Rat food contamination study questioned – Expert Reaction

A new study reports that the near-ubiquitous contamination of  ‘chow’ used to feed rats in experiments maybe causing spurious results in toxicology studies. However independent experts say the results may in fact show the opposite –  that there is no cause for concern.

chowThe study, published today in PLOS ONE, examined GMO content and the levels of pesticides, heavy metals and other contaminants in 13 varieties of laboratory rat food from around the world. To varying degrees, they identified contaminates in all samples.

Combining the potential health effects of the contaminants in a measure called a summed health quotient, the authors say many of the diets could be toxic to rats and causing misleading results in animals studies, especially toxicological studies.

However a number of independent experts say the measures used in the study and the conclusions drawn by the authors may be misleading, and that more experimental research is need to prove their hypothesis.

Expert commentary collected by the UK Science Media Centre and the US Genetic Expert News Service (GENeS):

Prof Cathie Martin, Group Leader, John Innes Centre, said:

“The conclusions of Seralini et al in this current manuscript are highly speculative. For their interpretations to be compelling, Seralini et al would need to do feeding experiments with ‘non-contaminated chow’ and show that there is a significantly lower level of cancer incidence in those animals on the non-contaminated chow (using an appropriate number of animals, over an appropriate amount of time) – this study has not done that.

“Seralini et al conclude that one cannot trust historical control data, but the tumour incidence of Sprague Dawley rats was published several decades ago (CANCER RESEARCH 33, 2768-2773, November 1973), when the contamination values for rodent chow were likely to be very different. Reports of the high cancer incidence of Sprague-Dawley rats were made long before the advent of GMO crops, so it cannot be concluded that GMOs are the cause of the high tumour incidence. The high tumour incidence in Sprague Dawley rats, which underpins the recommendation that they should not be used for studies beyond 18 months old, is most likely their high level of inbreeding.

“It is strange that the conclusion of Seralini et al is that one cannot trust any experiments done with ‘control’ rats because of dietary contaminants.  A more reasonable conclusion would be that no one can trust data purporting to show that glyphosate and glyphosate resistant corn cause tumors in Sprague-Dawley rats, when they have a high inherent incidence of tumors in animals older than the recommended 18 months (whatever the cause).”

Prof Andy Smith, Senior Scientist at the MRC Toxicology Unit, University of Leicester, said:

“The possible low level contamination of rodent diets and bedding has been recognised for decades in toxicological investigations but the data in the study of Mesnage and colleagues are useful additions to our background knowledge. With all the controls used, including historical controls, any such ‘contaminants’ are unlikely to affect conclusions drawn from most toxicity tests in rats.  In addition, it is very well established over many years that different strains of rats and mice develop particular diseases over their life, just as some people are more likely than others to develop some types of cancer.  In contrast to the suggestions of the authors, the use of this reproducible genetic variation, seen in historical controls and independent of diet, rather than being detrimental, has been exploited to greatly help our understanding of the risks to humans from the use of chemicals and consumption of novel foods.”

Prof Tamara Galloway, Professor of Ecotoxicology, University of Exeter, said:

“This is an interesting study that sets out to determine whether food chow intended for laboratory rodents is contaminated with residues of pesticides, metals and/or industrial chemicals. The chemical analysis has been performed by accredited laboratories and appears robust and of good quality. The authors found that a number of the test diets contained residues of pesticides, metals and industrial chemicals, and this is perhaps unsurprising given their widespread use, but it is reassuring to see that the concentrations of these compounds are below regulatory limits.

“In the paper, the authors have calculated a hazard quotient, which they then compare with acceptable daily intakes (ADI) calculated for human consumption. Since the ADI is calculated using a safety factor of 100, a direct comparison of the two estimates could give the wrong impression, since the safety factor is not taken into account.

“The authors do not measure the biological or health effects of the food and hence it is not possible to say anything from the paper on the likelihood that food contamination might have a role to play in the high incidence of health problems in laboratory rodents, as speculated by the authors. The discussion speculates beyond the evidence presented in the paper.

