French scientist Gilles-Eric Seralini caused quite a stir last week when he claimed he’d shown cancer in rats increased when they were fed genetically modified corn and/or water spiked with the herbicide Roundup. The paper, which seven of his colleagues co-authored, was published in the peer-reviewed journal Food and Chemical Toxicology.
France’s ministers for agriculture, ecology and health responded swiftly by commissioning the National Agency for Health and Safety to look into the claims. Depending on the findings, they could invoke an emergency suspension of imports of the Monsanto GM maize strain NK603, used in the study, into Europe. Now that’s what I call high impact.
But how did the authors come to their conclusion? And can such a significant claim be made using the study data?
The study focuses on cancers in rats. For this they use the Harlan Sprague-Dawley strain of rat, which is known to be predisposed to getting cancer. Lots of them. Over 70% of males and 87% of females from this strain reportedly get cancer during their lifetime, whether they have been fed GM corn or not. So it shouldn’t be a surprise that so many of Seralini’s rats were found with cancer.
To make sense of this study you have to ask the simple question: “does feeding rats GM corn and/or Roundup increase the frequency of cancers compared with rats that have been given non-GM food?”
To do this, the authors of the study split up 200 rats into ten groups. One “control” group (ten male and ten female) were fed non-GM corn and had access to plain water. The researchers monitored for the development of cancer over a period of two years.
Nine other groups of twenty rats (ten male and ten female) were also monitored, but this time, these groups were given food containing 11%, 22% or 33% of NK603 GM corn, 11%, 22% or 33% of NK603 GM corn treated with Roundup*, or just had Roundup spiked in their drinking water at different concentrations.
The male and female rats in the control group lived for just under two years. Other studies identified that these rats die from cancer or kidney failure around this time. But the authors don’t mention this. They simply write:
“ After mean survival time had elapsed, any deaths that occurred were considered to be largely due to aging.”
They have effectively chosen not look at – and therefore don’t have to report on – why rats in the control group died. This assumption alone is sufficient grounds for rejecting this paper from publication.
Treatment group vs the control
What do these mean? Well, not much because the authors failed to use a statistical test to tell if there was a difference between the control groups and treatment groups.
This is important, as all their claims relate to the incidence of cancers (and other “diseases”) in the “treatment group” compared to the “control group”. These comparisons can only be made if a statistical test shows that what you observe is not happening by chance.
Overstating the evidence
Still on Figure 1, we see that several “treatment groups” of male rats receiving GM NK603 corn (the 22% group and 33% group) actually had fewer cancers than the male control group at their arbitrarily determined point of assessment.
Similarly, a treatment group of male rats receiving 33% GM corn and Roundup had no difference to the control group, and two treatment groups receiving Roundup (A and C) had the same or less incidence of cancer compared with the control group.
By their perverted logic, they could equally claim that for male rats:
a) high percentages of GM corn (22% and 33%) was “protective” against getting cancer compared to group of control male rats
b) having 33% of GM corn with Roundup showed no difference to the control group and therefore wasn’t harmful to male rats, and
c) using 0.5% Roundup in the drinking water was protective against cancer in male rats compared to the the male control group.
But you can’t. You can no more make these statements than the claims about the increased incidence of cancers in the female rats in the various treatment groups. No statements can be made because no statistical test has been applied.
The full picture
One sentence that should set alarm bells ringing is the claim that “All data cannot be shown in one report.”
The retort to that statement is, “Oh yes it can. Please show it to me”. If you are reporting data, you need to show all the data.
Not enough space? Put it in the supplemental data.
In the data section, the authors show examples of pathology, histology and electron microscopy images of affected organs in the treatment groups and mention results from genetic testing of samples. All well and good, but for the genetic tests, they don’t show any data other than a statement of claim.
They also don’t present any biochemical data from the male rats – half of all their studied rats. In the legend for table 3 (which shows the “Percentage variation of parameters indicating kidney failures of female animals), they claim "Male kidney pathologies are already illustrated in Table 2” (which shows a “Summary of the most frequent anatomical pathologies observed”). But we’re not shown the raw, unmanipulated data, tested with standard statistical tests, for males and females.
The authors then go on to describe the cancers in detail. They state:
“ Up to 14 months, no animals in the control groups showed any signs of tumors whilst 10–30% of treated females per group developed tumors, with the exception of one group (33% GMO + R).”
Well done. They have just created a non-predefined outcome measure and made a biologically nonsensical statement.
Do they mean to imply that female rats eating the highest percentage of GM corn with Roundup are mysteriously no more affected than the female control group, compared to other female “treatment groups” which were somehow more affected?
Once again, no statistical test is applied and no conclusions can be drawn.
Further, they don’t describe diseases affecting the “control group”. At all. By neglecting to state if there were any changes in the “control group”, you cannot make any statement about the “treatment groups”. That’s why you have controls.
So, what have we learnt?
This study has shown that old Harlan Sprague-Dawley rats get cancers and other diseases. This has been shown before.
What this study does not show is that exposing these rats to GM corn and/or Roundup makes any difference to the frequency of cancers or other diseases. It can’t because no statistical tests have been applied, and perhaps most worryingly, the authors do not comprehensively report on why rats in the control group died.
This study can hardly be the basis from which any government should make policy decisions or draw conclusions about the safety of the NK603 GM maize or Roundup.
Read an article about the murky release of the paper – Modifying the message: how tricks masked home truths about anti-GM science
*A previous version of this article incorrectly stated this group had Roundup spiked in their drinking water at different concentrations.