In my last post, I gave a general reason why the American Association for the Advancement of Science (AAAS) Board was misleading in writing that a review by Snell and colleagues showed that genetically engineered (GE) foods are equivalent to non-GE counterparts.
Here, I want to discuss why the study does not lead to the conclusion that 90-day tests are generally sufficient to determine the safety of GE foods, and more reasons why the study says little about the long-term safety of engineered foods.
Not all that it seems
A major conclusion by the authors of the review is that long-term testing is rarely needed, because 90-day tests provide essentially the same results as the long-term tests. This paragraph from the discussion section sums up the view of Snell:
Despite the exploratory nature of the studies reviewed here, the step-by-step approach is supported by their results. Considering all of them, it is clear that GM food is not revealed to be harmful when the duration of feeding is increased to well over 90 days. Therefore, no evidence is available to show that a duration of 90 days is insufficient to assess the effects of GM food. Studies lasting two years, for example, do not seem necessary except when doubt remains after performing 90-day studies. [emphasis added]
If true, this would refute arguments that long-term tests are needed to detect possible long-term harm. To be fair, the authors do not entirely rule out the need for long term tests, but make them contingent on troubling results from 90-day tests. This means that 90-day tests need to be able to provide an indication of possible long-term harm that would require follow-up long-term studies to substantiate.
Let’s put aside that toxicologists do not generally accept 90-day subchronic studies as a stand-in for long-term or multi-generational tests. For example, agencies like EPA require long-term tests for chemical pesticides. Or go to the bible of toxicology, Caserette and Doull, which says on page 29 (in my 1996 edition), using the example of food additives, that long-term studies are typically needed to determine possible long-term harm: “…if the agent is a food additive with the potential for lifetime exposure in humans, a chronic study up to two years in duration is likely to be required.”
It is hard to predict how long any GE trait may stay on the market, and therefore how long exposure of the population may last. Glyphosate herbicide resistance and some Bt traits have now been in our food for about 16 years, and are not likely to disappear soon.
The basic problem of the Snell review is that it really does not address the adequacy of 90-day tests as a stand-in for long-term testing. For that, one would need to have an indication of risk in a 90-day test that later showed up as harm in a long-term test. But there were no risks detected in any of the studies that Snell considered to be reliable, so there was no real test of their hypothesis that 90-day tests are sufficient.
The Snell study is analogous to a proposed medical test, say a blood test for Alzheimer’s in middle-aged people to predict whether the disease would develop later in life, but where none of the test subjects went on to develop Alzheimer’s. The results of the blood tests during middle age would not tell you whether the test could detect the disease later, because there was never any Alzheimer’s to detect.
Similarly, the long-term studies reviewed by Snell either had substantial limitations or did not detect harm. We cannot determine whether the 90-day studies were sufficient to detect harm when the long-term studies detected no harm.
Few, if any, acceptable studies
Finally, very few of the studies that Snell and colleagues reviewed are without significant limitations, according to their analysis. For example, many of the 12 long-term and 12 multi-generational studies do not use the proper and universally accepted non-GE crop variety for comparison with the engineered crop (a so-called near-isogenic variety that is nearly identical to the engineered variety, except for the engineered gene). The reviewed research also had other substantial flaws, such as too few test animals.
In the end, they identify only six studies in total that used enough test animals according to OECD standards, including only three long-term studies, not the 12 that the AAAS Board noted. Use of too few animals means that the tests are not sensitive enough to reliably detect harm.
But when these six are examined more carefully, it turns out that only two used the proper near-isogenic control.
One of these so-called acceptable studies used salmon as the test species. But as Casarett and Doull note in their discussion of sub-chronic and chronic (long-term) tests, rodents and dogs are the standard test animals as stand-ins for humans. Fish may be used for ecotoxicology studies (to determine whether something will be harmful to fish in the environment), but they are too different from us to be good stand-ins for safety tests.
I should note that the authors of the salmon study did observe some differences, such as higher triglyceride levels in the GE-fed fish. The authors apparently dismiss this observation by saying that it may be due to genetic differences between the near-isogenic comparison variety (the control) and the GE soybeans. While that is possible, the variety they used is the proper and accepted control, and therefore significant differences should be considered legitimate unless shown by further tests to be in error. The apparent rationalization about this experiment is troubling.
The other test that seems to use the proper controls tested livestock species—chickens, sheep, and cows. Both birds and ruminants (cows and sheep) have very different digestive systems compared to people, which is problematic for food safety tests.
So really, there are no acceptable long-term or multi-generational tests that were reviewed by Snell and colleagues. It is hard to see how this study can be used to suggest that either 90-day studies are generally acceptable, or that they convincingly show that GE foods are safe.