SEARCH

SEARCH BY CITATION

Keywords:

  • transgenic plants;
  • natural enemies;
  • predators;
  • parasitoids;
  • laboratory tests;
  • life history;
  • stress factors;
  • genetic modification

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

We reviewed laboratory tests which studied the impact of genetically modified plants on arthropod natural enemies. A total of 18 species of predators and 14 species of parasitoids have been tested, most in only a few experiments. Certain groups (braconid wasps) or species (the green lacewing, Chrysoperla carnea) have attracted much effort, while representatives of others, including whole orders (e.g., Diptera), have never had a species tested. We conclude that laboratory tests are not the ‘worst case’ scenarios intended by the experimental designs, and are not often ecologically realistic: they typically provided ad libitum feeding, no prey choice, single prey type, no combination of stress factors and usually uniform temperatures. None of these are representative of field conditions, yet most could be easily mimicked in more complex laboratory tests. In most cases (94.6%), the studies were unable to indicate the level of power required to detect any impact. Small sample size and large variability are factors that mask all but very large differences in potential effects. For predators, 126 parameters were quantified, most commonly including survival/mortality (37 cases), development time (22), and body mass/size (20). For parasitoids, 128 parameters were quantified, the majority involving lectins or proteinase inhibitors. Most frequent measurements were: fecundity (23 experiments), adult longevity, extent of parasitism (17 each), body size, mortality, and larval development time. An aggregative scoring (summarising all quantified parameters) indicated that the laboratory tests quantified a remarkable number of cases (30% for predators, 39.8% for parasitoids), where the impacts of the genetically modified plant were significantly negative. These involve various parameters, organisms, test methods, and significance levels, but collectively they indicate that the use of genetically modified crops may result in negative effects on the natural enemies of crop pests.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

Agriculture is an important environmental quality driver (Hails, 2002), and its effect is not likely to diminish in the future (Tilman et al., 2001). Furthermore, it has been realised that life on Earth depends on the proper functioning of several large-scale ecological processes, many of which provide humanity with irreplaceable benefits, termed ‘ecosystem services’ (Daily et al., 1996). We know that loss of biodiversity threatens these benefits, but exactly what type and amount of biodiversity is necessary for unimpaired, sustained ecological functioning (and productivity) is unknown: the diversity–productivity relationship is currently hotly debated in ecology (Loreau et al., 2002). Considering these future scenarios and uncertainties, agricultural innovations are increasingly scrutinised for their potential environmental impacts (Hails, 2002). This also holds for the growing of genetically modified (GM) crop plants, a recent but fast-spreading (Shelton et al., 2002) agricultural innovation.

The first production of transgenic plants resistant to insects offered expectations as a means of pest control that could lead to a reduction in pesticide use in intensive cropping systems. On the other hand, claims of potential negative impacts were made, and the growing of transgenic crops was therefore linked to regulations and limitations, both before and after release (see Conner et al., 2003). The type of risk assessment, the formulation of the requirements, and the necessary level and complexity of the risk assessment procedure is variable, ‘in flux’, and often hotly debated.

The scope of the current review is to systematically and critically survey the published, peer-reviewed literature regarding the impact of GM plants on natural enemies. We restricted our attention to arthropod natural enemies tested in laboratory studies (there are virtually no studies on nonarthropod natural enemies of pests or weeds). This does not include any judgement about the importance of such studies, nor of the significance of these impacts relative to impacts from gene flow, resistance evolution, or socio-economic effects. We contend, however, that natural pest control is an important enough ecosystem service that the effects of any GM plant should be tested before the field release of such a plant. As a consequence, a stock-taking of our knowledge in this area 10 years after the start of the commercial growing of GM plants should be useful. We intend to devote special attention to the ecological realism of these laboratory studies, including the selection of species, study conditions, study duration, and statistical/experimental design.

Selection of studies

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

Previous reviews of empirical research have been varied, some only considering the ecological impacts of narrower or broader sense; others (Dale et al., 2002; Conner et al., 2003) aiming to be wider, also including economic, legal, and ethical aspects. The reviews sometimes include a significant portion of non-peer reviewed sources (e.g., Fontes et al., 2002), or even documents available on the Internet. In our view, the inclusion of non-peer reviewed results in this highly charged field is to the detriment of all involved. The advantage of accessing such studies before, rather than after the time lag that the peer-reviewed literature normally imposes on publication, does not counteract the peril of lacking or improper quality assurance.

Based on a review of these studies, it is evident that if generalisations were to be made under our current state of knowledge, they would risk outrunning the available data. No published laboratory study exists on major predatory groups such as predatory flies and staphylinids, polyphagous predators such as spiders, or social predators such as wasps and ants. We suggest feasible ways to improve the ecological realism of laboratory tests to check the impact of GM crops, and urge the consideration of elementary behavioural and ecological factors to create ecologically realistic ‘worst case scenarios’ that more closely approximate the most likely situation.

Evaluation methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

For laboratory tests, we surveyed and tabulated the arthropod species studied, their taxonomic affiliation, the test plant, the genetic modification, experimental conditions, number of true replications, the duration of the study, the inclusion of behavioural aspects, and the characters quantified. We sorted these characters into five classes, showing significant negative (P<0.05), non-significant negative (0.49>P> 0.05), neutral (P>0.5), non-significant positive, or significant positive effects. To arrive at a rough overall assessment, a ‘bean-counting algorithm’ was used. This simply summarised the number of cases (characters quantified) in each of the above ‘effect’ classes. Consequently, studies that quantified more parameters had a higher impact on the final scores.

