Learning to Analyze the Data


High prevalence rate of a risk factor:

In some areas in the world, for example in the U.S., Australia, New Zealand, England, France, Sweien, and the Netherlands, a very high percentage of women born after 1950 have taken OCPs (ie, between 80-95%) [21]. Will this affect the results of future studies? It can. If a very high percentage of the female population used OCPs, one may "fail to find" a significant difference between "cases" and "controls." It was noted above that if a factor was rarely found in a society, it would be difficult to measure and assess its risk. In a similar manner, if a factor is almost universally prevalent, it again becomes difficult to assess that factor's true risk. An easy way of conceptualizing this is to consider what would happen if every woman in the population used OCPs in her life. Obviously, both "cases" and "controls" would have a 100% usage rate and it would be impossible to tell the difference between them. It is ironic to note that as more information continues to document the strong risk of early OCP use (Chapters 8 and 9), the population of young women in some countries (eg, Australia) who have used the OCP is approaching 100% [21, p.36S].

Mixing older and younger women in the database:

Mixing women of various ages in a study may artificially reduce the risk of certain factors such as an abortion at a young age or early OCP use, if the older women never experienced "accessibility" to that factor. Let us illustrate by comparing two studies. In the first, we study 100 "cases" and 100 "controls" ages 20 to 40 years old at the time of interview, assuming they were all interviewed in 1990. This is illustrated in Table 5A. ("first full-term pregnancy" is abbreviated as FFTP).


Table 5A:
Example Demonstrating The "Mixing of Data" Effect
"Cases" "Controls"
Total Number in Group -> 100 100
Women age 20-40 who had an abortion prior to their FFTP 50 40
Women age 20-40 who had no abortions 50 60


Here one can see that the estimated relative risk (RR) is 50/50 divided by 40/60 = 1.50. That is, this study shows that young women 20 to 40 years old who had an abortion prior to their FFTP, suffered a 1.50 or 50% increased risk of breast cancer. Now what would happen if we expanded the study and added 100 "cases" and 100 "controls" who were ages 40 to 45 years old to each group? These women would have been very unlikely to have had an abortion prior to their FFTP because they would have been 23 to 28 years old in 1973ัthe year that the U.S. Supreme Court decriminalized abortion. If none of these women had an abortion prior to their FFTP, our second study would contain data as noted in Table 5B.


Table 5B:
Example Demonstrating The "Mixing of Data" Effect
"Cases" "Controls"
Total Number in Group -> 200 200
Women who had an abortion prior to their FFTP 50 40
Women who had no abortions 150 160


Note that the estimated relative risk is now: 50/150 divided by 40/160 = 1.33. We can see that by adding older women to the database, who had little chance of having an abortion at a young age, the estimated relative risk has been artificially decreased. This is an important point because many studies mix older women with younger women and then erroneously claim that abortion at a young age or OCP use has no effect.

The "real risk factor":

Certain studies have concluded that there is "no link" between early induced abortion and/or early OCP use and breast cancer in spite of failing to ask the proper question. For example, Dr. Brind et al in their meticulous metaanalysis published in October 1996 [1], noted 33 studies that reported on the risk of abortion and breast cancer but failed to distinguish between induced and spontaneous abortion (miscarriage). Often the authors of these studies would conclude that abortion either is or is not a risk factor for breast cancer without distinguishing between the two different types of abortion. This is critical as will be noted in detail in Chapter 6, because both estradiol, progesterone and hCG levels are far lower in pregnancies which eventually miscarry than in healthy pregnancies which are unnaturally aborted (ie, induced abortion). Because of this phenomenon, induced abortion carries a higher risk than does miscarriage, and failure to distinguish between them can easily result in failure to pick up the real risk of induced abortion.

Other studies claim to measure the risk of induced abortion and breast cancer but have failed to ask the women in the study whether their particular induced abortion was before or after their first full-term pregnancy (FFTP). This is important also, because Dr. Brind et al found that having an induced abortion before a FFTP was associated with a 50% increased risk, whereas having an induced abortion after a FFTP resulted in a 30% increased risk. Any study that claims either to show or deny a link between abortion and breast cancer must specifically examine the risk of abortion performed at a young age (ie, before a FFTP) and should also distinguish induced from spontaneous abortions.

