ecological risk assessment: implications of hormesis

JOURNAL OF APPLIED TOXICOLOGYJ. Appl. Toxicol.20, 131–139 (2000)

Ecological Risk Assessment: Implications ofHormesis†‡

William H. van der Schalie,1* and John H. Gentile2

1National Center for Environmental Assessment-Washington Office, US Environmental Protection Agency, Washington,DC, USA2Rosenstiel School of Marine and Atmospheric Science, University of Miami, Key Biscayne, FL, USA

Key words: hormesis; ecological risk assessment.

Hormesis is a widespread phenomenon across many taxa and chemicals, and, at the single species level,issues regarding the application of hormesis to human health and ecological risk assessment are similar.For example, convincing the public of a ‘beneficial’ effect of environmental chemicals may be problem-atic, and the design and analysis of laboratory studies may require modifications to detect hormesis.However, interpreting the significance of hormesis for even a single species in an ecological riskassessment can be complicated by considerations of competition with other species, predation effects,etc. Ecological risk assessments involve more than a single species; they may involve communities ofhundreds or thousands of species as well as a range of ecological processes. Applying hormeticadjustments to threshold effect levels for chemicals derived from sensitivity distributions for a largenumber of species is impractical. For ecological risks, chemical stressors are frequently of lessor concernthan physical stressors such as habitat alteration or biological stressors such as introduced species, butthe relevance of hormesis to non-chemical stressors is unclear. Although ecological theories such as theintermediate disturbance hypothesis offer some intriguing similarities between chemical hormesis andhormetic-like responses resulting from physical disturbances, mechanistic explanations are lacking.Further exploration of the relevance of hormesis to ecological risk assessment is desirable. Aspectsdeserving additional attention include developing a better understanding of the hormetic effects ofchemical mixtures, the relevance of hormesis to physical and biological stressors and the developmentof criteria for determining when hormesis is likely to be relevant to ecological risk assessments.

INTRODUCTION

Hormesis has been shown to occur widely.1 Althoughrecent discussions have focused on the role of hormesisin human health risk assessment,2–5 there has been lessemphasis on the relevance of hormesis to ecologicalrisk assessment. This paper explores the application ofhormesis to the ecological risk assessment process, asdescribed in guidelines recently published by the USEnvironmental Protection Agency,6 raises issues rel-evant to the consideration of hormesis at various stagesof ecological risk assessment and identifies wherehormesis may be most important. Relevant comparisonsbetween the application of hormesis in human healthand ecological risk assessment are also noted.

This paper does not debate the existence of hormesisor its specific form. For discussion purposes, it isassumed that hormesis follows the general formdescribed by Calabrese.1 As applied to the effects ofchemicals on organisms, this means that there is a

* Correspondence to: Dr. W. H. van der Schalie, National Center forEnvironmental Assessment (8623-D), US Environmental ProtectionAgency, 401 M Street, SW, Washington, DC 20460, USA.† The views expressed in this article are those of the authors andthey do not necessarily reflect the views or policies of the USEnvironmental Protection Agency.‡ This article is a US Government work and is in the public domainin the United States.

Published in 2000 by John Wiley & Sons, Ltd.

stimulatory effect on a given endpoint of 30–60%above the controls and that the stimulatory responseextends over a tenfold range below the no-observed-adverse-effect-level (NOAEL). The data from whichthis generalization is derived include many taxonomicgroups (bacteria, protozoa, fungi, plants and animals),endpoints (e.g. survival, longevity, growth, repro-duction and metabolic effects) and chemicals (e.g. pes-ticides, herbicides, metals and hydrocarbons).1,7–10

Hormesis appears to be widespread across manytaxa, and there are a number of commonalities inthe application of hormetic data to human health andecological risk assessments. However, there are alsosignificant differences in emphasis between the twodisciplines that are relevant to interpreting hormeticdata. Whereas human health risk assessments emphas-ize the protection of individuals, ecological risk assess-ments are most often concerned with effects at thepopulation level and higher (unless the assessmentis concerned with threatened or endangered species).Chemical stressors may affect ecological systemsindirectly (e.g. hormesis-related changes in one speciesmay affect another through changes in competition,predation, etc.) or cause complex cascades of effectsthat may be difficult to anticipate from the originalevent. In addition, ecological risk assessments fre-quently involve physical stressors (such as alterationsin habitat or in the flow regime of rivers) or biologicalstressors (such as non-indigenous species), for which

132 W. H. VAN DER SCHALIE AND J. H. GENTILE

the implications of hormesis are unclear. Finally, evalu-ating the ‘beneficial’ versus ‘adverse’ aspects of horm-esis can be complicated in ecological risk assessment.For example, is an increase in nutrients in a nutrient-poor lake adverse because it alters species compositionin the lake, or beneficial because it increases popu-lations of sport fish?

The following discussion of hormesis and ecologicalrisk assessment follows the phases of the ecologicalrisk assessment process: planning and problem formu-lation, analysis and risk characterization. In problemformulation, risk assessors evaluate goals and selectassessment endpoints, prepare the conceptual modeland develop an analysis plan. During the analysisphase, assessors evaluate exposure to stressors andthe relationship between stressor levels and ecologicaleffects. In the third phase, assessors estimate riskthrough integration of exposure and effects data anddescribe risk by discussing lines of evidence anddetermining ecological adversity. The interface amongrisk assessors, risk managers and interested partiesduring planning at the beginning and communicationof risk at the end of the risk assessment is critical toensure that the results of the assessment can be usedto support a management decision.