Statistical analyses

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

Only two of the 14 reviews published since 2000 discussed experimental design and data analysis. Marvier (2002), using a few field release applications prepared by the biotechnology industry, illustrated serious shortcomings concerning sample size, statistical power, and the duration of experiments. Statistical power analyses can enhance research planning, as well as clarifying the interpretation of the results. For risk-related research, type II errors may become more important than type I errors (Hill & Sendashonga, 2003), thus a power analysis is especially useful. The magnitudes of type I and type II errors are not independent. The relationship between the two types of errors can be simply expressed as a decreasing function: as α[RIGHTWARDS ARROW]0, β will grow for 0<α<1, meaning that the tendency to erroneously reject the null hypothesis will increase. It is common statistical practice to minimize type I errors. This is, despite the repeated protests of statisticians (e.g., Sokal & Rohlf, 1995), often automatically set by most statistical packages at values as low as P = 0.05. Therefore, if a type II error is not explicitly calculated via a power analysis (e.g., Steidl et al., 1997), the researcher has no control over this type of error. This very important point was rarely considered in the description of the experimental design in the research papers considered in the present review. Steidl et al. (1997) contend that the use of a retrospective power analysis calculated with experimental data is meaningless, as it yields no information other than that provided by the original hypothesis. Some authors (e.g., Lundgren & Wiedenmann, 2002; Romeis et al., 2004) published retrospective power analyses. This, however, is incorrect, as it is affected by the interdependence of type I and II errors (Hoenig & Heisley, 2001) and the chosen type I error rate automatically determines the power of the experiment. For this reason, the new specialist journal, Environmental Biosafety Research, does not accept the use of retrospective power analyses (Andow, 2003). An alternative use of a retrospective power analysis for experiments where the null hypothesis was not rejected, is in the calculation of effect sizes other than the observed effect size. This was chosen, for example, by Romeis et al. (2004), who tested detectable effects at a level of 20%. However, these results are to be taken cautiously, as they only respond to a question on a single hypothetical value and cannot answer broad questions about the validity of the null hypothesis. Pseudoreplication is another problematic statistical phenomenon that also occurs in the GMO biosafety research. Examples of pseudoreplication in laboratory studies with Chrysoperla carnea are discussed in Andow & Hilbeck (2004).

Various test methods were sometimes used by different authors to test effects on the same parameter (e.g., mortality). It might be useful to restate here the main difference between non-parametric and parametric tests. Parametric tests, such as least square differences or Student's t-test, require that the data have a ‘normal distribution’ defined by the parameters of their means and variance. Therefore, they represent a correct choice if two major prerequisites are satisfied: (1) the data are normally distributed and independent, and (2) their variance is homogeneous (homoschedasticity). In case of clearly different data structures, or even in cases of ambiguity, it is advisable to use non-parametric tests (Sokal & Rohlf, 1995). This was the choice in several papers evaluated in the present review. Nevertheless, parametric tests were sometimes used for measuring insect mortality (e.g., Couty & Poppy, 2001; Duan et al., 2002), and in a few cases it is not clear whether the assumption about normal distributions was properly considered (e.g., Sims, 1995).

The number of specimens tested varied widely among the studies (see Tables 1 and 3), with a greatest total of 240 lacewing larvae used by Hilbeck et al. (1999). The use of few replicates makes it less likely that an effect could be detected. Especially for parasitoids, the number of individuals used was sometimes extremely low (e.g., Schuler et al., 2003; Pruetz & Dettner, 2004). Increasing the sample size is the best means that scientists have to reduce the overall error rate of their experiments. The optimal sample size, however, can only be determined after a specific detection limit has been chosen (Marvier, 2002).

