|
ABORTION AND BREAST CANCER: ONLY FUZZY MATH CAN MAKE THE ABC LINK DISAPPEAR [Joel Brind, Ph.D. 8Apr04] A supposedly definitive study of immense statistical power, published in a top medical journal, has once again [tried to prove] the abortion-breast cancer link (ABC link) nonexistent.
This time [25Mar04] it was "a collaborative reanalysis of data from 53 epidemiological studies, including 83,000 women with breast cancer from 16 countries". It was authored by a prestigious group of Oxford researchers, and published in the Lancet, one of the most prominent medical journals in the world…To say that the Beral study is seriously flawed and that its conclusions do not stand up to close scrutiny is to understate seriously the magnitude of what is really going on here. For starters, the claim that this is a "full analysis" is flatly false. Let's just do the simple math. We start with 41 studies which showed data on induced abortion and breast cancer, dating as far back as 1957. Then how do we get to 53 studies? (Actually, the total is 52 studies.) We add 11 studies worth of unpublished data, right? That might be okay, but it wasn't what was done.
What Beral et al. actually did was:
**Throw out 2 studies for the scientifically appropriate reason that "specific information on whether pregnancies ended as spontaneous or induced abortions had not been recorded systematically for women with breast cancer and a comparison group." Specifically, one such study from Sweden in 1989 used general population statistics for comparison, instead of a control group, and one US study from 1993 ascertained abortions only indirectly, by subtracting the number of children from the number of pregnancies.**Throw out 11 more perfectly good studies for reasons such as: "Principal investigators ... could not be traced" (We can't find Professor Einstein, either. Does that mean we throw out relativity?); "original data could not be retrieved by the principal investigators", "researchers declined to take part in the collaboration", or investigators "judged their own information on induced abortion to be unreliable" (even though it had been published in a prominent medical journal).**Finally, 4 studies' worth of data (one on French women, one on Chinese women, One on Russian women, and one on African-American women) were simply not even mentioned, even though they had been previously published as abstracts or included in other reviews.**That brings the total down from 41 to only 24 studies. Now we add 28 studies worth of unpublished data, and we have 52 studies.
The fact that the majority of studies have not stood the test of peer review is troubling enough. But a closer look at the excluded studies is even more revealing. Of the 41 studies which have been previously published, 29 actually show increased risk of breast cancer among women who have chosen abortion. (Epidemiologists call this a "positive association".) 16 of these are statistically significant, which means there is at least a 95% certainty that the results cannot be explained by chance. Getting back to Beral's "full analysis", 10 of 16 significantly positive studies in the literature were excluded for one of the unscientific reasons cited above.
In fact, if we average all of the 15 studies Beral excluded for unscientific reasons, they show an average breast cancer risk increase of 80% among women who had chosen abortion. So if we just add up all the studies Beral's group decided selectively to include, we get no significant effect of abortion on breast cancer risk.
But we haven't even gotten to Beral's main argument yet. She actually divided the included studies into two types; those which used retrospective methods of data collection (i.e., interviews of breast cancer patients v. control subjects), and those which used prospective methods (i.e., medical records taken long before breast cancer diagnosis). The retrospective data-based studies are thought to be less reliable, because, as Beral told the Washington Post, women with breast cancer "are more likely than healthy women to reveal they had an abortion, leading to the conclusion that there are more abortions among this group".
Readers may recognize this "reporting bias" or "response bias" argument, used for over a decade now to dismiss the overwhelming majority of studies (which are retrospective data-based) which reveal an abc link. It is actually a hypothesis worthy of testing.
The trouble is, tests for such bias have proven negative over and over and over again in the published literature, in studies as far flung as Japan, the US and Greece. In fact, Beral still reaches back to a 1991 Swedish study, which was the only one ever to claim direct evidence of such "reporting bias". However, that study's conclusion depended upon the assumption -since publicly retracted by the original authors- that breast cancer patients had "overreported" abortions (i.e., reported abortions that had never taken place.)
That brings up another serious flaw in the Beral study, specifically, the exclusion of any published critiques of studies she found acceptable. She included uncritically, for example, data from a 1990 study on Norwegian women which study had found no link. However, in 1998 our own group published a rigorous, mathematical proof that those data were incorrectly compiled, and had actually indicated increased risk among Norwegian women.
Getting back to the reporting bias argument, Beral separately compiled all the studies that used prospective methodology (13 studies) and those that used retrospective methods (39 studies), and found the results to be significantly different. Specifically, the former showed a significant overall 7% decrease in risk with abortion, while the latter showed a significant overall 11% increase in risk. Beral's conclusion? "We have demonstrated that a certain group of studies (the ones with retrospective data) are unreliable and can't be trusted,", she told the Washington Post."
There are only two things wrong with that conclusion. First, it is completely illogical to leap to the conclusion that, just because there is |
Abortion 