PLANNING: MANAGEMENT GOALS, PUBLICVALUES AND HORMESIS

An important part of the planning phase of the ecologi-cal risk assessment process is to articulate the manage-ment goals for the risk assessment in terms of theecological values of concern. This process involves adialog among risk managers, risk assessors and inter-ested parties to ensure that the scientific, regulatoryand societal context for the assessment is madeexplicit.6 The National Research Council describes theparticipation by interested parties in a risk assessmentas an iterative process of ‘analysis’ and ‘deliberation’.11

In this process, interested parties communicate theirconcerns about the environment, economics, culturalchanges or other values potentially at risk fromenvironmental decisions. It is not uncommon for inter-ested parties to participate on advisory groups or evenbecome an integral part of the risk management team.The participation of the public and the regulatorycommunity in planning involves many issues beyondscientific questions regarding the existence of hormesis.

Public involvement in setting management goals forecological risk assessments means that social valuesmust be considered. Societal values, used in the broad-est sense, includes both the public’s value of ecologicalgoods and services and their perceptions and tolerancefor risk from certain activities. Factors that shape thepublic’s perception of the risks from chemicals thatare relevant to communicating the risks associated withhormesis include: cultural perceptions of chemicals,source of chemical (e.g., natural or man-made), use ofchemical (e.g. toxic or therapeutic), innate hazard (e.g.potency and type of harm) and sense of personalcontrol over exposure.12–14 Ultimately, it is the public’sunderstanding of toxicology and their tolerance for riskthat determines the success of incorporating hormetic

Published in 2000 by John Wiley & Sons, Ltd. J. Appl. Toxicol.20, 131–139 (2000)

principles into ecological risk assessments. (Althoughthis discussion focuses on chemical stressors, many ofthe same issues apply to physical and biologicalstressors.)

Although society has expended great effort andresources to improve health and safety for the publicand reduce pollution of the environment, a large seg-ment of the population has become more, rather thanless, concerned about risk.15 This persistent fear ofchemicals stems from widely held cultural premisessuch as the ‘law of contagion’, ‘once in contact, alwaysin contact’ and an ‘essence of harm’ syndrome thatbeing ‘contaminated’ is an all-or-none quality.12 Thisall-or-none quality is irrespective of the degree orlength of exposure and results in the observation byErickson16 that “To be contaminated by radiation orother toxins . . . is to be contaminated in some deepand lasting way, to feel dirtied, tainted, and corrupted”.The public perceives that environmental chemicals areinnately hazardous and that they, the public, lack theability to control their exposure to these chemicals.These culturally based perceptions lead to the demandfor regulatory agencies to set zero-risk levels as theirgoals.14 For example, current Clean Water Act (CWA)regulations have a long-term goal of zero discharge.Technology (BAT) and risk-based water quality criteria(WQC) are the two approaches currently being usedin the CWA to move toward this goal. The implicitassumption of the CWA is that these approaches willbe strengthened continually until the zero-dischargegoal is attained.

These views are relevant to environmental healthas well as public health. The public’s awareness ofenvironmental issues and of the intrinsic link betweenpublic and environmental health has increased dramati-cally since the 1970s. Specifically, the public has sup-ported large expenditures of resources for waste treat-ment facilities and has backed many legislative actionsto reduce the risk of man-made chemicals to fish andwildlife and to improve the water quality in our oceans,rivers and streams. This environmental ethic presentsa challenge to those who would propose anything butthe most stringent controls on the release of chemicalsinto the environment.

Given these public attitudes, regulators may find itdifficult to incorporate hormetic concepts that are con-sistent with the laws and regulations they administer.Taking the hormesis hypothesis seriously will necessi-tate substantive revision of regulatory philosophy andactions.5 If the intent of public policy is to improvethe health of the public and the environment andnot simply prevent negative effects, then the hormesishypothesis implies that public policy needs to beexpanded to ensure that exposure to small amounts ofchemicals is not prohibited.17 The environmental dia-logue would shift from how much pollution should beeliminated or removed to how much should be leftto enhance human health or the environment.18 Thisapproach, which contradicts the zero-risk philosophythat has dominated current regulatory policy, hasimportant implications for setting management goals inthe planning phase of the risk process. Further, itmay be difficult to represent hormesis by the single,quantitative numeric with associated uncertainty typicalof regulatory standards, criteria and benchmarks. Yet

133ECOLOGICAL RISK ASSESSMENT OF HORMESIS

this may be required if hormesis is to become anintegral part of the standards and criteria that are thecornerstone of regulatory policy.

Another dilemma for applying hormesis in a regu-latory setting is how to handle differentially sensitivesubpopulations. Forexample, regarding the healtheffects of pesticides, setting safe exposure levels forchildren will result in exposures below the hormeticrange and benefits for the adult population. Conversely,setting risk-based criteria at the hormetic range foradults will put children at risk. The public will likelytake the risk-averse approach and add the necessarysafety factors to protect children regardless of the lossof hormetic benefits to another segment of the popu-lation. Dealing with differential sensitivity is an evenmore important problem for ecological risks thatinvolve significant differences in both intra- and inter-specific sensitivity. This issue is addressed in moredetail in the discussion on the analysis phase.

