Members of the Steering Committee of BelTox agreed to write a letter to the editor of the journal “Food and Chemical Toxicology” on October 9th to express their concern about the very low scientific quality of a publication written by Gilles-Eric Séralini et al. entitled: “Long term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize” of which the results received a lot of publicity in Europe.


Despite the many protest letters that were sent by the scientific community to the editorial board of the journal, the decision was made to publish the article anyway. Below you can find the text of our letter. This is the first time that BelTox issued a scientific opinion on a toxicological issue.  

To Dr Wallace Hayes,
Editor, Food and Chemical Toxicology

9 October 2012


Dear Dr Hayes,


The publication of Séralini et al. (2012) came to our attention because of the shocking photographs of rats with huge masses that were shown in the media in Europe. This has a spectacular effect on the public opinion but is scientifically not common practice to correctly report the outcome of a toxicology study. As members of the Steering Committee of the Belgian Society of Toxicology and Ecotoxicology, a national scientific society, this triggered our curiosity and we all went through the publication with great attention.


The authors conducted a long term toxicology study (2 years) to compare the health status of rats fed genetically modified (GM) corn, GM corn treated with Roundup and several concentrations of Roundup in water with a control group fed near isogenic non-GM corn in the diet.


As recently published by Snell et al. in Food and Chemical Toxicology, the commonly accepted view is that “GM plants are nutritionally equivalent to their non-GM counterparts and can be safely used in food and feed”.

Challenging existing knowledge and paradigms is of course the basis of scientific progress, and re-visiting those current views might be appropriate and welcome.


Unfortunately, we came to the conclusion that these objectives were not met because of the poor scientific quality of the study. The study contrasts with the current best practices in the design, conduct, reporting and interpretation of long term toxicology studies as approved by national and international regulatory bodies. The conclusions drawn by the authors are not supported by the results generated by their bioassay. It was therefore a great surprise to us that this manuscript passed through the Journal’s review process.


We point out some of the numerous flaws in the paper of Séralini et al.:


   There were 9 test groups and one control group with only 10 rats/sex/group. This is a major deviation from the OECD test guidelines which recommend using at least 50 rats/sex/group to achieve a sufficient statistical power and to rule out chance findings


   The high incidence of mammary and pituitary tumors is a common feature in SD rats and can only be dealt with by using test groups of at least 50 rats/sex/group and a large control group if a solid in-lab historical database is not available


   No detail is provided on the nutritional balance of the diets in the test groups

   No determination of mycotoxin residues in the corn was made
   No determination of glyphosate and AMPA residues in the Roundup treated GM corn was made

   The animals exposed to the dilutions of Roundup received “control” diet, it is not clear whether this was standard lab chow (the European one?) or the diet containing 33% near-isogenic non-GM corn


   There is no mention that the animals were kept under SPF conditions


   It is not clear whether the test was conducted under GLP conditions


   It is difficult to understand why the rats were exposed to the lowest dilutions of Roundup (glyphosate + surfactant) since the ratio of glyphosate/surfactant changes after application on the corn due to different rates and pathways of degradation


   The authors state that “all data cannot be shown in one report”. That is true, but they could at least have shown summarized data tables, as normally reported in regulatory toxicology studies


   The proposal of a “threshold” effect for mortality at low doses is speculative and certainly not demonstrated by the data


   The reporting of the number of tumors per animal is not common practice in cancer studies. What is done is the calculation of the incidence (number of rats with the tumor vs. number of rats examined) per tumor type which is only determined histopathologically and the positive findings submitted to peer review by independent histopathologists


   By just looking at the bar charts with the total number of tumors per test group, we do not see any dose/response relationship, and it is difficult to follow the discussion and the conclusions of the authors


   Reporting of pathology data in Table 2 is very unclear


   The statistical analysis of the data is also problematic. The most critical data, i.e. mortality and tumor occurrence were not (could not be because of the small sample size) analyzed statistically and the conclusion of an increased incidence of tumors cannot be drawn from the presented data. In contrast, the statistical methods applied for the analysis of some of the other data (PCA, PLS, OPLS, …) are unusual and do not support the conclusion drawn by the authors.


To conclude, from a scientific point of view, it is difficult (if not impossible) to draw sound conclusions from this publication.


The controversy fueled by this publication is confusing for the consumer but also causes damage to the scientific community. The trust of the public and of competent national and European authorities in peer-reviewed scientific publications has been seriously affected by this episode.


Dr. Marie-Noëlle Blaude, Dr. Miranda Cornet, Dr. Arno Gutleb, Prof. Dr. Peter Hoet, Prof. Dr. Patrick Kestemont, Prof. Dr. Dominique Lison, Dr. Mark Martens, Dr. Francesca Tencalla, Dr. Koen Van Deun, Ir. Henk Vrijhof, Dr. Willem Penning

Members of the Steering Committee of the Belgian Society of Toxicology and Ecotoxicology