Prof Alan Boobis, Professor of Biochemical Pharmacology, Imperial College London, said:

“The paper of Mesnage et al provides a useful survey of the levels of a number of contaminants present in the feed of laboratory rodents.  However, I believe the title is potentially misleading in the use of the term “toxic levels”. Firstly, in general the levels of individual contaminants are extremely low. Note that a hazard quotient (HQ) of 1 would represent exposure at 1% of the critical no observed adverse effect level (NOAL).  Secondly, it is assumed that these low levels would all exhibit dose additivity (HQs were summed), regardless of mode of action or potential site of effect.  This is contrary to the approach that EFSA, and others, recommend for risk assessment of combined exposures.  However, even in this situation, the summed HQs are in all cases less than 100, indicating that total exposure is below the NOAEL. No information is provided on whether there is any relationship between the level of contamination found and background rates of pathology in the rodent populations.  However, such information would only be indicative, and to verify the conclusions of the authors, a specific intervention study would be necessary, using diets with a range of verified levels of contaminants. The use of laboratory animals for this purpose would be questionable.”

Dr Nick Plant, Reader in Molecular Toxicology, University of Surrey, said:

“A comprehensive assessment of potential contaminants in animal chow is welcome, and this study has been undertaken in a robust manner by nationally accredited laboratories. It is not unsurprising that some level of contamination exists given what is known about agricultural practices worldwide, and indeed it is pleasing to see that in the main the levels are only just above the limit of detection and below the regulatory levels. As such, this work is reassuring.

“The authors’ comment that such contamination may contribute to spontaneous pathologies in chronic studies may be justified, although this would need further study to prove. However, their assertion that this invalidates the use of historical data is questionable. Historical data provides a wealth of information on the background rate of pathology, even if this is driven (at least in part) by variable food contaminants. Not only that, animal studies always include a control group who will be fed the same diet as the treatment groups, proving an internal control for diet-induced effects.

“In the paper the authors refer to a hazard quotient (HQ).  This is a way of putting a single value on the level of several different hazards, in this case the hazards represented by the various components of the animal feed.  In feeding experiments, the hazard quotient should be compared to the level of feed at which no observed adverse effect is expected (the NOAEL).  However in this paper the authors compare the hazard quotient instead to a different measure, the acceptable daily intake (ADI), which is usually 100-fold lower than the level at which no adverse effect is expected.  This means that the hazard quotient calculated in this paper will substantially overestimate the cumulative risk of the feed.

“In summary, while this is a useful piece of work and adds to the body of evidence that allows us to interpret animal experiments correctly, it does not support the concerns raised by the authors on the use of historical data to interpret such studies.”

Prof Tony Dayan, Emeritus Toxicologist, said:

“Professor Mesnage and his colleagues have reported analyses of single samples of 13 different rodent diets. The nature of the data and interpretation of the results is not straightforward in this very complex area of experimental toxicology.

“It appears to have been assumed that the samples of diets were representative samples which had not been contaminated prior to analysis. It may be more important that the quality of the feed provided for animals in regulatory tests has to comply with legally binding Good Laboratory Practice regulations (GLP), which include regular analyses to demonstrate amongst other things that contamination with heavy metals, pesticides and other substances is kept below strict limits. No information is provided to show that the batches reported in the new work were of a quality compliant with GLP and so might have been used in carcinogenicity or other formal tests.

“Professor Mesnage and colleagues make many comparisons between their findings and ADI (‘Acceptable Daily Intake’) and HQ (‘Hazard Quotient’) values, suggesting that the results may indicate risks to rodents. Those concepts and the calculated values are empirical factors derived and successfully employed over many years specifically to assess potential risks to humans. There is no experimental justification or experience to apply them to rodents as done here.

“In attributing many of the diseases that may affect aging experimental rodents to the widespread chemical contamination of their diets claimed here Professor Mesnage and colleagues have not taken into consideration the considerable amount of evidence linking those disorders to the considerable over-nutrition that follows the ad lib feeding of the rodents. That is widely considered to be a major factor in the rising incidence of background lesions in elderly rats and mice.

“We rely to a large extent on the results of various experiments in rodents to predict and evaluate potential chemical hazards in humans and other species but many other aspects of the substances tested and the circumstances and extent of exposure are also evaluated in suggesting real risks in practice. The breadth and complexity of risk assessment has been overlooked in the conclusions suggested in this paper.”

