Share Button

Independent Review of the Seralini GMO-Roundup Rat Study

Citizens Oversight (2015-05-30) Ray Lutz

This Page:
More Info: Gmo Open Forum, March Against Monsanto


A lot as been said about the "Seralini Rat Study" AKA "Long term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize", published in the Peer-reviewed journal "Food and Chemical Toxicology" Volume 50, Issue 11, November 2012, Pages 4221-4231. This research was first approved, then retracted after the journal received a great deal of pressure to do so by the pro-GMO industry. Here, I will attempt a reasonable high-level review of the article and also the related "Hammond" research of 2004 which prompted the Seralini study.

The major complaints of this study were that the study groups were too small and improper study procedures were used, and that they used rats that were predisposed to getting tumors. We will look into this and compare their procedures with the reference study of Hammond.

First, some background. As a technical person from an unrelated field (I have a master's degree in electronics and computer engineering) I can hopefully provide a review of this without any ax to grind one way or the other. My experience ranges over many fields in practice, I have studied biotechnology as a side project, and have sold products into the medical device industry. Importantly, I did work as a quality assurance engineer where we employed sampling inspections all the time, and these are very much like the tests being run here.

Showing that something will not happen

It is difficult to "prove" that something will not happen. But there are standard practices to show that it is unlikely to happen using sampling inspections and testing. These methods are well understood and used everyday, but seem to be misinterpreted here.

If nothing happens, you have to make sure your sample size is large enough to show that you actually tested enough units to get past the sampling error issue. If you do get failures with small samples sizes, then it depends on how many, and you may have to increase your sample size to validate the result.

The Hammond Study

It is going to be useful to first review the Hammond Study of 2004 because this is the precursor of the Seralini study. This study was to assess the safety of GMO corn when fed to rats over a 90-day period.

Here is the full text of the Hammond study report: Hammond_et_al_2004.pdf

The test animals: Male and female Sprague–Dawley derived rats (Crl:CD 1(SD)IGS BR) from Charles River Laboratories (Raleigh, NC) were approximately 6 weeks of age at study start.

The first curious thing about this study is the large number of control groups compared with the number of treatment groups. There were two treatment groups, one getting 11% GMO corn in their food and the other getting 33%. Each treatment group included 40 rats, 20 of each sex. 80% of the rats were in control groups while only 20% were in treatment groups, and only 10% (40 rats out of the 400) were fed food with 33% GMO corn.
  • Hammond - Treatment Groups:

The food included up to 33% GMO corn.

But the corn did not have any glyphosate. This seems a bit artificial. Although they say it was grown using Round Up, it would seem that some residue would exist either on the surface or internal to the plant. This is a big contrast with the Seralini study which did include glyphosate at levels not exceeding those actually found in practice.

Very importantly, the Hammond study resulted in an early fatality of one male rat, almost at the end of the 90-day period.

Without any real reason for doing so, they say this fatality has nothing to do with the GMO diet. This is a big jump and I can't imagine that everyone in the industry did not worry about why this death occurred, and called therefore for more testing. Given the very small size of the high-dosage treatment group (only 10% of the lot) to have a death ONLY in this group is warrants more than the simple brush off they give it. This is a very big problem with this study, not so much that it happened, but because the experimenters ignored the results and promoted their research as showing that GMO corn was safe. In all honesty, you can't disregard that dead rat. It is not "missing data." It is the most important data point, signifying that GMO corn may be VERY dangerous indeed.

They also noticed that the urine chemistry of male rats had changed but was "within two standard deviations" of the control groups. But that could be significant and given the small size of the high-dosage treatment group (only 10% got 33% GMO food), this is cause for concern, but they decide to ignore it.

