputs. (Right) Physical simulation of a rye grass field in a TMC. Photo by J. Gillett.

effective, such application should be strongly advo- cated. Some earlier criticisms (Gillett and Witt 1979) are no longer valid, as efforts have reduced labor and skill requirements, minimized operational complexity, and provided comparability to field research.

Nevertheless, some reservations remain. These can be difficult to separate from certain deficiencies in ecological theory and practice. For example, the important questions of scale and the role of boundary conditions in the laboratory also are issues relevant to field testing. Duration of representativeness of the laboratory systems remains an operational and theoretical challenge, especially for aquatic systems. Although Van Voris and others (1983) found better comparability between soil core microcosms and equivalent field plots in the second year than in the first, Harte and others (1981) could maintain chemical and biological equivalence of a freshwater pelagic microcosm to the reservoir from which it was derived for only a few weeks. Pertinent to biotechnology impact assessment, the functional characteristics persisted beyond the point where significant changes in species composition occurred.

Table 1 shows some of the ecological processes quantified in microcosms. The simple excised systems (Perez 1983, Van Voris and others 1984) have been most successful. Mesocosms of field-engineered systems have shown better performance, but are too costly and probably inappropriate from the standpoint

of environmental safety. Microbial processes are readily tracked in even simpler systems (Bond and others 1976, Bourquin and others 1979, Wilson and others 1981, Lighthart and others 1982). Although some higher-level interactions involving larger verte- brates and invertebrates have been reported in microcosm studies, there is much opposition to use of these species. Thus, the model ecosystem most heavily used in pesticide and toxic substance assessments (the con- structed "farm-pond" system of Metcalf and others 1971) has been severely criticized for its ecological ir- relevance, although its use three decades earlier would have saved much grief regarding adverse ecological impacts of chlorinated hydrocarbon pesticides.

Two major categories of need for microcosms have been identified: (a) systems for study of ecological processes, and (b) systems for study of the survival, propagation, and transfer of manipulated genetic information. Whether the present systems in the first group are useful will depend, in part, on the degree to which they are applicable in the second. Fortunately, much of the work in exposure assessment can be performed in vitro or in the simplest microcosms [for example, the soil-litter system of Lighthart and others 1982), the gnotobiotic system of Taub and Crow (1980), or the excised aquatic system of Loeffler (1980)]. Traditional microbiological methods provide sufficient breadth for initial estimates of exposure potential and stability. Subsequently, hypotheses con-

526 J.W. Gillett

Table 1. Ecosystem processes and impacts measured in microcosms.

Process Impact System References

Nutrient leaching Loss of Ca, Terrestrial Van Voris and others NOa- soil core 1980 and 1983

Nutrient cycling Loss of N forms Freshwater Harte and others pelagic (1979)

Predator-prey interaction Increase or Terrestrial Gillett and others decrease microcosm (1983)

Primary production Loss Freshwater Harte and others pelagic (1979)

Loss or gain Terrestrial Van Voris and others soil core (1983)

Microbial respiration Loss Soil-litter Lighthart and others microcosm (1982)

Loss or gain Gnotobiotic Taub and Crow (1979) freshwater

cerning exposure could be confirmed in these simple microcosms. Our knowledge of more specific details of the microhabitats in the environment relative to questions of survival and transfer of genetic information is rather limited. Environments high in microbial activity - -decomposer communities, plant rhizospheres, wet- lands, and aquatic sediments--need more attention regarding such details in design and application of microcosm technology.

Overall Assessment Structure

ReLation to Risk Management

The complete assessment structure, as noted earlier, includes the following features: data development from specific tests and their evaluation with regard to exposure and hazard; characterization of risk from all available information; quantification of risk from realistic exposure scenarios affecting identified targets; cost-benefit analysis of the reasonableness or unrea- sonableness of the risks attendant to uses; and, ulti- mately, management of the risks by means involving all of the above, as well as activities extending into so- cial, economic, and political spheres.