The public’s association of ‘harm’ with chemicalsis so strong that hormesis may not be incorporatedinto the regulatory process even when there is strongscientific evidence of the validity of hormesis.18

However, although convincing the public that thereare benefits from exposure to low doses of chemicalswill be difficult, it is not necessarily impossible. Afirst step is clearly demonstrating the scientific basisfor hormesis and resolving the remaining technicalissues. As discussed below, this may be more difficultin the ecological than in the human health area.Assuming that scientific issues can be resolved, theconcept of hormesis will have to be clearly communi-cated to the public. Renn17 suggests a phased riskcommunication process, beginning with introducinghormesis in connection with natural agents familiarto most people. For example, starting with thephysiological requirements for small quantities ofminerals that are toxic at higher doses, the hormesisconcept could be expanded to include syntheticchemicals derived from natural products and, then tochemicals that are regarded by the public as pol-lutants. The last step is more likely to be successfulif hormesis hypothesis is not linked with vestedinterests in regulated community.

It is clear that adopting the hormetic hypothesis willrequire a major intellectual and emotional shift of focusfor the public from reducing pollution to zero or nearzero to some optimum level that is beneficial to healthand the environment. Unless society is able to acceptthat exposure to chemicals commonly perceived astoxic is not only safe but beneficial at some exposures,then the incorporation of hormesis into the regulatoryand policy arenas will remain problematic.

PROBLEM FORMULATION

The first phase of the risk assessment process is prob-lem formulation. Here, the purpose is articulated, thegoals are defined and a plan for analyzing and charac-terizing risk is developed.6 Initially, available infor-mation is used to identify the sources and stressors ofconcern, the ecosystem receptors that are at risk andthe endpoints to be used in the assessment. A concep-


tual model is constructed using these data to illustratethe potential causal relationships among sources, stres-sors, receptors and endpoints. The conceptual modeland assessment endpoints are the principal products ofproblem formulation and are used to complete ananalysis plan.

Stressor characterization and hormesis

Chemical stressors. There is considerable evidencethat several classes of chemicals (e.g. pesticides, herbi-cides, hydrocarbons, metals) as well as effluents pro-duce a hormetic response in a variety of phyla andlife stages.7–9 The descriptive evidence that there is arange of chemical exposure within an order of magni-tude below the NOAEL that is 30–60% stimulatorysupports the chemical hormesis hypothesis.1,8 A reviewof 4000 potentially relevant articles by Calabrese andBaldwin19 has identified approximately 350 studiesshowing qualitative evidence of hormesis. Of these,47% were shown to meet the rigorous quantitativecriteria necessary to demonstrate hormesis. Althoughthere is a body of evidence suggesting that individualchemicals and potentially mixtures of chemicals canproduce a stimulatory response at exposures within anorder of magnitude of the NOAEL, there is no singlemechanism to account for the observed phenomenonthat can be used to establish hormesis as a generaltoxicological concept.19

The lack of a general underlying mechanistic theoryfor hormesis limits the predictive value of the hormesishypothesis in problem formulation. One of theimportant aspects of stressor characterization is knowl-edge of the physical, chemical and biological(toxicological) properties of the stressor. This infor-mation is used to predict the transport, fate andexposure of the chemical and the potential ecologicaleffects. Given the spectrum of chemicals released intothe environment, having a mechanistic basis for pre-dicting which chemicals will display hormetic proper-ties would be invaluable in helping to define the prob-lem. A priori knowledge of the potential for hormesiswould ensure that the design of the risk assessmentfor both exposure and stress–response relationshipsaccounts for this possibility.

In problem formulation, it may be necessary toconsider hormetic effects resulting from exposurethrough bioaccumulation (e.g. by food chain transfer)as well as from direct exposure. Bioaccumulation typi-cally occurs with chemicals having logKow values$4(e.g. PCBs, DDT, etc.). For these types of chemicals,using an NOAEL based upon direct effects couldconceivably result in the selection of a hormetic con-centration that would result in bioaccumulation of thechemical to concentrations that would be unacceptableto the health of a population or its prey through trophictransfer. Given chemicals with a certain range ofKow,acute and chronic toxicity and hormetic potential,further investigation would be required to determinewhether the steady-state tissue residue values fromhormetic exposures would exceed ecological (e.g.wildlife) criteria.

Physical/biological stressors. Although there isincreasing evidence of hormesis in chemicals, physical


and biological stressors are even more important inshaping ecological communities and landscapes.Despite a paucity of literature that explicitly discusseshormetic propertiesper se of population, communityor ecosystem level responses to physical or biologicalstressors, there are parallels in ecological theory thatappear to fit the hormetic pattern.

The ‘intermediate disturbance hypothesis’20 predictsthat species richness will be highest in communitieswith moderate levels of disturbance and at intermediatetime spans following disturbance.21 This non-equilib-rium successional model was proposed to explain spec-ies richness in tropical forests and coral reefs. Thepremise is that soon after a severe disturbance, propag-ules of new species arrive and colonize the open space.If the interval between disturbances increases, diversitywill also increase because more time is available forthe invasion and growth of more species. However, ifthe interval between disturbance is too long then diver-sity begins to decline. Connell20 shows what is essen-tially a ‘hormetic-like’ pattern for the diversity fordifferent successional stages of the Budongo forest.In this example, changes in diversity illustrate theintermediate disturbance hypothesis while producing ahormetic-like response that is a function of disturbancefrequence, time after a disturbance and the size ofthe disturbance.

