York-Laval University Paper on Neonics and Bees Marred by Inconsistencies, Data Deficiencies, Dubious Bee Management and Weak Statistics


Two new research reports purporting to show serious negative effects of the neonicotinoid insecticides (neonics) on the health of honey bees were released in late June, 2017. One was one a major European study involving trials in Germany, Hungary and the United Kingdom (UK), and the other done by Tsvetkof et al at York and Laval Universities in Canada. (Tsvetkof et al paper is also available here.)

Reviews of the European study have been done by Iida Ruishalme and Jon Entine. In brief, a careful examination of the data (though not the authors’ conclusions) shows mostly no effects, positive or negative, of neonic exposure on bee health. Interestingly the senior author of this paper, B.A. Woodcock, was also senior author for a 2016 paper attempting to use a correlation between declining UK bee numbers and increasing neonic usage over a period of about two decades, to prove that the latter caused the former – see here, here and here for more.

This commentary is about the Tsvetkof et al paper. My conclusion, in brief:

This paper joins others in showing that honey bees exposed to high concentrations of neonics (in this case clothianidin) may demonstrate sub-lethal effects. However, the strength of this conclusion is weakened seriously by data inconsistencies and deficiencies, major questions about bee management, and dubious statistical analyses. The potential role of varroa mites and other pests and diseases is ignored.

To learn how I reached that conclusion, read on.

In 2014, the authors placed a total of 55 bee hives at 11 locations in Ontario (mostly) and Quebec, with five being within 0.5 km of a corn field and six located more than 3 km from the closest corn field. They sampled the level of several pesticides, including insecticides, herbicides, fungicides and a miticide, in living bees, dead bees, pollen and nectar in/near the hives at six times between early May and September. The data are shown for each location–by-time combination in a supplemental table provided by the authors.

Tsvetkof et al then took the average quantifiable concentration of the neonic clothianidin measured in pollen samples within hives in 2014 and used it as the basis for pollen concentrations of clothianidin to which bee colonies at York University in Toronto were artificially exposed (five colonies were exposed, five were not) in a study in 2015. The concentrations used as the base were not the average of all samples measured the previous year but, rather, the average of those high enough to be quantified (i.e., 15 of the 66 location x time combinations). Both the treated bees (fed clothianidin-laced pollen blocks) and the check (no clothianidin in pollen blocks) were also exposed to the fungicides dimethomorph and boscalid, preservatives in the pollen blocks. Worker bees from the five treated and five untreated colonies were collected, radio tagged, placed collectively in a single observation chamber, and the activity of these bees was monitored.

In another 2015 experiment, the authors fed five different dosage levels of clothianidin and thiamethoxam, in sucrose-acetone solutions, and in combinations with or without boscalid (497 ppb) and the herbicide linuron (7 ppb), to caged worker honey bees. They calculated acute LD50 values for the two neonic compounds (4 hr exposure to chemical mixtures; mortality measured after 24 hr). Relatively few details are provided on how this was done.

The authors reported that worker bees exposed to the neonic clothianidin in 2015 died younger and were less hygienic (propensity to remove dead bees from hives) than unexposed bees, that queen mortality/functionality was reduced with clothianidin exposure, and that the LD50 for clothianidin was reduced in the presence of a high concentration of boscalid. Some differences were observed in the foraging patterns of tagged worker bees.

I have reviewed the paper and the supplemental information which the authors have provided and have a number of major concerns about their experimental protocols and results.

Concentrations of neonics in bee-collected pollen

The concentrations of neonics in pollen which they collected from hives in 2014, and summarized in the following tables, seem very high.

Time of sampling Hives near corn Hives not near corn
Average clothianidin concentration, ppb (number of sites/total sites)
1 4.5 (2/5) 4.1 (1/6)
2 6.6 (4/5) 2.1 (3/6)
3 2.2 (2/5) 8.5(1/6)
4 2.6 (2/5) Not detected
5 Not detected Not detected
6 Not detected Not detected


Time of sampling Hives near corn Hives not near corn
Ave. thiamethoxam concentration, ppb (number of sites/total sites)
1 2.2 (2/5) 2.0 (1/6)
2 5.0 (3/5) 2.0 (1/6)
3 3.3 (2/5) 8.7(1/6)
4 3.0 (1/5) Not detected
5 1.1 (1/5) Not detected
6 Not detected Not detected

Note that these high concentrations don’t originate from the pollen of treated corn or soybeans. Corn pollen was found in only one of 21 neonic-positive samples and soybean pollen in only five of 21 and neither ever represented as much as 1 % of the pollen collected. The high concentrations came from other plants both near or more than 3 km away from, corn fields. The authors provide no information on proximity to soybean fields or other plants which may have been deliberately treated with neonics, other than to state that soybeans are grown near corn. Notably, two of the three “far corn” sites where concentrations of neonics were quantifiable in hive pollen (above 8 ppb for both clothianidin and thiamethoxam for one of these sites!) are within the cities of London and Toronto, Ontario and apparently remote from agricultural production. No explanation is provided is provided as to the source of this neonic exposure.