Risk management can be defined as all the actions involved in use of chemicals and other anthropogenic agents generating characterized or quantified risks. Valid risk assessments are the initial step toward effective risk management, but not the only basis for action. In addition to direct controls (restriction on manufacture, use, release, and disposal; requirements for labeling, product quality assurance, and user certifica- tion; import/export licenses and transportation classifi- cation/regulation), regulatory systems such as those employed by EPA require substantial stewardship on the part of the regulated industry and consumers/

users of the products. That is, people respect the laws, regulations, and policies so that relatively little direct action is needed. Education and informational programs support "enlightened self-interest" and sound enforcement. Monitoring of uses further improves testing criteria and simplification of assessments. Di- rect enforcement is enhanced by such features as commercial insurance availability, coverage, and restric- tions.

A decision to use a given product usually entails some choices as to the extent or nature of the risks involved, so that risk management covers more than simply reduction of risks for a given use. Efficient and effective risk management requires a network of both direct and indirect regulatory controls. In addition, it requires the ability to respond to emergencies created by spills, abuse, misuse, or accident, and the ability to correct such situations. Much of the information developed for risk assessments directly applies to these operations.

The biotechnology industry employing genetically engineered organisms for deliberate release to the environment is new, and both management and techno- logists may not be familiar with regulatory implications of risk in their efforts. Unlike the pesticide and drug industries, which have fostered product development, assessment, and management in a regulatory context, biotechnology industries have focused almost exclu- sively on utility and efficacy. Except for containment with regard to public and occupational health concerns, the industry has had to face few safety issues. It is thus unlikely that individual companies have staffs and managements sensitive to the ecological implications of environmental release of their products.

Successful pesticide, drug, and synthetic organic chemical manufacturers have developed an approach


in which research on health and environmental safety is interwoven with decision making on manufacture, development, and marketing. Very large amounts of capital (tens of millions of dollars per product) are required for this phase, and then perhaps hundreds of millions of dollars for plant construction and opera- dons before any profit is made. Timely and wise decisions make profits; recalled and litigious products do not. The very real contributions of the regulated industries to basic and applied sciences supporting risk assessment have been an important part of the risk management process.

The biotechnology industries do not require such large facilities or amounts of capital for product generation. Facilities may therefore be generally less subject to scrutiny in routine permitting, for which there are few applicable standards under the Clean Air and Clean Water Acts or state legislation and for which there are sparse resources for investigation. More important, these low capital requirements could have a disproportionately high liability, with few assets to seize if a problem develops. For example, in the late 1970s at the Hopewell, Virginia, site of Life Sciences, Inc., the manufacturer of kepone released extraordi- nary quantities to the environment and generated hundreds of millions of dollars of losses. In that case, the "parent" corporation (Allied Chemical Company) had assets that allowed affected individuals and governments to recoup some losses. Life Sciences facilities were an abandoned service station, but neither state nor federal governments can check ever structure to determine if it houses an obnoxious source of poilu- don.

Could such a situation arise, by premeditation or inadvertence, during biotechnology product development and manufacture? Most of the external and in- ternal controls that reduce societal risks from this type of activity are absent, so that such situations seem possible. The release of the gypsy moth occurred under unregulated private entrepreneurship, in attempts to attract a silk industry to the Northeast. The gypsy moth did not stay where released nor gradually decrease due to environmental losses. Instead, unlike the case with kepone, the moth continues to expand the range affected.

Risk management of biotechnology products may be as simple as that of agronomic crops or as complex as management of nuclear power. Genetic engineering may provide so many benefits that regulation mainly may keep uses from being explicitly counterproduc- tive. Many applications and trials involving releases to the environment will be unsuccessful and self-limiting. Yet at this moment, we do not know which uses will

not disrupt significant ecosystem processes, including those (such as evolution and adaptation) that may take many generations to reveal problems. Regulation, in the fullest sense of risk management of environmentally significant releases of biotechnology products, must receive attention while it is still feasible to be con- structive in devising a complete strategy to prevent dis- asters.

