Exposure to dioxin and endometriosis: the link that never was?
Exposure to dioxin and endometriosis: the link that never
by Sun-Wei Guo
It has been abundantly clear that dioxins, arguably the
most feared environmental contaminants worldwide, are associated with numerous
diseases and disorders such as cancer. However, most evidence for the
dioxin-endometriosis link actually came from a widely cited monkey study, which
was reported by Dr. Sherry Rier and her associates in 1993. Human data are, even
to this day, scanty, equivocal, and controversial. A careful reading of the
paper by Dr. Rier and her colleagues reveals that it suffers from many
methodological deficiencies, which seriously undermine its claims.
To understand that why the first piece of evidence is
actually rather thin, one has to understand how scientific research is
experiment/study design, and statistical analysis
In most cases, the scientist
first comes up with a specific hypothesis, say, exposure to dioxin can cause
endometriosis. With this hypothesis, the scientist sets out to design a study or
an experiment---how many people or animals are needed to be observed, for how
long, what kind of treatment to be given to the subjects, what outcome measure
to be used (say, laparoscopic examination of the pelvic, or incidence or
severity of endometriosis, etc).
Although ideally one would like
to observe/follow up as many as subjects as possible, practically and
logistically this would be a problem. Hence, scientists would have to strike a
balance between the constraints of time, resources, and logistics, and the
demand for enough subjects so that there won’t be wild fluctuations in chance.
By observing only a sample of subjects, called sampling, one potentially opens
up a floodgate of various biases and uncertainty. For example, in comparing the efficacy of a new drug N to a
standard drug S in treating, say, hypertension, one has to be very careful to
make sure that any patient willing to participate the trial is equally likely to
be assigned to one of the two drugs. Otherwise, the result will be very
difficult to interpret.
It is important to specify the
outcome measure well before the experiment/study is conducted. This is because
that there are potentially many outcome measures that could be compared between
the treatment group that receives, say, dioxin exposure, and the control group
that does not receive the treatment. Because of sampling, which introduces
statistical fluctuation or uncertainty, some measurements may end up quite
different between the two groups, simply because of chance events. Thus an error
could occur. By careful design of the experiment, a scientist can control the
chance of making errors. He cannot, of course, completely eliminate the error.
By specifying the hypothesis
ahead of the study/experiment, one can avoid the mistake of stating something
after the fact, or selective reporting. If a scientist does not specify his
outcome measure before the experiment/study is launched, or chooses an outcome
measurement to his liking after the study is finished, then any statistical
analysis that treats the result as if the measurement were defined beforehand
will have little validity. To see this, suppose a person is interested in
knowing whether a dice is fair. He throws it 10 times and gets “six” 10
times. With this result, he can pretty much conclude that the dice is loaded.
But he can still get 10 “six” in a row if he throws the dice many times.
With patience and time, 10 “six” in a row amid many throws is possible and
plausible. However, one cannot just report that he got 10 “six” in a row
and, based on this, claim that the dice is loaded. The conclusion or the
interpretation of the data really depends on how the study/experiment is
designed and executed.
In executing the
study/experiment, the scientist also has to pay attention to avoid potential
biases either in favor of or against his pet hypothesis, so that final data can
be judged fairly and appropriately.
Observational studies vs. designed experiments/studies
By requiring stating the hypothesis first and specifying
which outcome measure to use before the experiment/study is started, we do not
preclude the importance of observational studies, which are often conducted
without any hypothesis in mind or conducted after the fact. In fact,
observational studies are an important avenue for scientific discovery. In the
OB/GYN arena, there are ample examples. The link between the use of Thalidomide
and congenital birth defects, the link between in-utero exposure to
Diethylstilbestrol and vaginal clear cell adenocarcinoma, and the link between
the use of Tamoxifen and endometrial cancer, are all first reported from
observational studies. But all these successful findings share some common
features. First, both the exposure and the outcome are somewhat rare. This
contrasts sharply with dioxin and endometriosis: dioxin is almost ubiquitous in
industrialised countries, and the prevalence of endometriosis is also high.
Second, all these findings were later confirmed by independent studies.
As a Germen saying goes, “Eine Schwalbe macht noch keinen
Fruehling”. In scientific investigation, a finding reported in a single study
is usually viewed, at best, as tentative and preliminary, and its validity
should await confirmations from independent replications. This independent
replication requirement is what sets science from astrology, and is necessary to
rule out chance events, design flaws, biases of various kinds, errors, faulty
statistical analysis, and, in rare cases, alright scientific fraud such as
fabrication or doctoring of data.
Dioxin exposure and endometriosis
If we understand these
principles, it is not difficult to see why the evidence for the
dioxin-endometriosis link is surprisingly thin. Much of the evidence came from a
single monkey study published by Dr. Sherry Rier and her colleagues in 1993. In
this study, 24 rhesus monkeys were randomly assigned to 3 groups, one receiving
high-dose dioxin, one low-dose, and the third group serving as controls (ie. the
reference group against which the treatment groups can be compared). During 12
years of follow-up, 4 monkeys died of causes unrelated with endometriosis and
were excluded from the analysis. In the end, 2 animals out of 6 in the control
group, 5 out of 7 in the low-dose group, and 6 out of 7 in the high-dose group,
respectively, developed endometriosis. Thus, the prevalences of endometriosis
are 33%, 71%, and 86% in the control, low-dose, and high-dose groups,
respectively. Hence, at the first glimpse, the result appeared to be intriguing.
However, if we read the paper
carefully, we will see that the result is not that impressive as it appears.
