I approach this topic with a bit of trepidation. I say this not because I’m unsure that I’m correct in my assessment of the article that I’m about to apply some Respectful Insolence™ to. Rather, it’s because the last time I brought up anything having to do with abortion, it got ugly. The topic is such a polarized one that virtually anything one says is sure to attract vitriol. Regardless, though, this article by Dennis Byrne, a columnist for the Chicago Tribune and the “study” to which it refers, are so appallingly idiotic that even fear of touching the third rail of American politics will not deter me. The reason is simple: Regardless of what one’s position is regarding the morality of abortion, I would hope that we can all agree that making non-evidence-based claims about abortion is not a good thing, particularly when such non-evidence-based claims are clearly intended to influence women’s choice by frightening them with the specter of breast cancer. Byrne’s article does just that and is truly a case where the stupid really does burn. The issue I’m talking about is whether elective abortion increases breast cancer risk, and Byrne’s argument is classic crankery. (Indeed, I encourage you to count how many of the Denialist’ Deck of Cards you can spot.)
Byrne starts out, as über-crank Mike Adams did in attacking the “breast cancer establishment,” by appealing to Breast Cancer Awareness Month and what he perceives as horrible “disparity” in the reporting of breast cancer studies by the media:
During National Breast Cancer Awareness Month, it is fitting and proper that women be informed about any newly discovered dangers, even as the public groans under the weight of all the warnings surrounding the mere act of living.
For example, a well-researched Chicago Tribune story last week disclosed that women who have just a couple of alcoholic drinks daily increase their breast cancer risk by 13 percent. Coincidentally, a new study reported that abortion is an important breast cancer risk factor, yet I couldn’t find a word describing the research in mainstream media.
How to explain this disparity? I’ll be vigorously advised that “most” studies disprove an abortion-breast cancer link. Or that the study in question appeared in a “conservative” scientific journal. Or that the study is bogus or unimportant. Or, more rudely, that the whole breast cancer argument has been concocted by anti-abortion rights advocates to make women afraid to have abortions. The issue is dead, I’ll be notified. Kaput. Here I would remind critics that in science it’s not who says it or how many say it that counts. What does count are the data and the rigor with which they are collected, analyzed and held up to a scientifically credible hypothesis.
Here we have a veritable panoply of crank language, not unlike what we hear from “intelligent design” creationists when whining about why they are not taken seriously. I’ll concede that the science that has failed to show a link between abortion and breast cancer is not as well settled as the science supporting the theory of evolution, but the arguments against established science are disturbingly similar. As for the “disparity,” perhaps it could be because even the mainstream media has learned that the medical “journal” in which this dubious “study” appeared (and I use the terms “journal” and “study” very loosely) is a right wing propaganda rag masquerading as a medical journal. It’s also a “journal” with which regular readers of this blog should be very familiar. Yes, my friends, we’re talking about the Journal of American Physicians and Surgeons (JPANDS for short), and the study is entitled The Breast Cancer Epidemic: Modeling and Forecasts Based on Abortion and Other Risk Factors. More on the study and JPANDS in a moment. For now, let’s get back to Byrne’s oh-so-righteous indignation at the way the mainstream media is “ignoring” this amazing “science” that he considers so important. After briefly describing the study, he opines:
Some will object because the study is “only” epidemiological — meaning that it relies on a statistically significant relationship between the incidence of breast cancer and abortion to infer that one causes the other. The standard, but simple-minded, objection to epidemiological studies is that a correlation does not necessarily prove causation. That’s true, to some extent. But, epidemiologists use correlations in more complex ways, combining them with a range of medical, sociological, psychological and other information to lead their research in the right direction, to support or debunk hypotheses, and toward solutions for significant public health problems.
He might have a point–if that’s what the Patrick S. Carroll, the author of the study, did. Alas, Carroll did nothing resembling using correlations in complex ways, as we shall see. When Mark over at Good Math, Bad Math sees the abuse of statistics in this study, he may have a stroke. (Actually, he’s made of sterner stuff than that, but he won’t like it.) Not surprisingly, Byrne can’t resist finishing with the classic crank appeal to “science was wrong before.” Particularly risible is his claim that if we reject the JPANDS study because it is “epidemiological” we must also “dismiss a multitude of public health studies, including ones claiming a link between radon and lung cancer.”
No, dismissing a crappy study does not require us to dismiss epidemiology altogether. And, make no mistake, the JPANDS study to which Byrne refers is indeed a crappy study.
