The abortion-breast cancer link: a medical-legal nightmare on the horizon.

The Autumn 2005 issue of this journal contained an excellent article by Charles Francis, Q.C., on problems resulting from poor medical standards in abortion practice. One of the negative consequences for women undergoing abortion which Mr. Francis discussed was the link between abortion and breast cancer (known as the A.B.C. link). Mr. Francis highlighted a particular 2000 legal case in Australia --the first of its kind--in which the abortion practitioner settled for damages for failure to warn the plaintiff patient about the A.B.C. link.

The evidence in that case was largely derived from research in the United States In the United States, two other, similar cases have since been adjudicated in the patients' favour, also noted by Mr. Francis. However, being the venue for the first such case was not the only significant role for Australia vis-a-vis the A.B.C. link. In fact, it is research in Australia that has most clearly evidenced not only the link, but the suppression of the very fact of its existence.

Mr. Francis did not discuss at length the case for this suppression of scientific evidence, merely stating that "some rather dubious material has been produced suggesting that there is no such link". In fact, despite convincing evidence having been produced in medical journals dating back nearly half a century, the A.B.C. link has been the object of an ever intensifying cover-up for nearly a quarter century. Remarkably, this disinformation extends to the government public health agencies, medical societies and journals, and voluntary anti-cancer societies.

Between the late 1950s and the early 1980s, a number of epidemiological studies reported a higher incidence of breast cancer among women who had undergone abortion. These were, however, relatively few studies in number, and they typically included abortion as one of many variables that comprise a woman's reproductive history. The elements of reproductive history are necessarily included in breast cancer studies, because it is well established that, for example, the number of children (parity) and maternal age at first childbirth are known to influence the risk of future breast cancer. In such studies, the contribution of any factors toward increased or decreased risk are reported in terms of relative risk (R.R.). A R.R. exceeding unity signifies increased risk associated with the given factor; R.R. less than unity signifies decreased risk, and R.R. = 1 signifies no effect of the given factor on breast cancer risk.

In 1996, our research team published a comprehensive review and meta-analysis (a meta-analysis combines the data of many studies, to gain greater statistical power) of the A.B.C. link in the British Medical Association's Journal of Epidemiology and Community Health. At that time, 17 out of 23 studies worldwide had reported increased risk of breast cancer associated with abortion. While this is indeed a lopsided majority, and sure to compel warnings to patients for an elective medical or surgical procedure, no such sense of alarm was seen among purveyors of public health information. Indeed it is noteworthy--and disturbing--that among hundreds of studies seeking to identify risk factors for breast cancer, so few studies had reported data on the effect of abortion, even by 1996.

But what is particularly disturbing is that some of the evidence available by 1996 had been deliberately kept out of the published literature for quite a number of years. Specifically, in 1988 a study appeared in the American Journal of Epidemiology concerning the effect of potential dietary risk factors in women in South Australia. As noted above, a good study must also collect data on reproductive variables, since reproductive history is known to affect future breast cancer risk. Hence the 1988 study of T.E. Rohan, et al. included data on "gynecologic history (and), reproductive history" among dietary histories and the many other variables examined. In fact, the Rohan study was exceedingly well designed of a type in which cases (patients with breast cancer) are matched one-to-one with controls (women of similar age and history who do not have breast cancer). This pair-matched study, then, included 451 case-control pairs, with each subject interviewed individually by trained interviewers, in order to determine dietary, medical, reproductive and other important data.

The Rohan study results were rather unimpressive in terms of the effect of dietary factors. There was no observable effect of different levels of fat or protein intake, and only a slight hint --not statistically significant--of a protective effect (that is, decreased risk) of high fibre intake. The only significant dietary finding reported was a 24 per cent reduction in risk for the 20 per cent of women (known as a quintile) consuming the greatest amount of beta carotene, compared to the 20 per cent of women consuming the least amount of beta carotene.

Among reproductive variables, significant risk increases were noted among women with later age at first birth or with no children at all, and among women with later age at menopause, and decreased risk among women with later age at menarche (puberty) and younger age at menopause. All of these findings were in good agreement with what has been established, thus establishing the Rohan study population of greater Adelaide as representative of women generally. Importantly, the effects of all of these factors were within a twofold range, with none showing more than approximately a 50 per cent increase or a 50 per cent decrease in risk. Surprisingly, the word "abortion" did not appear at all in the Rohan paper, either in terms of miscarriage (medically known as spontaneous abortion) or induced abortion, even though one would presume such data would be collected as an important part of reproductive history.

Only seven years later was it revealed that Rohan et al. had indeed collected data on both induced and spontaneous abortion. This occurred when, in 1995, a small meta-analysis authored by French epidemiologist Nadine Andrieu et al. appeared in the British Journal of Cancer. The aim of the Andrieu study was to examine the possible connection between reproductive history and family history of breast cancer, and the risk of getting breast cancer. While it is not unusual for such a meta-analysis to include data from studies not previously published, the data on abortion was the most striking and significant data that had emerged from the 1988 Rohan study, from which study it had been omitted. Specifically, Rohan et al. found no significant effect of spontaneous abortion (in agreement with established worldwide data), but found a R.R. = 2.6, a 160 per cent increased risk of breast cancer among women who had had an induced abortion!

It is difficult to find words to describe the fact of a scientific research team, investing enormous amounts of time, effort and public funds, publishing their findings in a medical journal, yet deleting their strongest and most significant finding. Under any reasonable standard of scientific integrity, this simply is not done. But in the case of Australian breast cancer research, evidence has now emerged that it is done again and again.

