CEH Bee Study Shows No Conclusive Results

Selective use of data makes sensational headlines, but has little substance

Jul 11, 2017
Picture of oilseed rape field study

Picture of oilseed rape field study

As part of our company’s comprehensive approach to bee health, we at Bayer frequently work with independent scientists to help advance our collective knowledge of pollinators. In 2014, Bayer, working in collaboration with Syngenta, commissioned the UK-based Centre for Ecology & Hydrology (CEH) to conduct a large-scale field study to examine the potential impact of oilseed rape seed treated with the neonicotinoids, clothianidin and thiamethoxam, on foraging honey bee colonies.

The study was conducted in three different countries – in Germany, Hungary and the UK – and at each site the potential impact of neonicotinoid treatments on honey bee colonies was evaluated using several indicators of bee health, including colony strength and overwintering success. The basic study design was agreed upon with Syngenta and Bayer, but its implementation, execution and evaluation were the sole responsibility of CEH. Additionally, CEH scientists conducted parallel studies on a solitary bee (Osmia bicornis) and a bumble bee (Bombus terrestris) at the same test locations; however these studies were not commissioned by either company.

The CEH study found no consistent effects across the three countries regarding the key indicators of honey bee health examined. For example, in Germany, colony strength increased when hives were exposed to treated forage compared to the control colonies, while in the UK, and to a lesser extent in Hungary, colony strength was reported to have weakened. Such inconsistencies apparently played little role in influencing the report’s overall conclusions, nor did they prevent CEH from choosing a bold headline implying only negative effects of neonicotinoid pesticides on honey bees. A closer look at the study data suggests a far different conclusion, however.

Despite the claims made in the CEH press release, the data show no consistent differences between control and treated experimental colonies. Out of 254 different parameters measured to assess overall honey bee colony performance and health, 238 resulted in no significant differences between the colonies of the control and the treatment groups, 7 were more favorable in the treatment colonies, and only 9 were negative. Surprisingly, at no point in the publication did the authors acknowledge that 94 percent of all comparisons found no differences between colonies foraging on treated or untreated oilseed rape. It is difficult to understand how the authors could base their conclusions solely on 9 negative effects that were observed in only two countries and which comprised only 3% of the study’s data, while giving no consideration to the 238 parameters that showed no effects or the 7 parameters that showed positive effects. With such low levels of effects, it is impossible to rule out random error in this study. Drawing any reliable conclusion on such a small number of statistically significant effects is problematic.

One of the major problems of this study was the way a discrepancy in the initial strengths of the honey bee colonies used was handled. The strength of a bee colony throughout a study lasting several months is strongly influenced by the size of the colony at the start of a study. While it is expected to see some differences in colony strength during a field trial, greater care should have been taken to ensure a more homogeneous distribution of the colonies of different strengths over the different treatment groups, or, if not possible, steps should have been taken to correct for these differences. This was not done in this case. Indeed, when the differences in initial colony strength are appropriately accounted for, the “country specific effects” disappear entirely.
In the UK, the initial size of the honey bee colonies was very low (less than 5,000 bees) which significantly impaired the quality of the field study. The honey bees in the UK experienced strong Varroa mite pressure during the study, which led to losses far above the national average in the control groups. Such high colony mortality precluded formal statistical analyses of overwintering worker numbers, making it impossible to draw any reliable risk conclusions, as even the authors admit. Surprisingly, the CEH report did so anyway.

Likewise, the CEH data showed similar inconsistent results among the wild bee species examined. Neither bumble bee queens nor solitary bee reproductive cell production were directly affected by exposure to neonicotinoid seed-treated crops. Lacking consistent statistically significant effects in direct comparisons, the authors instead relied on data pooled across all three countries, to bolster their conclusion that crops grown from treated seed may have negative effects on the reproductive potential of both wild bees. However, when the data are examined within each country this conclusion falls apart. In the UK, for instance, there appears to be no such correlation – i.e., the levels of neonicotinoid residues in wild bee nests had no discernable influence on queen production of bumblebees or the number of reproductive cells of solitary bees, a finding that directly contradicts the study’s major conclusion of long-term effects.

When comparing the evaluation approach that CEH has chosen for honey bees on one hand, and solitary bees and bumblebees on the other, there is a conspicuous inconsistency: In the case of honey bees, the authors refer to “country-specific responses” to account for the variability observed between countries in environmental circumstances, but in the case of wild bees they used pooled data collected from all countries to show statistical differences when none were found in individual country comparisons. This raises the question – whether environmental influencers are not just as relevant for wild bees?

Conducting large-scale field trials is inherently difficult because of the many variables that must be properly managed. Because the results of this study were so inconsistent, and maybe even more important the vast majority of endpoints showed no difference between treated and control, in our view it is inappropriate for CEH to conclude that neonicotinoid seed treatments are harmful to wild bees or honey bees. Especially considering that in some cases the colonies of the treatment group fared better than did the ones of the control group. This lack of harmful effects is not surprising. There have been various large-scale field trials conducted in Europe and North America and the findings are essentially all the same: neonicotinoid seed treatments do not pose a risk to bee health under realistic agricultural conditions.

Back To Top