Prof Maurice Moloney, CEO of the Global Institute for Food Security, Canada (An Institute of the University of Saskatchewan), said:

“The underlying hypothesis of this paper is the contention that the development of tumorous growths and carcinomas on the widely used Sprague-Dawley rat is the result of the toxicological effects of contaminants carried over into the feed pellets of lab animals from agricultural cultivation. Special emphasis is placed on the GM nature of some of the source materials and to one particular herbicide, glyphosate, or when formulated ‘RoundUp’. If this hypothesis were upheld, the authors believe that it means that most of the toxicological studies using this strain of rats are invalid, because of the intrinsic pernicious nature of the feed itself rather than the chemical (or other treatment) being tested. The previous study from this group was retracted from the Journal of Food and Chemical Toxicology, based on serious flaws identified after publication that should have been corrected in peer review, but which rendered the work inconclusive. The current work is no less inconclusive than the previous study, but nevertheless does offer some new data not heretofore published.

“The actual measured data are probably all accurate and expected given the source materials. The group has monitored for a large number of common pesticides. It is unsurprising what was found based on the formulation of commercial rat chow. Perhaps the biggest surprise is that relatively few were found at toxicologically meaningful levels. The lumping together of GMOs or glyphosate with organophosphates or heavy metals is “guilt by association”. These results show no correlation, let alone causality with either GMOs or glyphosate. Linking these to pyrimiphos (an acetylcholine esterase inhibitor) or a heavy metal like lead, both of which do seem to have residues that could be toxicologically meaningful, is unwarranted and misleading in my opinion.

“The best evidence for interference of GM sources and glyphosate would be toxicological data with these rats in the absence of these agents. Although not presented here, those data do exist, because the GM trait and use of glyphosate is a post-1990 phenomenon. Yet, the literature is replete with reports of Sprague-Dawley rats decades earlier showing exactly the same pathologies when the rats are used in long-term studies for 18 months to 2 years. Clearly, this is not related to GMOs or glyphosate.  Whether the levels of organophosphates or certain heavy metals are pertinent is simply not shown here as no controls are provided. These controls could have been performed if the authors had availed themselves of pesticide-free ingredients and screened for low heavy-metal residues. Without this, the paper lacks even correlation and certainly does not establish causation. In short, it cannot displace the current well-supported hypothesis that these pathologies are essentially genetically controlled and which appear in each new generation of Sprague-Dawley rats.

“In summary, if this paper is an attempt to support the previous (now retracted) study by this group, it falls far short of doing so. The new paper suggests we ought to doubt an entire testing system but does not provide the evidence to back up such doubt.

“By analogy, this is like a football team that is consistently disallowed goals, because they fall prey to an offside trap, calling into question all the laws of football. They do this rather than simply imposing discipline on their strikers not to move until the ball is in play. This paper lacks discipline in its discussion of the data, even though in the case of organophosphates and heavy metals, it might have discovered something that warrants further investigation. Food with minimal known toxin residues would help to ensure that test animals are treated ethically. For Sprague-Dawley rats, an ethical prerequisite is to use them only as intended for 90 day trials. Beyond that, as we saw in the revoked paper in the Journal Food and Chemical Toxicology, the test animals are subjected to significant distress, most likely due to a well-known genetic predisposition to form tumours.”

Dr Graham Tobin, Technical Director, Harlan Laboratories, Europe/ROW, said:

“The authors bring needed attention to the often overlooked role of diet in laboratory animal safety and toxicology studies.

“Contaminant levels in diets are a function of their ingredient composition, and the geographical source of those ingredients. A primary source of variation in contaminant levels between the different diets in this study was likely related to the diet formulations. It is known that certain ingredients are more likely to contain specific contaminants.

“Diets fed in safety and toxicology studies must be monitored for contaminant residues in order to meet Good Laboratory Practices (GLP). In regards to the 13 diets analyzed for this study, no diet exceeded current established contaminant criteria in any of these standards used in the US and Europe.

