By L. LaReesa Wolfenbarger
A comprehensive, public database is now available that houses
information about studies on the non-target effects of Bt crops. The
database contains information on experimental design and results, and
facilitates a quantitative approach to synthesizing the studies on the
effects of Bt crops on non-target arthropods.
How complete is the database?

To locate all possible studies to include, Dr. Michelle Marvier and
her colleagues queried online databases of articles, such as Agricola,
BioAbstracts, PubMed, and ISI Web of Science; conducted general internet
searches; searched references within publications; and asked for additional
suggestions from over 100 researchers. Each study in the database meets four
criteria:
1) it involves a Bt field crop.
2) it measures an effect on a non-target arthropod.
3) its design includes a non-transgenic control or varies exposure
levels to Bt plants or their products.
4) it is in English.

What types of studies are in the database?

The database contains 5,758 experimental comparisons from 171 studies.
Studies vary in how many comparisons they contributed to the database with
50% of studies contributing 14 or fewer. Author affiliations for the studies
included academic institutions, government, corporations, and non-profit
organizations, with the largest contributor being academic institutions.

Publications on the non-target effect of Bt crops on arthropods began
in 1992 with the majority of studies published after 2000. The majority of
studies are either lab- (n = 81) or field- (n = 75) based, and a small
number contain both field and lab experiments (n = 8) or other approaches
(e.g., greenhouse).

What type of questions will the database address?

A wide variety of questions can be answered using the database given
how well and how detailed the authors coded information about each study's
experimental design, study species, and results. A few possibilities might
include questions about the effects on a particular species or group of
species, about whether field and laboratory studies produce similar results,
or about whether sampling protocols affect results.

As one example, Marvier et al. (2007) reported a meta-analysis on the
effects of Bt maize and Bt cotton on the abundance of non-target arthropods.
Meta-analysis is a quantitative approach to synthesizing results from
multiple studies that have similar experimental designs. For each
comparison, the analysis calculates an effect size, which is the difference
between the experimental treatment (Bt crop) and the control treatment. A
mean effect size is calculated by weighting each comparison by its sample
size; therefore, comparisons based on large sample sizes are given more
weight than comparisons based on small sample sizes. Similarly, the analysis
calculates a confidence interval by weighting the variance associated with
each treatment.

Marvier et al. focused on field studies that compared the abundance of
arthropods in Bt crops to controls. The authors eliminated non-independent
comparisons. For example, some studies report a comparison of abundance
calculated from seasonal means and from the peak day. These issues of
non-independence are coded in the database, and the user needs to go through
the data carefully to ensure that the data chosen are independent. Marvier
et al. report on the effects of Bt crops using three different types of
experimental and control comparisons:
a) a Bt crop vs. the non- transgenic crop.
b) a Bt crop vs. the non-transgenic crop + insecticides.
c) a Bt crop + insecticides vs. the non-transgenic crop +
insecticides.

The meta-analysis reveals positive and negative effects of Bt crops on
non-target arthropods, and the direction depends on the control or the
baseline for comparison. If a Bt crop replaces insecticides, then there are
more non-target arthropods in the Bt fields. There are fewer non-target
arthropods in Bt fields compared to non-Bt fields where no insecticide has
been used. The authors point out that both baselines reflect reality in US
maize and cotton production. For example, in 2005 insecticides were applied
to 23% of maize acreage and 71% of cotton acreage, according to Agriculture
Chemical Usage data published by the U.S. Department of Agriculture. The
paper cautions that statistically significant differences in arthropod
abundances do not easily translate into understanding what ecological
consequences these differences would have in agricultural systems or beyond.
However, there is tremendous value in being able to draw upon the cumulative
efforts and results of many studies to determine what, if any, changes may
occur and in what context those changes happen. This database, if used and
maintained, will contribute greatly to the advancement of what we have
learned about non-target effects of Bt crops and what we need to learn for
the future.

[www.isb.vt.edu]