2,4-D (Dacamine, Weed-B-Gon) Industry Task Force Report 5/92
2,4-D Dispatch May 1992
Industry Task Force on 2,4-D Research Data
NO LINK FOUND BETWEEN 2,4-D AND CANCER IN DOGS
Executive Summary: Review Panel Report on
"Case-Control Study of Canine Malignant Lymphoma: Positive Association With
Dog Owner's Use of 2,4-Dichlorophenoxyacetic Acid Herbicides"
Last September, the Journal of the National Cancer Institute published a
study suggesting links between lawn care uses of the herbicide 2,4-D and
malignant lymphoma in dogs. This study raised predictable concerns among lawn
applicators and home owners.
In response to these concerns, the Industry Task Force on 2,4-D Research
Data commissioned an independent panel of scientists to review the study and
its conclusions. After careful review, the panel found that due to
limitations in design, the study's data did not support an association between
2,4-D and malignant canine lymphoma.
Two primary reasons were given by the panel as to why the study was
unable to show an association. First. the study was based on responses of dog
owners, using a questionnaire that gave too little information about which
dogs were exposed to 2,4-D and how much exposure they had. Second, the panel
notes, that the effect reported by the study was small, so that it may have
been due to chance or might be linked to other factors in the dogs'
The findings of the dog study review panel are summarized below. These
findings are consistent with the weight of the evidence based on extensive
scientific testing which has not found a link between animal cancers and 2,4-
D. The complete text of the review is also available through the task force.
- The Editors
In December 1991, an independent scientific review panel was convened by
Drs. Ian Munro and George Carlo to evaluate critically the methodology and
findings of the study by Hayes et al. entitled, "Case-control study of canine
malignant lymphoma: Positive association with dog owner's use of 2,4-
dichlorophenoxyacetic acid herbicides."(1) The Study was published in the
Journal of the National Cancer Institute's September 4, 1991, issue. This
independent peer review was supported by the Industry Task Force II on 2,4-D
The scientific panel reviewed the published study, media releases that
accompanied the public release of the study, and the interview instrument
employed by the study investigators which was provided by the National Cancer
Institute. A meeting of the panel took place during December 1991. The panel
was charged with evaluating the degree to which the study by Hayes et al
supported the hypothesis that 2,4-dichlorophenoxyacetic acid (2,4-D)
herbicides are associated with canine malignant lymphoma.
OVERALL PANEL FINDINGS
Although the study, by virtue of its title, suggests a relationship
between exposure to 2,4-D and canine malignant lymphoma, this hypothesis was
not strongly supported by any of the results presented. The increases in risk
suggested by the odds ratios reported (e.g., OR=1.3) were small. Therefore,
it is possible that a small bias or confounder could be responsible for the
elevated odds ratio instead of a biological link to any of the potential
exposures studied. Exposure quantification was very poor, and it is therefore
difficult to discern what exposures were actually studied.
A relatively large sample size was used and the number of controls was
about twice the number of cases, which allowed the authors to conduct analysis
of many subgroups of dogs. Control dogs were selected from the same hospital
as the case dogs, which is important in addressing the potential problem of
selection bias. Two control pools were employed: one pool consisted of dogs
diagnosed with tumors other than lymphoma, the other pool consisted of dogs
that had died of other causes. This design should address the potential for
biased recall among the owners of the case dogs. The control dogs were
matched with case dogs by location, age and year of treatment. This matching
should control confounding by these variables.
Although the response rate was relatively low (about 45%) by mail, 70% of
those non-responders were interviewed, which is considered a good response
rate for a study of this type. Combining the methods, the total response rate
was about 83%. In addition, the questionnaire did not draw the attention of
the owners to the a priori hypothesis that 2,4-D exposure predisposed dogs to
canine malignant lymphoma. This approach should address the potential for
recall and interview bias.
The study was poorly designed to investigate the hypotheses that 2,4-D
causes cancer in dogs. Two sets of controls were combined in the final
analysis but analyses specific to each control group were not presented to
address the issue of potential recall bias. The authors should have analyzed
the control groups separately and, only if the risks were comparable, should
they have combined the two control groups.
Although the use of a hospital-based case-control design is generally
considered to be less rigorous than a population-based design because of the
greater opportunity for selection bias, the authors did not present what
proportion of local dogs were seen at the three hospitals to identify possible
differences between the groups. It is also likely that some of the dogs
diagnosed with fatal canine malignant lymphoma would be euthanized by the
local veterinarian rather than be taken to a hospital for treatment, thus,
raising questions about the representativeness of the case series as well as
the generalizability of the findings of the study.
There was substantial non-response to the mailed questionnaire. Forty-
five percent of the subjects had to be phoned in order to obtain information
and it is not clear whether these two groups were independently tested to
assess potential biases and determine internal consistency for the study.
Misclassification and Confounding
In this type of study, it is desirable to quantify exposure both
objectively and accurately. The questionnaire, however, is very non-specific
and broad with regard to use of pesticides, and a list of names of home-garden
products was not supplied. The design of several questions restricted
responses in such a manner that it is not clear whether 2,4-D was used alone
or in combination with other herbicides. Thus, the exposure quantification
data were of no value and the combination of the two groups (commercial and
homeowner) may not have been scientifically valid as there was no evidence
that these groups were exposed to the same chemicals. A market survey of
home-garden herbicide use in the U.S.A.(2) has shown that of those products
containing 2,4-D only 15% contained 2,4-D alone. The remaining 85% contained
other herbicides such as dicamba and MCPA in addition to 2,4-D. It is thus
very likely that a high percentage of the dogs in the exposed groups were
exposed to herbicides other than 2,4-D further complicating any link between
2,4-D and their cancer.