“The argument is flawed on the following basis:

  1. No data or compelling literature support was presented to demonstrate the contaminant levels found in the diets will contribute to these specific pathologies within rats.
  2. The safety assessment was predicated on human ADI levels that are not properly scaled to assess safety risks in rodents.
  3. The argument is too simplistic, ignoring other factors such as genetic predisposition, and practical experience. It overlooks the well-established effect of dietary restriction to improve survival and reduce pathology; this beneficial effect occurs in the absence of a reduction in contaminant intake per unit body weight.

“This paper addresses an important subject, but does a disservice to laboratory animal science by inflating the potential risk of typical dietary contaminant levels through use of an inappropriate metric.  On the basis of the evidence presented it certainly seems inadvisable and unwarranted to apply the adjective ‘toxic’ to the levels of contaminants reported here.”

Declared interests

Prof Cathie Martin: “I co-founded a spin-out company, Norfolk Plant Sciences, with the objective of commercialising health-promoting tomato products. These tomato products will, in some cases, have been produced by genetic modification. I am a holder of a patent (granted in the USA) giving me freedom to operate in producing tomatoes biofortified in flavonoids.”

Prof Andy Smith: Member of the Expert Committee on Pesticides, Member of Panel and Working Groups of EFSA.

Prof Tamara Galloway: “I’m a Professor of Ecotoxicology at the University of Exeter. I’m a member of the Hazardous Substance Advisory Committee which provides expert advice on how to protect the environment, and human health via the environment, from potentially hazardous substances.”

Prof Alan Boobis: “I have no direct or indirect financial interests related to the subject of the paper. I have a number of indirect interests, involving collaborative research and participation in scientific advisory committees, addressing the interpretation of results from toxicity studies in laboratory rodents.”

Dr Nick Plant: “I am a member of the UK Committee on Toxicity, which is an independent body that advises Government on the risks associated with chemicals in food, consumer products and the environment. As such, we have assessed the risk of contaminants in various products, although never animal chow.”

Prof Tony Dayan: “In the past 40 years I have been a toxicological adviser to many official agencies in Britain, Europe and WHO and to a large number of national and international chemical, food, pesticide and pharmaceutical companies. In my partial retirement over the past 10 years I have not been involved with practical carcinogenicity testing in animals for any specific company or industry.”

Prof Maurice Moloney: “The Institute’s core benefaction is from Potash Corp (a fertilizer, not pesticide, company) and the taxpayers (Govt) of Saskatchewan.  Our research is at arms- length from all donors, by its founding charter. Previously, I have directed Rothamsted Research and CSIRO’s life science portfolio. Both organisations have a history of collaboration with conventional agribusiness, biotechnology companies and agroecology proponents.”

Dr Graham Tobin has been involved in laboratory animal nutrition for many years in his work at Harlan Laboratories.



Further commentary collected by the US-Based Genetic Expert News Service (GENeS).

Dr. Carl Winter, Cooperative Extension Food Toxicologist, University of California, Davis (webpage):

Expertise: detection of pesticides and naturally-occurring toxins in foods, how to assess their risks and identify how to use the science in the regulatory decision-making process.

“I find the discussion of data presented in Table 2 to be misleading.  The authors contend that since the maximum dietary intake of the pesticide pirimiphos-methyl in seven of the diets exceeds the Acceptable Daily Intake, that animals fed such diets are being fed toxic levels of pirimiphos-methyl.  This conclusion is not supported for two reasons:

1) The maximum dietary intake represents an exaggeration of the true dietary intake.

2) (More important) Comparing maximum dietary intake with the Acceptable Daily Intake is not appropriate to demonstrate risk.  A more appropriate comparison would be the No Observed Effect Level (NOEL) from long-term animal toxicology studies, which represents the maximum amount given to laboratory animals on a daily basis that does NOT cause any noticeable toxicity.  In the case of pirimiphos-methyl, the lowest NOEL I was able to find was from long-term rat feeding studies, which represented a level of 500 micrograms of pirimiphos-methyl per kilogram of body weight per day.  Even the exaggerated maximum dietary intake levels are well below the rat NOEL level.

“ADI levels represent levels considered to be safe for daily human consumption.  Calculation of ADI levels typically takes the lowest NOEL found from animal studies and applies uncertainty factors that typically include a 10-fold factor that assumes humans are 10 times more sensitive than the most sensitive animal species studied which is multiplied by another 10-fold factor assuming some humans are 10 times more sensitive than other humans.  Thus, a typical uncertainty factor is 100, so the ADI is often the NOEL divided by 100. See this manuscript I wrote that discusses this in more detail.  Since the PLOS One study specifically considers the impacts of exposure to chemicals to laboratory animals, there is no need to apply the uncertainty factor.