Of course, one significant indication that the study is likely to be really crappy is the very fact that it appeared in JPANDS. As you may recall, I’ve discussed JPANDS before, as has Kathleen Seidel. The long versions of why it’s a highly dubious journal are contained in the links. The Cliff Notes version is as follows. Basically, JPANDS is an allegedly peer-reviewed “journal” published by the Association of American Physicians and Surgeons. AAPS is explicitly antiabortion (having published multiple dubious studies and editorials claiming to link abortion to breast cancer1, 2, 3, 4, 5), anti-immigration, and utterly opposed to Medicare. Moreover, it routinely publishes antivaccination screeds, characterizing mass vaccination as a “tool of the state” and a threat to physician autonomy. Most tellingly, if you want to know the real level of “science” published in JPANDS, know that JPANDS routinely publishes “studies” by mercury militia antivaccinationists Mark and David Geier. Indeed, JPANDs is so cranky that it is prone to publishing commentaries opposing evidence-based guidelines. Not surprisingly, MEDLINE doesn’t even index JPANDS.
Nonetheless, a stopped clock is correct twice a day. It’s possible that the Carroll study might somehow be halfway decent. (Of course, the wag in me can’t help but wonder why, if the study is so good, it wasn’t published in a reputable journal instead of JPANDS.) Be that as it may, let’s look at the study itself. One thing that should tip you off that this is a paper that’s probably more based more on ideology than science is the use of the term “epidemic” for breast cancer. It harkens back to the term “autism epidemic” in a most disturbing manner.
So what did Carroll do? Basically, he looked at national cancer registration data in eight European countries for which, it is asserted, there is comprehensive data on abortion incidence. The study claims to look at seven different reproductive risk factors for breast cancer, including induced abortion, reporting a correlation between abortion and risk of subsequent breast cancer. The risk factors were:
When a woman is nulliparous, an induced abortion has a greater carcinogenic effect because it leaves breast cells in a state of interrupted hormonal development in which they are more susceptible.
Alow age at first birth is protective.
Childlessness increases the risk.
A larger number of children (higher fertility) increases protection.
Breastfeeding gives additional protection.
Hormonal contraceptives are conducive to breast cancer.
Hormone replacement therapy (HRT) is also conducive to breast cancer.
This is sort of correct. I say “sort of” because not only is it not generally accepted by scientists that an induced abortion has a “greater” carcinogenic effect, much less any carcinogenic effect, but it’s very incomplete. Currently agreed-upon risk factors for the development of breast cancer are:
- Age (breast cancer becomes more common as women age)
- Sex (men can get breast cancer, but it is very uncommon)
- Personal history of breast cancer (a history of breast cancer in one breast is probably the strongest risk factor of all for the subsequent development of cancer in the contralateral breast)
- Family history of breast cancer and genetic risk factors (BRCA mutations, for example)
- Age at first menstrual period (earlier is worse)
- Race (breast cancer is slightly more common in Caucasians but African Americans are more likely to die of the disease)
- Age at menopause (older is worse)
- Age at first live birth (older is worse)
- Number of children (more is better, with nulliparity confering the highest risk)
- Previous chest irradiation (for childhood lymphoma in the chest, for example)
- Oral contraceptive use (this is a very weak and somewhat controversial risk factor; women who have not taken OCPs for 10 years are at no higher risk than those who have not taken them)
- Hormone replacement therapy
- Certain pathological findings on breast biopsy, such as atypical ductal hyperplasia, lobular carcinoma in situ, etc.)
- Breastfeeding (a weak protective effect)
There are others, but the data supporting them are less well accepted. One theme that runs through the risk factors due to reproductive history is that the more time a woman spends having menstrual cycles uninterrupted by completed pregnancies, the higher the risk. This is thought to be due to a higher lifetime exposure to estrogen unopposed by progesterone. This is the same reason that hormone replacement therapy is thought to increase breast cancer risk. It increases lifetime exposure to estrogen.
So right off the bat the Carroll study is ignoring a lot of potential risk factors. It’s worse than that, however. Here’s the model Carroll proposes:
Two explanatory variables are selected for modeling: x1 (abortion) and x2 (fertility). The trends for abortion and fertility are shown in Figures 8 and 9 for countries considered. The Mathematical Model is then:
Yi = a + b1x1i + b2x2i + ei
where Y represents cumulated cohort incidence of breast cancer within a particular age group; a is intercept, b1 and b2 are coefficients, and e is random error.
