James C. Lamb
Posted: Thursday, November 20, 2008 2:11 pm

Review of Zantek paper

At the request of Monsanto, I, James C. Lamb, IV, Ph.D., DABT, ATS, have completed a preliminary review of the report “Biological effects of transgenic maize NK603xMON810 fed in long term reproduction studies in mice” by Dr. A Velimirov, Dr. C. Binter, and Univ. Prof. Dr. J. Zentek. I have focused my efforts on the multigeneration (MG) study and the reproductive assessment by continuous breeding (RACB) study. The RACB study design was developed by me at the U.S. National Toxicology Program (NTP) in the early 1980s to improve the assessment of fertility in rodent toxicology studies. Work on this protocol was continued by Dr. Richard Morrissey and then by Dr. Robert Chapin of the NTP. The RACB has the potential to test certain biological indices that the MG study cannot address. The authors of this report made modifications to the RACB study design, but the basic elements remain the same. Male and female mice were divided into control and test groups and were house as breeding pairs for about 20 weeks. Each male and female pairing produced litters continuously through the cohabitation period, which allowed the pairs to produce as many as 4 litters per pair. As we originally developed this protocol at NTP, an effect on reproduction would be followed by a cross-over mating to help determine whether the effects might be attributed to males, to females or to both genders. That cross-over mating was not performed in this study.

The study report describes the design, conduct and interpretation of the study. My review is limited to the information provided in the report. The design is appropriate for the hypothesis that is being tested. The RACB study design can generate a large amount of data and analyses, but the report is not sufficiently detailed to interpret fully the study. Nevertheless, there are some significant issues raised by the results as presented and the interpretation that must be discussed. I have found already some significant errors in the data calculations that led to the important misinterpretations of the findings by the authors. Unfortunately, the study does not describe a quality assurance process that should have prevented these errors. Errors are clear in two of the most important tables in the report (Table 36 and Table 59). These errors directly impact the interpretation of the MG and the RACB studies.

The authors analyzed the MG study in an unconventional manner. The report is not clear whether the statistical tests and evaluation methods were defined before or after the data were developed. The methods were not consistent with the methods that we applied at the NTP. One important point is that the litter size was analyzed using the breeding pair as the denominator rather than the delivering breeding pair. The number of deliveries in the F3 (3rd generation) matings dropped off for both control and GM test groups and there was no explanation for this decline (see second row of Table 36). This should have been addressed. The decline in fertility resulted in an apparent decrease in litter size that is an artifact of the denominator that they used (see fourth row Table 36). This might best be illustrated by a fictitious example. If one cohabited 20 pairs and only one pair delivered a litter of 10 animals, the mean litter size would be 10; not 0.5 as would have been reported by the authors. If the proper denominator was used, the litter size did not change for either group compared to the other generations. The authors did not report the number of litters analyzed at weaning, so we cannot determine whether or not a similar error was made in the number of weanings/pair. Without the individual animal data, we cannot test the statistical findings either. In the end, the authors did not see a treatment effect in this study, but the computational errors raise concern about other examinations in such an important data set.

The authors also divided the MG study litters into less than 8 and greater than 8 pups per litter. I have never seen this approach to litter analysis and wonder why and when they chose to make this comparison. No rationale is provided for this approach. These data are very difficult to interpret and have dubious scientific meaning. The various charts lead to a good deal of discussion, but do not really aid in the data interpretation.

The RACB study has similar issues that do affect the authors’ interpretation of their study. In Table 59 they report data similar to the MG study. Instead of multiple generations, the table shows multiple litters across the study. One odd point is that they had one pair that never delivered any litters. There is no explanation for this, and that is inconsistent with the MG study. They probably should have analyzed the data with and without this pair in the study because it is not likely related to the treatment. This amplifies the error of using the wrong denominator because the litter size is understated for the treated group throughout the study because they again divided by the pairs cohabited without considering the actual litter size for those mated pairs that actually delivered litters. They have made the same mistake on the number of pups at birth per pair and on number of pups at weaning per pair.
The authors’ computational approach results in some key misstatements in Table 59. In 3rd litter for the GM group the mean number of pups per litter is incorrect and should be 9.41, not 9.06. For the 4th litter, the GM group should be 8.65 pups per litter, not 7.21. Similar mistakes are presented for the pups weaned and the numbers for the GM are consistently understated. For example, the 4th litter had an average of 9.11 weaned pups per litter, not 7.21 and the 3rd litter had 12.18 weaned pups per litter not 9.06.

These two tables illustrate that the computational errors resulted in an overstatement of the differences between the GM and the ISO groups. No historical control data have been presented or discussed. Such data could be very important in the interpretation of such differences, to the extent that any significant differences remain. The specific details of how computations were made should have been included in the report. It is not possible to reconstruct all of the data, but the ones that we have investigated indicted errors in the authors’ methods and therefore interpretations. The statistical methods and presentation should have been designed prior to the conduct of the experiments, and it is not clear whether or not this was done. The statistics cannot be tested appropriately without the individual animal data. The computational errors in such critical tables (Tables 36 and 59) raise serious questions about the other data in the report and the quality assurance methods that were or should have been applied before the conclusions were drafted and the report was released. When properly analyzed, these data do not appear to support an effect on fertility or reproduction from consumption of GM corn.

James C. Lamb, IV, Ph.D., DABT, ATS
Executive Vice President
The Weinberg Group
November 19, 2008