For these reasons, the analysis and recommenda- tions of Levin and Harwell (1985) are most appropriate and deserve consideration. They have considered the relevant issues and proposed a comprehensive approach for regulatory assessment and management, emphasizing the need to take certain steps prior to any implementation of environmental release.

Precedents and Principles for Management

Levin and Harwell (1985) draw upon the accumu- lated societal experience in six areas: laboratory experimentation with recombinant DNA and related technologies; agricultural plant and animal husbandry, including biological pest control; deliberate or accidental release of exotic species; the history of epizootics and epidemics; release and regulation of pesticides and toxic substances in the environment; and ecological and evolutionary theory. They identified five factors unique to deliberate release of genetically modified organisms:

�9 Modification of the ecological role of a microorganism may be the direct objective of biotechnology.

�9 Containment and survival are no longer controlled in the same sense as envisioned for earlier R-DNA research.

�9 Introduction of genuinely novel organisms has no direct parallel to transfer of existing organisms from one environment to another.

�9 Biological introductions differ from those of chemicals in that the former can grow and reproduce.

�9 Concern must include not only the spread of introduced organisms per se, but also the vertical (in- heritance) and horizontal (conjugation) movement of genetic information,

Although none of these can give complete, general guidance, the lessons learned within each area clearly underscore the necessity of:

1) A case-by-case assessment of each biotechnology product before use in the environment, including

528 J.W. Gillett

its fate and transport and possible effects on biota and ecosystem processes

2) A plan and established methodology for monitoring of the genetically engineered organism following release, including monitoring of processes anticipated to be affected

3) A plan for containment of the introduced organism within predefined geographical (and temporal) limits

4) Contingency plans for mitigation in case of unde- sirable side effects

These must be in place prior to the application in the environment.

Resource Requirements

Risk assessment of environmentally released genetically engineered organisms would appear to require substantial new resources, beyond those presently available for testing, evaluation, and assessment of conventional toxic chemicals, pesticides, drugs, hazardous wastes, waste streams, and other problematic materials. Diversion of these current, badly needed resources to biotechnology impact assessment would be unwise, since there are shortfalls in facilities, personnel, and revenues to support these on-going efforts. Moreover, the needs of biotechnology impact assessment (and the basic research described in chapters 4 and 5 as necessary to support assessment) would appear to be of even greater magnitude than those required for the assessment of conventional pollutants. No simple set of physical chemical tests, for example, provides information on transport and dispersion of genetically engineered organisms; rather, such testing involves examination in media held under appropriate sets of controlled environmental conditions. These environmental controls can be much more expensive (in terms of facilities and operating costs) than the costs of the measurements themselves.

Facilities. Under guidelines generated by the Na- tional Institutes of Health Recombinent DNA Advi- sory Committee (NIH RAC), a substantial number of private, academic, and governmental laboratories have been equipped to handle genetically engineered organisms safely from public and occupational health viewpoints. However, similar facilities for research on environmental aspects are much more limited. Labo- ratories with established microcosm technology capabilities and P2- and P3-1evel containment are ex- tremely scarce. The US Environmental Protection Agency, for example, has only the facilities at the Gulf Breeze Environmental Research Laboratory that might be readily adapted to the demands of this assessment process.

Fortunately, much of the basic research needed to develop appropriate tests and to generate data bases for evaluation of tests with genetically engineered organisms does not require actual experimentation with problematic organisms per se. At some point, however, these "safe" organisms will have to be compared in fate and effects tests with those believed (or even demonstrated) to be a problem. The research institutions involved in those comparisons will need facilities meeting criteria subsequently developed in consideration of environmental concerns. While growth in research facilities for biotechnology product development is proceeding apace, no similar growth of environmental research and testing facilities is apparent. Because of the capital costs and planning delays in- herent in such expansion of research and testing capabilities, this need may very well lag well behind other r e s o u r c e s .