“For all of the other seven pesticides detected, exposure at the exaggerated maximum dietary intake level was still below the ADI levels, (commonly 100 times lower than NOEL levels from animal studies), so it is difficult to make a valid case as to how such exposures would cause effects in the animals consuming feed containing pesticide residues.  It should also be pointed out that these were the only pesticides detected out of 262 analyzed.

“The process of comparing maximum dietary intake levels with ADI levels for other contaminants such as metals and PCDDs and PCBs is similarly misleading.”

Dr. Frédéric Y. Bois, Professor at Sorbonne UTC, and Research Director at L’Institut National de l’Environnement Industriel et des Risques (INERIS)(webpage):

Expertise: pharmacology, quantitative toxicology and risk assessment, with a strong emphasis on mathematical modeling and statistics

The authors have performed an independent assessment of the contamination of lab rat diets by pesticides, heavy metals and some persistent chemicals. Some producers already check the quality of the diets they produced, according to FDA guidelines for example, but a double check is always good. Not too surprisingly, rat diet, like most of ours by the way, is contaminated with those chemicals. It would actually be interesting to compare those results with the producers’ records.

“Whether the exposed rats are likely to suffer serious damage from such exposures is not clear though. The hazard indexes computed by the authors use acceptable daily intakes (ADI) calculated for humans. Those do not really apply to rats, because safety factors are used in their derivations to specifically protect humans (assumed to be more susceptible than animals), and among humans infants etc.  

“Furthermore the hazard indexes for various chemicals were summed to get an estimate of the effect of the multiple exposures. However, this is a very crude method: for example, take three chemicals that have indexes at 0.5 (no problem for them: the exposure level only half of the ADI). If you sum the indexes for those three chemicals you find a global index of 1.5 and we have a problem. But that makes sense only if the chemicals act without thresholds and on the same target, which are all assumptions. 

“Using hazard indexes and summing them maybe a reasonable conservative procedure for protecting human public health. However, when assessing the actual impact of food contaminants on the health of lab animals (which is a matter of scientific precision, not of rat health protection) they are likely to overestimate the risks. Animal specific ADIs should be used.

“As to estimating precisely the effect of mixtures of chemicals, this is a hot and debated topic in food safety (the European Union has recently given the green light to a large project, EuroMix on that topic) but we do not have yet good tools to do it precisely. 

“Given that, it goes very far to say that most rat feeding studies are flawed or that ‘historical control data’ are unreliable. In the case of low quality diets, some studies may have overestimated the “natural” disease background in the animals. The question is a good one, but should be answered with a more in-depth analysis of the actual risks for those animals. 

“The question also of the number of chemicals whose safety evaluation has been biased by the presence of food contaminants in lab animal diets needs to be addressed with precaution, without jumping too fast to conclusions. At least the authors should be congratulated for tackling an interesting question, which, by the way, also extends to the analysis of epidemiological cohorts. In essence, what is a control group? The question again is welcome.”

Dr. Richard E. Goodman, Research Professor in the Food Allergy Research and Resource Program, University of Nebraska (webpage):

Expertise: Refining methods and evaluation criteria for assessing the potential allergenicity of proteins in genetically engineered crops

“Se?ralini and colleagues have submitted an analytical study reporting on detection of traces of 262 pesticides, 4 heavy metals, 17 dioxins and furans, 18 PCBs and 22 GMOs in a small sample (13) of commercially prepared laboratory animal chow (diets) produced by a number of manufacturers.  The study is being reported in the open access journal PloS ONE, that has a generally favorable reputation.  But of course even the highest quality journals sometimes publish papers that report sensational findings, yet with little critical review and a lack of connectivity to biological outcomes, as appears to be the case in this study.

