Bringing policy relevance and scientific discipline toenvironmental risk assessment for geneticallymodified crops

Rod A. Herman1, Monica Garcia-Alonso2, Raymond Layton3, and Alan Raybould4

1 Dow AgroSciences LLC, 9330 Zionsville Road, Indianapolis, IN 46268, USA2 Estel Consult Ltd., Bracknell, Berkshire RG42 4HG, UK3 DuPont Pioneer, 7300 NW 62nd Avenue, Johnston, IA 50131, USA4 Syngenta, Jealott’s Hill International Research Centre, Bracknell, Berkshire RG42 6EY, UK

Forum: Science & Society

Although public opinion is important in deciding what isvalued by society, governments have determined thatscientific expertise is required to evaluate potential en-vironmental effects of genetically modified (GM) crops.We suggest how to evaluate rigorously the environmen-tal effects of GM crops in the context of a scientificinvestigation. Following a disciplined scientific approachto environmental risk assessment (ERA) for GM cropsshould help resolve controversy in identifying andaddressing risk.

Regulatory policy versus technology evaluationGovernment regulation of GM crops (or other agriculturaltechnologies) is designed to serve society. As such, societalconcerns and values are captured in regulatory policy.Protection of the environment is one objective of such policy.Although public opinion is important in determining what isvalued by society (i.e., what should be protected), regulatorswith scientific expertise evaluate specific technologies fortheir effects on the environment in the context of what isvalued and is to be protected. Governments have universallydetermined that scientific expertise is required to evaluatethe potential for environmental effects of new agriculturaltechnologies such as GM crops, and that public opinion for oragainst a specific technology is not necessarily a good mea-sure for what will benefit society. The majority of citizens donot have the training, time, or expertise to sort out facts frommisinformation, so dedicated teams of trained regulatoryscientists conduct these evaluations to ensure that thetechnologies that are brought forward to address the chal-lenges of agricultural production do not pose unreasonablerisk [1]. Here, we suggest how to evaluate rigorously theenvironmental effects of GM crops in the context of a scien-tific investigation.

Identifying risk and protection goalsProtection goals

A key element of an ecological or ERA for agriculturalpractices is deciding what constitutes a harmful effect. InERA jargon, this is often predicated on identifying protec-tion goals (e.g., which species or ecological functions are to

Corresponding author: Herman, R.A. (raherman@dow.com).

be protected). Although this concept may seem straight-forward and logical to many, this step in the ERA process isactually controversial in some circles [2]. An alternativeapproach is to measure an array of endpoints (typicallycounts of individuals within species that show up) in fieldstudies, and then decide if results support environmentalsafety. Here, we discuss the former approach to ERA forGM crops based on identified protection goals. This ap-proach follows the traditional scientific method of hypoth-esis formulation followed by experimental testing.

What to protect?

Whether we choose to protect certain species or ecologicalfunctions depends on our values. When our activities affectonly ourselves as individuals, our own values may be thesole determinant of what should be protected. When effectsof activities are spread widely, we require a collective set ofvalues. For the risks of agricultural practices, collectivevalues may be expressed as the objectives of agriculturaland environmental policy, and these objectives are thebasis of protection goals for regulatory risk assessments.In identifying regulatory protection goals, the intentions ofrelevant policies must be interpreted [3].

The objectives of policy are often broad; for example,protecting biodiversity or increasing agricultural production.From these broad objectives, specific attributes of the envi-ronment that will be the focus of the risk assessment must beidentified. These attributes might be the abundance of par-ticular species or the provision of an ecological function, suchas pollination. Deriving unambiguous and measurable envi-ronmental attributes from broad policy objectives is calledmaking the protection goals operational (Box 1).