Later, Hammond goes into a great deal of time and space to compare their study with other studies and to paint a picture that everything is fine, while dedicating almost no time to explain the huge problem of the dead rat. Such a review of other studies may be common, but is not required in a technical article documenting a study. Since almost no time was spent to explain the death, and as this is not in the abstract either, this underlines that they definitely wanted to help the reader reach a conclusion that GMO crops were safe. But as we continue to read, they also say that the concern is reduced because the percentage of crops that are GMO is low and so the feed is not concentrated. This whole passage leads one to believe that there really is a risk in eating too much GMO food. This is coupled with the fact that they did not find a reason for the dead rat, or perhaps they actively elected to suppress that information. Giving them the benefit of the doubt, it may be that the mechanism of death was simply not apparent using the inspection techniques used.

After looking over this study, there are some red flags:
  • Glyphosate (and related roundup ingredients) not at all on/in the food. This is artificial.
  • Many control groups (8) and very few (2) treatment groups -- only 10% at the high dosage.
    • standard practice is to have three dosage levels and an equal or perhaps double size control group. Thus the control group would be 20% to 40% of the total subjects rather than 80%.
  • Dosage limited to 33%. Why not have the rats eat 100% GMO food?
  • A fatality in the high-dosage test group that is blown off as nothing.
  • Changes in urine chemistry of the high-dosage test group is blown off as nothing because it is "within two standard deviations" of the control groups. Anything more than 1 standard deviation is outside the "normal" region. That's why they call it the standard deviation.
  • Extensive review of other studies within the study document... this is out of line and a distraction, but may indicate that they were pushing for an overall result.
  • Precautionary text at the end which explains that the "risk" is further reduced.

There are EPA Standards for toxicology tests using rodents for 90 days.

  • EPA requires 20 rodents in each group, 10 males, 10 females:

  • EPA requires 1 control group apparently of the same size as other groups:

So the fact that the Hammond study used only two dosages and had 80% controls (eating non-GMO food) makes you wonder what they were up to, and why the study was not rejected for not having sufficient test subjects, and now real explanation for the dead rat at 82 days into the study. If I were to read this, I certainly would not be very comfortable about the safety of GMO food.

The Seralini Study

With that as background, we move to the Seralini study.

Although retracted, the study report is available and you can read it at this link:

In the report, this important comparison table is presented to compare the Seralini study with the Hammond study, and the standard study requirements.

Although there is some consternation about the type of rats used, they are the same as those used in the Hammond study: "Virgin albino Sprague-Dawley rats at 5 weeks of age were obtained from Harlan (Gannat, France)."

The Seralini report also commented on the number of rats per group, and the existence of a control group. Seralini used three dosage levels, 11%, 22%, and 33%. They used groups of 20 rats each, with 10 males and 10 females, at various dosages and food compositions, and one control group.
  • Seralini -- Comment on Rats per Group:

The Seralini study makes an important comment on Combinations of factors that would reflect the real world situation. This is contrasted with the Hammond study where there was no glyphosate residue on the food.

The food used in the Seralini study is more like the real world. The Hammond Study did not include any Roundup R in the food, while the Seralini study did, although it states, "with no contaminating pesticides over the standard limits." Also, there is some glyphosate dissolved in drinking water, about what is expected in the environment.

Seralini presents startling results, with mortality in the treatment groups two to three time that of the control groups.

And interestingly, the results were not proportional to the GMO dosage in the food. It seems getting any at all was a problem

Many reviewers have said that the Seralini results are untrustworthy because the rats used were the variety that are prone to get tumors for cancer studies. The concern makes no sense, since there were 50% to 80% more tumors in the treatment groups:

Now, if we saw no harm to the rats that could be distinguished from the control group, then we might be concerned about sample size. But the differences are stark. Indeed, long term study design suggests that the control group should be twice as large as the treatment groups and more subjects per dosage (50 of each sex). Thus a standard long-term test would include 300 rats in treatment groups (50 of each sex) and 200 controls, or 500 rats. This would certainly allow better differentiation between normal deaths and fatalities but I find it hard to understand how the 50% to 80% larger tumors will change much just with more subjects.
  • Typical Long Term Study Design:

We wonder why such low dosages have been used, when normal procedures suggest much higher dosages are the norm. Thus, why not feed the rats 100% GMO food laced with glyphosate at 10x what anyone would ever encounter? That seems prudent, but we can tell from these results with VERY LOW DOSAGES that there are already problems.
  • Dosage, from Principles and Methods of Toxicology, pg 340:


Although the Seralini study could have been improved with 5x more rats in each study group, the results are so startling and clearcut, that increasing the numbers will hardly make the effect go away. A new follow-on study could easily be done with 50 rats of each sex in each study group and a larger control group. But this concern is only merited if there were no real differences between the groups at all. That is the time you need to up the numbers. If any study had too few test animals, it was the Hammond study, with 80% of the rats in control groups, and only 10% in a high-dosage test group, and that was only at 33% GMO food. And even with their super-biased study design, they had an early fatality and ignored it as nothing.

It is clear to me that the scientific community in the GMO field has been captured by the industry where it is almost professional suicide to posit that GMO crops and the associated hericides are dangerous. These studies indicate that there is reason for concern.

The Study we need

The science that should be done at this point is a similar study to the Seralini study, but with 100%, 50% and 10% GMO food, glyphosate (and other Round Up constituent) residue, with 100 rats in each dosage, and a control group with 200 rats. (all groups with sexes evenly represented.) If the rats show any problem (i.e. difference from the control groups) I assert such GMO food and round up should be systematically pulled from use.

The GRACE study

The GMO Skepti-forum (GMOSF) responded to this review by noting that the EU has started a new study "Ninety‑day oral toxicity studies on two genetically modified maize MON810 varieties in Wistar Han RCC rats (EU 7th Framework Programme project GRACE)". First of all, this study does not propose a two-year study, but only a 90-day study, although there is perhaps some other effort to also do a longer study.

The number of animals in the study is a bit of a concern. First, they calculate the number needed as 17, then they round it down, to 16, per sex per group.

  • GRACE study dials down the number of animals per group, then rounds down.:

The number should have been rounded up, for robustness and respect for rigor. Even the Hammond study had 20 rats per group per sex, with 10 groups, they had 400 rats (but as mentioned, 8 of the groups were controls). In that study, as we see above, there was an unexplained death, only in the high-dosage treatment group.

Here is a page with an overview of Power Analysis and the value of 17 for 80% power, 2-sided test and 95% confidence:
  • See this page:
  • Sample Size by Power Analysis:
  • Note that they recommend that in toxicology, they recommend a power of 90%.
  • To select the proper "minimum" number, you need to know the standard deviation of the effect, so as to determine the Signal to Noise ratio.

This is not a debate that the test may have enough "power" to detect the normal variations due to feeding the rats the GMO maize vs. controls. This is about the fact that
  1. We already saw an unexplained death in the Hammond study, and they also had a rat that had to be euthanized due to a broken nose. These things happen with live animals. We need a bit of a buffer between the absolute minimum number and the actual number used.
  2. It is prudent to include additional numbers in the study to avoid any claim that the study is invalid.
  3. Hammond had 20 animals per sex per study group. This test should be at least as strong.

This issue has been brought up to the GRACE project with a comment from Citizens Oversight. Anyone else can submit their own comment.

What is certainly clear is that this GRACE test IS NOT an attempt to redo the Seralini test. Instead, it is more like the test run by HAMMOND. We notice:
  • It does not include three dosage levels. There is a claim that dosage is irrelevant. That fact is very important if true. It would mean that we should avoid all contact with GMO foods to avoid the risk, Thus this study protocol should include three levels.
  • We are concerned that the highest dosage of GMO food is only 33%. The argument that it is unhealthy to feed the rats 100% corn, regardless of whether it is GMO.
  • GRACE does not use Round up during the cultivation of the crops. This is an artificial removal of any glyphosate and related products from the corn, and makes the study nearly worthless. Thus, this study does not directly address the issues addressed by HAMMOND and certainly not that addressed by Seralini.

