New NCI study finds--but covers up-- ABC link in North Carolina women


The US National Cancer Institute (NCI) never seems to run out of epidemiologists willing to sweep the ABC link under the rug. This time Pamela Marcus of the Biometry Branch,
Division of Cancer Prevention, also on the faculty at the School of Public Health of the University of North Carolina (UNC), published, with colleagues from UNC. the results of
the Carolina Breast Cancer Study19, in the August 1999 American Journal of Public Health.

In contrast to most studies—and the prevailing consensus—of the last three decades, the primary "Conclusion" of the Marcus study is: "Pregnancy during adolescence does not appear to influence breast cancer risk, but breast-feeding may." That Marcus et al. found no effect at all of a first full-term pregnancy before age 18 (compared to age 20 and over) is unusual but not unprecedented, Thus, for example, Daling et al.9 also found almost no effect of early full-term pregnancy. The significant (80%) reduction in
premenopausal breast cancer risk with lactation is a striking finding, but consistent with the widely observed protective trend. The study also reported a significant risk increase with oral contraceptive use among African-American adolescents, which trend the authors say "warrants further study". Considering the fact that this corroborates many earlier studies, it would seem to warrant more of a red flag than "further study", but this finding will not be further discussed here.

Actually, the study appears to be rather straightforward and correct. Appropriately, the analysis of the data with regard to induced abortion is restricted to women under age 50
(recruited into the study between 1993 and 1996), which translates to about age 30 or less in 1973, when abortion was legalized by Roe v. Wade.

As to the data reported for induced abortion, the relative risk is calculated with a comparison of adolescent women (age less than 20 years) with an induced abortion to all
women who never had an abortion as well as to women who had either a full-term pregnancy or a miscarriage under age 20. The reported data (adjusted for confounding
variables) are essentially the same with either comparison: a relative risk of 1.2 which is not statistically significant. An overall relative risk is also presented for women who had their first (or only) induced abortion at age 20 or older, and this is given as a non-significant 1.1. In the text, the results for the adolescents are described as follows: "Neither miscarriage nor induced abortion before 20 years of age, relative to no history of either event, conferred a meaningful increase in risk"

There is irony in the use of the word "meaningful" here, because that word has no in epidemiology: It is epidemiologically meaningless; a spin-word, as it were, used to dismiss a finding as a non-finding. Looking back at the data, we see indeed that the elevated relative risk is indeed non-significant. Why? Because it is derived from only 19 patients and 11 healthy controls who had an abortion under age 20. (It is also non-significant for the 20-and-olders, of whom there were also a relatively few 65 cases and 45 controls).

What would be a correct characterization of such data? It depends. Were the study done in isolation, such a small risk increase would hardly attract any attention in the absence of corroborating data. But such is hardly the context:

In the US alone, this study represents the thirteenth to show data on induced abortion and breast cancer risk, and it is the twelfth to show a risk increase. Moreover, the average
increase observed is about 30% overall, so an increase of 20% is exactly in the range of previous studies. Couple this to the fact that induced abortion is overwhelmingly an elective exposure, and that a single abortion makes a difference, and the characterization of "not meaningful" is seen to be glaringly inaccurate and misleading.

But there is more. A closer examination of the data presented shows that the statistical adjustments used to arrive at the listed relative risks are unusually large: The raw (unadjusted) relative risks for women with an adolescent abortion are calculated to be 1.74 and 1.43, compared to the adolescent "full term-or-miscarriage" group and the "never had an induced abortion" group, respectively. No other data given in the entire paper reflect such a large adjustment. What are the data adjusted for? The table’s footnote says: "adjusted for race and age at diagnosis-selection, as well as sampling design" From the text, this would include all the design variables, including oral contraceptive use, which the study found to be a risk factor. If oral contraceptive use correlates with induced
abortion, some of the risk attributed to oral contraceptive use might actually be due to induced abortion.

In any event, the real bottom line of the Marcus study is the observation of increased risk of breast cancer of a magnitude consistent with most earlier studies (Computing the overall raw relative risk for all the women in the study aged under 50 years, we obtain 1.24, or a 24% risk elevation.). The risk was also higher for adolescents. The sample size was too small, however, to achieve statistical significance.

But one could hardly guess that, when the only mention of induced abortion in the abstract is: "Results: Miscarriage, induced abortion, and full-term pregnancy before 20 years of age were not associated with breast cancer."

This is of course disappointing, but not surprising, in light of the continuing role of the NCI in preventing public knowledge of the ABC link. It is also not unexpected for the ‘politically correct’ American Journal of Public Health, which ran an editorial20 in 1996 which was nothing less that a call to arms to "provide a vigorous and organized public health
defense against the attacks on abortion." (See Spring 1998 ABCQ Update.) -JB-