Table 1.  A summary of the laboratory test conditions and parameters measured on predatory natural enemies
Order/familySpeciesPlantaGM/ analoguen/testTemperatureFeeding type: single/ multiplePrey quantity: ad libitum/ limitedChoice yes/noDuration in days/no. generationsParameter measuredRef.
Variable?°C
Heteroptera
AnthocoridaeOrius tristicolorCottonBt 10–30No25Singlead lib.No1 gen.Longevity, mortality 1
O. tristicolorPotato/plantBt 20No25Singlead lib.No1–13 daysMortality 2
O. insidiosusMaize/pollenBt 39No26Singlead lib.No1 gen.Dev. time, mortality, adult mass 3
O. insidiosusDietBt 67No24Singlead lib.No23 daysMortality, mass, body size, plant feeding, mortality, dev. time 4
O. majusculusMaizeBt  3No25D:20NSinglead lib.No1 gen.Mortality, dev. time 5
LygaeidaeGeocoris puncticepsPotato/plantBt 20No25Singlead lib.No1–13 daysMortality 2
G. punctipesCottonBt 10–30No25Singlead lib.No1 gen.Longevity, mortality 1
G. pallensPotato/plantBt 20No25Singlead lib.No1–13 daysMortality 2
Lygus hesperusPotato/plantBt 20No25Singlead lib.No1–13 daysMortality 2
Zelus renardiiCottonBt 10–30No25Singlead lib.No1 gen.Longevity, mortality 1
MiridaeCyrtorhinus lividipennisRiceBt 80Yes28D:23NSinglead lib.No Mortality, dev. time 6
NabidaeNabis sp.CottonBt 10–30No25Singlead lib.No1 gen.Longevity, mortality 1
Nabis sp.Potato/plantBt 20No25Singlead lib.No1–13 daysMortality 2
PentatomidaePerillus bioculatusPotatoOCI PI 30–35Yes25D:15NSinglead lib.No10 daysEnzyme activity, mortality, moult/dev. time, mass, mass w/prey, survival w/prey, moulting w/prey 7
P. bioculatusPotatoOCI PI 60No??Singlead lib.No5 daysEnzyme activity 8
P. bioculatusDietOCI PI 12No22SinglelimitedNo28 daysMortality, consumption, egg laying, days to maturity, fecundity, egg mass, egg viability, satiation level, enzyme activity 9
Podisus maculiventrisSpiked preyGNA 30–40No20Singlead lib.No1 gen.Mass, larval growth, mortality, dev. time, egg prod., egg viability10
P. maculiventrisPotatoGNA/ CpTI 20–27No20Singlead lib.No1 gen.Adult mass female – male, dev. time female – male, fecundity, CpTI, fecundity, GNA, viability 10
Neuroptera
ChrysopidaeChrysoperla carneaMaizeBt 20No22–25Singlead lib.YesN/APreference, prey consumed, feeding time11
C. carneaMaizeBt 60No25Singlead lib.No1 gen.Mortality, dev. time, mass12
C. carneaMaizeBt200Yes25D:20NSinglead lib.No1 gen.Mortality, dev. time13
C. carneaDietBt150Yes25D:20NSinglead lib.No1 gen.Mortality, dev. time14
C. carneaDietBt240Yes25D:20NSinglead lib.No1 gen.Mortality, dev. time15
C. carneaMaizeBt 70No25Singlead lib.No1 gen.Mortality, dev. time16
C. carneaMaize/pollenBt 43No24Multiplead lib.No1 gen.Mortality, dev. time, adult mass3
Coleoptera
CarabidaeNebria brevicollisDietBPTI 97No23Singlead lib.No1 gen.Enzyme activity, mortality, mass, consumption17
Harpalus affinisDietBPTI 54No23Singlead lib.No1 gen.Consumption 24–48 h18
Lebia grandisPotatoBt  8–10No25Singlead lib.No3 daysPrey mass eaten,% prey offered eaten19
CoccinellidaeAdalia bipunctataPotatoGNA 12–24No20 (?)Singlead lib.No1 gen.Consumption, egg production, egg viability, adult longevity20
A. bipunctataDietGNA 25No21Singlead lib.No1 gen.Mortality, dev. time, larval mass, consumption21
A. bipunctataDiet/potatoGNA 25No21Singlead lib.No1 gen.Aphid consumption, L1 duration, L2 mass, L1 survival, larval survival, adult mass, no. egg/female, egg viability, early adult mortality, adult mortality22
Coleomegilla maculataMaize/pollenBt 30No27Singlead lib.No1 gen.Mortality, dev. time, adult mass, no. eggs laid, adult mortality23
C. maculataMaize/pollenBt 36No25Singlead lib.No2 gen.Mortality, dev. time, mass, adult mobility, adult survivorship, fecundity24
C. maculataMaize/pollenBt 45No24Singlead lib.No1 gen.Dev. time, mortality, adult mass3
C. maculataPotatoBt  4–15No26S/Mad lib.No1 gen.Prey consumption, mass, % prey offered, dev. time, adult mass, % mortality, % pupal mortality, adult fecundity25
Harmonia axyridisOilseed rapeOCI PI 30No20Singlead lib.No2 gen.Mortality, mass, prey consumption, enzyme activity, fecundity, viability, dev. time26
Table 3.  A summary of the laboratory test conditions and parameters performed on parasitoid natural enemies
Hymenopteran familySpeciesPlant/ artificial dietaGM/ analoguen/testTemperatureFeeding type: single/ multipleQuantity: ad lib./ limitedChoice: yes/noDuration: days/no. generationsParameter measuredRef.
Variable?°C
AphelinidaeAphelinus abdominalisDietGNA12No23Singlead lib.No1 gen.Fecundity, longevity, dev. success 1
A. abdominalisDietGNA6–15No23Singlead lib.No?% parasitism, dev. time, sex ratio, mass, resistance to starvation 2
A. abdominalisDietGNA10No23Singlead lib.No1 gen.Mass, longevity, fecundity, sex ratio 3
BraconidaeAphidius erviDietGNA15No23Singlead lib.No1 gen.Mass, dev. time, mortality, adult resistance to starvation, sex ratio 4
Aphidius nigripesPotatoBt Cry3A12/?Yes12–22SingleN/ANo2 gen.Mortality, host acceptance, dev. time, adult mass, fecundity, life span, sex ratio 5
A. nigripesPotatoOCI-PI12/?Yes12–22SingleN/ANo2 gen.Mortality, dev. time, size, fecundity, life span, sex ratio 5
A. colemanniSucroseGNA?/28–156No23Singlead lib.No1 gen.Host acceptance, time feeding, longevity, fecundity, emergence, sex ratio 6
Cotesia flavipesMaizeBt Event 1765No27Single?No1 gen.Mortality, no. par./host, par. mass (cocoon/host), food consumption by host 7
C. flavipesSugarcaneGNA2/8/100No22–25N/A or SinglelimitedYes1 gen.Behaviour, % parasitism, no. adult par., sex ratio 8
C. flavipesSugarcaneGNA?No30Singlelimited?No2 gen.Host acceptance, suitability, no. par/cocoon, emerging adult/host, dev. time, female longevity, egg load, sex ratio 9
C. glomerataSucroseGNA?–157No23Singlead lib.No1 gen.Behaviour, longevity, time feeding 6
C. marginiventrisCottonBt Event 53140No27Single?No2 gen.Mortality, dev. time, longevity, size, sex ratio, fecundity10
C. plutellaeOilseed rapeBt4No26Singlead lib.Yes1 gen.Choice, % parasitism, emergence11
Parallorhogas pyralophagusMaizeBt59No25Singlead lib.No2 gen.Mortality, adult longevity, dev. time, brood size, egg load, size, sex ratio12
P. pyralophagusSugarcane/ dietGNA40–80No25?Singlead lib.No/Yes1 gen.% parasitism, host selection13
P. pyralophagusSugarcane/ dietGNA20–46No25Singlead lib.No2 gen.Adult longevity gen. 1, gen. 2, size, fecundity, gen. 1, gen. 2, dev. time, mortality14
TrichogrammatidaeTrichogramma brassicaeSucroseGNA?/30–160No25Singlead lib.No1 genBehaviour, time feeding, longevity, fecundity, emergence, sex ratio 6
EncyrtidaeCopidosoma floridanumCottonBt Event 531?No27Single?No1 gen.Development, pupal mass, size, % parasitism10
EulophidaeE. pennicornisDiet/ potatoCpTI20No20–25Singlead lib.No1 gen.% parasitism, no. offspring, no. par. host, mortality, sex ratio, longevity, egg load, size15
E. pennicornisPotatoGNA2Yes15–30SinglelimitedNo17 days/F1% parasitism, no. pupae/host, no. adult/host, sex ratio, emergence success, size, longevity, fertility, viability16
E. pennicornisDietGNA/ CpTI/ConA30No25Singlead lib.No1 gen.Longevity, fecundity, potential fecundity17
E. pennicornisDietGNA6–20No25Singlead lib.No2 gen.% hosts parasitised, no. par./host, dev. time, egg load, female size, longevity, egg viability18

A final point worth considering is the acceptable level of baseline mortality in the control treatments of laboratory experiments. In any experiment, a certain level of mortality, caused by the artificial laboratory conditions, is unavoidable. Such mortality in the experimental studies we reviewed was quite variable. A mortality rate of around 15–20% (e.g., Zwahlen et al., 2000; Burgess et al., 2002) can be considered acceptable, while several experiments reported a much higher mortality in the control groups (e.g., Down et al., 2003). Differences can even be found when the same biological system was used. In laboratory experiments with larvae of the lacewing Chrysoperla carnea, first instar mortality when feeding on Spodoptera littoralis from control plants varied between 16% (Hilbeck et al., 1998a) and 27% (Dutton et al., 2003). Apart from the biological significance of possible suboptimal experimental conditions, a high level of mortality in the control groups could affect the outcome of the statistical analyses and mask a significant effect. In laboratory feeding tests, restricting the pre-test variability in size or mass in the experimental populations could be considered, in order to improve the powers of the procedure to detect treatment effects.

