Review of "A Comparison of the Effects of Three GM Corn Varieties on Mammalian Health"

After a brief vacation in the land of poutine, ketchup flavoured chips, Tim Horton's, and my awesome niece and nephews, I'm back in California where it still hasn't rained. The spouse and I have decided to install artificial grass in our yard, but dang!! That stuff is pricey!! Someone should work on implementing some sort of tax break for getting it installed.

This week, I received a request to look over a paper entitled "A Comparison of the Effects of Three GM Corn Varieties on Mammalian Health". The paper was published in 2009, and one of it's authors is Giles-Eric Séralini (you can read about the "Séralini affair" on Wikipedia. My previous posts on his work can be found in the index under "Seralini Study").

The paper (I'll be referring to it as the "de Vendômois paper") is freely available so you should be able to view it. Monsanto wrote a response, but I'm going to go over the paper and my own impressions before reading Monsanto's. I will then provide an overview of those critiques and my conclusion.

When a GMO goes through the regulatory process with the FDA and other agencies, a feeding study is performed where the GMO is fed to rats for 90 days and different parameters are measured. The goal is to determine if there are any health impacts during this time period. Since the rats used in these studies live 2.5-3.5 years, 90 days represents approximately 7-10% of their lifespan.

de Vendômois obtained data from Monsanto's 90 day feeding studies for three strains of corn: one strain was Round-up Ready (NK603) and two strains contained the Bt-toxin (MON810 and MON863 - please see this excellent post from Biofortified on how Bt works). de Vendômois outlines that he had to get court orders to obtain some of the data, and worked with Greenpeace to get these.

The 90-day feeding studies performed by Monsanto consisted of 200 male and 200 female Sprague-Dawley rats for each feeding study. There were two doses of GM-feed, meaning that some rats got more GM corn in their feed than others (11% and 33%).  Of the 400 rats in each study, only 80 were fed the GM diet, while 320 were controls. de Vendômois states that the large number of controls is because there were several types of control feed: some were from the non-GM variety of the GM corn, but other non-related corn were also included as controls. de Vendômois was not pleased with the fact that there were way more control rats than rats being fed the GM stuff.

I understand the use of different types of controls, as well as wanting equal number of treatment and control rats. In having so many controls, I imagine that Monsanto wanted to figure out if any health impact observed was due to the corn itself or due to the transgenic protein. Here's a completely fictional example: if I take a Granny Smith apple and make it glow in the dark, I'd do a feeding study to determine the impact of the glowing gene on rats. As a control, I'd use a regular Granny Smith apple. But I might also want to include a Red Delicious apple as a second control. That way, if I see something odd happening, I'd be able to determine if it was due to the glowing gene or something that's unique to the Granny Smith apple (such as higher acidity). Yet I also understand the perspective of wanting to have a more equivalent number of rats. But doing the logical math, I'd assume that the 80 treatment rats get broken down to 40 per dose, and further broken down to 20 per gender. That's pretty good (I thought that this criticism was pretty ironic, since 20 per gender per treatment is more than what Séralini used in his own study, done a few years later). However, for some reason, apparently only 10 of the 20 rats were chosen each time for the blood and urine collection. Additionally, blood and urine were taken at only 2 time points, and de Vendômois thought that this was far too low for any useful statistical analysis. I'm REALLY curious to see what Monsanto's response is to this! You'll just have to keep reading!

de Vendômois points out that 2 dose levels is not standard and that 3 is the recommendation. I double-checked the OECD guidelines and this statement is accurate (please see point 17 in the link).

The Sprague Dawley Rat
From Wikipedia

Then, the paper starts going downhill. The description of the statistical tests used makes no sense. They run the data-set through a variety of tests, jumping from one to the next, and the only reason I can think of is that they're just fishing for some type of significance, regardless of how. Based on my understanding of stats, you select a statistical method/test for the type of data that you're analyzing and the comparison you're doing, and you stick with it throughout the entire data-set. If you find significance, great! If you don't, then you can't just say "alright, now I'll run it through this separate test to see what I can find," unless there's a very good reason for doing so. No such reason is provided.

Spouse, I can hear your demands for a better explanation about this in my head. This article from Wikipedia gives a great description of why this is important: "When large numbers of tests are performed, some produce false results, hence 5% of randomly chosen hypotheses turn out to be significant at the 5% level, 1% turn out to be significant at the 1% significance level, and so on, by chance alone. When enough hypotheses are tested, it is virtually certain that some falsely appear statistically significant, since almost every data set with any degree of randomness is likely to contain some spurious correlations. If they are not cautious, researchers using data mining techniques can be easily misled by these apparently significant results."

This is a big flaw and basically makes their analysis meaningless. I scanned through the rest of the paper and nothing really jumped out. The measurements where they find significance are not maintained between sexes, and that doesn't make sense for things like liver or kidney function (I double-checked with my brother and sister-in-law, who are MDs, and they agreed that after balancing for weight, a toxic compound should impact males and females equally).

Alrighty... Moving on to Monsanto's response, which is readily available here.

In addition to the issues pointed out before, a solid point that Monsanto makes is that de Vendômois doesn't examine whether or not the "statistically significant" values are within the normal range for the strain of rat. Meaning, is it biologically relevant?

Monsanto also states that their study has enough data for the standard statistical test used for these types of studies. Their point is that if de Vendômois had done his analysis with the "normal" stats methods, there might have been enough data.

Monsanto's response concludes with statements from 3 international regulatory agencies who reviewed the data: the European Food Safety Authority, the Food Standards Australia New Zealand, and the French High Counsel on Biotechnology. All the agencies conclude that the reanalysis of Monsanto's data was performed with crummy statistics and the results don't mean anything.

However, Monsanto does not address the issue of 2 dosages.

Here are my conclusions: regarding the safety of these specific traits, I think there's a LOT to suggest that they're safe. One of the common comments that I've read from those who defend Séralini's work is that his studies indicate that follow-up and longer-term analysis is needed. I completely understand that argument, but the work has already been done. For example, look at this paper which consists of a short-term feeding study of MON810 in pigs (note that I've only read the abstract): its conclusion is that there were a few differences that merited a long-term study. Then, they performed the long term study and found that any differences observed were not biologically relevant. Séralini has been at this for quite some time and he hasn't found anything that is biologically relevant. Feeding studies for NK603 can be found here, here, here, and here (the last link is a study by Monsanto).

The second issue is that of standardization of feeding studies. I've written about this before: I think that there's need for standardized tests as well as their analysis. de Vendômois's paper highlights the point, particularly for the analysis! Imagine how much money and animal lives would be saved if crummy feeding studies weren't performed. If the OECD guidelines are the ones that are supposed to be followed for feeding studies, then everyone should stick to them, whether its Monsanto or Séralini.

I want to be abundantly clear about this: I think that crops should be regulated based on the trait, not on the method used to generate that trait. So standardized tests should be used for conventional crops, too, when necessary. It doesn't make sense to have huge feeding studies on a trait that has already been tested: other experiments and data may be required, but a feeding study to determine the effect of the transgenic (i.e. "added") protein seems redundant. It seems ironic that hybrids such as nectaplums or broccoflowers aren't regulated when so many different genetic events occur during their creation.

Anyhoo, that's it for this paper. The next few posts I have planned are to finish up my series on sequencing papers examining GMOs, and my former work-spouse asked me to write about GMOs and butterflies. If you have any requests, questions, suggestions, or corrections to provide, please comment below!