Extrapolating non-target risk of Bt crops from laboratory to field

Abstract

The tiered approach to assessing ecological risk of insect-resistant transgenic crops assumes that lower tier laboratory studies, which expose surrogate non-target organisms to high doses of insecticidal proteins, can detect harmful effects that might be manifested in the field. To test this assumption, we performed meta-analyses comparing results for non-target invertebrates exposed to Bacillus thuringiensis (Bt) Cry proteins in laboratory studies with results derived from independent field studies examining effects on the abundance of non-target invertebrates. For Lepidopteran-active Cry proteins, laboratory studies correctly predicted the reduced field abundance of non-target Lepidoptera. However, laboratory studies incorporating tri-trophic interactions of Bt plants, herbivores and parasitoids were better correlated with the decreased field abundance of parasitoids than were direct-exposure assays. For predators, laboratory tri-trophic studies predicted reduced abundances that were not realized in field studies and thus overestimated ecological risk. Exposure to Coleopteran-active Cry proteins did not significantly reduce the laboratory survival or field abundance of any functional group examined. Our findings support the assumption that laboratory studies of transgenic insecticidal crops show effects that are either consistent with, or more conservative than, those found in field studies, with the important caveat that laboratory studies should explore all ecologically relevant routes of exposure.

Figure 1.

Effect sizes measured in the field versus laboratory for major functional groups of non-target invertebrates. Laboratory studies for parasitoids and predators used either direct or tri-trophic exposure. Positive mean effect sizes (Hedge's d) for (a) Lepidopteran-active and (b) Coleopteran-active Bt Cry proteins indicate improved survival or increased abundance when exposed to Bt plant tissues or purified Cry proteins relative to a non-Bt control. Error bars represent unbiased, bootstrapped 95 per cent confidence intervals. Numbers denote total observations per column. Medium grey bars, field; dark grey bars, laboratory (direct exposure); light grey bars, laboratory (tri-trophic).

Figure 2.

Effect sizes measured in laboratory (direct and tri-trophic studies pooled) versus field studies for species with five or more observations per study type. Data are pooled across Lepidopteran-active and Coleopteran-active Cry protein studies. P, predator; H, herbivore. See figure 1 for description of chart elements. Lighter grey bars, field; darker grey bars, laboratory (tri-trophic and direct exposure).