Home Page Link AgBioWorld Home Page
About AgBioWorld Donations Ag-Biotech News Declaration Supporting Agricultural Biotechnology Ag-biotech Info Experts on Agricultural Biotechnology Contact Links Subscribe to AgBioView Home Page

AgBioView Archives

A daily collection of news and commentaries on

Subscribe AgBioView Subscribe

Search AgBioWorld Search

AgBioView Archives





July 2, 2000


Marcus Williamson's misunderstandings


Can someone pass my reply along to Williamson?


>>>>If it's so "environmentally friendly" perhaps you can explain this
>>>>article :
>>>>which indicates that beneficial arthropods may be impacted by
>>>>glyphosphate use.

There is no threat to arthopods from glyphosate use except where it removes
their cover or food, as dozens of studies have shown. I suspect that the
study to which Williamson refers is one in which the doses used were so
high that the insects got stuck to the glass testing surface by the
surfactants used! This story was carried by the media last year and then
dropped when the "leaked study" was actually made available to other people
to review.

>>>>How about the increased risk of non-Hodgkin lymphoma? Is that
>>>>"environmentally friendly"?

>>>>>>Glyphosphate has already been shown to cause lymphoma in tests in
>>>>>>Sweden. The substance is banned in several European countries.

In actual fact, the study found that the association with Roundup was no
closer than chance, and that there was a stronger relationship of the
cancer to glass wool! What countries now ban the use of Roundup on the
basis of this study?

I show at the end of this message my review of this paper, which I sent to
the authors several months ago and to which they have not responded

>>>>>>If Glyphosphate is intended to _reduce_ herbicide use, why have US
>>>>>>UK governments _increased_ the permitted application of glyphosphate
>>>>>>by 200 times?

They didn't. The level has not changed in the US, where it has been
20mg/kg in soybeans since long before GE crops. In countries that didn't
use glyphosate on soybeans, like Australia, no level had ever been applied
for, so the limit was set at 0.1 mg/kg, essentially a detection level. To
allow imports of soybeans, as required under trade treaties, Australia and
at least some other countries now accept the Codex limit of 20 mg/kg.

Contrary to popular belief, GE herbicide tolerant crops have reduced
herbicide use overall, and have especially reduced the use of persistent
herbicides in favour of safer and less persistent herbicides. A USDA
report in 1999 stated that the
"technology significantly reduced herbicide treatments for
soybeans and, to a lesser extent, for cotton"
(see also new USDA report at http://www.ers.usda.gov/epubs/pdf/aer786/).

Farmers often use less than labelled herbicide rates, sometimes applying
only 1/2 to 1/3 the labelled rates now. The idea that farmers will drench
their crops with herbicide is not only economically silly, it is belied by
current practice with the herbicides to which crops are naturally tolerant.

A Case-Control Study of Non-Hodgkin Lymphoma and Exposure to Pesticides.
Lennart Hardell
Mikael Eriksson

Cancer 1999;85:1353-1360

Contrary to popular claims, the paper provides NO statistically
significant links between glyphosate and cancer. By "statistically
significant", I mean that there is no evidence that the results differ from
what might occur from random chance.

The paper itself does NOT "clearly" make claims that glyphosate is linked
cancer. For example, phenoxy herbicides and fungicides are identified in
the abstract as being associated with higher cancer risks (for me, no new
surprises there), but glyphosate is not even mentioned. Elsewhere in the
paper: "Futhermore, due to low numbers of exposed subjects in some of the
categories, definite conclusions cannot be drawn for seperate chemicals,
such as MCPA and glyphosate, from the multivariate analysis" (page 1358).

And in the conclusions, "Other much used pesticides, e.g., glyphosate, also
might be of concern....Glyphosate deserves further epidemiologic studies"
(page 1359).

With respect to the statistics, the measure of association used was the
odds ratio (OR), a statistic that estimates the ratio of disease rates (in
this case non-Hodgkin's lymphoma or "NHL" rates) for exposed and unexposed
populations. An OR of 1.0 implies that the disease rate is the same for
exposed and unexposed and indicates no association between exposure and
disease. An OR greater than 1.0 implies that the disease rate is greater
for the exposed population than for the unexposed population. An OR of 2.0
is consistent with a disease rate among exposed persons that is twice the
disease rate for unexposed persons.

This assumes that the differences are not due to chance, and that is the
issue here. The authors used a method that allows us to accept that the
results are unlikely to be due to chance only when an OR of 1.0 (that is,
no effect) is NOT included within a 95% confidence interval, which they
show in their tables as "CI". In no case does the CI for glyphosate
EXCLUDE 1.0. In fact, in the only result that they give, the lower CI for
glyphosate is 0.4. This study was based on just 4 cases with NHL and
glyphosate exposure (assuming accurate recall of the subjects or their

By comparison, the authors also state that "Glass wool turned out to be a
risk factor for NHL in this study, an association that has not been
reported previously. It may be a random finding, which is supported by the
lack of dose response effect (page 1358)." For the record, the CI for
glass wool is 1.0-2.3, and there wasn't a dose rate effect for glyphosate
either. In other words, the authors appear to accept that glass wool is a
random finding, yet it has a stronger confidence interval than glyphosate!

At least as importantly, the claims in this paper are not consistent with a
much larger literature on the cancer risks of glyphosate. The authors
mention some reports of mutations from glyphosate, but at least five of
their references (44-48) are not about glyphosate at all but about
sulfosate, a compound with a different toxiciology. Another study of theirs
(reference 50), on hairy cell leukemia, was also based on a confidence
interval that included 1.0 (95% CI 0.8-12.0) and was also based on just 4
cases who reported use of glyphosate

There are other possible limitations, including recall bias and lack of
controls for potential confounding factors (eg., exposure to other
herbicides), but it wouldn't surprise me if phenoxy herbicides and
fungicides really do have a significant cancer risk. However, the authors
have not excluded those as causes of NHL in the 4 (!) cases associated with
glyphosate exposure.

In sum, any reasonably careful study of this paper does not support the
popular claims that have been made for it.