Carroll’s “model” is hilariously silly. You might want to know how he came up with it. Carroll includes a number of “science-y”-looking graphs to justify which risk factors he does and does not include in his model. In essence, he looked at a number of breast cancer risk factors, did what as far as I can tell is a linear regression analysis to try to correlate these factors to breast cancer incidence in the age 50-54 cohort, and then threw out the ones that didn’t have a high enough correlation coefficient. This left him with only fertility (Figure 5) and abortion (Figure 3). No, I’m not making this up. He threw out known risk factors for cancer, factors supported by better cohort or prospective studies, just because the graphs he found didn’t fit his linear regression. Perhaps most telling is Figure 3:
Holy correlation, Batman! But as we all know, correlation does not equal causation, as anyone who’s looked at the correlation between global temperatures and the declining number of pirates knows.
Indeed, what this graph purports to show is the abortion rates in women born all the way back in 1926. Let’s look at this. Until 1967, there was no law explicitly permitting abortion in the U.K, and court rulings only first stated in 1938 that abortion could be legally carried out to save the life or mental health of the mother. Moreover, even now, technically, women in the U.K. don’t truly have the right to choose abortion for any reason; the 1967 law, although giving doctors wide latitude to make exceptions as they see fit for medical and mental health reasons, still requires a woman to obtain signatures from two doctors before the NHS will do the procedure. (That requirement may be loosened soon, though.) Be that as it may, women born in 1926 would be in their reproductive age between the early 1940s and the late 1960s to maybe the early 1970s. Does anyone really think that the data for abortions for these women is accurately reported? For one thing, the reference cited (reference #17) contains statistics from 1968 to 2001. In 1968, women born in 1926 would have been 42 years old. Does anyone see how this methodology almost certainly inherently underestimates how many abortions women born between 1928 to 1950 had undergone, probably by a large degree? Think: A 42-year-old woman in 1968 would have spent nearly all of her reproductive years under a legal system that did not permit abortion other than for medical or mental health reasons. Moreover, the government report only goes back to 1968, which means that, unless women were polled in 1968 and beyond about whether or not they had ever had an abortion (which would also have underestimated the number of abortions), the 1926 abortion rate on Carroll’s graph includes reliable data only for women who underwent an abortion at 42 years of age and older, because the report does not include statistics earlier than 1968. Given the time period involved, this report can’t even begin to provide an accurate estimate of the true abortion rate until the cohort of women born around 1950, when the 1968 data would begin to capture abortion rates for all women age 18 and above. Oh, wait. 1950 just so happens to be the birth year of the last cohort Carroll included in his graph.
What a surprise.
In any case, on the basis of this highly dubious graph, Carroll decides that abortion rates needed to be included in his model. Worse, when Carroll tries to apply this model to other parts of the U.K, he presents howlers like this;
Abortions in England on women resident in Northern Ireland as reported in English abortion statistics are used to derive abortion rates for Northern Ireland. The trends in abortion and fertility in Northern Ireland are shown in Figures 8 and 9. Abortion rates in Ireland, where abortion is illegal, are much lower than in Great Britain. By smoothing the graph of cohort cancer incidence for Northern Ireland it was possible to fit the model and make estimates.
And trying to measure abortions in done in England on women from Ireland reflects a sufficiently accurate estimate of abortion rates in Northern Ireland exactly…how?
I could go on, but why bother? Carroll basically does the same thing as he did in Figure 3 for known, well-accepted risk factors for breast cancer and, because he doesn’t get a correlation coefficient adequately high, decides to leave them out. Another serious problem is that the variables he looked at are not independent variables; yet no valid attempt is made to control for possible interrelations between variables or to control for confounding factors. What was needed was a rigorous multivariate analysis to see if abortion was an independent risk factor for breast cancer, but what we got was in essence a series of poorly done univariate analyses. Moreover, it appears that Carroll cherrypicked the data that fit his hypothesis, took this simplistic model based on dubious assumptions, and used it to “predict” a 50% increase in breast cancer incidence in 25 years due to the increase in abortion rates. The bottom line is that this article is steaming, stinking pile of horse feces, and Byrne is either too blinded by ideology or too ignorant to realize how bad it is. Or maybe he does realize how bad it is but doesn’t care, as long as it supports what he wants it to. It’s impossible to tell.
Finally, we are always instructed by our alternative medicine colleagues to pay close attention to the funding sources of research. It turns out that Carroll was funded by a the UK charity LIFE, which is–you guessed it–an antiabortion charity. Indeed, its antiabortion views are so extreme that it opposes abortion even in the case of rape or severe fetal deformities. The other charity that supported the “research” is the Medical Education Trust, a group whose activism is not limited to abortion but does include a strong antiabortion viewpoint. In and of itself, the funding sources shouldn’t necessarily sway you one way or the other in assessing a study; it should just be one factor to be considered among many. However, combine the funding source with a lousy study like this, which twists epidemiology into a pretzel to support an antiabortion agenda, and its publication in the house journal of an organization (the AAPS) that also explicitly opposes abortion as a matter of principle, and you have a perfect storm of religiously and politically motivated junk science.