“The authors are using a controversial method of estimating cumulative risks (Hazard Quotients) of minor contaminants in animal feeds that have not been validated to demonstrate realistic risks, with biological proof of harm at levels of contamination that the authors found in this PLoS ONE study.  As far as I know, there have not been publications demonstrating that the method of cumulative hazard assessment, even though referenced by a paper by the US EPA and one in the EFSA Journal (11:3313), have suggested this approach might be useful for understanding risks of mixtures of chemical toxicants including pesticides, dioxins and heavy metals have accurately predicted risks or outcomes.

“The fact is that no studies I am aware of, other than the authors’ previous retracted study in Food and Chemical Toxicology in 2012 on NK603 maize and glyphosate, has shown high levels of tumors, cancers or other deleterious effects in control animals fed normal commercial rodent diets.  The authors have not cited published studies that have shown high levels of adverse effects in control animals fed commercial rodent diets.  I have been involved in monitoring rodent studies and in conducting studies of 60 and 90 days in rats and mice and am aware of many studies that are longer that have not shown anything other than occasional adverse effects in control animals that are not associated with an unintended genetic or infectious trait carried in the test animals, and those are sporadic issues that provide some of the background that the authors cite as issues.

“It is quite disturbing to me as a scientist involved in food safety evaluations of GMOs and novel food ingredients as well as all kinds of processed foods that the authors of this study are implying that food raw materials deemed acceptable and without evidence of risks for humans by scientists and regulators in the US, EU, Japan, Australia and a number of other countries are being touted as risk factors for rodents consuming those approved GM events in any toxicology study. 

“All agricultural crops, including organically grown crops are ‘contaminated’ with at least low levels of metals, such as arsenic; with some level of chemical herbicide or insecticide and other compounds.  If those compounds are fed at high concentrations to animals or humans there would likely be some adverse effects.  Drinking too much water in a day (~ 4 gallons) will lead to death in a typical human.  Consumption of too many calories over time will increase the rate of developing certain types of tumors or cancers.  Consumption of some vitamins at a high dose over time will cause adverse health effects.  The authors report detection in the range of parts per billion of a number of compounds that are deemed safe by regulators at the levels found in the rodent diets.

“It seems the authors are trying to prove that their earlier reported tests with high tumor incidence in the 2012 study was due to “incidental contamination of control diets” rather than the genetic strain of rats used in their 2 year study.  Or that they want to invalidate the safety tests on many GM crops.  Yet the same tests and lab chow has been used safely to demonstrate convincing risk or safety of a number of pharmaceuticals, pesticides and other compounds.

“While their data is superficially interesting, the value really is to show that all food and feed contains small amounts of chemicals and metals, and that we should not be afraid to consume commodity crops that are grown with a wide variety of methods of cultivation, by farmers who follow appropriate guidelines for the use of chemicals. Even organic farmers use some metals legally,  and many non-organic farmers use chemicals that allow the safe and efficient production of food and feed and if they follow labeled approved dose schemes, the residues of pesticides on the food materials will be far below hazardous levels. 

“If we allowed the data of these authors to scare us from using the animal safety testing methods we have used for decades, or if we allow them to “invalidate” years of testing, we will not be able to make safe decisions about our food, our pharmaceutical products or even drinking water. As with the 2012 study, the authors have chosen to put out press releases and announce their findings in controversial and alarming terms, this time before the public release of the study by the journal.  I hope the sensationalistic press conferences and hype used by the authors go out of vogue. They do not help the public or regulators decide what is safe and what is not safe.”

Declared interests (see GENeS register of interests policy):

Dr. Richard Goodman: Prior to my current position I worked at Monsanto from 1997-2004 on the safety evaluation of genetically modified crops (GMOs), primarily evaluating potential allergenicity, also potential transfer of DNA from GMOs to other organisms including gut microbes and livestock and to some extent on the potential toxicity of GMOs. 

I am paid by the University of Nebraska from grant money and contracts that I perform for a variety of companies, governments and non-profit organizations.  I do not receive any money from companies to defend their products.

I served as an Associate Editor for the journal Food and Chemical Toxicology from late 2012 until March, 2015.  That was just after publication of a paper by Se?ralini describing the 2 year rat feeding study which reportedly showed tumor incidence in rats fed herbicide tolerant maize (NK603), or with glyphosate in the water, or a combination of both. After a year review, which I was not involved in, the journal (FCT) withdrew that paper.