Personnel resources. To meet national needs in both the areas of product development and risk assessment, specific training programs need to be established and staff added to agencies charged with responsibility for taking regulatory action. These cover three broad areas of education and training for which present resources are either unavailable or inappropriate: basic training of specialists, retraining of scientists from other areas, and upgrading of current regulatory staffs.

The complexity of assessing impacts of biotechnology on the environment requires rethinking of not only the regulatory and testing mechanisms, but also the educational requirements of those involved in risk assessments and product development. An understanding of both the basic and applied sciences must be linked to an understanding of ecology and ecotoxicology. This, in turn, requires dialogue among those trained in the supporting disciplines of microbiology, biochemistry, genetics, molecular biology, and toxicology. Educational programs need to be established which specifically address the interrelationships of these disciplines and provide integrated knowledge. Currently, none are available. The first academic biotechnology programs were established at educational institutions with strong molecular biology/genetics programs. These generally have not had substantive interaction with ecology, environmental biology, or environmental toxicology programs. Few institutions are in a position even to attempt interfacing biotechnology and ecotoxicology.

There is a need for basic training of graduate students, and additional training or retraining for established scientists, in the basic sciences of these inter- locking disciplines. Besides regulatory agency personnel and those involved in product research and


development, lawyers and management personnel need special training in the underlying sciences and potential difficulties of environmental risk assessment of genetically engineered organisms.

Currently, many students have been educated in areas supporting biotechnology development under funding by NIH, NSF (National Science Foundation), and PHS (US Public Health Service) grant programs in biochemistry, genetics, and molecular biology. Many more will receive such training as USDA program funds, state funds, and industrial sources are rushed to support the potential growth of this field. However, toxicology training grants and N1EHS Environmental Health Sciences Centers (EHSC) are essentially health focused. Gaps in ecotoxicology education and training (identified by Kimball 1981) are unlikely to be met within any of these programs without specific redirec- tion and additional resources. Furthermore, it is not in the national interest to divert needed resources from areas of chemical toxicology to biotechnology impact assessment. Both are crucial.

The most appropriate institutions for furthering education supporting biotechnology impact assessment would be those with suitable combinations of disciplines and the capability to implement quality programs immediately directed at an interrelated, multi- disciplinary education. This rationale points to existing EHSCs and EPA-sponsored Centers of Excellence. Such centers usually are oriented toward multidisci- plinary and interdisciplinary research. Institutions and staff are accustomed to working closely with personnel from a variety of federal and state agencies. These institutions are distributed regionally, but have a certain 'focus or theme which results in a relative concentration of specialists. Often they are called upon for rapid responses to national needs, and part of the process by which such institutions have been selected must consider the adaptability of the institution and its staff to these needs.

Over the long term, however, it would be advanta- geous to increase these capabilities across the nation, depending upon the needs of the technology and the relative difficulty of biotechnology impact assessment. The size of the data base needed, the enormous demand for and scope of test method development, and the great variety of critical habitats distributed over the nation all demand increased support for a host of disciplines heterogeneously distributed at institutions large and small.

Risk Assessment Structure

On the basis of the foregoing, a tentative structure (Table 2) for the risk assessment of biotechnology products involves the following sequence of informational elements:

1) Purpose (use) of and rationale for genetic engineering of product organism

2) Process of creation of "new" organism 3) Identity of product, its function, and basic fate

characteristics (survival, growth) 4) Process of manufacturing product (including con-

taminant identity) 5) Manner, quantity, site, and conditions of use and

other releases 6) Fate of manufactured product under conditions

of release, including transport, dissemination, or transference of organism or its altered genetic material

7) Acute effects of product, congeners, contaminants, and environmental derivatives (pathogenicity and so forth)

8) Chronic effects on ecosystem processes of product and derivatives

9) Containment, mitigation, and emergency response procedures

10) Monitoring procedure s for efficacy and identity

These appear to provide the minimum information needed to construct a critical path to a target, identify the target(s) under circumstances of likely introduction into the environment, and identify potential responses which in themselves might involve risk. Assessment is the process of tying together pieces of information. The degree of necessary detail, the methods for obtaining the information, and the criteria by which that information is to be judged must be defined. Finally, a data base must be developed against which to assess the relative importance of these data.