Although this hypothesis has gained support fromempirical field studies and mathematical modeling,22–24

the evidence is not always consistent. For example,comparison of unburned, annual and 4-year fire fre-quencies in tall grass prairies suggests that there is amonotonic decline in species richness with increasingdisturbance frequency but that maximum richness wasreached at an intermediate time interval since the lastdisturbance.21 However, as we know from studies ofchemical hormesis, failure to detect hormesis in anexperiment may mean that hormesis is absent or simplythat the study design was inadequate to detect it. Thusit is conceivable that the tall grass prairie field studydid not include an appropriate range of fire frequenciesto demonstrate the intermediate disturbance hypothesis.Nevertheless, the intermediate disturbance hypothesisand its analogs suggest the possibility of hormetic-likeresponses from physical and/or biological disturbanceregimes.

Assessment endpoints and hormesis

Assessment endpoints, which are explicit expressionsof the actual environmental values to be protected, arecritical to problem formulation because they structurethe assessment to address management concerns andare central to the development of the conceptualmodel.6 Assessment endpoints are drawn from the fullrange of ecological organizational levels: from individ-uals and populations to communities, ecosystems andlandscapes. Considering the array of possibilities, thechallenge is in selecting those ecological attributes thatneed to be protected to meet management goals. Thethree principal criteria used to select assessment end-points are: ecological relevance; susceptibility to knownor potential stressors; and relevance to societal valuesand management goals.6


Because hormesis has its origins in toxicology, thefocus of studies of hormesis has been on measuringbiological responses at the individual and populationlevels of ecological organization. To date, a variety ofbiological responses (e.g. survival, growth, repro-duction, and metabolic effects) have been shown toexhibit a hormetic response.1,7–10 Stebbing recentlyreviewed findings on the dynamics of growth hormesisin biological systems, principally microalgae, protozoa,yeast and colonial hydroids.8 Calabrese describes sev-eral examples involving approximately 40 evaluationsof hormetic-like reproductive effects below theNOAEL. Unequivocal demonstration of a hormeticresponse for reproductive effects will be particularlyimportant given the importance of reproductive func-tion as an assessment endpoint in many ecologicalstudies. Survival, growth and reproduction areimportant assessment endpoints that can be used at theindividual level of assessment or to project popu-lation effects.

The apparent lack of studies to assess hormeticeffects at higher levels of organization (e.g. community,ecosystem or landscape) may be the result of omissionor semantics. Whatever the reason, this is a fruitfularea for further study, as evidenced by the intermediatedisturbance hypothesis.20,21 Community attributes (e.g.diversity, species richness) are dependent upon intra-and inter-specific competition for resources and spaceand thus are qualitatively different from individualsurvival, growth and reproduction. Anthropogenicchemical and natural physical stressors (e.g. fire, habitatfragmentation) operate by altering the intra- and inter-specific patterns of competition.20,21 The chemical stres-sors will operate through the differential susceptibility(e.g. comparative toxicology) of community membersand the natural physical stressors through the alterationof successional patterns within the landscape mosaic.Selecting the appropriate spatial and temporal scalesand the range of stressor intensity and durationbecomes critically important in detecting hormesis atthe community, ecosystem and landscape scales. Thispoint is clearly demonstrated in the studies of theintermediate disturbance hypothesis for the tall grassprairie.21 Thus there is at least some evidence indicat-ing that hormetic-like responses can be detected acrossthe range of ecological scales and responses thatcharacterize assessment endpoints.

Conceptual models and hormesis

The conceptual model is a visual representation andwritten description of the potential causal relationshipsamong sources, stressors and ecological effects. Con-ceptual models can be hierarchical, i.e. nested, to showincreasing details. Depending on the level of detail andaggregation, conceptual models can illustrate processesfor linking land use to sediment and nutrient run-off with its consequences to downstream ecologicalcommunities and habitats that support migrating water-fowl. Conceptual models can be used to describemultiple exposure pathways that converge on anendangered or threatened species. In all cases, theconceptual model describes the linkages among thesources, stressors, ecological receptors and the assessmentendpoints. These relationships are the basis for


developing a suite of risk hypotheses that posit thepotential ecological consequences of stressor exposureor disturbance regime. If sufficient data are available,a relative ranking of the importance of specific path-ways and assessment endpoints can be conducted andcandidate benchmarks can be determined for the assess-ment endpoints. This information would provideimportant guidance for the design of the analysis planfor the assessment.

There are several points of intersection between theconceptual model and hormesis. First, by describingthe types of ecological responses/endpoints and stressorcategories, the conceptual model can be used to identifythose stressor–response pathways for which hormesishas been demonstrated in the literature and would havethe highest probability of contributing to the assess-ment. Second, if predictive criteria can be developedfor discriminating between hormetic and non-hormeticstressors, then applying these criteria to the conceptualmodel would provide useful guidance to the managerand risk assessor and minimize the unnecessary expen-diture of resources. Third, the conceptual model canbe used to identify the potential indirect and non-lineareffects that might be subject to hormesis. Fourth, theconceptual model can provide meaningful guidance forsetting performance criteria and/or benchmarks for theassessment endpoints.18,25 Thus, the conceptual modelprovides qualitative guidance by identifying the riskhypotheses (e.g. stressor–effect relationships) for whichthere is a plausible hormetic response.

Once the decision has been made to consider horm-esis in an assessment, the analysis plan must addressthose elements necessary to detect and quantify theresponse for the assessment endpoints. The implicationsfor a quantitative risk assessment when hormesis existsare far reaching and are discussed elsewhere in thispaper and the literature.5 The analysis plan specifiesthe design and analyses requirements for quantitativeinterpretation of exposure–response data and theenvironmental measures required to detect hormesis fora particular chemical or physical stressor. These designfeatures may require departures from current practice.

