Rotten in Denmark: Melbye Study flawed due to incomplete cohort

back to the abortion files            back to abortion and breast cancer page

Dr Joel Brind and others have postulated that if ALL the women who had had abortions had been counted in the Melbye study, the study would have found a very high incidence of breast cancer among peri menopausal and post menopausal, aborted women.  Dr Melbye admitted in the conclusion of the study that the data may not be accurate because many women having abortions before computerization were not included in the cohort. One wonders how Dr Melbye could triumphantly tell the media that the matter was settled (no link) when he knew most of the women at menopausal age had not been counted??  Roy Spencer, former NASA climatologist explains that Scientists are human and if they don't come up with eye catching headlines, they may not get funding. You can hear Dr Brind talk about the abortion/breast cancer link in the video below. (his article follows)



Article by Dr. Joel Brind


The Culture of Life Foundation -
Published in the Jan. 9 issue of the prestigious New England Journal of Medicine2, (NEJM) the most recent study on abortion and breast cancer was hyped most brazenly by itís own lead author, Dr Mads Melbye of the Statens Serum Institut in Copenhagen. "I think this settles it. Definitely--there is no increased risk of breast cancer for the average woman who has had an abortion", Melbye told The Wall Street Journal 3.

Headline news stories trumpeted these conclusions worldwide, and lead editorials from the NEJM4 itself to The New York Times 5 echoed the reassuring counsel that women "need not worry about the risk of breast cancer" when considering abortion 4.

What the Melbye study claimed to find was an overall breast cancer risk increase of exactly 0.0% among Danish women who had had at least one abortion. Actually, a few other studies (five6-10 out of twenty-nine, to be exact) had also found no tendency toward increased risk.

Even if the Melbye study were entirely valid, it would still be outrageous to claim that one single study, on a population of women from another continent, could somehow prove that women from the U.S. and elsewhere were not put at increased breast cancer risk by choosing abortion. After all, ten studies11-20 out of eleven on American women have shown increased risk, eight of them13-20 statistically significant on their own! This fact alone belies the political agenda behind claims that the "issue should now be settled" 21, that "the notion that abortions cause breast cancer (has been) largely disproved" 5.

Such claims, however, were reportedly based on two aspects of the Melbye study which were supposed to put it head and shoulders above the rest, namely: 1) It covers an extremely large population, i.e., all 1.5 Danish women born from April 1, 1935 to April 1, 1978, and 2) It relies entirely on prospective (see glossary, page 7) computerized records, which are considered more accurate than data based on womenís own recollections, on which most other studies were based. But this studyís flaws are many and ultimately, fatal.

The errors in the Melbye study may be grouped into three categories, as detailed below:

1) The inappropriate selection of computerized data from Danish birth, abortion and breast cancer registries

a) The use of birth records back to 1935 meant that women in the study were as old as 38 when the computerized abortion registry began, so records of abortions among them are largely missing. This is reflected in the extremely atypical age distribution of the patients in the study who had any abortions. As shown in Figure 1, the majority of these patients are on record as having had their abortions over the age of 35, whereas the average abortion client in Denmark is only 2722 (in the US, she is only about 2223). The acknowledgement of this source of error by Melbye et al. epitomizes understatement: "we might have obtained and incomplete history of induced abortions for some of the oldest women in the cohort." In fact, since abortion was legalized in Denmark in 1939 (although Melbye et al. misrepresent this as well: "In 1973, the legal right to an induced abortion through 12 weeksí gestation was established for women with residence in Denmark."), the published record of legal abortions from 1940 to 197322 reveals that 60,000 women in the Melbye study cohort listed as not having had any abortions, actually did have abortions! That means Melbye et al., who listed 281,000 women as having abortions, misclassified more than one of every six women in the study cohort who had an abortion. Importantly, these are also the oldest members of the cohort, which makes them the majority of the women who have gotten breast cancer.

b) The inclusion of women born as recently as 1978 inflates the population size studied by over 350,000 subjects. That is the number of women born between 1968 and 197822, who were therefore under 25 at the termination of the study in 1992. Almost no one gets breast cancer that young. The published Danish records show that these young women account for a grand total of only 8 cases of breast cancer25-28, which amount to less that one tenth of one percent of the 10,246 cases of breast cancer in the study. In contrast, these young women account for over 40,00022, or 11% of the 371,000 induced abortions reported in the study.

c) The inclusion of women who got breast cancer between 1968 and 1973 is patently ridiculous: Of course every one of these over 300 patients25 is listed as having no abortions, since abortions before Oct. 1, 1973 are not included in the study! This error, of course, serves to lower, artificially, the calculated relative risk.

2) Invalid statistical adjustment of the raw data

The consequence of selecting such an inappropriately large segment of the Danish population is the need to adjust substantially the calculated value of relative risk to account for the large age differences between the average woman who had an abortion and the average woman who got breast cancer. This adjustment tends to raise the relative risk estimate. Interestingly, the raw relative risk (more precisely, the rate ratio) is not shown in the paper, although enough data are shown to calculate it (1.44, or a 44% increased breast cancer risk). As a round figure, the real, adjusted breast cancer risk increase attributable to abortion for this study population would seem to be around 100%! (Indeed, an earlier study on part of this population29 reported an almost 200% risk increase with abortion among childless women.) Yet somehow, Melbye et al. end up with an adjusted relative risk estimate of exactly 0.0%!