Byrne’s article and Carroll’s dubious “study,” it turns out, are all of a piece. There is a concerted effort on the part of right wing ideologues opposed to abortion to link abortion with breast cancer, when in fact science does not support the hypothesis that elective abortion causes cancer:
The Collaborative Group on Hormonal Factors in Breast Cancer, based out of Oxford University in England, recently put together the results from 53 separate studies conducted in 16 different countries. These studies included about 83,000 women with breast cancer. After combining and reviewing the results from these studies, the researchers concluded that “the totality of worldwide epidemiological evidence indicates that pregnancies ending as either spontaneous or induced abortions do not have adverse effects on women’s subsequent risk of developing breast cancer.”
Although there are studies that claim a link between abortion and breast cancer, they are almost all weaker case control studies, which are prone to recall bias. It’s been shown that healthy women are less likely to reveal that they have had an abortion to an interviewer, while women with cancer are more likely to do so, primarily because they are searching for causes of their cancer. (This is not unlike the problem with Generation Rescue’s dubious vaccination phone survey.) Nearly all the better designed prospective studies have found no link. Indeed, now there are numerous studies that have failed to find a link between breast cancer and abortion1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17. Given the preponderance of evidence, although it is still possible that there may be a link between abortion and breast cancer, it is highly unlikely that there is a real association. In this, it is not unlike the state of evidence regarding vaccines and autism. Current evidence does not support such a link, and there are enough studies to allow us to conclude that there probably is none. That’s as good as it gets in epidemiological studies, which, unfortunately, can never entirely eliminate the possibility of a correlation. They can only conclude that the chance of a significant correlation is very, very low.
Despite these studies, however, if you do a Google search for “abortion and breast cancer” what you will find is a preponderance of websites pushing a link that is not scientifically supported. The purpose, as far as ideologues go, of trying to push a link between breast cancer and abortion, is clearly to scare women. However, it’s also being pushed to make physicians who perform abortion afraid of malpractice lawsuits by women who have abortions and later develop breast cancer, the basis failure to inform them of the risk. In other words, if scaring women won’t work, then maybe threatening doctors with malpractice suits will. Meanwhile one breast surgeon who is a prominent speaker and activist supporting the dubious link between abortion and breast cancer compares herself and fellow activists to Ignaz Semmelweis and makes the overblown and unsupported claim that a single abortion during a woman’s teen years will leave her with at least a 30% lifetime risk of breast cancer.
Whatever you think of abortion, whether it’s murder, a necessary evil that’s not murder, a morally neutral act, or a moral good (the last of which is a view that I do not hold), I would hope that we could all agree that attributing risks to the procedure that are not supported by science does not contribute anything to the debate and serves primarily as a means of frightening women with the specter of breast cancer if they choose to terminate a pregnancy. Dubious studies such as the Carroll study are a transparent attempt to abuse epidemiology to find a link between abortion and breast cancer that probably does not exist. They only serve to obfuscate the the issue.
ADDENDUM: And what’s happened to The Chicago Tribune in the eight years since I last lived in Chicago? First it’s picked up the credulously woo-friendly antivaccinationist Julie Deardorff, and now Dennis Byrne.
ADDENDUM #2: Ah, just like the good old days! Mark Chu-Carroll at Good Math, Bad Math has looked at the statistics in more detail. Guess what he found. Oh, OK, I’ll give you a brief taste:
The model is wrong. Invalid models to not produce valid results. Stop. Do not pass go. Do not collect $200. Do not get your paper published in a decent journal. Do get laughed at by people who aren’t clueless jackasses.
ADDENDUM #3: Mark Hoofnagle points me to an even more idiotic article than Byrne’s. Not surprisingly, it comes from WorldNetDaily. Particularly telling is this quote that I came across by Karen Malec:
“We have had sufficient evidence to support an abortion-breast cancer link, through biological, epidemiological and experimental evidence … since the 1980s,” Malec told WND.
“They knew it back then, but nobody told women that this research was even going on,” she said…
“They’re very wealthy. It’s a money industry, just like tobacco,” she told WND.
ADDENDUM #4: The Arch Pundit is very amused by Carroll’s use of correlation coefficients.
ADDENDUM #5: Karen Malec responds to this post. It ain’t pretty.