GMO Skeptic Forum

This was posted on the GMO skeptic forum at this link:

Out of rudeness, they placed the link in the "Explanations for a 5 year old" and then banned me from the group, they said because

"Ray, I regret to inform you that you have been removed from the GMO Skepti-forum for failing to provide evidence that Chad Niederhuth was requesting. If you choose to return to GMO Skepti-Forums, you will need to provide this evidence via PM to me and we will go from there.

Evidence required >>> I am going to call you out right now. Provide evidence that 16 rats per sex per group is less than the minimum required for normal oral acute toxicity studies. If you cannot, then admit that GRACE has enough power


Goto Discussion

Media Form edit

Title Independent Review of the Seralini GMO-Roundup Rat Study
Publisher Citizens Oversight
Author Ray Lutz
Pub Date 2015-05-30
Media Link
Remote Link
Embed HTML
Forum Link
Keywords Gmo Open Forum, March Against Monsanto
Related Keywords
Media Type Article
Media Group News, Research, Blog Entry
Curator Rating Plain
Author Name Sortable
Thumbnail Link
I Attachment Action Size Date Who Comment
Dosage,_from_Principles_and_Methods_of_Toxicology,_5th_Ed,_pg_340.JPGJPG Dosage,_from_Principles_and_Methods_of_Toxicology,_5th_Ed,_pg_340.JPG manage 36 K 31 May 2015 - 01:57 Raymond Lutz Dosage, from Principles and Methods of Toxicology, pg 340
EPA-HQ-OPPT-2009-0156-0010-90-day-oral-toxicity-in-rodents.pdfpdf EPA-HQ-OPPT-2009-0156-0010-90-day-oral-toxicity-in-rodents.pdf manage 31 K 30 May 2015 - 23:55 Raymond Lutz EPA standards for toxicology tests in rodents
EPA-one_control_group_required.JPGJPG EPA-one_control_group_required.JPG manage 51 K 31 May 2015 - 00:01 Raymond Lutz EPA requires 1 control group apparently of the same size as other groups
EPA_standards_on_numbers_of_animals_for_90-day_toxicology_test.JPGJPG EPA_standards_on_numbers_of_animals_for_90-day_toxicology_test.JPG manage 19 K 31 May 2015 - 00:00 Raymond Lutz EPA requires 20 rodents in each group, 10 males, 10 females
GRACE-FeedingTrials_AB_ArchToxicol_2014.pdfpdf GRACE-FeedingTrials_AB_ArchToxicol_2014.pdf manage 1 MB 31 May 2015 - 18:55 Raymond Lutz GRACE Study: Ninety‑day oral toxicity studies on two genetically modified maize MON810 varieties in Wistar Han RCC rats (EU 7th Framework Programme project GRACE)"
GRACE-NoRoundUpUsed.jpgjpg GRACE-NoRoundUpUsed.jpg manage 55 K 31 May 2015 - 19:15 Raymond Lutz GRACE study does not apply Round Up to RR crops.
Hammond-DietaryExposure-ExtraptoHuman-BasedOnPercentCrops.JPGJPG Hammond-DietaryExposure-ExtraptoHuman-BasedOnPercentCrops.JPG manage 86 K 31 May 2015 - 01:56 Raymond Lutz Hammond - Explains that low percentage of crops makes it safe
Hammond-FoodForTest.JPGJPG Hammond-FoodForTest.JPG manage 50 K 30 May 2015 - 23:41 Raymond Lutz Hammond - Food for Test