What Melbye et al. did was to adjust the data for birth cohort. A birth cohort refers to a population of people born during the same time period, and it differs from the adjustment for age. Age adjustment corrects for the fact that breast cancer risk increases with a womanís age. Therefore, one must adjust the relative risk estimate upward if oneís study compares women with cancer who are, on average, older than the cancer-free women in the study. The birth cohort adjustment is supposed to correct for the fact that breast cancer risk differs for women who are the same age when they get breast cancer, but who were born at different times. For example, the breast cancer incidence among 50-year-old women born in 1940 was higher than that for 50-year-old women born in 1930, since the incidence of breast cancer has been steadily on the rise for most of this century for women of any given age.

All this seems quite reasonable except for one extremely crucial fact: The reason why breast cancer has been on the rise for most of this century is unknown. If abortion is one of the reasons, adjusting for this birth cohort effect, as Melbye et al. did, necessarily eliminates the effect of the very factor under study (abortion) and virtually guarantees the null result they obtained! In fact, the pattern of breast cancer risk with time25-29 in women in Denmark, is seen to move in striking parallel to the pattern of induced abortion exposure22. Figure 2 shows the close correlation of age-adjusted breast cancer risk and frequency of induced abortion, examining the data on all Danish women in 5-year (the standard interval used) birth cohort groupings.

3) Mischaracterization of their own findings and of other published research in the field.

a)Even on the paperís face, the "Conclusions" section of the abstract of Melbye et al. is noteworthy for its brevity: "Induced abortions have no overall effect on the risk of breast cancer." Yet even with the paperís eggregious flaws, it still evidenced a statistically significant trend of increased breast cancer risk with increasing gestational age at abortion. Thus, as their data table shows, they found a 12% risk elevation at 11-12 weeks (late first trimester) rising to an 89% risk increase for abortions later than 18 weeks. This didnít show up in their "overall" relative risk figure, because most abortions in Denmark (as in the US) are done somewhat earlier (7-10 weeks) in the first trimester, and their effect on breast cancer risk was statistically eliminated (see section 2 above).

b) Melbye et al. also make no mention in the abstract or the discussion of the fact that they found a 29% increased breast cancer risk among women who had an abortion as a teenager. The reason the authors were able to avoid mentioning this statistic is the fact that it was not statistically significant. The reason it was not statistically significant is the small number of women in this category in their study, due to the recent inception (1973) of the computerized abortion registry they used.

c) The authors disparage and dismiss the findings of the worldwide literature compiled in our 1996 "Comprehensive review and meta-analysis"1 with the blanket statement: "it is not unreasonable to assume that many of them (the 23 worldwide studies we compiled) were inherently biased, making the pooled conclusions biased as well." (See the "Brief guide to epidemiological study design and interpretation", page 7 of this issue.) Two important points should be made here: i) The potential problem of bias in case-control control studies, is, as Melbye et al. point out, that it would, if it exists, result in the "differential misclassification" of women who had abortions as not having had abortions. The irony here is that the very study of Melbye et al. takes the proverbial cake for misclassification by orders of magnitude, considering their misclassification of 60,000 women (see 1a above)! ii). The overall average 30% risk increase attributable to induced abortion which we reported in our meta-analysis, and which Melbye et al. take exception to, is virtually the same statistic they obtained for women who had abortions as teenagers (see 3b above).

d) Melbye et al. misrepresent their findings as essentially confirming the findings of other researchers who investigated the ABC link using prospective, computerized data: "This result is very much in line with the results of previous ... cohort studies". To back this statement up, they cite four studies, three of which30-32 concern only (or mostly) spontaneous abortion (miscarriage), and are therefore entirely irrelevant. (See article in this issue for why miscarriages donít usually increase breast cancer risk.) They also conveniently ignore the fact that one of the studies showing significantly increased risk (90%) also used prospective computerized data14.

It is of course outrageous that the Melbye study, is being used to spread the claim that the ABC link does not exist. This is not good news for women, who are still actively being kept in the dark by the very agencies who should warn them about avoidable cancer risks. In this regard, Dr. Patricia Hartge, author of the NEJM editorial4 which accompanied the Melbye study, is only the most recent National Cancer Institute researcher disparaging the link. However I find it encouraging that the ABC debate is finally out in the open. I have already published a brief rebuttal in The Wall Street Journal33, and coauthored another which is in press in the NEJM34. The fact that Melbye et al. did such a bad job will only hasten the day when the ABC link will be common knowledge. -jb-

Dr. Joel Brind, Ph.D.of Baruch College in New York City, is one of the leading researchers in the area of Abortion-Breast Cancer link. He is also on the Advisory Board of the Life Research & Communications Institute.

NOTE: Joel Brind's personal website is down at this time and the footnotes have been lost. I am searching for these at the present time.  Apologies (no one seems to think footnotes are important - I got this from a reprint on another site).