Laboratory tests on predators

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

The 26 published papers considered in this section included 35 experiments on 18 species in three insect orders (Table 1). Most of these (17 studies) were on Heteroptera (involving 11 species), seven on Neuroptera (all on C. carnea), and 11 on beetles (eight on Coccinellidae, three on Carabidae, involving three species each in the two families). Chrysoperla carnea (of which only the larvae are predatory) is the most often studied predator species. Apart from this, more than a single study was done on only a handful of species: two coccinellids (Adalia bipunctata, and Coleomegilla maculata, three and four studies, respectively) and a few heteropterans (Perillus bioculatus– three studies, Geocoris puncticeps, Orius tristicolor, O. insidiosus – two each). Typically, a laboratory test on predators was of short duration, performed at a constant temperature, with unlimited access to a single type of prey, under a no-choice feeding regime (Table 1). The majority of studies (18) involved Bt-toxin, either in artificial diet, or in GM plants. Six of the Bt-related studies involved plant- or (mostly) pollen-feeding species.

A total of 126 parameters were quantified in the laboratory tests. Most of these were connected to development, general biology, or fitness, which we classified into nine major groups (Table 2). Most commonly, survival/mortality (37), development time (22), and body mass/size (20) were measured. Surprisingly, prey consumption was measured in only 13 cases. Reproduction-related measurements were taken in 12 cases.

Table 2.  The relative distribution of reaction classes, from significantlya negative to significantly positive, of the different parameters quantified in laboratory tests of GM plant impacts on predatory insects
ParameterRelative no. of cases (%)bTotal no. of tests
Negative significantNegative non-significantNeutralPositive non-significantPositive significant
  • a

    Significance level was set at P<0.05.

  • b

    The lines are percentage of cases of the number of tests that quantified the given parameter. Overall total refers to all the parameters in all tests.

Survival/mortality15.5185610 39
Development time281256 4 25
Body mass/size221748 9 4 23
Prey consumption3358 8 12
Reproduction42 842 8 12
Enzyme activity4466  9
Longevity6040  5
Egg viability8020  5
Behaviour5050  4
Total301147.5 6 6135

Employing a ‘bean-counting algorithm’ (see above) for these admittedly incomplete data, 135 assessments could be categorised (Table 2), with the majority (47.4%) showing no significant response (neutral). However, a positive effect was found in 16 cases (12% of total tests, half of them significant), whereas a negative impact was registered for 55 (41%) cases (30% of all tests were significantly, and 11% non-significantly negative). The relative distribution of the five classes per parameter was typically highest in the ‘neutral’ class, and was skewed towards the negative (Table 2). The characters with the lowest frequency in the ‘non-responsive’ class that can be considered the most sensitive were longevity, reproduction, and egg/progeny viability. Enzyme activity, while always sensitive, was an equivocal character: the activity of certain enzymes decreased, while that of others increased, both significantly (Table 2).

Laboratory tests on parasitoids

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

We analysed 18 papers published between 1999 and 2004. All except one of these were studies on Hymenoptera, involving 14 species, mostly Braconidae (13 studies on eight species). The genus Cotesia was the most frequently studied, with C. flavipes in three studies, and three other species in one study each (Table 3). Four studies were done on Eulophus pennicornis (Eulophidae), and three on Aphelinus abdominalis (Aphelinidae). The only non-hymenopteran involved was Lydella thomsoni, a parasitoid fly of the European corn borer, Ostrinia nubilalis (Manachini, 2003). However, this last-mentioned work relied on studying parasitoids from field-collected hosts and thus we did not include it in the detailed analysis. The larger number of studies on a small number of species was probably due to experimental difficulties, and the more specific host requirements of parasitoids vs. predators. Interestingly, the plants used in the tests were rarely commercially available varieties: only four studies (three using maize and one cotton) used commercially available plant lines. The use of plants with proteinase inhibitors and lectins (13 studies), often using artificial diets instead of plants (11 studies, Table 2), differed from the predator tests, in which Bt-plants or the Bt-toxin were mostly used.

A total of 128 parameters were quantified, 31 on Bt-plants and 97 on plants/diets containing lectins or protease inhibitors. More than 15 different reference variables were used, which we classified into nine classes. Most frequently, some measure of fecundity (23 experiments) was quantified, followed by adult longevity, extent of parasitism (17 each), body size, mortality, and larval development time (Table 4). Host acceptance or behavioural characteristics were also studied (Table 4), but choice was not frequently investigated. Tests were mostly done under a constant temperature; only two studies (Ashouri et al., 2001; Bell et al., 2001) used variable temperatures.

Table 4.  The relative distribution of reaction classes, from significantlya negative to significantly positive, of the different parameters quantified in laboratory tests of GM plant (or GM product in artificial diet) impacts on parasitoids
ParameterRelative no. of cases (%)bTotal no. of tests
Negative significantNegative non-significantNeutralPositive non-significantPositive significant
  • a

    Significance level was set at P<0.05.

  • b

    The lines are percentage of cases of the number of tests that quantified the given parameter. Overall total refers to all the parameters in all tests.

Fecundity481330 4 4 23
Longevity59 6121212 17
% parasitism/no. parasitoids emerged35182418 6 17
Body mass/size331347 7 15
Host acceptance, feeding, diff. behaviours276013 15
Mortality/emergence5043 7 14
Development time503317 12
Viability6633  3
Sex ratio173342 8 12
Overall total39.814.833.6 7.0 5.5128

The majority of parameters in both plant classes (Bt-plants and other GM plants) showed significant negative impacts (57% and 32% for Bt-plants and other GM plants, respectively). The next largest group was that of the nonaffected (neutral) parameters (27% and 35%, respectively), followed by the non-significant negative impacts (13% and 22%, respectively). Overall, 54.6% of the parameters examined indicated a significant (39.8%) or non-significant (14.8%) negative impact of the examined transgenic plant or trait on the organism studied. Only 12.5% of the parameters showed a positive response (Table 4).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

Species selection vs. taxonomic composition of natural enemies

Van Driesche & Bellows (1996) listed the main groups of arthropod natural enemies. Among parasitic/parasitoid groups, there are 12 families of Diptera, the most important being the Tachinidae. Hymenoptera is a hyperdiverse order, with eight superfamilies and >36 families containing parasitoid species, among them economically important families such as Encyrtidae, Aphelinidae, Trichogrammatidae, Ichneumonidae, and Braconidae (Godfray, 1994). There are nine orders of predators, with Hemiptera, Neuroptera, Coleoptera, Diptera, and Hymenoptera being the most numerous and important. Among the arachnids, mites (27 families) and spiders (60 families) contain predators (van Driesche & Bellows, 1996).