Hammond-NeitherDeathConsideredTestArticleRelated-Why.jpgjpg Hammond-NeitherDeathConsideredTestArticleRelated-Why.jpg manage 13 K 30 May 2015 - 23:51 Raymond Lutz Hammond - Ignores the dead male rat.
Hammond-NoglyphosateInDiets.JPGJPG Hammond-NoglyphosateInDiets.JPG manage 23 K 30 May 2015 - 23:48 Raymond Lutz Hammond did not include glyphosate or Roundup in diets.
Hammond-OneMaleDeadOnDay82.JPGJPG Hammond-OneMaleDeadOnDay82.JPG manage 28 K 30 May 2015 - 23:50 Raymond Lutz Hammond - One male dead on day 82
Hammond-TreatmentGroups.jpgjpg Hammond-TreatmentGroups.jpg manage 38 K 30 May 2015 - 23:44 Raymond Lutz Hammond - Treatment Groups
Hammond-UrineChemistryAltered.jpgjpg Hammond-UrineChemistryAltered.jpg manage 33 K 30 May 2015 - 23:52 Raymond Lutz Hammond - Urine Chemistry Altered - Ignored
Hammond_et_al_2004.pdfpdf Hammond_et_al_2004.pdf manage 470 K 30 May 2015 - 23:37 Raymond Lutz Hammond, et al, 90-day GMO test on rats (no glyphosate)
Ingredients_of_Lab_Rodent_Meal.jpgjpg Ingredients_of_Lab_Rodent_Meal.jpg manage 35 K 31 May 2015 - 19:08 Raymond Lutz Ingredients of Lab Rodent Meal, typical, from Lab
ProtocolComparison.JPGJPG ProtocolComparison.JPG manage 72 K 30 May 2015 - 23:22 Raymond Lutz Protocol Comparison Table from Seralini Report
Quote_from_GRACE_study_calculating_17_rats_per_group.jpgjpg Quote_from_GRACE_study_calculating_17_rats_per_group.jpg manage 40 K 31 May 2015 - 18:41 Raymond Lutz GRACE study dials down the number of animals per group, then rounds down.
RetractedSeraliniStudyReport.pdfpdf RetractedSeraliniStudyReport.pdf manage 2 MB 30 May 2015 - 23:21 Raymond Lutz Retracted Seralini Study Report
SampleSizeByPowerAnalysis.jpgjpg SampleSizeByPowerAnalysis.jpg manage 134 K 31 May 2015 - 20:59 Raymond Lutz Sample Size by Power Analysis
Sarelini-MortalityResults-2-3xMoreInTreatmentGroups.jpgjpg Sarelini-MortalityResults-2-3xMoreInTreatmentGroups.jpg manage 115 K 30 May 2015 - 23:23 Raymond Lutz Seralini Results -- Mortality
Sarelini-NotProportionalToDose.JPGJPG Sarelini-NotProportionalToDose.JPG manage 71 K 30 May 2015 - 23:26 Raymond Lutz results not proportional to dose
Sarelini-Results-80MoreTumors.JPGJPG Sarelini-Results-80MoreTumors.JPG manage 70 K 30 May 2015 - 23:27 Raymond Lutz Seralini - Results, 50% to 80% more tumors
Sarelini-Test-Food.JPGJPG Sarelini-Test-Food.JPG manage 45 K 30 May 2015 - 23:24 Raymond Lutz Seralni -- Test Food
Seralini-CommentOnCombinations.jpgjpg Seralini-CommentOnCombinations.jpg manage 76 K 30 May 2015 - 23:24 Raymond Lutz Seralini - Comment on Combinations
Seralini-RatsPerGroupComment.jpgjpg Seralini-RatsPerGroupComment.jpg manage 52 K 30 May 2015 - 23:25 Raymond Lutz Seralini -- Comment on Rats per Group
TypicalLongTermStudyDesign-AnimalToxicologyStudies.JPGJPG TypicalLongTermStudyDesign-AnimalToxicologyStudies.JPG manage 35 K 30 May 2015 - 23:29 Raymond Lutz Typical Long Term Study Design
Topic revision: r6 - 24 Dec 2017, RaymondLutz
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Cops? Send feedback