Statistical power analysis of a 2-year field study and design of experiments to evaluate non-target effects of genetically modified Bacillius thuringiensis corn

1. A 2-year field study was conducted in 2000 and 2001 at a farm in Monmouth, Illinois to evaluate the effect of MON 863, a transgenic corn event that expresses the Cry3Bb1 protein for control of corn rootworms, on diverse groups of arthropod populations in a corn ecosystem. 2. The field study employed a split-plot design with MON 863 and conventional corn as the main plots and several insecticide treatments as subplots; each main plot was replicated four times (i.e. n = 4 block replications for each corn variety). A large number of arthropod taxa were trapped, taxonomically characterised, and counted in all plots. A mixed linear repeated-measures model was used for analysis of each taxon collected to determine if there was an effect of MON 863 or insecticide regime on that taxonomic group. 3. This report describes the results of a power analysis to determine the ability of the study design to detect an effect of MON 863 corn on the abundance of different groups of arthropods. A 50% difference in average taxon abundance between MON 863 and conventional corn plots was the criterion used in assessing the power. 4. A statistical power analysis of the taxonomic group comparisons made in the 2-year field study (with independent assignment of treatments each year) showed that only about 28% (22 of 56) of the statistical comparisons made with single years of data achieved >= 80% power to detect a 50% difference in population density. When data from both years of the study were analysed jointly, this level of power was obtained for 86% (24 of 28) the comparisons made. 5. Graphical examination of the mean abundance and coefficient of variation (CV) for different taxa indicated that high abundance (> 5 trap captures per plot per sampling period) and low CV (< 100%) contributed to high statistical power (> 80%). Taxa with extremely low abundance (mean trap captures < 1 individual per plot per sampling period) generally had high CV (> 100%), and statistical comparisons for those taxa had lower power (< 80%). 6. The relevance of these findings for designing field studies on the effects of genetically modified crops on non-target arthropods with desired levels of power is discussed.