Summary

By combining hazard assessment of effects of a potential biotechnology product with exposure assessments based on study of the genetically engineered organism's fate, conclusions may be reached about the risk involved in release of the product to the environment. In order to make this risk assessment, criteria (including regulatory endpoints) must be established and then developed further against a data base from well-accepted tests. Other aspects requiring research and development include test evaluation, quality assurance, statistical procedures, and methods of identi-

f y i n g and monitoring not only the nominal organism(s) in the products, but also any contaminating material or organisms to which the genetically engineered components may be transferred in the environment.

Application of microcosm technology to testing of genetically engineered organisms is expected to be im-

530 J.W. Gillett

Table 2. Overall assessment structure. Key questions posed for each part of the assessment process, followed by listing of data needs and uses relevant for those questions. Not all of the possible interactions and components can be illustrated.

Component Key questions Data needs

Exposure assessment What is the nature of the genetically engineered, manufactured product?

What numbers of such organisms might be present to affect a valuable and vulnerable organisms or system?

Hazard assessment

Risk characterization

Exposure scenarios

Risk assessment

What organisms, :populations, communities, or ecosystem processes might be affected by the product organism?

What numbers of an organism are required to affect a valued and vulnerable species or process?

How substantial and imminent are the opportunities for exposure and the consequences of the hazard(s)?

What are the critical steps providing effective numbers of a released genetically engineered organism impacting the activity or function of a sensitive species or process?

Does the production and use of this organism present unreasonable risk of hazard or exposure?

Risk management What actions, if any, are appropriate to lessening the nature and extent of risk?

Purpose and rationale of genetic engineering Source organism Process for creating new organism

(contamination? by-products?) Identity of product Fate of organism:

Survival conditions Growth conditions Dispersal mechanisms

Anticipated uses: manner, quantity, site, and conditions of use

Fate of product under conditions of release (genetic stability?)

Acute effects (pathology; direct action) Chronic effects (ecosystem processes and

structure) (competition?) Relationship to numbers of organism present Relationship to regulatory endpoints

Time-to-effect Nature of exposure Relative seriousness of possible outcomes

Conditions of release Mechanisms of fate and transport Transport processes Organism/system behavior

Interactions of dosage-response relationships with exposure scenarios

Impacts on regulatory endpoints Mechanisms of mitigation or regulation available Benefits vs costs

Effectiveness of control procedures Cost of control, monitoring, and/or mitigation Availability of alternatives (products, use

patterns, controls, and so on)

portant, since these systems may be used safely to un- derstand fate and effects prior to (or in place of) testing the product in the environment. Limitations in the use of microcosms may be offset by the cost-effectiveness and incisiveness of results, as has been shown for other pollutants.

Risk management for biotechnology products currently lacks an adequate background, but components o f the process exist or can be developed. New resources, in terms of personnel, training, facilities, and funding, will be needed in order to apply the risk assessment paradigm used for toxic chemicals and pesticides. We will need to know:

�9 The identity, purpose, and process for engineering the new organism as a product

�9 The survival growth, transport, and dispersal of the product and any contaminants in the environment

�9 The transfer of genetic information to other species and their subsequent fate

�9 Acute (pathologic) and chronic (ecologic) effects of the product on all levels o f biota and ecosystem processes which might be affected

�9 Procedures for monitoring, mitigation, and emergency response

Literature Cited

Auer, C. 1983. The pre-manufacturing review process of TSCA. Seminar presented at Corvallis Environmental Re- search Laboratory, Corvallis, Oregon (March).


Bond, H., B. Lighthart, R. Shimabuku, and L. Russell. 1976. Some effects of cadmium on coniferous forest soil and litter microcosms. Soil Science 121:278-287.