To begin derivation of operational protection goals,organisms may be grouped into beneficial species and char-ismatic species. Beneficial species often include biocontrolagents that are valuable in controlling agricultural pests(e.g., predatory insects such as lady beetles). Beneficialspecies may also include species that serve more thanagriculture, such as bats that eat biting mosquitoes. Char-ismatic species may include iconic species such as monarchbutterflies, which provide beauty to the environment. Inaddition to individual species, goals may involve protectingenvironmental function such as soil fertility, where benefi-cial microbes and earthworms might be important. Finally,

493

Box 1. Making protection goals operational

Operational protection goals: A key step in bringing effective

scientific discipline to ecological risk assessment is the development

of operational protection goals. This is done by translating a loosely

defined policy protection goal such as ‘protect biodiversity’ into

relevant and measurable outcomes. There are three steps in the

translation: establishing value, understanding context, and defining

the endpoint. An example of a value is an important biological

function where disruption would cause long-term harm. The context

describes the protection boundaries. Spatial context defines the

area to be protected – within a field, a field edge, or a region.

Temporal context defines the time element for protection – a short

interval within a growing season versus effects lasting beyond a

growing season. Taxonomic or functional boundaries define the

assessment level – populations, communities, or ecosystems. Once

the value and context are defined, then finally the endpoint can be

linked to the value and defined as a hypothesis: a species population

critical to providing a valued function will not be adversely affected

in a certain way over a specified space and time.

Endpoint selection: High quality regulatory decisions can only be

made when hypotheses are constructed such that experiments can

be designed to measure changes in appropriate endpoints.

Useful questions to ask when selecting endpoints include:

(i) Can it be measured and is the endpoint useful in making

decisions regarding harm? Not all things that can be measured

and tested statistically are of equal value in determining the

potential for biological significance that might indicate the

potential for future harm.

(ii) Is what is being measured representative of the temporal and

spatial area requiring protection? A short-term depression in a

population of a globally distributed highly mobile insect with

short recovery time within a single agricultural field would have

a lower level of biological significance compared with a regional

population decrease of the same magnitude in a longer-lived

organism restricted to a small geographic area.

(iii) Does the endpoint represent a valued function in the ecosystem

(or in the case of charismatic species, a value to society)?

Although all species have value, most ecosystem functions may

be carried out by numerous different species. It is highly unlikely

that decreases in populations of a single species within a field will

have any biologically significant effect on ecosystem function;

especially when viewed within the broader context of the

potential effects of agricultural practices on the environment.

Forum: Science & Society Trends in Biotechnology September 2013, Vol. 31, No. 9

species richness or biodiversity may be a protection goal inwhich the number of species or taxa rather than the numberof individuals or their environmental function may be thefocus [4]. Biodiversity is often seen as a measure of environ-mental stability and as a natural resource, and may also beof aesthetic value [5].

It is important to distinguish the aforementioned oper-ational protection goals from vague goals like protectingthe environment that often lead to experiments withoutclear hypotheses addressing the likelihood of harmfuleffects, resulting in data with dubious value in an ERA.One should be able to identify with some reasonable level ofspecificity what is desirable to protect and why, beforeexperiments intended to support risk assessment are con-ducted and data are collected. In this way, interpretation ofthe data can be directed toward evaluating the impact ofany observed effect on what is desirable to protect. Withoutthis scientific discipline, experiments are descriptive rath-er than tests of hypotheses relevant to risk.

Where do the effects occur?

An additional aspect of defining protection goals is aconsideration of whether effects are spatially and

494

temporally restricted to the agricultural field where theGM crop is grown, or whether the effects extend beyond thefield margins and cropping cycle. A localized effect mightbe an impact on a population of relatively immobile pred-atory insects, which is reduced due to an insecticidal traitthat significantly reduces a pest that is a primary food itemfor the biocontrol agent. The former is an indirect effect,but it is also possible for an insecticidal trait to have adirect adverse effect on a beneficial species. For example, ifan insecticidal trait is directly toxic to a beneficial insectspecies and it is present in pollen, species that eat pollen(e.g., some lady beetle species) could be directly affected.Direct and indirect effects beyond field margins are alsopossible. If an herbicide tolerance trait allows farmers toobtain higher levels of weed control in their crop comparedwith other methods, less weed seed may be available forspecies that eat this seed (e.g., some birds). If the greaterlandscape cannot adequately support populations of thesehighly mobile bird species, then the agricultural practicewill adversely affect populations [6]. As with in-fieldeffects, it is also possible for there to be direct toxicityto highly mobile species and such effects should also beconsidered. It is noteworthy that landscape level benefitsfrom GM crops are also possible such as reduced soilerosion through greater adoption of conservation tillageusing herbicide-tolerant crops, and increased populationsof beneficial insects, on a landscape scale, whereinsect-tolerant crops have replaced broad-spectruminsecticides [4,7].

