Appropriate analytical methods are necessary to assess nontarget effects of insecticidal proteins in GM crops through meta-analysis (Response to Andow et al. 2009)

We strongly restate our criticisms of the report by Lövei et al. (2009). (1) They failed to account for the critical importance of well-described prey/host-quality mediated effects in the studies included in their analyses. Studies that failed to delineate toxicity of Bt proteins from poor prey quality should have either been eliminated from the analysis or coded so that heterogeneity analysis could have been conducted to reveal true treatment effects. (2) They included multiple nonindependent measures of various life history and behavioral traits in their analyses. (3) They used a distribution approach that negates much of the power of a meta-analysis and the subsequent inferences possible. (4) They lumped together proteins that have entirely different modes of action and host ranges into a single category (i.e., proteinase inhibitors, lectins) (5) They failed to provide any ecological context for their assessments and they disregarded actual field studies that have shown the lack of harm to natural enemies in environments in which Bt plants have been grown. To reiterate, the suggestion by Andow et al. that we have fundamental criticisms of meta-analysis is a red herring that diverts attention away from the real debate over the merits of different meta-analytic approaches. Our criticism is directed to the meta-analysis by Lövei et al. (2009) and not to all meta-analyses per se. Additionally, the seven findings added to the end of Andow et al. (2009) with the phrase that that they were not disputed by Shelton et al. (2009) works counter to a full and objective debate in the scientific literature. Our initial rebuttal (Shelton et al. 2009) was limited by page length, as is this letter. Because we did not address each of these issues does not mean we agree with them or find them without fault.