Written by Joel Brind
Prominent abortion practitioner and promoter David Grimes bemoans that bumper stickers still warn that abortion increases the risk of breast cancer, even though, he asserts, that “theory . . . was debunked long ago.” So begins Grimes’s recent piece on the Huffington Post’s blog Healthy Living. “Long ago” was, though Grimes doesn’t say so, 1997 to 2008, when there flowed a stream of “debunking” publications — largely studies that were methodologically flawed — reporting that no abortion–breast cancer (ABC) link existed. They were effective in fading the ABC link from public consciousness. But now the ABC link has returned, stubbornly, provoking renewed efforts to debunk it.
Being real, the ABC link is showing up, conspicuously, as millions of women worldwide who have had abortions over the past several decades are coming down with breast cancer at alarmingly increased rates. Dozens of papers are being published that show the trend. Grimes does not acknowledge the recent studies, however, relying rather on the discredited arguments of “long ago” — and some clever sleight of hand — in his shoddy attempt to disprove the link.
The first epidemiological study to show a link between induced abortion and breast cancer was published in 1957. In 1996, a research team I headed up published a comprehensive review and meta-analysis of worldwide research on the subject — some 23 studies. Our finding of a statistically significant, 30 percent increase in breast cancer among women who had had an abortion prompted a major, decade-long backlash from many mainstream medical organizations, medical journals, and government public-health ministries.
Grimes’s case rests almost entirely on two studies from Scandinavia in the 1990s. One of them is a prominent 1991 study, using data from Sweden’s Karolinska Institute (a committee there appoints the laureates for the Nobel Prize in Physiology or Medicine), that supposedly exposed the main methodological flaw in earlier epidemiological studies that found an ABC link.
Typically, epidemiological studies rely on a retrospective design. That is, they identify a group of patients with the disease in question and a matched group of control subjects without it. The subjects are then interviewed or asked to answer a questionnaire. They are instructed to review their lifestyle and relevant medical history. Differences between patients and control subjects with respect to the frequency of the exposure in question — here, abortion — provide a measure of “relative risk.” So, for example, if 30 percent more of the patients report a history of abortion than do the control subjects, the relative risk would be 1.3, the figure that our group found when we combined worldwide data in 1996.
But the 1991 Swedish study claimed to find evidence of “response bias” in the interview-based data, which they compared with computerized medical-record-based data on the same Swedish women. It is well established that, in interviews, women tend to underreport their abortions. But the Swedish researchers posited that women with breast cancer would be more honest about their abortion history and would underreport less. Hence, in a retrospective study in which more past abortions were reported by the breast-cancer patients than by the control subjects, the finding of increased risk was deemed a possible “artifact” — that is, a false finding — due to “response bias.”
Grimes even shows tables as evidence of how such differential “underreporting” “caused an apparent 60 percent increase in risk!” among women who had an abortion. He uses the simple past tense when explaining the data, as if they came from a real study. Trouble is, the tabular data Grimes shows are 100 percent fictitious, and he never tells us they are hypothetical.
The tables from the 1991 Swedish study compared the survey data with the data from computerized records and documented the “overreporting” of abortions among patients, and those tables formed the basis for the finding of significant response bias among breast-cancer patients. “Overreporting” means that patients imagined abortions that they never had — that is, they reported abortions that did not appear on their computerized medical record, the assumption being that the computerized record was correct. So preposterous is the notion of overreporting that, in 1998, the authors of the 1991 Swedish study, led by Olav Meirik of the World Health Organization, publicly retracted the claim, in correspondence published by the British Medical Association. They admitted that the phantom abortions were real but “not recorded as legally induced abortions.”
Although the Meirik study, the only published study that claimed to report direct evidence of the response-bias hypothesis, was debunked in 1998, response bias continues to this day to be invoked as a matter of fact by Grimes and other safe-abortion advocates. He says it explains why “abortion foes got it wrong” about the ABC link.
Pursuing the response-bias argument, Grimes and other researchers claim that studies that are based on medical records and other prospective data are immune to response bias and that they show no increased risk of breast cancer among post-abortive women. The largest study that Grimes relies on to make this point is “a landmark prospective study,” by Mads, Melbye, et al. (1997), based on medical records of 1.5 million Danish women. But Grimes neglects to tell us that this study was debunked by published correspondence that demonstrated that 60,000 women in the study who had had legal abortions on record were misclassified as not having had an abortion. Also, the fundamental rule of temporality was violated by the authors’ inclusion of breast-cancer diagnoses since 1968 but of abortions only since 1973. The exclusion of all the pre-1973 abortion data was based on the Danish authors’ false claim that abortion was legalized in Denmark in 1973, when in fact it had been legalized for reasons other than medical necessity in 1939, and only further liberalized in 1973.
All of Grimes’s rehashing and misrepresentation of bad science might be funny were the ABC link not devastating so many women’s lives. In this regard it is hardly funny that Grimes is silent on the veritable tsunami of ABC-link evidence that has poured in from Asia in just the past few years. A 2014 meta-analysis of 36 studies from mainland China reported a 44 percent overall increase in breast-cancer risk among women who had had an abortion. But the strongest evidence comes from South Asia — India, Pakistan, Bangladesh, Sri Lanka — where the typical woman marries young, has several children and breastfeeds them all, and never drinks alcohol or smokes cigarettes. In such populations, where there is little else besides abortion to cause breast cancer, relative risks for abortion average greater than fourfold and as high as twentyfold, according to at least a dozen South Asian studies in the past five years alone.
Contrary to Grimes’s claim that the ABC link was long ago “debunked,” the epidemiological evidence has grown tremendously stronger. The inference of a causal association between abortion and breast cancer has become all the more compelling, with our advancing knowledge of the hormonal changes during pregnancy and of how such changes during interrupted pregnancies dovetail with the susceptibility of cells in the breast to become cancerous.
But Grimes makes mention of no data from this century at all, only of data from “long ago.”
With over a billion women in China and India alone, a very conservative prediction would be that in the coming decades, millions there will die of breast cancer that can be attributed to abortion. No wonder Grimes is not interested in the recent data: It’s devastating to his “safe abortion” agenda.
Joel Brind is a professor of biology and endocrinology at Baruch College, City University of New York, and a board member and co-founder of the Breast Cancer Prevention Institute. Originally posted at National Review.com.