Against this background, the list of families and species on which laboratory tests have been performed (Tables 1 and 3) is limited. Several important groups of natural enemies have hardly (Diptera), or never (ants, Formicidae) been tested. The same organism in different countries on different plants or in different situations has rarely been examined. Some groups or even species have been studied frequently, while on others hardly any reports can be found. In the published tests, Neuroptera, represented by a single species – the green lacewing, Chrysoperla carnea (note that the species identity is not always ascertained, see Henry et al., 2002) – makes up nearly 20% of all predator tests. Hemiptera (11 species) make up close to 50% of species on which tests have been published. There have been very few studies on beetles (but see Jørgensen & Lövei, 1999; Burgess et al., 2002). There are virtually no published tests on very important groups such as parasitoid flies (but see Manachini, 2003), and several wasp families. Among predators, the lack of work on spiders, social predators, and flies is the most striking omission. Reasons to choose particular species are often not well substantiated, and seem governed by opportunity, ease of access, or ease of culturing rather than the systematic screening of several potential candidates, even though this has repeatedly been suggested (e.g., Cowgill & Atkinson, 2003).

Test conditions

Worst case scenarios have been suggested for laboratory testing (e.g., Dutton et al., 2003), where the organisms are exposed to more extreme conditions than they would encounter under field conditions, i.e., in ‘reality’. At the same time, in order to show some relevance to the real world, such tests should also be ecologically realistic (Lövei, 2001). For example, early tests using the green lacewing, C. carnea (Hilbeck et al., 1998a,b, 1999), have been criticised for their choice of prey organism, i.e., for lack of ecological realism (Hilbeck et al., 1998a; Dutton et al., 2002). However, published laboratory experiments can be considered ‘worst case scenarios’ only with respect to the candidate toxin concentrations applied. This is the strict consequence of a restricted, ecotoxicological view that has limited relevance to GM ecological impact assessment studies (Ervin et al., 2003; Andow & Hilbeck, 2004). Numerous studies on C. carnea (Hilbeck et al., 1998b; Romeis et al., 2004) and those on A. bipunctata (Birch et al., 1999; Down et al., 2003) reinforce that significant impacts cannot be fully explored by following an ecotoxicological conceptual framework. This is a serious shortcoming of a recently suggested testing approach for non-target impacts of GM crops (Dutton et al., 2003). For example, predators under field conditions experience variable temperatures, prey choice, several different prey types, food shortage, and predation risk. All these factors constitute variable stresses, and the impact of an additional potential stress (toxic or suboptimal prey, as a consequence of feeding on a GM plant) is modulated by them. A ‘worst case’ should include the important elements of such field conditions as they arguably contribute to a ‘less than optimal’ environment. Some of these conditions are difficult to simulate under controlled conditions, but several (variable temperature, prey shortage, or mixed feeding) are not, and we recommend that these should be included in the test conditions.

Selection of response parameters and the evaluation of their responses

A wide range of response parameters were employed for measuring the potential impacts of GM crops, but only a few have been used frequently. We can assume that all response parameters were selected because of a direct link to the fitness of the test organisms. Due to the restricted range of the organisms, we would strongly emphasise that generalisations should only be made with caution. What is not clear is how sensitive the selected parameters are to changes in conditions, and whether a lack of response (such as the majority of parameters quantified in predator tests) means a real lack of risk, or a lack of sensitivity. It is also extremely difficult to assess how important a certain parameter is for fitness. If two such parameters indicate opposite impacts, which one should be accepted? If no binary (accept/reject) decision is to be taken, how could they be reconciled? We feel that our current database is simply too sparse to provide us with a valid basis to answer this important question.

We have chosen to ‘score’ published tests. If several criteria were examined, all parameters were classified as +/– and an ‘aggregate score’ was calculated (see Evaluation methods). This approach has several tacit assumptions. These include that the selected species are equally important as natural enemies and all studied parameters have the same influence on fitness. This is clearly not the case. The simple ‘aggregating’ evaluation technique is just a first, very crude approximation. We consider this only a temporary and not a final nor a completely satisfactory technique. Nevertheless, the overall skew towards negative impacts (Tables 2 and 4) is a signal that we ought to consider seriously. The negative impacts are too numerous to just explain them away as non-significant or non-relevant. It should be remembered that all of them were selected because they were assumed to be fitness-related. We cannot retrospectively de-couple them from fitness, arguing that they are not ecologically important.

Physiological and behavioural parameters are probably more sensitive in showing impacts than parameters such as mortality or population growth rate. Another type of sensitive measure is the ‘integrative’ type, in which an accumulated impact is quantified. Progeny viability over a representative period of time can be considered such a case. Several processes in the mother must operate at (reasonably) normal levels and intensity to produce viable offspring. As a consequence, a number of conditions and their impact are ‘cumulatively expressed’ in such traits. At least in some situations, the inclusion of such parameters should be considered a standard requirement for the pre-release testing of non-target impacts of GM crops.

Immature natural enemy stages are usually more sensitive and have narrower tolerance limits than adults. For example, ground beetle larvae have weakly chitinised, predatory larvae that typically live in the soil (Lövei & Sunderland, 1996). They are more sensitive to moisture and feeding conditions than adults. Moreover, they often represent the overwhelming majority of individuals of a given species under field conditions. Methodological difficulties can pose practical problems for culturing immature stages and using them for biosafety tests, but in several situations they can and should be considered as test organisms and preferred over the less sensitive adults.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

In conclusion, we emphasise the need for a clear, consistent approach for selecting laboratory test species, which considers their ecological role in the agro-ecosystem including the GM crop (Birch et al., 2004). We emphatically stress that ecological communities have a structure, and that not all species are equally important (Lawton, 1992). There are neither theoretical nor practical grounds for demanding all-inclusive tests. Our current situation is analogous to the results presented by Malcolm (1992). This author measured the growth and mortality of the larvae of nine species of aphid predators raised on two host species, viz., an alkaloid-accumulating aphid (Aphis nerii) and a non-accumulating (Acyrtosiphon pisum) one, kept on the same host plant, Nerium oleander. Included were three species each of ladybirds (Coccinellidae), lacewings (Chrysopidae), and hover flies (Syrphidae). Malcolm (1992) obtained only three types of response: no effect, a partial effect, or a fatal, toxic effect. The taxonomic affinity of test organisms, however, did not predict the outcome. Likewise, we can expect a finite number of impact types of GM plants, but with our current level of knowledge we are unable to predict the impact on a selected natural enemy group or species without testing. Therefore, we infer that we are in great need of a larger body of empirical data. These should be systematically collected, including species in taxa not studied thus far. They should be tested under ecologically more realistic laboratory ‘worst case’ scenarios, choosing sensitive and reliably measurable response parameters over realistic time scales. We should consider that multiple stresses are the norm under field conditions, not the exception, that organisms often react in non-linear ways to combined stresses (Stamp et al., 1997), and we should at least attempt to mimic these conditions in laboratory tests. These altered practices, several of which are easily achievable, would hopefully improve our powers of prediction regarding the potential ecological impacts of growing GM crops.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References

