[ryanatkn]

Seems the main criticism is sample size - what was it?

> showed “no evidence of fraud or intentional misrepresentation of the data”, said a statement from Elsevier, which publishes the journal. But the small number and type of animals used in the study means that “no definitive conclusions can be reached”.

The type of animal is the same as used in the compared Monsanto study, so that seems shaky.

[grammaton]

For starters, the breed of rat used - Sprague-Dawley - is one that routinely develops cancer and tumors, and is, in fact, used in cancer research for exactly this reason. So the fact that they developed tumors is about as surprising as a sunrise, and the study in question did not have any controls or statistical compensation whatsoever for this fact. Then, of course, there's also the sample size.

[josephlord]

http://ac.els-cdn.com/S0278691512005637/1-s2.0-S027869151200...

I haven't read the full paper but the abstract does mention controls. I have no idea if they were for some reason inadequate or there were methodological errors but you are incorrect to say that there were no controls.

[grammaton]

Let me rephrase: by control, what I mean is not "control group," but "statistical control." Their statistical analysis did not properly account for the immensely increased likelihood of cancers given the breed of rat they chose.

[josephlord]

OK, I wouldn't dispute that claim without reading the full paper and the criticism of it properly. Thanks for the clarification.

[darkmighty]

From their abstract, they don't cite the breed of mice they use, but they do cite that control groups had tumors and cites the tumor rate as compared to control in addition to absolute rates. To me that makes this a non-issue considering the target audience.

[grammaton]

Stop reading the abstract and go read the paper.

"but they do cite that control groups had tumors"

Yes, of course they did - they were the same breed of rat, and as I've already mentioned, they almost invariably get tumors.

"cites the tumor rate as compared to control in addition to absolute rates"

So what? The problem is that with a single trial, and such a small sample size, there is no evidence that any differences in observed cancer rates were not simply due to random variation.

"To me that makes this a non-issue considering the target audience."

No offense, but no one really cares how it seems to you. The whole purpose of the scientific method is to get around individual biases. And what do you mean by "the target audience?" What does that have to do with anything? It doesn't matter who the target audience is - that doesn't excuse a poorly designed study. Seralini's study had a small sample size - one sixth the minimum recommended by accepted guidelines for studies of this type using this kind of animal. He's repeatedly refused to make his full data set publicly available. There's some evidence that rat feed may have been contaminated with GMO derived soy products, which would, of course, invalidate the results completely. There's also some question of multiplicity effects as well. In short, there are plenty of problems with this study, and your cursory examination of an abstract isn't going to dig up some smoking gun that invalidates all of the well deserved criticism.

[yapcguy]

Hold on a second.

The publisher was more than happy to publish the study after it had been scrutinized even though they knew it might be controversial.

Suddenly they change their mind. Why? Does this mean that all papers published by Elsevier or the journal involved should be pulled until further review?

No pun intended, don't you smell a rat?

"the retraction derives from the journal's editorial appointment of biologist Richard Goodman, who previously worked for biotechnology giant Monsanto for seven years."

In a nutshell, after publishing the paper, the journal hired a former Monsanto employee who wrote papers saying GM crops were safe. Don't you think he would want to retract a study which contradicted his own life's work?!

"Richard E. Goodman is professor at the Food Allergy Research and Resource Program, University of Nebraska. But he is also a former Monsanto employee, who worked for the company between 1997 and 2004. While at Monsanto he assessed the allergenicity of the company's GM crops and published papers on its behalf on allergenicity and safety issues relating to GM food (Goodman and Leach 2004)" http://www.greenmedinfo.com/blog/monsanto-targets-heart-scie...

[feelthepain]

Goodman has since responded to this allegation (it's in the story): "I did not review the data in the Séralini study, nor did I have anything to do with the determination that the paper should be withdrawn from or retained by the journal." The decision was already made by the previous editor. Indeed, despite peer review, papers are withdrawn every day. That's because peer review isn't perfect but the best possible system that's available.

[aristus]

Iirc it was 26 animals total. I read the paper when the blogosphere was hyperventilating about it. The data looked very shaky. If they had stopped a few months earlier they could have concluded that gmo maize prevents cancer. Way too small a population, in other words.

[rainforest]

According to §2.3, there were 200 (100 of each sex)[1]. These 200 were divided into groups of 10. One group per sex was control (standard feed, water), 6 per sex had different amounts of GMO food and 3 per sex had normal feed + Roundup-treated water.

I don't feel confident attempting to interpret or criticise the results, but I did notice there weren't any null hypothesis tests; could anyone with more experience in biostatistics say whether or not their analysis is the norm?

[1] : http://www.sciencedirect.com/science/article/pii/S0278691512...

[tedunangst]

There are some weird results in there. Table 2 has the results for males in a format that's pretty easy to eyeball.

The pituitary results are particularly weird. The more GMO you eat, the fewer anomalies you have. The more Roundup you drink, the fewer anomalies you have. Actually, that last comment holds true for every category in the last three (R) groups. Drinking Roundup is always worse than the control group, but the more Roundup you drink the closer to the control group the results become.

[1337biz]

Aren't even medical trials having most of the time not a much larger sample size than 20-50 participants?

[Sanddancer]

Depends on the test. Full-life trials, like this one, tend to have many more than that because you need to take into consideration a lot of the random factors of being alive. However, there other things to keep in mind here. The breed of rat used has a predisposition to get cancer in a real short time; more than half of Sprague-Dawley rats will get cancer within two years. As a result, they're used much more often in research centered around having cancer than getting cancer. Finally, the rats were allowed to eat food whenever they wanted, and however much they wanted. So, there's the possibility that one or more of the groups through chance just had particularly gluttonous rats. Because of this, you have a lot of questions as to what could have led to the problem -- was it the GMO corn/roundup, or did the rats feeding on GMO/Roundup corn eat more, or was it just (un)luck of the draw, etc?

[bayesianhorse]

One big difference is the cost of the studies. A study with 20-50 human participants would be orders of magnitudes more expensive than using rats.

But there just is no practical reason why this particular study lacks a control group or the statistical power to show its conclusion.

[jules]

Actually, in many countries it is really really hard to be allowed to do a study on a significant amount of animals due to animal rights groups, which have lobbied for committees who decide how many animals are allowed. Most times this number is way too low to do good science (numbers like 16 -- 8 main and 8 control mice, are not uncommon).

[grammaton]

Some studies that purport to show certain effects do, and are routinely criticized for the same reason. Proper medical trials, particularly for, say, a new drug use far larger sample sizes.