While not central to the theme of this paper (since the authors state that this compound is not of concern for bee well being) a very high concentration of 329 ppb of acetamiprid, another neonic, was measured for one “near corn” pollen sample in Quebec, but with no explanation provided as to source.

I have compared those reported values for clothianidin and thiamethoxam with others in the literature.

Blacquière et al (2012) provide a summary of published measurements of neonic concentrations in bee pollen including results of one extensive survey over three years in France. In it, imidicloprid was found at concentrations (where detected) at 0.9 to a maximum of 3.1 ppb though with no information available as to the source of the insecticide. Other European data cited by Blacquière et al involving imidicloprid, clothianidin, and imidicloprid (the three neonics generally used as seed treatments), generally show lower concentrations in pollen collected by bees, typically under 1 ppb and quantifiable in less than 1% of the assayed samples.

Stewart et al (2014) measured the concentration of neonics in pollen/flowers of seed-treated corn, cotton and soybean fields in Tennessee, Mississippi and Arkansas, in adjacent wild flowers, and in pollen collected by honey bees for colonies positioned an average of 180 m from the treated fields. Only 23% of samples (of whole plants, not just pollen) had neonic concentrations about 1 ppb, but the others ranging up to 257 ppb. (The highest sample came from a place where the farmer filled the planter with treated seed.) The arithmetic mean for all samples was 10 ppb. Neonics were measured in corn pollen at average concentrations of 0.4 to 5.6 ppb (and as high as 23 ppb for an individual sample) depending on compound used (higher for clothianidin than thiamethoxam) and the seed-treatment rate, but only at much lower levels for cotton pollen (almost all under 1 ppb) and not at all in soybean flowers. Most significantly, neonic concentrations above 1 ppb were detected in only 2 of 74 pollen samples collected from foraging bees despite their nearness to treated fields.

Long and Krupke (2015) in Indiana measured a mean concentration of clothianidin of 0.64  ppb (maximum 9.4) in 32 samples of pollen collected by bees for hives located immediately beside a treated corn field.

Botias et al (2015) in Sussex England measured the neonic concentration in the pollen of wild flowers adjacent to an oilseed rape (OSR) fields sown using thiamethoxam-treated seed and adjacent to wheat field grown from clothianidin-treated seed. They also measured neonic levels in the pollen collected by bees colonies positioned on the same farms. They found thiamethoxam concentrations averaging 15 ppb in wild flowers next to the OSR, but almost no clothianidin in flowers next to wheat fields. However, the maximum concentration in pollen collected by bees at the time of OSR flowering was 1.8 ppb for thiamethoxam and 1.2 for clothianidin (averages 0.2 and <0.1, respectively). Curiously, pollen collected by bees at time of OSR flowering contained an average of 2.5 ppb imidicloprid even though this had not been applied to either field for three years and field-adjacent wild flowers averaged less than 1 ppb.

Cutler and Scott-Dupree (2007) measured the concentrations of clothianidin in pollen collected by bees colonies located in the middle of four Ontario canola fields grown from clothianidin-treated seed. The maximum concentration of clothianidin measured in any pollen sample was 2.6 ppb. A similar though larger experiment, which was reported on in 2014, yielded a maximum concentration of 1.9 ppb.

How does one get concentrations in bee pollen of 8.5 ppb for clothianidin plus 8.7 ppb for thiamethoxam in colonies located in the city of London Ontario, remote from agriculture, but only a maximum of 2.6 ppb for hives in the neonic-seed-treated canola fields in bloom?

While the results of Tsvetkof et al generally do show more neonicotinoids in colonies nearer corn fields, they do not show that corn fields are the dominant source – indeed, they cannot be for the three within-city sites remote from corn fields, where neonics were measured in bee pollen at sizable concentration.

I do note that high neonic concentrations have been measured by other researchers in pollen collected by bees near agricultural crops. Pohorecka et al (2013) measured an average concentration of 27 ppb of clothianidin in pollen collected in pollen traps for bees returning to hives positioned in a corn field in Poland, though the concentration of clothianidin in “pollen bread” found within those hives was below the level of detection. Pilling et al (2013) also found much lower concentrations of neonics in “bee bread” than in pollen collected from returning worker bees.  Dr. Schaafsma at the Ridgetown campus of the University of Guelph has told me in personal communication that he has measured levels of neonics in pollen collected by bees at least comparable to those measured by Tsvetkof et al.

Bee health and management

The authors state “we did not chemically treat the colonies to control hive pests or diseases” in describing the 2014 research protocol and the same for the research in 2015. This is despite the fact that varroa and other pests/diseases are the dominant cause of poor health in Ontario/Canadian bees (see here for example), and chemical treatment for control of one or more pests/diseases, at least for varroa, is standard practice. Not only were no varroa-control treatments applied, but there is no mention of this potential complication for the research results – other than a simple statement that bees appeared visually to be healthy at the beginning of the trials.