This work was partially funded by a grant from D.J.F. to G.L.L. We thank Ms J.-Y.Guo for help with collecting material for the review and obtaining Chinese publications, and three anonymous reviewers, whose comments have improved the manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Selection of studies
  5. Evaluation methods
  6. Statistical analyses
  7. Laboratory tests on predators
  8. Laboratory tests on parasitoids
  9. Discussion
  10. Conclusions
  11. Acknowledgements
  12. References
  • Al Deeb MA, Wilde GE & Higgins RA (2001) No effect of Bacillus thuringiensis corn and Bacillus thuringiensis on the predator Orius insidiosus (Hemiptera: Anthocoridae). Environmental Entomology 30: 625629.
  • Andow DA (2003) Negative and positive data, statistical power, and confidence intervals. Environmental Biosafety Research 2: 7580.
  • Andow DA & Hilbeck A (2004) Science-based risk assessment for non-target effects of transgenic crops. Bioscience 54: 637649.
  • Armer CA, Berry RE & Kogan M (2000) Longevity of phytophagous heteropteran predators feeding on transgenic Bt-potato plants. Entomologia Experimentalis et Applicata 95: 329333.
  • Ashouri A, Michaud D & Cloutier C (2001) Recombinant and classically selected factors of potato plant resistance to the Colorado potato beetle, Leptinotarsa decemlineata, variously affect the potato aphid parasitoid Aphidius nigripes. Biocontrol 46: 401418.
  • Ashouri A, Overney S, Michaud D & Cloutier C (1998) Fitness and feeding are affected in the two-spotted stinkbug, Perillus bioculatus, by the cysteine proteinase inhibitor, oryzacystatin I. Archives of Insect Biochemistry and Physiology 38: 7483.
  • Baur ME & Boethel DJ (2003) Effect of Bt-cotton expressing Cry1A(c) on the survival and fecundity of two hymenopteran parasitoids (Braconidae, Encyrtidae) in the laboratory. Biological Control 26: 325332.
  • Bell HA, Down RE, Fitches EC, Edwards JP & Gatehouse AMR (2003) Impact of genetically modified potato expressing plant-derived insect resistance genes on the predatory bug Podisus maculiventris (Heteroptera: Pentatomidae). Biocontrol Science and Technology 13: 729741.
  • Bell HA, Fitches EC, Down RE, Ford L, Marris GC, Edwards JP, Gatehouse JA & Gatehouse AMR (2001a) Effect of dietary cowpea trypsin inhibitor (CpTI) on the growth and development of the tomato moth Lacanobia oleracea (Lepidoptera: Noctuidae) and on the success of the gregarious ectoparasitoid Eulophus pennicornis (Hymenoptera: Eulophidae). Pest Management Science 57: 5765.
  • Bell HA, Fitches EC, Down RE, Marris GC, Edwards JP, Gatehouse JA & Gatehouse AMR (1999) The effect of snowdrop lectin (GNA) delivered via artificial diet and transgenic plants on Eulophus pennicornis (Hymenoptera: Eulophidae), a parasitoid of the tomato moth Lacanobia oleracea (Lepidoptera: Noctuidae). Journal of Insect Physiology 45: 983991.
  • Bell HA, Fitches EC, Marris GC, Bell J, Edwards JP, Gatehouse JA & Gatehouse AMR (2001b) Transgenic GNA expressing potato plants augment the beneficial biocontrol of Lacanobia oleracea (Lepidoptera; Noctuidae) by the parasitoid Eulophus pennicornis (Hymenoptera; Eulophidae). Transgenic Research 10: 3542.
  • Bell HA, Kirkbride-Smith AE, Marris GC, Edwards JP & Gatehouse AMR (2004) Oral toxicity and impact on fecundity of three insecticidal proteins on the gregarious ectoparasitoid Eulophus pennicornis (Hymenoptera: Eulophidae). Agricultural and Forest Entomology 6: 215222.
  • Bernal CC, Aguda RM & Cohen MB (2002) Effect of rice lines transformed with Bacillus thuringiensis toxin genes on the brown planthopper and its predator Cyrtorhinus lividipennis. Entomologia Experimentalis et Applicata 102: 2128.
  • Birch ANE, Geoghegan IE, Majerus MEN, McNicol JW, Hackett CA, Gatehouse AMR & Gatehouse JA (1999) Tri-trophic interactions involving pest aphids, predatory 2-spot ladybirds and transgenic potatoes expressing snowdrop lectin for aphid resistance. Molecular Breeding 5: 7583.
  • Birch ANE, Wheatley R, Anyango B, Arpaia S, Capalbo D, Emana Getu D, Fontes E, Kalama P, Lelmen E, Lövei GL, Melo IS, Muyekho F, Ngi-Song A, Ochieno D, Ogwang J, Pitelli R, Schuler T, Sétamou M, Sithanantham S, Smith J, Son N-V, Songa J, Sujii E, Tan TQ, Wan F-H & Hilbeck A (2004) Biodiversity and non-target impacts: a case study of Bt maize in Kenya. Environmental Risk Assessment of Transgenic Organisms: a Case Study of Bt Maize in Kenya (ed. by AHilbeck & DAAndow), pp. 117185. CAB International, Wallingford, UK.
  • Bouchard E, Cloutier C & Michaud D (2003a) Oryzacystatin I expressed in transgenic potato induces digestive compensation in an insect natural predator via its herbivorous prey feeding on the plant. Molecular Ecology 12: 24392446.
  • Bouchard E, Michaud D & Cloutier C (2003b) Molecular interactions between an insect predator and its herbivore prey on transgenic potato expressing a cysteine proteinase inhibitor from rice. Molecular Ecology 12: 24292437.
  • Burgess EPJ, Lövei GL, Malone LA, Nielsen IW, Gatehouse HS & Christeller JT (2002) Prey-mediated effects of the protease inhibitor aprotinin on the predatory carabid beetle Nebria brevicollis. Journal of Insect Physiology 48: 10931101.
  • Conner AJ, Glare TR & Nap JP (2003) The release of genetically modified crops into the environment – Part II. Overview of ecological risk assessment. Plant Journal 33: 1946.