A similar phenomenon exists in the literature regarding OCPs. In the largest meta-analysis ever performed concerning OCPs, the authors of the Oxford study concluded: '11Vomen who are currently using combined contraceptives or have used them in the past 10 years are at a slightly increased risk of having breast cancer diagnosed. . . There is no evidence of an increase in the risk of having breast cancer diagnosed 10 or more years after cessation of use. . ." (The Lancet48). The study, however, failed to study the real risk factor, namely, OCP use prior to a first full-term pregnancy (FFTP) in premenopausal women. This is extremely important, because women today are using oral contraceptives more often and for longer periods of time prior to their FFTP, than they did in the 1960s and 1970s. Thus, millions of women may be at risk [21, Table 14]. Unfortunately, when a study of this magnitude reaches the medical community and the public it often gets translated into the message that "oral contraceptives pose no long-term risk." If the authors of the Oxford study had pursued the real question, namely, "Does OCP use prior to a FFTP cause breast cancer in premenopausal women?" They may have found their results to correspond closely with those of another author's meta-analysis in which she accounted for OCP use before a woman's FFTP and found that women who had taken OCPs for 4 or more years prior to their FFTP had a 72% increased risk in breast cancer49. Clearly, one must ask about and measure the real risk factor (ie, OCP use prior to a FFTP) if one is to honestly study this issue.

The death factor:

In most retrospective studies the researcher(s) will choose a number of women who have breast cancer, from a certain database, and then go on to interview them at a later date. If there is a period of even a few months (eg, 6 months) between the choosing of subjects and the subsequent interview, some of them, especially those with more aggressive breast cancer, may already have died, and thus cannot be interviewed. This is referred to as "the death factor." A good example of this is found in the example of a study performed in Oxford called the United Kingdom National Case-Control Study by Chilvers et al. All women in the study who had breast cancer were diagnosed with it between the dates of January 1, 1982 and December 31, 1985, but they were interviewed between the dates of January, 1984 and February, 1988. During this time 16% of the "cases" died [30, p.974]. This factor becomes critical when one notes that several papers have noted that women who have had an abortion at a young age or had taken OCPs early in life, developed a more aggressive type of cancer ([25, p. 267; 50, 51]. If this is true, then the women with breast cancer who died during the 6 months between identification and interview may well be disproportionately represented by those women who had an abortion at a young age and / or used OCPs early in life. In other words, studies in which a significant number of women have died before they were interviewed may well be giving results which we refer to as "false negatives." These studies may well claim that there is no relationship between abortion at a young age/OCP use, and an increased risk in breast cancer, when in reality the study could fail to pick up the risk factor(s) because the women who died were more likely to be those who had an abortion at a young age and/or had early OCP use.

Let us look at an example. Suppose a researcher interviewed two groups of women, 1,010 women with breast cancer and 1,000 without. Here 65 of the "cases" had an abortion at a young age versus 40 of the "controls." The estimated relative risk (RR) can be calculated below as:


Table 5C:
Sample Study of the Death Factor
1010 women with breast cancer 1000 without breast cancer
had an abortion pFFTP 65 (A) 40 (C)
no abortion pFFTP 945 (B) 960 (D)


The estimated relative risk (RR), as noted earlier, would be A/C diviied by B/D = (65) (960)/(945) (40)= 1.65 RR.

Now let us see what would happen if only 1% (10 "cases") had died before the scheduled interview and if 5 out of those 10 women had had an abortion at a young age. The table would now look as follows:


Table 5D:
Sample Study of the Death Factor
1010 women with breast cancer 1000 without breast cancer
had an abortion pFFTP 60 (A) 40 (C)
no abortion pFFTP 940 (B) 960 (D)


The estimated relative risk (RR) now becomes: (60) (960)/(940) (40)= 1.53, which is about a 7% increase from 1.65. In statistical terms, this difference could be the difference between a statistically significant or insignificant result. Janet Daling, a prominent researcher from the Fred Hutchinson Cancer Research Center in Seattle, Washington, is one of the few researchers to acknowledge this effect*: (*This author knows of no researcher who has pointed out this effect, except Daling.) ". . tit could be argued that those women with breast cancer whom we were unable to interview because of serious illness or death may have been more likely to have had an induced abortion than the women we did interview. If this bias were present, we would have underestimated the risk of breast cancer that is associated with induced abortion." [29, p.1589].