Addressing protection goalsIn-field effects

Farmers have experience balancing in-field direct effectsof agricultural practices such as the use of broad-spec-trum pesticides which may kill both pests and beneficialarthropods. For example, use of pyrethroid insecticidesto manage pests is known to increase pest mite popula-tions in apple orchards due to predator reduction [8].Farmers need to balance this negative effect with thebenefits of using these products. If the effects of anagricultural practice are restricted to the farmer’s field,it would seem that market pressures will address thecost–benefit ratio of the practice. Of course, farmersshould be made aware of any known downsides to theuse of a product as well as its benefits, but if they areaware of the trade-offs, it would seem that the farmerswill incur the benefits and risks, and thus the decision touse the technology, or not, should rest with them. If theydo not see a net benefit, they will not continue to use thetechnology. A similar scenario occurs when populationsof relatively immobile predators are reduced due to fewerpests in a field on which the predators can feed. In thiscase, the ecological function of biological pest controlmay be reduced and the farmer will need to weigh thisloss of function against the benefit of the reduced pestpopulation due to use of an insecticidal trait. Ultimately,this will be driven by costs of production. GM crops arenot unique in causing potential ecological effects, soexisting literature should be useful in understandingthe spatial localization of many effects that could becaused by these crops.

Forum: Science & Society Trends in Biotechnology September 2013, Vol. 31, No. 9

Direct landscape effects

For effects that extend outside of the field, the risks (andbenefits) may be shared by others besides the farmer usingthe practice, so government regulation is often used toaddress these risks equitably. Direct effects, such as tox-icity to a highly mobile and beneficial species (e.g., polli-nating honey bees), represents a straightforward example.Again, a pesticidal case provides a clear path to a solutionfor addressing the concerns of multiple stakeholders. Forbroad-spectrum insecticides toxic to honey bees, applica-tions may be restricted during the bloom period (when beesare attracted to the crop), thus limiting exposure [9].Similarly, programs are often in place to lessen soil erosionand fertilizer run-off that could affect waterways (e.g.,sediment control practices) [10].

Indirect landscape effects

Actions to address indirect effects that extend outside of afarmer’s field are less universally accepted or applied. Aclassic example is the case of herbicide-tolerant crops thatoften allow high levels of weed control within agriculturalfields (analogous to using highly effective herbicides onnon-GM crops). This can adversely affect populations ofbirds that feed on weed seeds [4,7]. Two approaches havebeen adopted by different government regulatory agencies.In the US, farmers are given incentives to set aside landspecifically for wildlife, whereas in some European coun-tries, farmers may be restricted from using highly effectiveweed control measures so that some weeds will grow intheir agricultural fields [11]. Although one can advocateeither approach in the absence of objective data, thiscontroversy can be informed by clearly defining protectiongoals and designing experiments to determine the efficien-cy of each approach. Based on the crop, environment, andprotection goals, it should be straightforward to compareexperimentally the loss of agricultural production efficien-cy due to inferior weed control within agricultural produc-tion areas with the production loss due to setting asidededicated wildlife land that provides the same level ofbenefit to the environment. As such, alternativeapproaches with equivalent benefits could be presentedas options for mitigating adverse effects.

Agricultural context

An additional concept that is important to consider whenaddressing protection goals is the context of agriculture[4]. GM crops represent one more tool that farmers haveat their disposal to increase productivity and efficiency.For the majority of existing GM crops, a long history ofuse with alternative technologies to achieve the samebenefits exists. For example, GM herbicide-tolerant cropssimply expand the herbicide options previously availablebased on the endogenous tolerance of crops to certainherbicides compared with many weeds. In the area ofinsect-tolerant crops, pesticides (as well as insect-toler-ant crops developed by traditional breeding techniques)have been used for many decades to combat insect pests.There is a long history of developing operational protec-tion goals for these alternative tools (although somealternative tools are not regulated such as pest-resistantand herbicide-tolerant non-GM varieties), and these

should serve as a template for evaluating new tools toachieve the same ends. Likewise, the risk–benefit ratio ofthese existing tools should serve as a benchmark forassessing the safety of new tools that serve the samepurpose. If new standards for protection are to be imple-mented for GM crops, a clear rationale for their develop-ment and implementation should be articulated based onrational scientific principles grounded in the context ofexisting agricultural practices.