ANALYSIS

Once planning and problem formulation have beencompleted, the analysis phase of ecological risk assess-ment can begin. Both exposure and effects data areevaluated in the analysis phase, but this discussionemphasizes the possible influence of hormesis on thecharacterization of ecological effects. Important aspectsinclude experimental design limitations of both labora-tory and field studies, extrapolating from measuredeffects to the environmental values of concern(assessment endpoints) and the treatment of chemicalmixtures. The discussion of laboratory effects datafocuses on chemical stressors, whereas the discussionof field effects data includes both chemical and non-chemical stressors.

Laboratory effects data

As in many human health studies, hormesis is notalways detectable using traditional ecotoxicological


tests. As others have noted, the size of the hormeticeffect is relatively small, the effect occurs at low levelswhere there may be insufficient numbers of samplesto define the response and many studies focus onhigher, more easily detected chemical levels.9,26,27 Lab-oratory tests for acute lethality typically focus on thedetermination oflc50s, ec50s or ld50s values, so testconcentrations may be well above levels that causehormetic responses. Tests measuring sublethal end-points (typically growth and reproduction) are morelikely to encounter hormetic effects, but their abilityto detect such effects depends on the appropriate spac-ing of test concentrations/doses and the statisticalpower of the tests.

Chronic test data are usually analyzed either byusing regression analysis (to generate an effect levelsuch as anec25) or by hypothesis testing to calculateNOAELs and the lowest-observed-adverse-effect levels(LOAELs). Experiments relying upon hypothesis test-ing may have insufficient statistical power to detecthormesis. For example, Suter and Rosen28 found that intests with fish, on average, the threshold for statisticallysignificant effects ranged from 12% for hatching to42% for adult fecundity. If hormesis is to be evaluatedin ecotoxicological testing, tests will have to bedesigned to consider the statistical power required,given the typical variability found for the variousendpoints and species used in traditional laboratorytests.

Another option for analyzing laboratory test data isregression-type models. Bailer and Oris9 note that,given a consistent form for the hormesis response,there are concentration–response models that can beused to describe the phenomenon.29,30 Analytical pro-cedures assuming a monotonic pattern in the datashould be avoided if hormesis is to be evaluated,because such methods may artificially elevate the con-trol response level by pooling controls and lowerchemical levels associated with hormetic responses.30

A related issue is whether measures such as anec50

should be based on half the control mean or half themean at the hormetic peak.9,31 The answer depends onwhether one views the hormetic response as an optimallevel (as it might be for an essential micronutrient) ornot. In either case, pooling responses from the controland low-level exposures is inappropriate.9

Characterizing ecological effects usually involvesextrapolation from measures of effect (e.g. laboratorytoxicity data) to the environmental values of concern(assessment endpoints, e.g. effects on a population ofconcern in a field situation). For example, modelsmay be used to forecast population impacts based onlaboratory data on individual survival, reproduction andgrowth. Forbes and Callow32 suggest that hormesismay exert metabolic costs on organisms, and that theimpact of effects could be evaluated through modelslinking energy budget measurements for individualswith population dynamics. They point out, however,that such linkages may be complex and do not takeinto account interspecies factors such as predation orcompetition. In his discussion of the use of populationmodels to estimate ecological risks to aquatic com-munities, Bartell33 describes the application of concen-tration–response functions (CRFs) for describing thesublethal effects of toxic chemicals. Recognizing that


detailed CRFs are never available for all species andchemicals of interest, he suggests using linear approxi-mations instead. He notes that these approximationswill not address hormetic effects and will also beinaccurate at concentrations exceeding thelc50. Giventhe widespread nature of hormesis, it would be interest-ing to repeat Bartell’s aquatic toxicity model evalu-ations using CRFs that follow the generalized hormesispattern that has been described.1

Ecological effects data from laboratory toxicity testswith a number of species are frequently summarizedas point estimates (e.g. water quality criteria) or asdistributions of species sensitivities.34 Consider themethods used by the US Environmental ProtectionAgency’s Office of Water for generating water qualitycriteria.35 To generate an acute toxicity benchmarkvalue (the ‘criterion maximum concentration’), 48–96 hlc50 or ec50 data (depending on species) are obtainedfor organisms representing a range of taxa, includingfish, aquatic invertebrates and aquatic plants. Meanacute toxicity values are determined for each genus oforganism, and these mean values are used to constructa species sensitivity curve. A chemical concentrationis read from the curve at the 5% level (i.e. below thelc50 or ec50 for 95% of the species tested), and halfthat level is used as the ‘criterion maximum concen-tration’. A similar procedure is used to calculate achronic effect benchmark concentration using chronicno-effect levels. It is important to note that the rangeof sensitivities among species can be several orders ofmagnitude for acute effects and at least 1–2 orders ofmagnitude for chronic effects.

This methodology emphasizes a fundamental dif-ference between human health and ecological riskassessments: human health deals with only one spec-ies (albeit with a range of subpopulations) whereasecology deals with many species. Frequently, eco-logical risk assessors must extrapolate from data ona limited number of species to many other species,communities and ecosystems. As illustrated by thewater quality criteria example, a ‘criterion’ or‘benchmark’ level is derived that is deemed ‘safe’for 95% of the tested species (and, by inference,many more untested species in the environment).Applying hormesis in this context is difficult.Assuming that the hormetic range falls within afactor of 10 below the NOAEL for the most sensitivespecies, choosing a hormetic value for a criterioncould ‘benefit’ those species whose NOAELs arewithin the designated range. However, the other 95%of species will experience no benefit from setting ahormetic benchmark. Conversely, setting a hormeticbenchmark above the 95% level will clearly jeop-ardize the more sensitive range of the species distri-bution, provide marginal benefit to some subsets ofthe species and provide no benefit to the remainder.Gaylor25 reached a similar conclusion when he exam-ined the feasibility of setting reference doses (RfD)under a hormetic scenario for five human subpopula-tions with varying sensitivities. His conclusion isthat there is no general recommendation for alteringRfDs when hormetic effects exist and that eachsituation should be considered on a case-by-casebasis.25