Unfortunately, even some of the larger studies which have studied early OCP use have had huge "death factors" in which the "cases" have had a far higher incidence of death or serious illness than the "controls" and thus could not be interviewed. Some examples include: the CASH study52, 9% of "cases too ill" before interview (BI); the New Zealand study [15], 1.9% of "cases" died versus 0.5% of "controls"; the United Kingdom study [30] by Chilvers, 16%"cases" died BI; and White's study [38] in the Journal of the National Cancer Institute, 6.5% "cases" died BI. All of these studies have had large percentages of "cases" who either died or were too ill and so could not be interviewed.

A similar phenomenon has occurred in the studies concerning abortion at a young age. In her well-known study, Janet Daling [29] noted that 5.7% of the "cases" died before interview (BI); Newcomb53 noted that 5.4% died BI; and Pike54 noted that about 15% had died BI. This effect, which has rarely even been acknowledged, may well be responsible for underestimating the real risk of both having an abortion at a young age and OCP use. Hopefully in the future, researchers will compensate for this effect by interviewing women soon after they are chosen to be in a study or by obtaining the medical histories from the women who have died before interview by looking at their medical records.


References:

1 Brind J, Chinchilli M, et al. Induced abortion as an independent risk factor for breast cancer: a comprehensive review and metaanalysis. J Epidemiol Community Health. 10/ 1996; 50: 481-496. [Back]

2 Brind J. ABC down under. Abortion Breast Cancer Quarterly Update. Summer, 1997. [Back]

3 Rohan, et al. Am J Epidemiol. 1988; 128: 478-489. [Back]

4 Leishman K. On the abortion front line. Health and Medicine for Physicians. March 1997: 22-26. [Back]

5 Huse D. Childhood Vaccination Against Chickenpox: An Analysis of Benefits and Costs, Pediatrics. June, 1994; 124: 869-873. [Back]

6 Rosenberg L, Palmer JR, et al. Case-control study of oral contraceptive use and risk of breast cancer. Am J Epidemiol. 1996; 143: 25-37. [Back]

7 Rosenberg L, Palmer JR, et al. A case-control study of the risk of breast cancer in relation to oral contraceptive use. Am J Epidemiol. 1992; 136: 1437-1444. [Back]

8 Rosenberg L, Palmer JR, et al. Breast cancer in relation to the occurrence and time of induced and spontaneous abortion. Am J Epidemiol. 1988; 127: 981-989. [Back]

9 Miller D, Rosenberg L, et al. Breast cancer before age 45 and oral contraceptive use: new findings. Am J Epidemiol. 1989; 129: 269-279. [Back]

10 Jick SS, Walker AM, et al. Oral contraceptives and breast cancer. Br J Cancer. 1989; 59: 618-621. [Back]

11 La Vecchia C, Decarli A, et al. Oral contraceptives and cancers of the breast and of the female genital tract. Interim results from a case-control study. Br J Cancer. 1986; 54: 311-317. [Back]

12 Kay CR, Hannaford PC. Breast cancer and the pill -- A further report from the Royal College of General Practitioners' oral contraception study. Br J Cancer. 1988; 58: 675-680. [Back]

13 Somerville S. Before you choose: The link between abortion and breast cancer. Purcellville, VA. Ann Intern Med. 1993. [Back]

14 Laarsson KS, et al. Predictability of the Safety of Hormonal Contraceptives from Canine Toxicology Studies. In: Michal, F. ed. Safety Requirements for Contraceptive Steroids. Cambridge: Cambridge University Press; 1989: 203-269. [Back]

15 Paul C, Skegg DC, et al. Oral contraceptives and risk of breast cancer. Int J Cancer. 1990; 46: 366-373. [Back]

16 Thomas DB, Noonan EA. Breast cancer and combined oral contraceptives: results from a multinational study [The WHO collaborative study of Neoplasia and steroid contraceptives]. Br J Cancer. 1990; 61: 110-119. [Back]

17 Lee NC, Rosero-Bixby L, et al. A case-control study of breast cancer and hormonal Contraception in Costa Rica. J Natl Cancer Inst. 1987; 6: 1247-1254. [Back]

18 Liniefors-Harris BM, Edlund G, et al. Risk of cancer of the breast after legal abortion during the first trimester: a Swedish register study. Br Med J. 1989; 299: 1430-1432. [Back]

19 Liniefors-Harris BM, Eklund G, et al. Response bias in a case-control study: analysis utilizing comparative data concerning legal abortions from two independent Swedish studies. Am J Epidemiol. 1991; 134: 1003-1008. [Back]