First, the difference in
prevalence among the three groups is not statistically significant. There are
two possibilities: dioxin exposure is linked with endometriosis, but the sample
sizes are still small that the results are not statistically significant; or,
alternatively, there is no such link. With this finding, the link cannot be
Second, the study actually was
not an experiment designed specifically to test the hypothesis that exposure to
dioxin increases the risk of endometriosis. In fact, the study was initially
designed for investigation of reproductive toxicity of dioxin. No specific
hypothesis regarding exposure to dioxin and endometriosis was set before the
experiment was even designed. Hence, the study reported by Dr. Rier and her
colleagues was, at most, an observational study, which should be confirmed or
refuted by other studies.
Lastly, the study was launched
after it had been noticed that monkeys in the treatment groups had more and
severe endometriosis than the control group. Therefore, all the data were
collected after-the-fact, and thus the statistical validity is questionable
because of this practice. At best, the finding should have been reported as
preliminary or anecdotal, awaiting for further, independent confirmations.
Unfortunately, 11 years after the 1993 report, there has been no study that
replicates the finding. In fact, a larger study by a Canadian group did not
confirm the finding.
Six epidemiological studies have so far been published or
presented at professional meetings. Some are negative, some positive. As with
many epidemiological studies, controversy is the hallmark. It is interesting to
note that the results from all studies in support of the link are not
statistically significant. A very recent study by Dr. Brenda Eskenazi and her
colleagues reported a tantalizing and moderate, yet statistically
non-significant, increase in risk of endometriosis following an accidental,
high-dose exposure of dioxin. This study, by far the largest and
methodologically most rigorous, takes advantage of a “natural experiment”
following a factory explosion which released a large quantity of dioxin to its
environ in Seveso, Italy, in 1976, which resulted in the highest known
population exposure to dioxin. After 20 years of follow-up, with measured blood
concentrations of dioxins, the study found that women with a blood concentration
of dioxin between greater than 20 but less than 100 ppt (which would amount to a
similar exposure level as the high-dose group in Dr. Rier’s monkey study) have
about 20% increased risk of developing endometriosis as compared with those with
blood concentration of dioxin less than 20 ppt. This increase in risk, however,
is not statistically significant (translation: statistically speaking, there is
no difference in risk between the two groups). For women with blood
concentration of great than 100 ppt, the risk is 100% higher than the unexposed
women. Again, the result is not statistically significant.
Synthesizing all published human
studies, there appears to be a publication bias toward reporting the link. The
publication bias refers to the phenomenon that editors of scientific journals
tend to publish research papers that report positive finding, say, the link
between dioxin and endometriosis, and reject papers that report negative
Summary and remarks
The strongest evidence for the
link between dioxin exposure and endometriosis is based mostly on a single study
with small sample sizes and considerable methodological deficiencies. The study
so far has never been replicated independently. Human data lend little support
for the link, and there is a titillating sign that there is bias in favor of
publishing positive studies.
It should be noted that poking holes of a published study
is by no means splitting hairs. The link between dioxin and endometriosis, if
true, has important scientific and public health ramifications. If not, we
should put our energy and resources to investigate other possible causes for
Science thrives through discourse and debates. For a
weighty matter such as the dioxin-endometriosis link, taking stock and
re-appraisal of the available data are long over due, especially when there has
not been any independent replication.
It should be noted that, prior to the publication of Dr.
Rier’s paper, a group of Canadian investigators reported, at a professional
meeting in 1985, that there might be a link between dioxin and endometriosis,
based on some preliminary data that monkeys exposed to chlorinated biphenyls (a
member of the dioxin family) had higher incidence of endometriosis. However,
when the study was completed and finally reported in 1996, the difference
evaporated. This example highlights the importance of setting the hypothesis
beforehand and the strict adherence to experimental protocols.
It also should be noted that the validity of findings in
Dr. Rier’s paper had been questioned before by a few investigators. Dr.
Anthony Scialli and Dr. Thomas D’Hooghe both raised some concerns about the
paper. But an in-depth analysis of the original data and a comprehensive
reappraisal have been done only recently.
For the full paper please see:
S-W. The Link between Exposure to Dioxin and Endometriosis: A Critical Reappraisal of Primate Data.
Gynecol Obstet Invest 2004;57:157-173
10 July 2004: I have always wondered about this theory from its inception by Donnez and others and and
their claim that adenomyotic nodules are "a new disease" caused by
Our clinical experience in Tehran, Iran certainly does not support that and in this part of the world we see all forms of disease, peritoneal, Ovarian, RV septum and intestinal endometriosis in a single patient drastically not only as a very active form (red lesions) but also invasive fashion involving the entire abdomen.
I lectured about this issue at the recent ISGE congress in Kuala Lumpur and showed photos and video clips extensively.
I personally believe the environment we live in along with the numerous polutant that our patients are exposed to are the main reason, by way of causing different genetic mutations in different towns, cities and even countries with different levels of its related invasiveness.
Comparing the photos of what kind of disease I have treated in the USA (Illinois) till year 2000 and what I am treating in Tehran since then certainly depicts this matter very vividly.
I believe as long as the complexity of our industries change we will notice revolving forms of this disease as we have noticed so far.
I hope the United Nation, our governments, environmentalists and health organizations pay direct attention to this matter and study these mutagens carefully and delineate its inherent impact in different cities, towns and countries not only in regard to this disease but also many others.
We have a lot to know and learn about the environment we live in, as long as we are in dark we will never be able to control this disease.
Adel Shervin MD, FACOG, FACS
Chairman dept. of MIS surgery Tehran Clinic Hospital, Tehran, Iran