Wednesday, June 20, 2012

No More Crying About Your Stats Class and Cry1Ab: An Application of the Coefficient of Variation

Lots of times students complain that either their statistics classes used silly examples that were too simple to ever be realistic, or that their course was too complicated and thus they leave the class without the capability of  any practical application.  A recent study looking at the safety of GMO corn provides a great case study for the practical application of the coefficient of variation (CV). 

In ‘Maternal and fetal exposure to pesticides associated to genetically modified Foods in Eastern Townships of Quebec, Canada’ the authors claim to have identified the toxin Cry1Ab in the blood of pregnant women.  Cry1Ab is a protein produced by the bacteria Bacillus thuringiensis (Bt) that is toxic to certain insect pests. Cry1Ab is just one version (event) of this Bt toxin.  Bt toxins have been used extensively by organic farmers and biotechnology has enabled seed companies to develop corn plants that express Cry1Ab proteins giving them a built in defense mechanism against insects susceptible to the toxin, while preserving the biodiversity of friendly insects.  Bt genetics have also been incorporated into cotton. The economic, environmental, safety, and health benefits have made this a very popular  tool used by the majority of family farmers. 

One of the major criticisms of the article was the use of the test used to identify the Cry1Ab protein. In the article the authors state:

‘Cry1Ab protein levels were determined in blood using a commercially available double antibody sandwich(DAS)enzyme-linked immune sorbent assay.’

In previous research, the enzyme-linked immunosorbent assay or ELISA test has been shown to be one of the most unreliable tests for detecting Cry1Ab proteins.  Recall, the CV is relative measure of variation measuring the standard deviation relative to the mean.   It can be used as a metric for risk and reliability (such a consistent yield performance or stock returns).  In the article ‘Comparison and Validation of Methods To Quantify Cry1Ab Toxin from Bacillus thuringiensis for Standardization of Insect Bioassays’ the authors investigate procedures commonly used to identify Cry1Ab.  The authors explain:

“We compared three methods of quantification on three different toxin preparations from independent sources: enzyme-linked immunosorbent assay (ELISA), sodium dodecyl sulfate-polyacrylamide gel electrophoresis and densitometry (SDS-PAGE/densitometry), and the Bradford assay for total protein....The Bradford method resulted in statistically higher estimates than either ELISA or SDSPAGE/ densitometry but also provided the lowest coefficients of variation (CVs) for estimates of the Cry1Ab concentration (from 2.4 to 5.4%). The CV of estimates obtained by ELISA ranged from 12.8 to 26.5%, whereas  the CV of estimates obtained by SDS-PAGE/densitometry ranged from 0.2 to 15.4%....we conclude that standardization of Cry1Ab production and quantification by SDS-PAGE/densitometry may improve data consistency.”

If we look at their reported statistics, we can see for ourselves just how high the CV is on the ELISA test (and therefore how unreliable it is as a method for quantifying Cry1Ab)  compared to other proven methods of quantification.

So there you have it. A practical example of an application of a very basic statistic, the coefficient of variation.


Maternal and fetal exposure to pesticides associated to genetically modified foods in Eastern Townships of Quebec, Canada. Reprod Toxicol. 2011 May;31(4):528-33. Epub 2011 Feb 18.
Aris A, Leblanc S.

Comparison and Validation of Methods To Quantify Cry1Ab Toxin from Bacillus thuringiensis for Standardization of Insect Bioassays. Andre´ L. B. Crespo,1 Terence A. Spencer,1 Emily Nekl,2 Marianne Pusztai-Carey,3 William J. Moar,4 and Blair D. Siegfried1* APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Jan. 2008, p. 130–135 Vol. 74, No. 1

A Meta-Analysis of Effects of Bt Cotton and Maize on Nontarget Invertebrates. Michelle Marvier, Chanel McCreedy, James Regetz, Peter Kareiva Science 8 June 2007: Vol. 316. no. 5830, pp. 1475 – 1477

Comparison of Fumonisin Concentrations in Kernels of Transgenic Bt Maize Hybrids and Nontransgenic Hybrids. Munkvold, G.P. et al . Plant Disease 83, 130-138 1999.

Indirect Reduction of Ear Molds and Associated Mycotoxins in Bacillus thuringiensis Corn Under Controlled and Open Field Conditions: Utility and Limitations. Dowd, J. Economic Entomology. 93 1669-1679 2000.

Impact of Bt cotton on pesticide poisoning in smallholder agriculture: A panel data analysis. Shahzad Kouser, Matin Qaim. Ecological Economics
Volume 70, Issue 11, 15 September 2011, Pages 2105–2113

Communal Benefits of Transgenic Corn. Bruce E. Tabashnik  Science 8 October 2010:Vol. 330. no. 6001, pp. 189 - 190DOI: 10.1126/science.1196864