I have fallen into the genetically modified organism ‘discussion’ by accident. I have a friend who is violently anti-GMO. This friend regularly posts the lies and fake ‘studies’ that populate the anti-GMO movement. I started by just looking up a few stats and now, I’m into full-blown deep research mode on this stuff.
The research, really is quite interesting and I may go into more of it. But the one thing I wanted to bring up here is the concept of cherry-picking. Why, I keep asking my friend, do you keep bringing up non-peer reviewed studies and ignore the peer-reviewed studies that I present to you? The ready answer is this massive global conspiracy between Monsanto and the bought government.
Honestly, I don’t like Monsanto and their practices. I’m not a huge fan of Roundup Ready plants, because that does tend to increase the use of herbicides. But that doesn’t mean that ALL GM food must be eliminated.
So, finally, a study appears that seems to support the Anti-GM crowd’s claims. GM food causes cancer. Well, I was skeptical. Even though this paper was published in a peer-reviewed journal, it was pre-published alongside 3 articles that all concluded that GM food was not harmful.
I contacted the lead author and got a copy of the paper, which is now available here.
Honestly, after reading the abstract, I could see that this might be a good paper. After reviewing the graphs (here), I had some concerns. After reading the paper I have some major concerns.
First, I want to talk a bit about the authors. This isn’t an ad hominem attack. If you are talking to a car salesman about which car is best and he describes the virtues of a particular car, oh and by the way I have one for sale here… perhaps you should consider the source.
Now, the lead author, Gilles-Eric Séralini, is also the author of a couple of books. One of which is title “Genetically Incorrect” and another is “We Can Clean Up”. These are books talking about how genetic manipulation of organisms is wrong (ironic considering the lab rats he used for this research, but more on that in a bit) and are part of the organic food revolution in France. Do you honestly think he’s unbiased?
One other point. Séralini has been censured before for incorrect analysis of data and incorrect procedures.
Another author, Joël Spiroux de Vendômois, is a homeopath. I’m sorry, but the fundamental principles of homeopathy (like cure like, dilution, rejection of the germ theory of disease) are all wrong. I’ve talked briefly about this before and I’ll happily go into it again if anyone wants me to. But the principles of homeopathy aren’t even wrong, they are impossible.
So, there’s two of the authors of this paper. One seems to be committed to defeating genetic manipulation of organisms and the other (despite being a doctor) doesn’t believe that science works. Now, would you buy a car from these two salesmen?
Again, this isn’t an ad hominem attack. These people freely admit that they think these things, but they don’t state so in their paper under the “conflict of interests” section. That’s not to say that these people can’t do good science. We should examine the science… so let’s do that.
Here’s the results graph
Fig. 1. Mortality of rats fed GMO treated or not with Roundup, and effects of Roundup alone. Rats were fed with NK603 GM maize (with or without application of Roundup) at three different doses (11, 22, 33% in their diet: thin, medium and bold lines, respectively) compared to the substantially equivalent closest isogenic non-GM maize (control, dotted line). Roundup was administrated in drinking water at 3 increasing doses, same symbols (environmental (A), MRL in agricultural GMOs (B) and half of minimal agricultural levels (C), see Section 2). Lifespan during the experiment for the control group is represented by the vertical bar ± SEM (grey area). In bar histograms, the causes of mortality before the grey area are detailed in comparison to the controls (0). In black are represented the necessary euthanasia because of suffering in accordance with ethical rules (tumors over 25% body weight, more than 25% weight loss, hemorrhagic bleeding, etc.); and in hatched areas, spontaneous mortality.
They are reporting in the graphic on page 4 that 10 males of the group that was fed Roundup sprayed GM maize died. Ten animals. That sounds bad. Then we learn that there were only 10 animals in that group. Holy crap, that’s a 100% death rate for those animals.
Now look at the control group. All animals were fed non-GM maize with no pesticides. Seven out of 10 died. Holy crap! That’s a 70% death rate.
Wait… what?
First let’s look at the actual groups of animals in this study.
- 10 male rats fed normal diet
- 10 male rats fed 11% GM Maize, not sprayed with round-up
- 10 male rats fed 22% GM Maize, not sprayed with round-up
- 10 male rats fed 33% GM Maize, not sprayed with round-up
- 10 male rats fed 11% GM Maize, sprayed with round-up
- 10 male rats fed 22% GM Maize, sprayed with round-up
- 10 male rats fed 33% GM Maize, sprayed with round-up
- 10 male rats fed normal diet, but their water had environmental levels of round-up in the water
- 10 male rats fed normal diet, but their water had half agricultural levels of round-up in the water
- 10 male rats fed normal diet, but their water had agricultural levels of round-up in the water
The females had similar groupings. So, there were 100 male rats and 100 female rats divided into 10 groupings. Then the control group of 10 was compared to each experimental group of 10.
