Abortion and Breast Cancer
Oxford Group Adds New Wrinkle to Disinformation

By Joel Brind, Ph.D.

NRL News readers are likely as familiar with the many and varied forms of public denial of the abortion-breast cancer link (ABC link) as they are with the link itself. As night follows day, knee-jerk reactions follow whenever evidence for the ABC link is published, whether it appears in the medical or the popular press.

But disproving the old adage you can't teach an old denial new tricks, a group of Oxford University epidemiologists has just published an ABC study that is a novel new addition to the disinformation genre. The forum is this April's issue of the British Medical Association's Journal of Epidemiology and Community Health.

The new Oxford study, by M.J. Goldacre et al., appears to be a backlash by the professional British medical establishment against the credibility attained by the ABC link last year.

Last spring, the Royal College of Obstetricians and Gynaecologists (RCOG) publicly acknowledged, although it still considered all the evidence inconclusive, that the comprehensive review and meta-analysis of the ABC link that I wrote in collaboration with colleagues from the Penn State College of Medicine "could not be disregarded." (See May 2000 NRL News.) When the British popular press learned of the RCOG statement late in the summer, the issue surfaced briefly - - very briefly - - in the London tabloids. (See September 2000 NRL News.)

On the surface, the Oxford study appears to be the ideal type of study which includes two key components only rarely found together: a case-control record-linkage study.

Most epidemiological studies include a group of breast cancer patients ("cases") who are compared to a group of women of similar age and other important characteristics, but who do not have breast cancer ("controls"). But the "record linkage" part means that a study does not depend upon information provided by the study's participants gleaned from interviews or questionnaires. Rather, information has already been recorded as part of the subjects' medical records.

Consequently, there is not even the possibility of so-called " response bias." This hypothetical scenario imagines that women with breast cancer are more honest about a past abortion than are women who do not have breast cancer, thus skewing the results. The possibility of such bias has often been invoked by the nay-sayers to explain away the repeated finding of an ABC link in case-control studies which do not rely on record linkage.

The results of the Oxford study would certainly appear to refute the ABC link: Overall, breast cancer occurred only 83% as often in women who'd had an induced abortion (as opposed to spontaneous abortion, a.k.a. miscarriage, for which the authors obtained similar results), compared to women without a history of induced abortion. This finding was also statistically significant, that is, it was more than 95% certain not to be due to chance.

Moreover, this number did not vary significantly with the time period after the abortion. This leads to the authors' conclusive literal bottom line: "The fact that the relative rates did not increase with increasing time intervals between abortion and breast cancer indicates that interruption of pregnancy, whether spontaneous or induced, does not increase the risk of breast cancer."

Hold the phone. A careful look at the "Methods" section reveals a huge deficiency.

True, all the abortions took place between 1968 and 1998 - - after elective abortion had been legalized in the U.K. in 1967. However, whereas almost all abortions in the country are performed in freestanding clinics, the only abortions which appear in the medical records of the study subjects were those done in hospitals. Moreover, the authors admit, "It was common for abortions to be coded without qualification of whether they were induced or spontaneous."

Thus, one key question here is, how many induced abortions are recorded in the Oxford database? Out of a total of 28,616 breast cancer cases, only 300 - - just over 1 percent - - had a recorded, induced abortion in the hospital over the 30-year span of the study.

Considering that the abortion rate in England over the same period has averaged about 2.5 percent of the women of child- bearing age per year, we can conservatively estimate that at least 95% of induced abortions done on this population of women - - both in and out of the hospital - - do not appear on their records at all.

Translation: over 95% of the women who did have an induced abortion - - both cases and controls - - were misclassified as not having had an abortion. What this really means is that the Oxford database is useless for determining anything about the effect of abortion, since the vast majority of abortions in the study population were unrecorded.

To give a parallel example, suppose you were conducting a study on cigarette smoking as a risk factor for lung cancer. But suppose you only had the smoking histories of subjects who smoked "Brand X" that represented no more than 5% of the smoking population.

Then suppose - - since you didn't know who else was a smoker or non-smoker - - you classified everybody who didn't smoke "Brand X" as a nonsmoker. Obviously, such a study would be totally useless as a measure of the effect of cigarette smoking, since almost all the smokers would be [mis]classified as nonsmokers. That's basically what Goldacre et al. did with induced abortion.

Then what really tips the Oxford group's hand is its misrepresentation of the published record. The group begins its " Discussion" section with this statement: "None of the cohort or record-linkage studies have shown a significant increase in breast cancer risk after exposure to induced abortion."

The only problem with this statement is that it is a patent falsehood: The only other record linkage case-control study - - the New York state study by Howe et al. published in 1989 in the International Journal of Epidemiology - - reported a statistically significant, 90% increased risk with induced abortion.

The tack taken by the Oxford group - - packaging an unsuitable database as a refutation of the ABC link - - has a familiar ring to one familiar with the ABC literature. When we look back to 1982, in fact, we find a group at Oxford, including D. Yeates, one of the authors of the current Oxford study, published a study on the ABC link in the British Journal of Cancer (BJC).

That study was explicitly in response to the 1981 study by Malcolm Pike and colleagues at the University of Southern California, also published in the BJC. It was the first U.S. study of the ABC link, and it reported a 140% increase in breast cancer risk for young women who'd had an abortion before having a child.

In contrast, the Oxford study of 1982 claimed to be "entirely reassuring," reporting a relative risk of 0.84, i.e., a 16% decrease in breast cancer risk with abortion in women who had not yet had any children.

But how many of the women in the 1982 study had had any induced abortions? "Only a handful," admitted the authors in their "Results" section!

Some things, it seems, never change. Nineteen years after we are "reassured" that there is no ABC link, we are given a similar message from the folks at Oxford, again with data that contribute nothing meaningful to the debate about the ABC link.

But, wait, there is a ray of hope. Last summer, writing in The Times of London, Dr. Thomas Stuttaford came down on the side that the evidence for the ABC link was "inconclusive." But on May 17 of this year, he wrote an article in The Times which focused on new drug treatments for breast cancer.

Dr. Stuttaford seems to have done some homework since last summer. He begins his article, entitled "Fresh line of attack," with the following statement: "Breast cancer is diagnosed in 33,000 women in the UK each year; of these, an unusually high proportion had an abortion before eventually starting a family. Such women are up to four times more likely to develop breast cancer." His next paragraph succinctly explains why, biologically.

No fanfare; no bells and whistles; just the truth. That suits me fine.

Joel Brind, Ph.D., is a professor of biology and endocrinology at Baruch College, City University of New York.