Field effects data

Evaluating hormesis under field conditions shouldbe much more difficult than in the laboratory, for anumber of reasons. The statistical power of obser-vational field studies is relatively low. For example,Peterman36 cites the work of de la Mare37 that dem-onstrated only a 69% chance that a 50% decline inwhale abundance over 20 years could be detectedowing to large sampling variability. Peterman notesthat statistical power is rarely reported in field stud-ies, which can give the false impression that noeffect is occurring. Experimental test systems in thefield may also have limited power to detect hormesis.Kraufvelin38 found that for a land-based brackishwater mesocosm, half of the 50 population and com-munity variables studied during five separate testyears had coefficients of variation greater than 60%.For these variables, detection of hormetic influenceswould be a challenge.

Landis39 disputes using traditional approaches,such as comparing field control and reference sites,for evaluating ecological risks. Landis notes thatecological systems (or structures) are complex, withnon-linear interactions and stochastic events. Accord-ing to the Community Conditioning hypothesis,40

historical events are held in a number of compart-ments in an ecological structure; the information ispersistent and determines the response of a com-munity to subsequent events. One consequence isthat an ecological structure cannot return to a pre-existing state and reference or control sites cannotexist. Given this situation, hormesis becomes anadditional variable may influence the trajectory ofthe ecological structure but may be extremely diffi-cult to evaluate as a separate entity in a field system.Forbes and Callow32 note that if hormesis-relatedmetabolic costs of stress resistance occur generallyacross species, changes may occur in the relation-ships between community/ecosystem production, res-piration and biomass levels. However, they alsoobserve that the complex relationships in such sys-tems make it very difficult to draw any conclusionsabout potential community or ecosystem effects fromobservations on individual species.

Interpretation of hormetic effects in field situationsis further complicated by the presence of multiplestressors. Our focus here is on chemical stressors,although there may be biological and physical stres-sors as well. Foran18 raises the issue of whetherthere are cumulative hormetic effects for chemicalsdemonstrating hormesis and having the same mech-anism of toxic action. Published information41–44 sug-gest that a strictly additive model can predict effectsof such chemicals for a wide range of chemicals,organisms and endpoints. Thus, there appears to belittle evidence of cumulative hormetic effects forchemical mixtures. Nevertheless, this area deservesfurther investigation.

Exposure considerations

The analysis phase of ecological risk assessmentincludes evaluation of exposure as well as effects.Exposure considerations for chemicals causing


hormetic effects are generally the same as those forall chemicals, with the exception of the need tofocus on the lower end of the dose– or concen-tration–response curve, as noted above. In addition,as observed by Sielken and Stevenson,5 dose scalesor metrics should reflect age or time dependency ofthe exposure.

RISK CHARACTERIZATION

Risk characterization is the final phase of ecologicalrisk assessment and is the culmination of planning,problem formulation and analysis. Risk assessors usethe results of the analysis phase to develop an esti-mate of the risk posed to the ecological entitiesincluded in the assessment endpoints identified inproblem formulation. After estimating the risk, theassessor describes the risk by considering anticipatedadverse effects and lines of evidence supportingtheir likelihood. Finally, the assessor identifies andsummarizes the uncertainties, assumptions and quali-fiers in the risk assessment and reports the con-clusions to risk managers.

There are many different approaches for combin-ing exposure and effects information to estimaterisks, including field observational studies, categ-orical rankings, comparing single point estimates ofexposure and effects, incorporating variability intoexposure and effects estimates (e.g. using distri-butions of data) and process models.6 Aspects of anumber of these approaches relevant to hormesishave been discussed in the previous section. Here,we briefly discuss some aspects of describing therisk estimates: are the hormetic effects real?; whatis the adverse or beneficial nature of hormeticeffects?; and how do hormetic data fit into a weightof evidence consideration?

In risk estimates that involve a hormetic response,it is important to evaluate the nature as well asthe magnitude of the response. Suter10 cites severalexamples where toxicity test results may show a‘beneficial’ hormetic response that is irrelevant ornot beneficial in the real world. Examples includereduced aggression due to a narcotizing effect or anincrease in the growth of larval fish that may berelated to increased food from microbes growing onthe test chemical or solvent. Davis and Svendsgaard2

describe a number of possible mechanistic expla-nations for hormesis, such as statistical artifacts,confounding factors, interactive effects (e.g. ofchemical mixtures), adaptation by an organism(homeostasis) and compensatory and protectivemechanisms. Stebbing8 proposes a mechanisticexplanation for growth hormesis. In risk characteriz-ation, every effort should be made to identify theunderlying cause of the response.

Risk characterization also requires a considerationof ecological adversity—the significance of hormesisfor the ecological entities identified in the assessmentendpoints. Forbes and Callow32 discuss how the rela-tive effects of hormesis on a population may varydepending on whether the hormetic effect isinducible or constitutive. If the assessment endpoint


involves a particular species of concern, decidingwhether hormesis is beneficial or adverse dependson understanding not only the mechanism causingthe observed response but also on how changes inthe population of the species of concern may interactwith other species or components of the ecologicalsystem. Davis and Svendsgaard2 note that chemicalconcentrations causing a positive effect on one spec-ies may have very different effects on other species.Even in the simplest ecological systems, an increasein one population may have complex and variableeffects on other populations. Add in compensatorypopulation responses, and prediction can quicklybecome very difficult.

