Is the Long-Term Safety of Genetically Engineered Food Settled? Not by a Long Shot.

Bookmark and Share

One of the most contentious issues surrounding the controversy about genetically engineered (GE) foods is whether there may be long-term safety risks, and whether current regulations are sufficient to prevent such risks from occurring.

As I briefly discussed in my last post, major science organizations have said that some GE foods produced by current methods could be harmful, and have provided some examples of the kinds of harm that might occur.

The current situation

Most corn, soybeans, and cotton in the U.S. are engineered to contain one or several genes for insect or herbicide resistance. Many more types of engineered genes are in the works. Photo by danellesheree.

No long-term safety tests in animals are required by any regulatory agency. In some circumstances, 90-day, so-called sub-chronic tests may be required in Europe. But 90 days is far short of the one to two years that usually satisfy long-term safety test requirements.

Long-term experiments are required for products like drugs and chemical pesticides, and sometimes for food additives. They are considered important or necessary for determining harm that may take years to develop, such as cancers, Parkinson’s disease, and so on.

Recently, the American Association for the Advancement of Science (AAAS) Board of Directors cited a review of several long-term and multi-generational studies by Snell and colleagues in support of their claim that GE foods are safe and well tested.

The study cited by the AAAS Board has made the rounds in recent months, being used to claim that long-term studies show that GE is safe, and that shorter-term tests are sufficient.

The study authors extrapolate from the reviewed research that GE crops can be safely used in foods, based on currently required tests. For example, at the end of the paper’s abstract they write: “The studies reviewed present evidence to show that GM plants are nutritionally equivalent to their non-GM counterparts and can be safely used in food and feed.” They also conclude that 90-day tests are usually sufficient, and even these may not always be needed.

The study makes some useful contributions, but a careful reading shows that it contains some serious flaws. These limitations — some of which involve the interpretation of the results by the authors — eliminate the value of this study for drawing general conclusions about the safety of GE foods, or the adequacy of current shorter tests to reveal long-term risks from engineered foods.

Missing the fundamentals

The most glaring limitation is that regardless of the findings, generalization about safety and testing of GE crops is not scientifically justified based on the review of several studies.

The safety of the very few currently available engineered traits and foods is not the only issue. There are many novel engineered genes in the pipeline, coding for a plethora of traits, from pest resistance to drought tolerance, to nutritional alterations and industrial products.

We therefore depend on the regulation of engineered traits to ensure the safety of all GE foods — current and future. We must therefore consider whether our regulations are adequate for current and future GE crops and foods.

This is important because most scientists on all sides of the debate about GE safety agree that the risk of each crop and gene combination must be considered on its own merits. In other words, just because one, or ten, or even one hundred genes are shown to be safe does not assure the safety of the next one.

This is easy to understand if we consider the wide range of functions that genes affect — virtually everything that an organism is composed of. Take a Brazil nut tree, for example. The majority of genes code for harmless things like enzymes that make sugars or amino acids in the plant, or make the chlorophyll involved in capturing light energy. But, as I have noted before, one gene also codes for the major allergen from Brazil nuts, which was unsuspectingly engineered into soybeans (and never commercialized), and which can be deadly if consumed by some sensitive people.

One could run safety tests of all of those genes that code for benign functions in the Brazil nut, and according to the review study thereby conclude that, based on current tests, GE foods can be safely used. But what if the next engineered gene was the major Brazil nut allergen? Would this broad conclusion still be valid? Certainly not.

There are obviously harmful genes (or groups of genes) from non-food plants and other organisms that we know about and would therefore avoid engineering into foods, such as the insecticide nicotine produced by tobacco, or the insecticide rotenone, produced by the roots of several tropical plants.

These are generally understood, and therefore avoidable. Rotenone is a particularly interesting example, because it has relatively low immediate (acute) toxicity in mammals, but has been implicated for its possible connection with Parkinson’s disease, which usually occurs later in life, and which was not understood until recently.

These are also extreme examples. But like most phenomena in biology, there are likely to be a range of possible health effects that genes can produce, from innocuous to extreme, and including those that are intermediate and not easily predicable or detectable with short-term tests.

Less fundamental, but important, is that the number of genes reviewed in the Snell paper is actually very small, and not representative of what may be put into crops in coming years. Of the dozen long-term studies reviewed, 10 tested the gene for glyphosate herbicide resistance (EPSPS) in soybeans, one was a Bt insecticidal gene in corn, and one a cedar pollen gene in rice.

Even if generalizations about safety were possible, that is a tiny representation of the possibly hundreds of genes that may be used in coming years, and virtually meaningless for drawing general conclusions.

For the dozen multi-generational studies, six tested Bt in corn, three were for glyphosate in soybeans, two for another herbicide resistance gene (glufosinate) in the grain triticale, and one for a similar gene in potato. This is not likely to be representative of engineered genes or crops in coming years.

For these reasons alone, the Snell paper says little about the safety of GE foods generally, or the need for long-term testing. It is therefore misleading to use the study to make broad claims that GE foods are safe based on existing long-term tests in animals, or, as the AAAS Board claimed, are equivalent to their non-GE counterparts.

But there is more, and in my next post I will discuss specific limitations of the Snell research, including why it does not demonstrate that short-term or 90-day studies are generally sufficient to determine the safety of engineered genes.

Posted in: Food and Agriculture Tags: , , , , ,

About the author: Doug Gurian-Sherman is a widely-cited expert on biotechnology and sustainable agriculture. He holds a Ph.D. in plant pathology. See Doug's full bio.

Support from UCS members make work like this possible. Will you join us? Help UCS advance independent science for a healthy environment and a safer world.

Comments are closed. Comments are automatically closed after two weeks.

