Commentary: Statistical aspects of environmental risk assessment of GM plants for effects on non-target organisms

¹ Oaklands Barn, Lug's Lane, Broome, Norfolk NR35 2HT, UK
² Biometris, Plant Research International, Wageningen University and Research Centre, P.O. Box 100, 6700 AC Wageningen, The Netherlands
³ Department of Statistics, 120 Snedecor Hall, Ames IA, 50011-1210, USA
⁴ USDA ARS Beneficial Insects Introduction Laboratory, Newark, Delaware 19713, USA
⁵ Centre for Ecology and Hydrology, Mansfield Rd, Oxford OX1 3SR, UK
⁶ Julius Kuehn Institute, Federal Research Centre for Cultivated Plants (JKI), Institute for Biosafety of Genetically Modified Plants, Messeweg 11/12, 38104 Braunschweig, Germany
⁷ INRIA Saclay, Université Paris-Sud, Bât. 425, 91405 Orsay Cedex, France
⁸ Dept. of Biology & Environmental Studies Institute, Santa Clara University, Santa Clara, CA 95053, USA
⁹ Department of Biology, “Tor Vergata” University, Via della Ricerca Scientifica, 00133 Rome, Italy
¹⁰ BioOK GmbH, Schnickmannstrasse 4, 18055 Rostock, Germany
¹¹ Department of Ecology, University of Debrecen; Debrecen, P.O. Box 71, 4010, Hungary
¹² Leibniz Universität Hannover, Fakultät Naturwissenschaften, Institut für Biostatistik, Herrenhaeuser Str. 2, 30419 Hannover, Germany
¹³ Biometris, Wageningen University and Research Centre, P.O. Box 100, 6700 AC Wageningen, The Netherlands

Corresponding author: joe.perry@bbsrc.ac.uk

Abstract

Previous European guidance for environmental risk assessment of genetically modified plants emphasized the concepts of statistical power but provided no explicit requirements for the provision of statistical power analyses. Similarly, whilst the need for good experimental designs was stressed, no minimum guidelines were set for replication or sample sizes. Furthermore, although substantial equivalence was stressed as central to risk assessment, no means of quantification of this concept was given. This paper suggests several ways in which existing guidance might be revised to address these problems. One approach explored is the `bioequivalence' test, which has the advantage that the error of most concern to the consumer may be set relatively easily. Also, since the burden of proof is placed on the experimenter, the test promotes high-quality, well-replicated experiments with sufficient statistical power. Other recommendations cover the specification of effect sizes, the choice of appropriate comparators, the use of positive controls, meta-analyses, multivariate analysis and diversity indices. Specific guidance is suggested for experimental designs of field trials and their statistical analyses. A checklist for experimental design is proposed to accompany all environmental risk assessments.