The conclusions of an ecological risk assessmentmay be strengthened by evaluating risks using sev-eral different lines of evidence. For example, demon-stration of hormesis in multiple species, in laboratoryand field tests and/or in modeling studies couldimprove confidence in its importance. In consideringthese various elements, the risk assessor should con-sider factors such as the adequacy and quality ofthe data, the degree and type of associated uncer-tainty and the relevance of the data to the assessmentendpoints and risk assessment.6

CONCLUSIONS AND RECOMMENDATIONS

Ecological risk assessors should be cognizant ofhormesis, given its widespread occurrence acrosstaxonomic groups, endpoints and chemicals, but itis unclear whether hormesis also applies to physicaland biological stressors. Understanding the signifi-cance of hormesis for ecological risks will requirebetter understanding of the underlying mechanismsof the phenomenon as well as an appreciation of theconsequences of hormesis at levels of biologicalorganization above the individual. Compensatory andadaptive population responses combined with inter-species interactions and ecosystem level processeswill make this a difficult task. Given the level ofuncertainties in the overall ecological risk assessmentprocess, hormesis may not be a highly importantfactor. Nevertheless, we offer the following rec-ommendations for considering hormesis in the con-text of ecological risk assessment:

(i) Standard laboratory toxicity tests and fieldobservational studies should be flexible enoughto accommodate and fit the observed shape ofthe data and not be forced to be linear.5 Statisti-cal methods assuming monotonic dose–concen-tration–response relationships should be usedwith caution. As noted by Bailer and Oris,9

‘. . . experimentalists should design studies withspacing of concentrations that capture the areawhere subtoxic stimulation might be observed,and then fit statistical models that allow forthis phenomenon’. However, increasing samplesize and utilization of lower concentration/doseranges has implications for the cost of studiesthat will need to be addressed.27


(ii) It is impractical to apply hormetic effects to ecologi-cal risk-based benchmarks or criteria derived fromdistributions of species sensitivity data.

(iii) A better understanding is needed of the hormeticeffects of chemical mixtures.18

(iv) The relevance of hormesis to non-chemical stres-sors should be explored.

(v) The existence of hormetic effects at the com-munity level and above, where the assessmentendpoints involve appropriate attributes of multi-species assemblages (e.g. species diversity, rich-ness, etc.), should be explored. Also, hormeticmechanisms at this organizational level should beevaluated. The intermediate disturbance hypoth-esis20 offers a point of departure for dealing withthe hormetic effects of chemical, physical andbiological stressors at community and ecosystemscales.

(vi) Given the myriad of issues regarding the inherentassumptions, degree of universality, delectability

REFERENCES

1. E. J. Calabrese and L. A. Baldwin, The dose determinesthe stimulation (and poison): development of a chemicalhormesis database. Int. J. Toxicol. 16, 545–559 (1997).

2. J. M. Davis and D. J. Svendsgaard, U-shaped dose–response curves: their occurrence and implications for riskassessment. J. Toxicol. Environ. Health 30, 71–83 (1990).

3. E. A. McKenna, Hormesis: considerations and implicationsfor human health risk assessment. Int. J. Environ. Pollut.9, 90–107 (1998).

4. C. J. Paperiello, Risk assessment and risk managementimplications of hormesis. Belle Newsl 7, 28–30 (1998).

5. R. L. Sielken, Jr. and D. E. Stevenson, Some implicationsfor quantitative risk assessment if hormesis exists. BelleNewsl 6, 13–17 (1998).

6. US Environmental Protection Agency, Guidelines for eco-logical risk assessment. Fed. Reg. 63, 26846–26924 (1998).

7. A. P. Appleby, The practical implications of hormeticeffects of herbicides on plants. Belle Newsl. 6, 23–24(1998).

8. A. R. D. Stebbing, A theory for growth hormesis. BelleNewsl. 6, 1–12 (1997).

9. A. J. Bailer and J. T. Oris, Incorporating hormesis in theroutine testing of hazards. Belle Newsl. 6, 2–5 (1998).

10. G. W. Suter, II, Ecological Risk Assessment. Lewis Pub-lishers, Chelsea, MI (1993).

11. National Research Council, Understanding Risk: InformingDecisions in a Democratic Society. National AcademyPress, Washington, DC (1996).

12. P. Slovic, If hormesis exists. . . implications for risk percep-tion and communication. Belle Newsl. 7, 9–10 (1998).

13. R. Renn and D. Levine, Trust and credibility in risk com-munication. In Communicating Risks to the Public, ed.by R. Kasperson and P. J. Stallen, pp. 175–218. Kluwer,Dordrecht (1991).

14. L. Sjoberg. Perceived Risk versus Demand for RiskReduction. RHIZIKON: Risk Report No. 18. Center for RiskResearch: Stockholm School of Economics, Stockholm,Sweden (1994).

