Modelling in the light of uncertainty of key parameters: a call to exercise caution in field predictions of Bt-maize effects

Perry et al. [1] developed a model to simulate the field model is based; however, this was not quantified in exposure and adverse effects for three European nontarget Lepidoptera species (Inachis io L., Vanessa atalanta L., and Plutella xylostella L.) to pollen of the Bacillus thuringiensis (Bt) maize MON810 containing lepidopterantargeting Cry1Ab toxin. Perry et al. explicitly modelled the worst case scenario and came to hard quantitative predictions. However, the incomplete and uncertain input data cause a higher uncertainty than Perry et al. indicate, and we are specifically concerned with the possibility that the effects might be worse than they predict. Here, we specify this uncertainty by addressing some of the basic model assumptions and input data regarding the toxic effects of Cry1Ab to lepidopteran larvae. We do not address the hypotheses of the model regarding maize pollen dispersal and deposition, or population-dynamic effects. A key problem of the modelling study of Perry et al. is that virtually no studies exist in the peer-reviewed literature reporting on the dose–response effects of MON810 pollen to European non-target Lepidoptera [2]. In consequence, the authors had to substitute data obtained from experiments with another transgenic event expressing Cry1Ab, the Bt176 maize [3,4]. In our view, the use of these studies is problematic for several reasons. First, the authors set the toxicity of MON810 pollen to 31-fold less than Bt176 maize pollen, but did not provide data supporting the implied assumption of a linear relationship between Cry1Ab dose and adverse effects. The dose–effect relation may be nonlinear, possibly resulting in a disproportionally higher effect of low Cry1Ab concentrations [5,6]. Second, the assumption of a 31-fold difference in Cry1Ab concentration of pollen of MON810 and Bt176 is an average value, which Perry et al. seemingly derived from four papers (but three of them refer to the same source, the US Environmental Protection Agency). However, Cry1Ab content in pollen varies substantially even within the same event depending on the year, the site, the plant and possibly the cultivar [7]. For example, Cry1Ab concentration has been recorded to be as low as 389 ng g in Bt176 pollen and as high as 97 ng g in MON810 pollen, which is only a fourfold difference [8]. In consequence, the relevant information would be the Cry1Ab toxin concentration of the pollen used in the studies upon which the