Bourquin, A. W., R. L. Garnas, P. H. Pritchard, F. G. Wilkes, C. R. Cripe, and N. I. Rubenstein. 1979. Interdependent microcosms for the assessment of pollutants in the marine environment. International Journal of Environmental Studies 13:131-140.

Cairns, J., Jr. 1983. Are single species toxicity tests alone adequate for estimating environmental hazard? Hydrobiologia 100:47-57.

Cairns, J., Jr., M. Alexander, K.W. Cummins, W.T. Ed- mondson, R. Goldman, J. Harte, A. R. Isensee, R. Levin, J. F. McCormick, T. J. Peterle, andJ. H. Zar. 1981. Testing for effects of chemicals on ecosystems. National Academy Press, Washington, DC, 103 pp.

Clements, R. 1984. Adverse ecological impacts which might drive environmental regulation. Draft document, Environ- mental Effects Branch, Office of Toxic Substances, US En- vironmental Protection Agency, Washington, DC.

[EPA] Environmental Protection Agency. 1982. Test Guide- lines. (a) Health effects (EPA-560/6-82-001); (b) Environ- mental effects (EPA-560/6-82-002); (c) Chemical fate (EPA-560/6-82-003). US Environmental Protection Agency, Washington, DC.

[EPA] Environmental Protection Agency. 1983a. Revised test guidelines. EPA-560/6-83-000. US Environmental Protec- tion Agency, Washington, DC.

[EPA] Environmental Protection Agency. 1983b. Toxic substances control: good laboratory practice standards; final rule. Federal Register 48:53922-53944.

[EPA] Environmental Protection Agency. 1984. Subpart M, Biorational pesticides. Data requirements for pesticide registration. Federal Register 49:42900-42902.

Foundation on Economic Trends vs. Margaret M. Heckler, US District Court, Washington, DC, Civil Action no. 83-2714. J.J. Sirica, Memorandum and Order (16 May 1984).

Giesy, J. P., Jr. 1980. Microcosms in ecological research. DOE Symposium no. 52. National Technical Information Ser- vice, Springfield, Virginia, 1008 pp.

Gillett, J. W., J. D. Gile, and L. K. Russell. 1983. Predator- prey (vole-cricket) interactions: effects of wood preserva- tives. Environmental Toxicology and Chemistry 2:83-93.

Gillett, J. W. and J. M. Witt (eds). 1979. Terrestrial microcosms. NSF-RA-79-0034. National Science Foundation, Washington, DC, 34 pp.

Hammonds, A. (ed.). 1981. Ecotoxicological test systems (ORNL-5709) (EPA-560/11-80-004). Office of Toxic Sub- stances, US Environmental Protection Agency, Wash- ington, DC.

Harte, J., D. Levy, J. Rees, and E. Saagebarth. 1980. Making microcosms an effective assessment tool. Pages 105-107 in J.P. Geisy, Jr. (ed.), Microcosms in ecological research. DOE Symposium no. 52. National Technical Information Service, Springfield, Virginia.

Harte, J., D. Levy, J. Rees, and E. Saagebarth. 1981. Assess- ment of optimum aquatic microcosm design for pollution impact studies (RP-1224-1). Electric Power Research Insti- tute, Palo Alto, California.

Kimball, S. 1981. Graduate programs in ecotoxicology. Eco- systems Research Center, Cornell University, Ithaca, New York (ERC-007), 10 pp.

Levin, S. A., and K. D. Kimball (eds.). 1984. New perspectives in ecotoxicology. Environmental Management 8:375-442. Also appears as ERC Report no. 14a, Ecosystems Research Center, Cornell University, Ithaca, New York.

Levin, S.A., and M. A. Harwell. 1985. Environmental risks and genetically engineered organisms. Pages 56-72 in S. Panem (ed.), Biotechnology implications for public policy. Brookings Institution, Washington, DC. Also appears as ERC-067.