4 Responses

  1. Jonathan Gressel says:


    You missed a huge two year study feeding rats transgenic glyphosate resistant soybeans, well performed, excellent statistical and anatomical analyses. The paper text is in Japanese, but there is an English abstract and the figures and tables are in English. Sakamoto, Y., Tada, Y., Fukumori, N., Tayama, K., Ando, H., Takahashi, H., Kubo, Y., Nagasawa, A., Yano, N., Yuzawa, K., Ogata, A., 2008. A 104-week feeding study of genetically modified soybeans in F344 rats. J. Food Hyg. Soc. Japan 49, 272–282.

    They found no significant effects of the transgenic material.

    If you or anyone else wants a pfd, I’d be glad to send it.

    There was also a large meta-analysis that you missed -available online at the journal: “Food and Chemical Toxicology” prior to publication entitled: Assessment of the health impact of GM plant diets in long-term and multigenerational animal feeding trials: A literature review

    Seems like your blog is suffering from “citation amnesia”, which is rather unbecoming of UCS.


    • Doug Gurian-Sherman says:

      Hi Jonny, Nice to hear from you.

      Actually, the study that I discuss in this blog post, Snell and colleagues (see the link), is the same “large meta-analysis” that you suggest that I missed. Snell et al. did review the Sakamoto et al. study that you refer to. But apparently Sakamoto did not use a near-isogenic control, which as you know, is not acceptable. Although I will give you that it is probably less of an issue when no-differences are found between the control and the test GE food.

      But the broader point of this blog post is that based on the logic and biology of the need for case-by-case analyses of risk, due the the potentially very different properties of different transgene/host/environment combinations, one cannot generalize about the safety of GE foods based on several studies.

      In a blog post that should go up today, I go into more detail about why the Snell et al. study is seriously flawed. Normally, I would not devote blog posts to such a technical subject, but it was a primary citation of the AAAS Board in support of their assertion that GE foods are equivalent to non-GE foods. It is also widely cited by vocal GE supporters as an important study showing that long-term tests show that GE foods are safe. So, given the serious limitations of the study, I think it deserves my attention.

      To a minor error in your comment, the Snell et al. study is not really a meta-analysis, which requires the use of statistical methods to aggregate the data from several studies to gain statistical power, Snell just performs a qualitative assessment of the studies they review.

  2. Doug Gurian-Sherman says:

    Thanks for the comment Mark. All of the National Research Council (NRC) reports acknowledge that increases in allergenicity or toxicity are real possibilities with GE (as well as breeding), and supported regulation of GE for those reasons. So unless you feel that our major science institutions are “mad sciencey,” I don’t think that you are right about that.

    The point about the Brazil nut allergen is a good one, but misses an important aspect of the science. First, though, I do not base anything I write on what Jeffrey Smith says, and frankly do not pay attention to his positions. So I can’t help you there.

    The Brazil nut allergen situation is quite different from a possible allergen from outside of the food supply, that is introduced into food through GE. The Brazil nut allergen could be easily detected using an immunoassay based on the blood of people allergic to Brazil nuts, even without previously knowing what the gene and protein were. That is not possible for proteins new to the food supply, where we now have to rely on GE tests that are not as reliable as that immunoassay. This is one reason that the NRC strongly recommended that FDA take steps to improve its risk assessment for allergens from GE foods, as I noted in a previous post( ), and which the FDA has never done.

    I do not agree that GE scientists would have to be incredibly stupid to miss possible harms from GE foods or crops. As a former GE research scientist, I would take exception to that! It is not at all straightforward to determine the safety of whole foods and transgenes/gene products. That’s why testing is needed in the first place. And it a legitimate issue as to how much testing, and what testing, should take place. I, and many other scientists, believe that the current testing requirements, or lack thereof for FDA, have several weaknesses that should be remedied.

    You are right that there is risk in everything. I do not believe GE foods are necessarily harmful. But the issue is not merely the safety of the few current GE foods, but also the many genes that will be used in the future, and are already being introduced. Since food is so integral to our lives, I think we should be a little more cautious about its safety than many other products.

    Beyond food safety, there are also environmental issues and social issues that are relevant, but beyond this post. As for current GE, the glyphosate herbicide resistant crops have led to an epidemic of resistant weeds that have likely increased herbicide use, and will lead to increased use of older and nastier herbicides from the 1940s. And while resistant weeds are nothing new, the ability to spray the herbicide onto the crop has led to its overuse, which greatly exacerbated the resistant weed problem, unlike with previous herbicide technologies. This is largely a management or regulatory issue, but that is my point–our regulations are inadequate.

  3. MarkH says:

    So, your major safety concern is that some company will splice in a gene for a human toxin or allergen at some point? Sounds a little mad-sciencey to me. And the brazil nut example isn’t such a great one for your case because it was found in pre-market testing, and the transfer occurred before the 2s albumin was known to be the allergen in Brazil nuts. Isn’t this just Jeffrey Smith’s talking point, and should the UCS be repeating the arguments of a crank new age yogi with no scientific training?

    The GM mods so far widely employed like Bt or RR really are very benign modifications. You’re talking about expressing a protein used in organic farming, harmless to humans, that had previously been sprayed on crops since the 20s, and with RR a modification of an enzyme plants already have to be resistant to a single herbicide. Neither of them presents a plausible threat to human health.

    I don’t find this criticisms plausible because it suggests (1) that GM scientists are incredibly stupid (2) that testing for allergens does not occur with new GE proteins (they are tested) and (3) you’re talking point is the same as one from Smith, a non-scientist and anti-GMO crank. All technology comes with some risks and every human endeavor is subject to error, but these risks strike me as being very low, as well as well anticipated by GM researchers. We shouldn’t allow irrational risk assessments guide our response to new technologies.