Concluding remarksA balance between risks and benefits should be obtainedwhenever evaluating the environmental risk of a newtechnology, and GM crops are no exception. We advocateallowing farmers to address the environmental risks andbenefits of agricultural practices for protection goals thatare largely restricted to their fields such as maintainingadequate soil fertility and populations of in-field beneficialinsects. In these cases, farmers are the main stakeholdersand production efficiency should dictate the best manage-ment of these practices. For protection goals that extendbeyond field margins, and for which multiple stakeholdersexist, government regulation may be a good approach. Foragricultural practices that directly affect protection goals,straightforward approaches are often available. However,for indirect effects on protection goals, differing approachesare common and are often based on philosophical percep-tions rather than objective science. In these cases, weadvocate clear identification of protection goals followedby hypothesis-driven experimentation to inform the pre-ferred approach (or approaches) to most efficiently meetingthese protection goals while maximizing agricultural pro-ductivity. Finally, existing agricultural practices shouldserve as a model for formulating operational protectiongoals and assessing the risks and benefits of GM crops.Following a disciplined scientific approach to ERA for GMcrops should help resolve controversy in identifying andaddressing risk. Through rigorous application of the triedand true scientific method, the risks and benefits of specificGM crops can be objectively evaluated and equitable regu-lation can be implemented.

Disclaimer statementR.H., R.L., and A.R. are employed by companies thatdevelop and market transgenic seed. M.G.-A. is an inde-pendent consultant and has clients that develop and mar-ket transgenic seed.

References1 Romeis, J. et al. (2013) When bad science makes good headlines: Bt

maize and regulatory bans. Nat. Biotechnol. 31, 386–3872 Herman, R.A. and Raybould, A. (2013) Invoking ideology in the

promotion of ecological risk assessment for GM crops. TrendsBiotechnol. 31, 217–218

3 Sanvido, O. et al. (2012) Evaluating environmental risks of geneticallymodified crops: ecological harm criteria for regulatory decision-making.Environ. Sci. Policy 15, 82–91

4 Carpenter, J.E. (2011) Impact of GM crops on biodiversity. GM Crops 2,7–23

5 Mace, G.M. et al. (2012) Biodiversity and ecosystem services: amultilayered relationship. Trends Ecol. Evol. 27, 19–26

6 Watkinson, A.R. et al. (2000) Predictions of biodiversity response togenetically modified herbicide-tolerant crops. Science 289, 1554–1557

495

Forum: Science & Society Trends in Biotechnology September 2013, Vol. 31, No. 9

7 Lu, Y. et al. (2012) Widespread adoption of Bt cotton and insecticidedecrease promotes biocontrol services. Nature 487, 362–365

8 Hall, F.R. (1979) Effects of synthetic pyrethroids on major insect andmite pests of apple. J. Econ. Entomol. 72, 441–446

9 Riedl, H. et al. (2006) How to reduce bee poisoning from pesticides.Extension publication, Oregon State University (http://scholarsarchive.library.oregonstate.edu/xmlui/bitstream/handle/1957/20772/pnw591.pdf?sequence=1)

496

10 Rao, M. et al. (2007) A web-based GIS Decision Support System formanaging and planning USDA’s Conservation Reserve Program(CRP). Environ. Model. Software 22, 1270–1280

11 Herman, R.A. (2010) Ecological risk assessment for transgenic crops:separating the seed from the chaff. Trends Biotechnol. 28, 159–160

0167-7799/$ – see front matter � 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.tibtech.2013.05.012 Trends in Biotechnology, September 2013,

Vol. 31, No. 9