Data from the Hayes et al. study suggest that 33% of owners reported
using 2,4-D only, which is more than twice the percentage reported in the U.S.
market survey. This raises questions about the reliability (or possible
misclassification) of the responses to the questionnaire.
Also supporting the possibility of misclassification is the reported
response that some of the case owners applied the herbicide three and more
times per year. Given the geographical locations of this study, it is not
recommended and very unusual that 2,4-D would be used more than two times per
year. The significant trend test for number of applications per year is
driven by a significant number of dog owners who reported using 2,4-D four or
more times per year. Since it would be surprising that even a small
percentage of dog owners used 2,4-D four times per year, this raises serious
questions about the reliability of the questionnaire with respect to
quantifying this variable, as well as the reliability of the trend test.
The questionnaire is also inadequate with respect to gathering
information about the potential confounder of viral causes of lymphoma.
Therefore, the potential impact of viruses could not be tested for, or
In addition to the poor designs of the study and the questionnaire, the
results did not support the authors' hypothesis. The only statistically
significant observation of a relationship between canine malignant lymphoma
and lawn chemicals was observed when the data from three exposure groups and
the two control groups were combined. When each of these sub-groups was
analyzed separately, no statistical significance was observed.
The authors claim to have tested a number of associations in their study,
listing some 35 factors in the tables and text. It is to be expected that one
or more of these tests will show statistical significance purely on the basis
of chance, particularly when the OR is small (e.g., OR=1.3). One would expect
an OR of 2 or greater if a strong relationship existed.
Although the authors claim that a statistically significant trend was
observed for the number of owner applications per year, this is not a
meaningful dose-response as the "0" application or unexposed group should not
be included in the trend test. With this point excluded, the trend is not
significant. In addition, there was no statistically significant trend
observed for number of commercial applications of lawn chemicals per year,
which is not in concordance with the authors' claim of a statistically
significant trend for owner applications.
Finally, other studies on humans(3) and animals(4)(5)(6)(7) exposed to
2,4-D do not support the findings of this study.
As a result of limitations in the design of the study, it may be
concluded that it did not show an association between dog owner's use of 2,4-D
and canine lymphoma.
George L. Carlo,
Ph.D., M.S., J.D. (Co-chair)
Chairman, Health & Environmental
Sciences Group, Ltd.
Adjunct Faculty, George Washington University
School of Medicine and Health Sciences
Philip Cole, M.D., Dr. P.H.
Professor and Chairman
Department of Epidemiology
School of Public Health
University of Alabama at Birmingham
Anthony B. Miller, M.B., FRCP, FFPHM
(UK), FRCP (C), FACE
Professor, Department of Preventive
Medicine and Biostatistics
Faculty of Medicine
University of Toronto
Toronto, Ontario Canada
Ian Munro, Ph.D., FRC Path. (Co-chair)
Director, Canadian Centre for Toxicology
Guelph, Ontario Canada
Keith Solomon, Ph.D.
Canadian Centre for Toxicology
Professor, Department of
University of Guelph
Guelph, Ontario Canada
Robert Squire, D.V.M., Ph.D.
Professor, Division of Comparative Medicine
Johns Hopkins University School of Medicine
(1) Hayes HM, Tarone RE, Cantor KP, Jessen CR, McCurnin DM, Richardson RC.
Case-control study of canine malignant lymphoma: positive association with
dog owner's use of 2,4-dichlorophenoxyacetic acid herbicides. J Natl Cancer
Inct 1991; 83:1226-1231.
(2) Page D. Personal communication. Dow, 1991.
(3) Harris SA, Solomon KR. Bowhey C, Stephenson GR. Human exposure to 2,4-D
following controlled activities on recently sprayed turf. J Environ Health Sci
(4) Arnold EK, Beasley VR, Parker AJ, Zachary JF, Lovell R, Sawchuck S,
Richardson B, Butcher M. The toxicological evaluation of the dimethylamine
salt of 2,4-dichlorophenoxyacetic acid (2,4 D) in the dog. Report of the
College of Veterinary Medicine, University of Illinois, Urbana, IL May 1987.
(5) Arnold EK, Lovell RA, Beasley VR, Parker AJ, Stedelin JR. 2,4-D toxicosis
III: An attempt to produce 2,4-D toxicosis in dogs on treated grass plots. Vet
Hum Toxicol 1991: 33:457-461 .
(6) Hansen WH, Quaife ML, Haberman RT. Chronic toxicity of 2,4-dichloro-
phenoxyacetic acid in rats and dogs. Toxicol Appl Pharmacol 1971: 20:122-129.
(7) Hazelton Laboratories. Combined toxicity and oncogenicity study in rats,
2,4-dichlorophenoxyacetic acid. Final report. Volume 1. Hazelton Laboratories
America, Inc., Vienna, VA. May 29,1986. Project No. 2184-103.