This neglect is in marked contrast to the approach used by other researchers. For example, Pohorecka et al (2013) in Poland were very careful to monitor the prevalence of varroa mites and of several bee viruses in their research and bees were treated with Apivar for mite control. The Polish researchers measured no significant effect of exposure to either clothianidin or imidicloprid treated corn fields on any measure of colony health. Perhaps their diligence in controlling varroa is why.

Interestingly, the most common chemical found in bees, pollen and nectar in 2014 results of Tsvetkof et al was coumaphos, an organophosphate compound used for mite control in bees.  Coumaphos was measured in 91 samples in total as compared to clothianidin in 26 samples. Coumaphos was once commonly used for varroa control in Canada though I understand not so much lately because the mites have developed resistance. The origin of the coumaphos detected in this study is unknown; perhaps the hives were treated with it before the experimental period began. According to this Cornell University document, “Coumaphos poses a moderate hazard to honey bees and a slight hazard to other beneficial insects.”

There is no mention of the presence of other miticides in this study – likely because they were not in the list of chemicals assayed.

There is a good amount of published information showing that bees are more vulnerable to insecticides, including neonics, when weakened by varroa and other bee pests/diseases (example here). Were varroa mites, bee diseases and other pests the cause of the sub-lethal effects reported by Tsvetkof et al? No one knows because the researchers ignored this possibility and provide no relevant information.

The authors present a graph from their 2015 research purporting to show that bee queen health was worse for colonies artificially exposed to clothianidin versus those which were not. It’s graph D in the following figure from their paper. The blue dots and line are for the control treatment; yellow for treated. First, it’s doubtful, with so few data points, whether those lines really are different at P<0.05. In fact only four data points are graphed for the treated hives though I suspect that at least one additional data point may be hidden under an identical value for the control. More pertinent to this discussion is the question: how does a check treatment with only 20% of hives having functioning queens (i.e., the check treatment at about 55 days) represent proper bee management? I’m told that this would be major cause for concern if it happened in a commercial bee yard. One suspects that there was a serious management problem with these colonies at York University (perhaps poor control of varroa or diseases?).

Zayed Fig 2a

Zayed Fig 2

Statistical analyses

While I don’t profess to be a good statistician myself, I wondered about the validity of some of the statistical analyses performed by Tsvetkof et al. As a result, I checked with statistician Bill Price at the University of Idaho. Among his comments he noted the small number of data points (two) and low degrees of freedom (n=4) used distinguish the two lines in Fig 2B above.

More problems exist with graph D above. Firstly, it is highly unlikely that the two lines shown differ significantly given the scatter of the data especially for the control treatment. But there is also a problem in credibility of the basic data. The numbers for each of the treated and control treatments come from data from five hives. With five samples, how can you have a value of 0.75 for one control measurement? It’s mathematically impossible. (The 0.25 values for the treated hives are explained by the authors; a queen was inadvertently killed by the researchers so that hive was excluded for consideration for the remainder of the trial.)

He also, noted that treated and untreated bees were mixed together in one observation chamber for measurements of the respective flight behaviour and longevity of worker bees in 2015. In effect the two treatments were not imposed independently which makes statistical interpretation difficult.

There are also questions about the procedures and statistical analyses used to determine LD50s (no basic data used in calculations are provided by the authors) but I will not enlarge on that here.

Lack of colony data

A major question with this paper is the lack of colony data. Why did the authors provide no information on colony performance of the 55 hives in 2014? More importantly, why none for the 10 hives located on the York University campus on 2015 as that should have been easy to do? The authors measured a parameter called hygienic behavior (propensity to remove dead bees) in treated and untreated colonies, but there are no data on numbers of workers or dead bees – or on colony size, honey production and subsequent hive survival. (Note that the authors’ use of the term, hygienic behavior, differs from the definition often used by bee keepers in Ontario, which is the propensity to removed varroa mites from other bees.) This void contrast with the work of Pohorecka et al (2013) who measured various colony-level effects and found none – this despite the even larger exposures to clothianidin in pollen in their study.

The authors’ one-sentence conclusion: “Our findings indicate that chronic [neonicotinoid] exposure reduces the health of honey bee colonies near corn crops,” is belied by the fact that they provided no real measurements of colony health (queen vitality being the exception).

In summary

This paper joins others in showing that honey bees exposed to high concentrations of neonics (in this case clothianidin) may demonstrate sub-lethal effects. However, the strength of this conclusion is weakened seriously by data inconsistencies and deficiencies, major questions about bee management, and dubious statistical analyses. The potential role of varroa mites and other pests and diseases is ignored. Also, no consideration is given to the question of what the effect would be if/when neonics are replaced by other insecticides for pest control on food-producing farm crops.


I thank Dr. Bill Price Director of Statistical Programs, College of Agricultural and Life Sciences, University of Idaho, and Dr. Chris Cutler, Associate Professor, Department of Plant, Food and Environmental Sciences, Dalhousie University for their advice on aspects of the Tsvetkof et al paper. However, the comments made in the preceding review are attributable solely to me.