The 168 pigs in the study were fed a diet of either GM or non-GM corn and soy over the course of their normal commercial lifespan (just over five months) at a pig farm in the United States. They were then autopsied by veterinarians after slaughter and tested for any significant differences across a wide range of measures.
The study found evidence that stomach inflammation classified as ‘severe’ was higher in pigs fed the GM diet. It also found that pigs in the GM feed group had, on average, a heavier uterus. There were no other significant differences between the disease status or weight of the organs of pigs in the two groups.
The results further showed no significant differences in blood biochemistry, weight, illness, veterinary interventions or mortality between the two groups.
The full study is available here.
A Before the Headlines analysis of the paper is available here.
The Science Media Centre approached experts for reaction to the study. We will update our website with further comments as they are received.
Assoc Prof Peter Dearden, Director, Genetics Otago, University of Otago, comments:
“This study investigates is GM feed has adverse affects on pigs, and shows that both the rates of severe stomach inflammation and uterine weight appear to differ between the GM and non-GM fed pigs.
“I have a few issues with the outcomes of this paper. The first is that the procedure has been to feed the pigs the differing diets (and it must be noted that the diets between the groups differed more than just due to the GM) and then measure a vast array of parameters and report only those that are significant. In this kind of statistical analysis it is important to note that tests can have a positive result by chance. thus the more tests you, the more times a difference will be seen just by chance. Statisticians have known this for a while and there are a range of ways to account for multiple testing; none of which have been applied here. This is a problem.
“Secondly, the author report that GM fed pigs have more severe inflammation of the stomach than non GM fed pigs. But non-GM fed pigs have more moderate inflammation than GM fed pigs, reversing this trend. I would expect that the more GM you gave the more inflammation you would get, if there was a direct affect of GM food. We would expect, I think, to see a relationship between dose and response. The data from the paper doesn’t show this, which is puzzling.
“In summary, there may be something in this paper, but it needs repeating, and a more robust study needs publishing in a high impact, international journal. That would encourage me to think that that the weak effects seen here are not just an artifact.”
Prof Thomas Lumley, Professor of Biostatistics, University of Auckland, comments:
“This study is much better designed and analysed than the controversial French GM/glyphosate study from last year. It has just two groups and a reasonable sample size in each group. There’s a good level of detail given about the study process, and the assessments were carried out in a way that would reduce bias, with approximately matched numbers of pigs from each diet at each occasion, and with measurements made by someone who didn’t know which diet each pig had received. I can’t comment on how appropriate these diets are either for pigs or as a guide to human effects, but enough detail is given for this assessment to be made by a knowledgeable person. The researchers do comment that the finely-ground feed used in the pig industry is known to lead to stomach inflammation. This inflammation would be common to the GM and non-GM diets, so it would not cause bias, but it may make the results less generalisable to other animals and to humans.
“The statistical analysis is reasonable, and follows standard practice. The results are consistent with a harmful effect of the GM corn or soy. The results provide only modest evidence of harm, even for pigs on a 100% GM diet, because many possible effects were tested, making it quite possible that some difference would turn up just by chance. The researchers give a plausible explanation for an effect on stomach inflammation, but would likely have been able to give equally plausible explanations for other findings. The study should not change policy on its own, but it is worth taking seriously for future research. If the same results were found in a replication they might signal a harmful effect of high doses of the specific insecticidal proteins in these GM feed varieties.”
Further comments gathered by the UK Science Media Centre
Prof Patrick Wolfe, Professor of Statistics at University College London, said:
“I am not an expert on animal health, husbandry, toxicology etc, and therefore I cannot comment on these aspects of the study. As a statistical methodologist I can however comment on the data analysis undertaken and presented in the article.
“The biggest issue is that the study was not conducted to test any specific hypothesis. This means that the same sample (in this case nearly 150 pigs) is, in effect, being continually tested over and over for different findings.
“The statistical tests employed assume that a single test is done to test a single, pre-stated hypothesis; otherwise the significance levels stemming from the tests are just plain wrong, and can be vastly over-interpreted.
“Thus there is a higher-than-reported likelihood that the results are due purely to chance. The number of pigs being in the low hundreds (instead of, say, the thousands, as is often the case in large medical studies) can make this effect even more prominent.
“Bottom line: a better-designed study would have hypothesized a particular effect (such as changes in stomach size), and then applied a statistical test solely to check this hypothesis. Perhaps another independent team of researchers will go down this path. Until then, this study definitely does not show that GM-fed pigs are at any greater risks than non-GM fed pigs.”
Prof Tom Sanders, Head of the Nutritional Sciences Research Division at Kings College London, said:
“It does not look like a convincing adverse effect as it was a minor incidental finding. There were no differences in growth and mortality rates and pigs at sacrifice appeared in similar health. The dietary groups were also not identical as different supplements had been used. If you do not specify outcomes at least one in 20 will come up as being statistically significant by play of chance.
“The probabilities for the abnormal findings were low. Gastric inflammation is often due to infections and uterine weight can be influenced by factors such as rate of weight gain. The number of pigs with mild to severe inflammation did not different between groups with 69/73 showing inflammation in the control group and 64/79. The authors have unwisely highlighted differences in the number with severe inflammation. It seems unlikely that the effects observed were treatment-related.”
Prof David Spiegelhalter, Winton Professor of the Public Understanding of Risk at the University of Cambridge, said:
“The study’s conclusions don’t really stand up to statistical scrutiny. The authors focus on ‘severe’ stomach inflammation but all the other inflammation categories actually favour the GM-diet. So this selective focus is scientifically inappropriate.
“When analysed using appropriate methods, the stomach inflammation data does not show a statistical association with diet. There are also 19 other reported statistical tests, which means we would expect one significant association just by chance: and so the apparent difference in uterus weight is likely to be a false positive.”