  • Couty A, Clark SJ & Poppy GM (2001c) Are fecundity and longevity of female Aphelinus abdominalis affected by development in GNA-dosed Macrosiphum euphorbiae? Physiological Entomology 26: 287293.
  • Couty A, De La Vina G, Clark SJ, Kaiser L, Pham-Delegue MH & Poppy GM (2001a) Direct and indirect sublethal effects of Galanthus nivalis agglutinin (GNA) on the development of a potato-aphid parasitoid, Aphelinus abdominalis (Hymenoptera: Aphelinidae). Journal of Insect Physiology 47: 553561.
  • Couty A, Down RE, Gatehouse AMR, Kaiser L, Pham-Delegue MH & Poppy GM (2001b) Effects of artificial diet containing GNA and GNA-expressing potatoes on the development of the aphid parasitoid Aphidius ervi Haliday (Hymenoptera: Aphidiidae). Journal of Insect Physiology 47: 13571366.
  • Couty A & Poppy GM (2001) Does host-feeding on GNA-intoxicated aphids by Aphelinus abdominalis affect their longevity and/or fecundity? Entomologia Experimentalis et Applicata 100: 331337.
  • Cowgill SE & Atkinson HJ (2003) A sequential approach to risk assessment of transgenic plants expressing protease inhibitors: effects on nontarget herbivorous insects. Transgenic Research 12: 439449.
  • Daily GC, Ehrlich PR & Alberti M (1996) Managing earth's life support systems: The game, the players, and getting everyone to play. Ecological Applications 6: 1921.
  • Dale PJ, Clarke B & Fontes EMG (2002) Potential for the environmental impact of transgenic crops. Nature Biotechnology 20: 567574.
  • Down RE, Ford L, Woodhouse SD, Davison GM, Majerus MEN, Gatehouse JA & Gatehouse AMR (2003) Tritrophic interactions between transgenic potato expressing snowdrop lectin (GNA), an aphid pest (peach-potato aphid; Myzus persicae (Sulz.) and a beneficial predator, 2-spot ladybird; Adalia bipunctata L.). Transgenic Research 12: 229241.
  • Down RE, Ford L, Woodhouse SD, Raemaekers RJM, Leitch B, Gatehouse JA & Gatehouse AMR (2000) Snowdrop lectin (GNA) has no acute toxic effects on a beneficial insect predator, the 2-spot ladybird (Adalia bipunctata L.). Journal of Insect Physiology 46: 379391.
  • Duan JJ, Head G, Mckee MJ, Nickson TE, Martin JW & Sayegh FS (2002) Evaluation of dietary effects of transgenic corn pollen expressing Cry3Bb1 protein on a non-target ladybird beetle, Coleomegilla maculata. Entomologia Experimentalis et Applicata 104: 271280.
  • Dutton A, Klein H, Romeis J & Bigler F (2002) Uptake of Bt-toxin by herbivores feeding on transgenic maize and consequences for the predator Chrysoperla carnea. Ecological Entomology 27: 441447.
  • Dutton A, Romeis J & Bigler F (2003) Assessing the risks of insect resistant transgenic plants on entomophagous arthropods: Bt-maize expressing Cry1Ab as a case study. Biocontrol 48: 611636.
  • Ervin DE, Welsh R, Batie SS & Carpentier CL (2003) Towards an ecological systems approach in public research for environmental regulation of transgenic crops. Agriculture, Ecosystems & Environment 99: 114.
  • Ferry N, Raemaekers RJM, Majerus MEN, Jouanin L, Port G, Gatehouse JA & Gatehouse AMR (2003) Impact of oilseed rape expressing the insecticidal cysteine protease inhibitor oryzacystatin on the beneficial predator Harmonia axyridis (multicoloured Asian ladybeetle). Molecular Ecology 12: 493504.
  • Fontes EMG, Pires CSS, Sujii ER & Panizzi AR (2002) The environmental effects of genetically modified crops resistant to insects. Neotropical Entomology 31: 497513.
  • Godfray HCJ (1994) Parasitoids. Behavioural and Evolutionary Ecology. Princeton University Press, Princeton, NJ, USA.
  • Hails RS (2002) Assessing the risks associated with new agricultural practices. Nature 418: 685688.
  • Henry CS, Brooks SJ, Duelli P & Johnson JB (2002) Discovering the true Chrysoperla carnea (Insecta: Neuropterea. Chrysopidae) using song analysis, morphology, and ecology. Annals of the Entomological Society of America 95: 172191.
  • Hilbeck A, Baumgartner M, Fried PM & Bigler F (1998a) Effects of transgenic Bacillus thuringiensis corn-fed prey on mortality and development time of immature Chrysoperla carnea (Neuroptera: Chrysopidae). Environmental Entomology 27: 480487.
  • Hilbeck A, Moar WJ, Pusztai-Carey M, Filippini A & Bigler F (1998b) Toxicity of Bacillus thuringiensis Cry1Ab toxin to the predator Chrysoperla carnea (Neuroptera: Chrysopidae). Environmental Entomology 27: 12551263.
  • Hilbeck A, Moar WJ, Pusztai-Carey M, Filippini A & Bigler F (1999) Prey-mediated effects of Cry1Ab toxin and protoxin and Cry2A protoxin on the predator Chrysoperla carnea. Entomologia Experimentalis et Applicata 91: 305316.
  • Hill RA & Sendashonga C (2003) General principles for risk assessment of living modified organisms: Lessons from chenical risk assessment. Environmental Biosafety Research 2: 8188.
  • Hoenig JM & Heisley DM (2001) The abuse of power: the pervasive fallacy of power calculations for data analysis. American Statistician 55: 1924.
  • Jørgensen HB & Lövei GL (1999) Tri-trophic effect on predator feeding: consumption by the carabid Harpalus affinis of Heliothis armigera caterpillars fed on proteinase inhibitor-containing diet. Entomologia Experimentalis et Applicata 93: 113116.
  • Lawton JH (1992) There are not 10 million kinds of population-dynamics. Oikos 63: 337338.
  • Loreau M, Naeem S & Inchausti P (2002) Biodiversity and Ecosystem Functioning. Oxford University Press, Oxford, UK.
  • Lövei GL (2001) Ecological risks and benefits of transgenic plants. Proceedings of the 54th Conference of the New Zealand Plant Protection Society 54: 93100.