That’s a major flaw in the study. Most people won’t recognize it because it’s very subtle, but there is a huge statistical problem here. The potential for cherry-picking is super strong here. That’s not to say it happened, but then, the complete results of every animal are not published here.
There are simply too many variables. A proper study would have been 50 male rats as a single test group and 50 male rats as a control group. The test should also be done using a couple of different breeds (again, more on why that’s a concern in a minute). I have since learned that 50 animals is the standard for long-tern studies.
Let’s talk for a minute about correlation and causation.
Correlation is the concept that two variables are moving at the same time. Causation is the concept that one of the variables is moving because of the other one. An example will serve. We note that over the last 100 years, the global temperature of the Earth has increased. We also note that over the same time period, that the number of pirates decreased. Those two numbers are correlated. Does anyone think that the lack of pirates is causing global warming (causation)?
Yeah, I didn’t think so.
Scientists use statistical measures to analyze data and determine if two variables are correlated. There’s no (that is zero) statistical analysis in this paper. Saying that 70% of a group died is not a statistical analysis. In fact, one statistician is going to us this paper as an example of how NOT to do studies like this.
Now, let’s look at the causation. Did the GM, non-sprayed maize group have tumors and death? Did the GM, sprayed maize group have tumors and death? Did the normal feed, drinking water with pesticide group have tumors and death?
The answer is yes to every single one of those. But note that I left one group out. The control.
Did the control group have tumors and death.
That’s also a yes. In fact, if you look at the charts, out of 6 comparisons with the control group, two of the control groups had more tumors and death than the experimental groups.
So we have to ask, what exactly is causing the tumors and deaths in the control group? Since the paper doesn’t say, we don’t know. Since we don’t know what is causing the issue in the control group, we cannot say that the same issue isn’t causing the cancer and death in the experimental groups.
That’s the point of a well crafted scientific experiment, to eliminate all but one variable that is then manipulated by the researchers to observe the changes that happen. This is not a well crafted experiment. We can’t say that the GM food causes cancer until we can show that the cause of cancer in the control animals does not also cause cancer in the experimental animals. I’ll talk more about the animals in a moment and, I think, all will become clear.
Statistically speaking, it’s impossible to say that there’s even any correlation between the tested chemicals and crops and the tumors and deaths. Why do I say that? Because in every single category, the rats fed the most amount of GM Maize, sprayed GM maize, and pesticide laden water had a higher survival rate than the other two test groups.
This is extremely significant. It means that something weird is going on. The authors ‘conclusions’ about a threshold amount of pesticide or mitigating affects of the pesticide on the chemical pathways don’t hold up under this weird condition that the fewest deaths in three case and second fewest deaths in three cases were in the rats fed the most chemicals.
That’s roughly equivalent of saying that you have a better chance of survival if you get shot twice instead of shot once.
Now, I promised a comment on the rats. The strain used in this experiment were all genetically engineered (by selection)* to be lab rats. These are highly inbred animals. The particular strain used in this test were virgin albino Sprague-Dawley rats at 5 weeks of age.
Here’s a very interesting report on the fact that 81% of Sprague-Dawley rats end up with multiple tumors after two years (which is the length of time of the study). So, the study’s authors have used a strain of rats specifically designed to develop tumors and then they are shocked when the rats develop tumors.
This leads me to another point. In another paper I referred to, my friend pointed out that the same strain of rats was used. My response was, so what? In the other paper, the authors were testing for the presence of GM DNA in the rats’ blood and organs, not whether they were getting cancer or not.
Finally, a comment on the nature of the methods. One method routinely employed in science is called the ‘blind’ or ‘double blind’ study. These two processes are used to ensure that there is no bias in experiments. To do a blind study, the person actually making the measurements is not told whether the subject is in the experimental group or in the control group. Why? Because of personal bias.
If one is not told which group is test and which group is experiment, then bias, even unconscious bias is impossible. In this case, a researcher might decide that one tumor is different because it’s in a test animal (especially considering the background of the lead researcher).
Now, to my conclusion.
Does this paper show that round-up and GM Maize are toxic?
No.
Do the points in this blog article mean that round-up and GM maize are safe?
No.
What it means is the state of round-up and GM maize is not changed by this poor study and we need to stick with previous studies that used the proper methodology and the proper analysis.
This could have been a very valuable, useful study. But it just isn’t.
________________________
* Keep in mind that maize is a genetically engineered product to (via artificial selection), it doesn’t exist in the wild.