As noted above, the Rohan data emerged in time to include it in our 1996 meta-analysis, wherein we pooled all the data that had been published specifically in regard to the effect of induced abortion and breast cancer risk. The combined result was a statistically significant, overall 30 per cent increase in breast cancer risk among women who had chosen abortion. But rather than provoking alarm among medical societies and public health agencies that such a common elective surgical procedure could increase the risk of such a common life-threatening disease, it provoked a counter-offensive of studies which claimed to prove that the A.B.C. link did not exist.

Within three months of the publication of our meta-analysis, a large study specifically devoted to the question of the A.B.C. link was published in perhaps the most influential medical journal in the world, the New England Journal of Medicine. In fact, this study by Danish epidemiologist Mads Melbye and colleagues, was--and still remains--the largest single study of its kind. The Melbye study included all 1.5 million women born in Denmark born between 1935 and 1978, comprising over 300,000 abortions and over 10,000 cases of breast cancer, diagnosed through 1992. Moreover, since the study was based entirely on medical records of abortions--rather than on the potentially less reliable information provided by women at interview years later--it was widely touted as definitive. Melbye's "definitive" conclusion was: "Induced abortions have no overall effect on the risk of breast cancer."

Without going into excruciating detail, suffice it to say that the methodological flaws in the Melbye study were breathtaking. For example, the medical records they used turned out not to be so reliable, as they included no abortions at all that took place before 1973. A simple check of the published national statistics of Denmark, however, reveals that some 60,000 of the women in the study who did have abortions, were misclassified as not having had any abortions. Most glaring of all was the fact that the Melbye study included data on breast cancer dating back to 1968--over five years before any of the abortions they included. If there is a most fundamental rule of any scientific research--epidemiological or otherwise--it is that the exposure in question (in this case, abortion) must always precede the outcome (breast cancer.) Yet even with such clear violations of scientific methodology, Melbye et al. still detected a significant risk-increasing effect of second trimester abortions, although they neglected to include this fact among their official "conclusions".

Since the Melbye study's publication in 1997, there has been a steady stream of studies published on the A.B.C. link, mostly based on medical records and all claiming no significant A.B.C. link. These studies included women from Sweden, the United Kingdom, the United States and China. These studies are plagued by a host of methodological flaws sufficient to render their findings invalid. We have detailed these flaws in published letters in the medical journals in which they appeared. The most common flaw is similar to the most serious flaw of the Melbye study, that is, the existence of huge gaps in the database resulting in the misclassification of women who had had abortions as not having had any abortions.

In 2003, an international "workshop" was convened by the U.S. National Cancer Institute, specifically addressing the A.B.C. link, among other "early reproductive events". The summary finding agreed upon by almost all of the 100 participating scientists was that it was now "well established" that "induced abortion is not associated with an increase in breast cancer risk". While claiming to be a thorough review of the extant data, no such review was conducted at the "workshop", a fact I can attest to as an invited participant. In fact, I submitted a "minority report", which receives scant mention on the N.C.I.'s website (www.cancer.gov), without naming the author or where to obtain the full report. (It is available on the website of the Breast Cancer Prevention Institute, www.bcpinstitute.org).

Nevertheless, controversy about the existence of the A.B.C. link continued, and in 2004, yet another study appeared which has been widely touted as definitive. Published this time in the prestigious British medical journal The Lancet, this study purported to be a comprehensive meta-analysis of extant data, a "collaborative reanalysis" comprising data from 52 individual studies. Again, while it is not unusual to include unpublished data in a meta-analysis, the Lancet "reanalysis" was more noteworthy for the published data that it did not include. In fact, lead author Valerie Beral of Oxford University and colleagues excluded some fifteen studies on non-scientific grounds. Specifically, eleven were excluded on such grounds as "principal investigators ... could not be traced", or "original data could not be retrieved", or "researchers declined to take part in the collaboration" or "principal investigators judged their own information on induced abortion to be unreliable", and four other studies were excluded by simple omission, without any mention at all. Disturbingly, most of these were studies which had reported a significant A.B.C. link.

Meanwhile, 28 studies' worth of unpublished data--more than the number of published studies included --were thus combined to result in the authors' conclusion: "Pregnancies that end as a spontaneous or induced abortion do not increase a woman's risk of developing breast cancer."

It is most interesting, however, to examine some of the studies which were included, for we find again, several cases of abortion data from previously published studies, but wherein the abortion data did not appear in the original publication. Of particular interest here are two studies on Australian women which had been published in 1998, one by M.R.E. McCredie et al. and one by J.L. Hopper. The results of these two studies are presented in the Beral "reanalysis" as a combined statistic, that statistic being a significant R.R. = 1.41, that is, a 41 per cent increased breast cancer among women with a history of induced abortion (but no increased risk with spontaneous abortion).

The A.B.C. link is therefore emerging as a classic cover-up of a common cancer-causing exposure, much as was the case with cigarettes in decades past. Australia has the dubious distinction of being the only nation so far known in which every published study (now known to be three) to consider the A.B.C. link has not only found significantly increased risk but has knowingly kept the data out of the public view for at least six years (and then, only buried it within larger summary reviews). It is a good sign that Australia is also the first to score a legal victory against those practitioners who would so harm women by withholding such vital information as the increased risk of breast cancer.