Liang, L. N., J. L. Sinclair, L. N. Mallory, and M. Alexander. 1982. Fate in model ecosystem of microbial species of potential use in genetic engineering. Applied and Environmental Microbiology 44:708-714.

Lighthart, B., Baham,J., and V. V. Volk. 1982. Microbial respiration and chemical speciation in metal-amended soils. Journal of Environmental Quality 12:543-548.

Loeffler, J. 1980. Microcosmology: theoretical applications of biological models. Pages 14-29 in J. P. Geisy, Jr. (ed.), Mi- crocosms in ecological research. DOE Symposium no. 52. National Technical Information Service, Springfield, Vir- ginia.

Metcalf, R.L., G.K. Sangha, and I.P. Kapoor. 1971. A model ecosystem for the evaluation of pesticide degrad- ability and ecological magnification. Environmental Science and Technology 5:709-713.

[NAS] National Academy of Sciences. 1983. Risk assessment in the federal government: managing the process. National Academy Press, Washington, DC, 191 pp.

Neuhold, J., and L. Ruggerio. 1975. Ecosystem processes and organic contaminents (NSF-RA-76008). National Science Foundation, Washington, DC, 44 pp.

(OSTP) Office of Science and Technology Policy. 1984. Pro- posal for a coordinated framework for regulation of biotechnology: notice for public comment, Cabinet Council Working Group on Biotechnology. Federal Register 49:50855-50907.

Perez, K. T. 1983. Experimental marine microcosm test protocol and technical support document [draft]. Ecosystems Analysis Branch, US Environmental Protection Agency, Environmental Research Laboratory, Narragansett, Rhode Island, 44 pp.

Perez, K.T., G.M. Morrison, N.F. Lackie, C.A. Oviatt, S. W. Nixon, B. A. Buckley, and J. F. Heltsche. 1977. The importance of physical and biotic scaling to experimental simulation of a coastal marine ecosystem. Helgolaender Wis- sen~chaftliche Meereuntersuchungen 30:114-162.

Rohrmann, G. F., R. H. McPartland, M. E. Martignoni, and G. S. Beaudreau. 1978. Genetic relatedness of two nncleo- polyhedrosis viruses pathogenic for Orygia pseudotsugata. Vi- rology 84:213-217.

Schafer, M. P., G. F. Rohrmann, U. Heine, and G. S. Beau- dreau. 1979. DNA from two Orygia pseudotsugata baculo- viruses: molecular weight determination by means of elec- tron microscopy and restriction endonuclease analysis. Vi- rology 95:176-184.

Sun, M. 1985. Monsanto may bypass NIH in microbe test. Science 227:153.

532 J.w. Gillett

Taub, F. B. and M. E. Crow. 1980. Synthesizing aquatic microcosms. Pages 69-104 in J. p. Geisy, Jr. (ed.), Micro- cosms in ecological research DOE Symposium no. 52. Na- tional Technical Information Service, Springfield, Virginia.

Van Voris, P., R. V. O'Neill, W. R. Emanuel, and H. H. Shu- gart. 1980. Functional complexity and ecosystem stability. Ecology 6:1352-1360.

Van Voris, P., D.A. Tolle, M.F. Arthur, J. Chesson, and R. W. Brocksen. 1983. Terrestrial microcosms: validation, applications, and cost-benefit analysis. Presented at So-

ciety of Environmental Toxicology and Chemistry, Ar- lington, Virginia, November.

Van Voris, P., M. F. Arthur, and D. A. Tolle. 1984. Terres- trial soil core microcosm test protocol and technical support document. Draft document submitted to the Office of Toxic Substances, US Environmental Protection Agency, Washington, DC.

Wilson, J., C. G. Enfield, W.J. Dunlop, R.L. Cosby, D. A. Foster, and L. B. Baskin. 1981. Transport and fate of selected organic pollutants in a sandy soil. Journal of Environ- metal Quality 10:501-506.

Risk Assessment

Documents

Transcript of Risk Assessment