  • Lövei GL & Sunderland KD (1996) The ecology and behaviour of ground beetles. Annual Review of Entomology 41: 231256.
  • Lozzia GC, Furlanis C, Manachini B & Rigamonti IE (1998) Effects of Bt corn on Rhopalosiphum padi L. (Rhynchota Aphididae) and on its predator Chrysoperla carnea Stephen (Neuroptera Chrysopidae). Bollettino di Zoologia Agraria e Bachicoltura 30: 153164.
  • Lundgren JG & Wiedenmann RN (2002) Coleopteran-specific Cry3Bb toxin from transgenic corn pollen does not affect the fitness of a nontarget species, Coleomegilla maculata DeGeer (Coleoptera: Coccinellidae). Environmental Entomology 31: 12131218.
  • Malcolm SB (1992) Prey defence and predator foraging. Natural Enemies (ed. by M JCrawley), pp. 458475. Blackwell, Oxford, UK.
  • Manachini B (2003) Effects of transgenic corn on Lydella thompsoni Herting (Diptera Tachinidae) parasitoid of Ostrinia nubilalis Hb. (Lepidoptera Crambidae). Bollettino di Zoologia Agraria e di Bachicoltura 35: 111125.
  • Marvier M (2002) Improving risk assessment for nontarget safety of transgenic crops. Ecological Applications 12: 11191124.
  • Meier MS & Hilbeck A (2001) Influence of transgenic Bacillus thuringiensis corn-fed prey on prey preference of immature Chrysoperla carnea (Neuroptera: Chrysopidae). Basic and Applied Ecology 2: 3544.
  • Pilcher CD, Obrycki JJ, Rice ME & Lewis LC (1997) Preimaginal development, survival, and field abundance of insect predators on transgenic Bacillus thuringiensis corn. Environmental Entomology 26: 446454.
  • Ponsard S, Gutierrez AP & Mills NJ (2002) Effect of Bt-toxin (Cry1Ac) in transgenic cotton on the adult longevity of four heteropteran predators. Environmental Entomology 31: 11971205.
  • Pruetz G & Dettner K (2004) Effect of Bt corn leaf suspension on food consumption by chilo partellus and life history parameters of its parasitoids Cotesia flavipes under laboratory conditions. Entomologia Experimentalis et Applicata 111: 179187.
  • Riddick EW & Barbosa P (1998) Impact of Cry3A-intoxicated Leptinotarsa decemlineata (Coleoptera: Chrysomelidae) and pollen on consumption, development, and fecundity of Coleomegilla maculata (Coleoptera: Coccinellidae). Annals of the Entomological Society of America 91: 303307.
  • Riddick EW & Barbosa P (2000) Cry3A-intoxicated Leptinotarsa decemlineata (Say) are palatable prey for Lebia grandis Hentz. Journal of Entomological Science 35: 342346.
  • Romeis J, Babendreier D & Wackers FL (2003) Consumption of snowdrop lectin (Galanthus nivalis agglutinin) causes direct effects on adult parasitic wasps. Oecologia 134: 528536.
  • Romeis J, Dutton A & Bigler F (2004) Bacillus thuringiensis toxin (Cry1Ab) has no direct effect on larvae of the green lacewing Chrysoperla carnea (Stephens) (Neuroptera: Chrysopidae). Journal of Insect Physiology 50: 175183.
  • Schuler TH, Potting RPJ, Denholm I, Clark SJ, Clark AJ, Stewart CN & Poppy GM (2003) Tritrophic choice experiments with Bt plants, the diamondback moth (Plutella xylostella) and the parasitoid Cotesia plutellae. Transgenic Research 12: 351361.
  • Setamou M, Bernal JS, Legaspi JC & Mirkov TE (2002a) Parasitism and location of sugarcane borer (Lepidoptera: Pyralidae) by Cotesia flavipes (Hymenoptera: Braconidae) on transgenic and conventional sugarcane. Environmental Entomology 31: 12191225.
  • Setamou M, Bernal JS, Legaspi JC & Mirkov TE (2002b) Effects of snowdrop lectin (Galanthus nivalis agglutinin) on fitness of Cotesia flavipes (Hymenoptera: Braconidae), a parasitoid of the nontarget pest Diaeterella saccharalis (Lepidoptera: Crambidae). Annals of the Entomological Society of America 95: 7583.
  • Shelton AM, Zhao JZ & Roush RT (2002) Economic, ecological, food safety, and social consequences of the deployment of Bt transgenic plants. Annual Review of Entomology 47: 845881.
  • Sims SR (1995) Bacillus thuringiensis var. kurstaki (CryIA (C)) protein expressed in transgenic cotton: effects on beneficial and other non-target insects. Southwestern Entomologist 20: 493500.
  • Sokal RR & Rohlf FJ (1995) Biometry, 3rd edn. Freeman & Co., New York, USA.
  • Stamp NE, Yang YL & Osier TL (1997) Response of an insect predator to prey fed multiple allelochemicals under representative thermal regimes. Ecology 78: 203214.
  • Steidl RJ, Hayes JP & Schauber E (1997) Statistical power analysis in wildlife research. Journal of Wildlife Management 61: 270279.
  • Tilman D, Fargione J, Wolff B, D’Antonio C, Dobson A, Howarth R, Schindler D, Schlesinger WH, Simberloff D & Swackhamer D (2001) Forecasting agriculturally driven global environmental change. Science 292: 281284.
  • Tomov BW & Bernal JS (2003a) Effects of GNA transgenic sugarcane on life history parameters of Parallorhogas pyralophagus (Marsh) (Hymenoptera: Braconidae), a parasitoid of Mexican rice borer. Journal of Economic Entomology 96: 570576.
  • Tomov BW, Bernal JS & Vinson SB (2003b) Impacts of transgenic sugarcane expressing GNA lectin on parasitism of Mexican rice borer by Parallorhogas pyralophagus (Marsh) (Hymenoptera: Braconidae). Environmental Entomology 32: 866872.
  • Van Driesche RG & Bellows TS Jr (1996) Biological Control. Chapman & Hall, New York, USA.
  • Zwahlen C, Nentwig W, Bigler F & Hilbeck A (2000) Tritrophic interactions of transgenic Bacillus thuringiensis corn, Anaphothrips obscurus (Thysanoptera: Thripidae), and the predator Orius majusculus (Heteroptera: Anthocoridae). Environmental Entomology 29: 846850.