15. P. Slovic, Perception of risk. Science 236, 280–285 (1987).16. K. Erickson, Toxic reckoning: business faces a new kind

of fear. Harvard Bus. Rev. Jan./Feb., 118–126 (1990).17. R. Renn, Implications of the hormesis hypothesis for risk

perception and communication. Belle Newsl. 7, 2–9 (1998).18. J. A. Foran, Regulatory implications of hormesis. Belle

Newsl. 7, 11–13 (1998).19. E. J. Calabrese and L. A. Baldwin, A quantitatively-based

methodology for the evaluation of chemical hormesis.Hum. Ecol. Risk Assess. 3, 545–554 (1997).


and measurement uncertainty of chemically-induced hormetic responses, it may be useful todevelop criteria for determining when hormesisis likely to be relevant to ecological risk assess-ments. This could then alert the risk assessor ofthe need to address this issue in all phases ofthe assessment.

(vii) Applications of hormetic principles to risk man-agement decisions are likely to be limited byboth public attitudes and the strength of ourscientific understanding of hormesis. A key scien-tific issue is understanding the underlying mech-anisms for hormetic responses to chemical, physi-cal or biological stressors. The more difficultissue may be convincing the public that somelevel of exposure to stressors can be beneficial.

Acknowledgements

We thank Jennifer Orme-Zavaleta for her comments on the mansucript.

20. J. H. Connell, Diversity in tropical rain forests and coralreefs. Science 199, 1302–1310 (1978).

21. S. L. Collins, S. M. Glenn and D. J. Gibson, Experimentalanalysis of intermediate disturbance and initial floristiccomposition: decoupling cause and effect. Ecology 76,486–492 (1995).

22. R. Abugov, Species diversity and phasing of disturbance.Ecology 63, 289–293 (1982).

23. M. Houston, A general hypothesis of species diversity.Am. Nat. 113, 81–101 (1979).

24. D. Tilman, Resource Competition and Community Struc-ture. Princeton University Press, Princeton, NJ (1982).

25. D. G. Gaylor, Safety assessment with hormetic effects.Belle Newsl. 6, 6–8 (1998).

26. A. R. D. Stebbing, Hormesis—the stimulation of growthby low levels of inhibitors. Sci. Total Environ. 22, 213–234 (1982).

27. T. E. Johnson and H. Bruunsgaard, Implications of horm-esis for biomedical aging research. Belle Newsl. 6, 17–19 (1998).

28. G. W. Suter, II and A. E. Rosen, Endpoints for responsesof fish to chronic toxic exposures. Environ. Toxicol. Chem.6, 793–809 (1987).

29. A. J. Bailer and J. T. Oris, Estimating inhibition concen-trations for different response scales using generalizedlinear models. Environ. Toxicol. Chem. 16, 1554–1559(1997).

30. P. H. Van Ewijk and J. A. Hoekstra, Calculation of theec50 and its confidence interval when subtoxic stimulus ispresent. Ecotoxicol. Environ. Safety 25, 25–32 (1993).

31. D. J. Svendsgaard, Commentary. Belle Newsl. 2, 13–16(1993).

32. V. E. Forbes and P. Callow, Costs of living with contami-nants: implications for assessing low-level exposures.Belle Newsl. 4, 1–13 (1996).

33. S. M. Bartell, R. H. Gardner and R. V. O’Neill, Modelingsublethal toxic effects. In Ecological Risk Estimation, ed.by S. M. Bartell, R. H. Gardner and R. V. O’Neill, pp. 65–92. Lewis Publishers, Chelsea, MI (1992).

34. R. D. Cardwell, B. R. Parkhurst, W. Warren-Hicks and J. S.Volosin, Aquatic ecological risk. Water Environ. Technol.5, 47–51 (1993).

35. C. E. Stephan, D. I. Mount, D. J. Hansen, G. A. Gentile,G. A. Chapman and W. A. Brungs, Guidelines for DerivingNumerical Water Quality Criteria for the Protection ofAquatic Organisms and their Uses. EPA 822/R-85-100.Office of Water, US Environmental Protection Agency,Washington, DC (1985).


36. R. M. Peterman, The importance of reporting statisticalpower: the forest decline and acidic deposition example.Ecology 71, 2024–2027 (1990).

37. W. K. de la Mare, On the power of catch per unit effortseries to detect declines in whale stocks. Rep. Int. WhalingComm. 34, 655–662 (1984).

38. P. Kraufvelin, Model ecosystem replicability challenged bythe ‘soft’ reality of a hard bottom mesocosm. J. Exp. Mar.Biol. Ecol. 222, 247–267 (1998).

39. W. G. Landis, R. A. Matthews and G. B. Matthews, Thelayered and historical nature of ecological systems andthe risk assessment of pesticides. Environ. Toxicol. Chem.15, 432–440 (1996).

40. R. A. Matthews, W. G. Landis and G. B. Matthews, Com-munity conditioning: an ecological approach to environmen-


tal toxicology. Environ. Toxicol. Chem. 15, 597–603 (1996).41. J. Hermens, H. Canton, N. Steyger and R. Wegman, Joint

effects of a mixture of 14 chemicals on mortality andinhibition of reproduction of Daphnia magna. Aquat.Toxicol. 5, 315–322 (1984).

42. L. S. McCarty and D. Mackay, Enhancing ecotoxicologicalmodeling and assessment: body residues and modes oftoxic action. Environ. Sci. Technol. 27, 1719–1728 (1993).

43. T. W. Sawyer and S. Safe, In vitro AHH induction bypolychlorinated biphenyl and dibenzofuran mixtures: addi-tive effects. Chemosphere 14, 79–84 (1985).

44. S. J. Broderius, M. D. Kahl and M. D. Hoglund, Use ofjoint toxic response to define the primary mode of toxicaction for diverse industrial organic chemicals. Environ.Toxicol. Chem. 9, 1591–1605 (1995).

ecological risk assessment: implications of hormesis

Documents