In Steinkes paper in 2010 on the effect of GM corn in particular MON810 on lactating cows, it is noted that 9 cows from each group were removed and replaced. It is claimed that there was no analysis done on these 18 cows. Is this true? If so, why is it used in so many reviews as evidence that GMOs are safe? I know I could have just read the article, but it was getting too technical. It seemed like they might have done an analysis but I was not sure.

Expert response from GMOAnswers Admin_1

Friday, 17/06/2016 11:48

Thanks for your question. Other than what was published, I do not have first-hand knowledge of this study, which was not funded by Monsanto or another biotech company. But, having overseen some of the largest dairy studies conducted, I think I can help you understand the challenges of running and interpreting multi-lactation dairy studies.

Let’s start with some terminology. Primiparous cows are cows that are in their first lactation. This lactation is initiated at about two years of age after delivering the cows’ first calf. Multiparous cows are cows that are in their second or greater lactation. Lactation curves of primiparous cows peak lower and are flatter than multiparous cows so typically cows are assigned to treatments and data are analyzed by these parity groups. Although not specifically stated in the paper, they likely assigned cows to treatments within these parity groups resulting in nine primiparous cows fed control diet, nine primiparous cows fed test diet, nine multiparous cows fed control and nine multiparous cows fed test. This study is a relatively small one for a two-lactation study.

Dairy cow studies can be designed with the primary focus of examining animal health or animal productivity. Milk yield is a continuous variable (numerical, i.e. kg/d) and health is often a discrete variable (observations such as mastitis: yes/no or pregnant: yes/no). In general, due to the type of data, milk yield is easier to detect differences that might exist rather than health variables. Also, health data are often accompanied by blood variables (continuous) or necropsy data (discrete). Steinke et al. (2010) is designed for animal productivity data; however, stressed or sick cows don’t produce milk as well as healthy cows so if a health problem existed, milk production is a place that I would expect this would show up. Their data do not indicate this.

In a two-lactation study such as this one, there are three at-risk periods for removing cows from the study. Obviously cows can be removed due to a health problem at any point in the study; however, calving and for the first couple weeks afterwards is the most risky time to have significant health problems for cows that might result in death or removal from the study. This is a period that commercial dairy farms use significant labor to manage cows to prevent health maladies. There are two of these periods in this design. The third critical period for losing cows is when they complete the first lactation. Cows that do not get pregnant won’t initiate a subsequent lactation. Removal of approximately 40 percent of cows in a lactation study has been observed from commercial herds in Wisconsin and Minnesota and the WI data saw herds as high as 60 percent (Nordlund and Cook, 2004).

While more information would be helpful, it is not unreasonable that there would be removal of cows and the authors indicated that there were the same numbers for each treatment group. The fact that they did not statistically analyze these removals is not surprising because they would need a larger study to determine real health effects. Furthermore, an analysis would have to be split among reproduction issues, injuries, metabolic health, mastitis, etc. because these do not share a common etiology. Health data should be analyzed as discrete variables and that type of analysis would be complicated and not likely be meaningful for this size study.

There are more than 400 studies of animals fed GM crops and the weight of evidence is that GM crops are as nutritious as conventional crops. Toxicology studies with laboratory animals (e.g., rats, mice) typically follow internationally specified guidelines (e.g.,OECD) that assess well defined endpoints related to identifying hazards (e.g., liver toxicity). By comparison, livestock studies, that typically test nutrition-related endpoints, do not have similar OECD guidelines, therefore their design is often influenced by budget, facilities, competition for other researchers for cows, feed availability, etc. Steinke et al. is testing effects on milk production over a long-term and is not designed to be a health study. It seems appropriate that they did not make conclusions about health. Many reviews provide a comprehensive list of animal studies but the interpretation of GM safety comes from the weight of evidence of all studies, which includes what we know about effects of the transgenic protein on metabolism, composition studies, environmental studies, rodent tox studies and livestock studies. Making conclusions on any single study greatly increases the risk for error. Therefore it is most appropriate to base conclusions about the safety and nutritional quality of foods and feeds from GMO crops on the sum total of the studies that have been conducted.

Nordlund KV, Cook NB. 2004. Using herd records to monitor transition cow survival, productivity, and health. Vet Clin North Am Food Anim Pract. 20(3):627-49.