THE FIRST epidemiological study to show significantly higher rates of breast cancer in women who had any abortions was published in Japan in 1957. The tally now stands at 25 out of 31 studies showing this link between breast cancer and abortion (but not miscarriage), a link demonstrated in women of European and African, as well as Asian ancestry.

The distinction between abortion and miscarriage is absolutely crucial to understanding the abortion-breast-cancer link, for it goes to the heart of the question of how abortion raises breast-cancer risk.

In medical terms, what is commonly called "abortion" or "termination" is known as "induced abortion". Importantly, what is commonly called "miscarriage" is medically referred to as "spontaneous abortion".

The potential for confusion here is obvious. For example, when Dorothy Broom had Wayne Smith (CT, November 28, page C3) cite the results of American cohort studies in which "there was no difference in the history of spontaneous abortions between those with and those without breast cancer", only a reader well versed in cancer epidemiology knows that this statistic is entirely irrelevant to the issue at hand, for spontaneous abortions are known not to increase the risk of breast cancer.

Why should it matter how and when a pregnancy ends? The answer is in the nature of the pregnancy itself. Usually, miscarriages occur because of insufficient production of the pregnancy hormone progesterone by the mother’s ovaries. They also make another hormone called estradiol—the main chemical form of estrogen — from progesterone, and so its concentration in the mother’s blood parallels that of progesterone.

Consequently, just days after conception, when progesterone production starts to rise steeply, so does that of estradiol. By the end of the first trimester, the concentration of estradiol in the mother’s blood is 20 times higher than at the time of conception, but not in most pregnancies that miscarry. Estradiol prepares the breasts for the job of milk production by making the breasts grow. In fact, most known breast-cancer risk factors operate by increasing a woman’s exposure to some form of estrogen. Thus, women who have more menstrual cycles by entering puberty earlier or menopause later, or who take oral contraceptives or estrogen-containing medications over long periods of time, also have measurably increased breast cancer risk.

Why then, doesn’t childbirth also confer increased breast cancer risk? During the third trimester — if there is one —of a normal pregnancy, other hormones (whose actions are not fully understood) make the breast cells differentiate from growing (and therefore, potentially cancer-forming) cells, into milk-producing cells, thus leaving breasts with fewer immature cells capable of ultimately becoming cancerous.

In contrast, an artificial termination during the first or second trimester leaves the breasts with more such cells. It is undisputed that a young, childless woman who gets pregnant will lower her risk of future breast cancer if she carries the pregnancy to term.

Discredited studies

But even after a woman has any children, the increased exposure to estradiol she gets with a subsequent pregnancy, without the growth-stopping effect of third-trimester hormones, will also increase her breast-cancer risk.

Incredibly, such public-health officials as Broom and Smith continue to rely on discredited evidence of something called "response bias", to explain away the data.

For example, they cite a 1991 Swedish study which compared the responses given in interviews of women without breast cancer with computerised records obtained years earlier, at the time of the abortion.

They found that (according to Broom and Smith) "controls [without breast cancer] under-reported their previous abortions significantly more often than cases [with breast cancer]". What Broom and Smith fail to note is that the "significant comparison is with cases who supposedly "over-reported" their abortions — i.e., reported having abortions that never took place (which was assumed to be the case if the computer had no record of the abortions the women reported). Is this assumption preposterous? Of course! That is why the Swedish group which made the claim finally retracted it in March of this year.

The Dutch study cited by Broom and Smith as providing "compelling evidence" of "response bias" also borders on the bizarre.

Yes, there were significant differences in accuracy of responses between different regions of the country, but the study design nullified this source of error by matching cases and controls by region.

And both regions showed an increased risk (average 90 per cent) with abortion. That was precisely the same magnitude of increased risk found in women who had chosen abortion in a 1989 New York State Department of Health study based on prospective, cohort data, i.e. ruling out response bias. Why don’t Broom and Smith cite this study?

Other studies showing a significant link have also ruled out response bias. In fact, credible evidence of response bias in this area of research has never been produced.

Yet Broom and Smith would have us believe that only cohort studies are reliable. That way they can ignore the only study on Australian women, which showed a 160 per cent increased risk with abortion, and cite instead, a 1997 Danish study which showed no overall link

What they don’t say is that this latter study suffers from the researchers’ own bias in selecting a data base including women born since 1935, but only including abortions performed since 1973, thus missing 80,000 abortions, and vastly under-estimating the link. Other crippling flaws In the Danish study have also been documented in the medical literature.

Sorry Dorothy and Wayne, but the overwhelming weight of the epidemiological and biological evidence does show a link between abortion and breast cancer. And that cannot be wished away by public-health officials, nor by the women considering abortion.

Joel Brind Is professor of human biology and endocrinology at Baruch College of the City University of New York.