20 Ellery C, MacLennan R, et al. A case-control study of breast cancer in relation to the use of steroid contraceptive agents. The Medical Journal of Australia. 1986; 144: 173-176. [Back]

21 Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormonal contraceptives: further results. Contraception. 1996; 34: Sl-S106. [Back]

22 Anonymous. Another look at the pill and breast cancer. The Lancet. November 2, 1985: 985-987. [Back]

23 Colton T, Greenberg ER, et al. Breast cancer in mothers prescribed diethylstilbestrol in pregnancy. JAMA. 1993; 269: 2096-3000. [Back]

24 McPherson K, Vessey MP, et al. Early oral contraceptive use and breast cancer Results of another case-control study. Br J Cancer. 1987; 56: 653-660. [Back]

25 Olsson H, Borg A, et al. Early oral contraceptive use and premenopausal breast cancer -- A review of studies performed in southern Sweien. Cancer Detection and Prevention. 1991; 15: 265-271. [Back]

26 Hulka BS. Oral contraceptives, the good news. JAMA. 1983; 249: 1624-1625. [Back]

27 Melbye M, Wohlfahrt J, et al. Induced abortion and the risk of breast cancer. N Engl J Med. 1997; 336: 81-85. [Back]

28 Malone K, Daling J, et al. Oral contraceptives in relation to breast cancer. Epidemiologic Reviews. 1993; 15: 80-94. [Back]

29 Daling J, Malone K, et al. Risk of breast cancer among young women: relationship to induced abortion. J Natl Cancer Inst. 1994; 86: 1584-1592. [Back]

30 Chilvers C, McPherson K, et al. Oral contraceptive use and breast cancer risk in young women (UK National Case-Control Study Group). The Lancet. May 6, 1989: 973-982. [Back]

31 Rookus MA, Leeuwen FE. Oral contraceptives and risk of breast cancer in women ages 20-54 years. The Lancet. 1994; 344: 844-851. [Back]

32 Newcomb PA, Longnecker MP, et al. Recent oral contraceptive use and risk of breast cancer (United States). Cancer Causes and Control. 1996; 7: 525-532. [Back]

33 Pike MC, Bernstein L. Oral contraceptives and breast cancer. The Lancet. July 15, 1989: 158. [Back]

34 Pike MC, Bernstein L. Oral contraceptives and breast cancer. The Lancet. March 18, 1989: 615-616. [Back]

35 Olsson H, Ranstam J, et al. (Letter on CASH) The Lancet. November 23, 1985; 1181. [Back]

36 Staiel BV, Lai S, et al. Oral contraceptives and premenopausal breast cancer in nulliparous women. Contraception. 1988; 38: 287-299. [Back]

37 Thomas DB. Oral contraceptives and breast cancer. J Natl Cancer Inst. 1993; 85: 359-364. [Back]

38 White E, Malone K, Weiss N, Daling J. Breast cancer among young U.S. women in relation to oral contraceptive use. J Natl Cancer Inst. 1994; 86: 505-514. [Back]

39 Harris JR, Lippman ME, et al. Breast cancer. N Engl J Med. 1992; 327: 319-328. [Back]

40 American Academy of Pediatrics. The adolescent's right to confidential care when considering abortion. Pediatrics. 1996; 97: 746-751. [Back]

41 Howe H, et al. Early abortion and breast cancer risk among women under age 40. Int J Epidemiol. 1989; 18: 300-304. [Back]

42 Gabutti G, et al. AIDS related neoplasms in Genoa, Italy. Eur J Epidemiol. 1995; 11: 609-614. [Back]

43 McPherson K. The pill and breast cancer: why the uncertainty? Br Med J. Sept. 20, 1986; 293: 709-710. [Back]

44 Tietze C. Induced Abortion: A World Review. 1983. 5th ed. Population Council. [Back]

45 Henshaw SK. Induced abortion: a worldwide perspective. Family Planning Perspectives. 1986; 18: 250-254. [Back]

46 Anonymous. Oral contraceptives in the 1980s. Population Reports. May-June 1982; X: A189-A222. [Back]

47 Olsson H, Moller TR, Ranstam J. Early contraceptive use and breast cancer among premenopausal women: Final report from a study in southern Sweien. J Natl Cancer Inst. 1989; 81: 1000-1004. [Back]

48 Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormonal contraceptives: collaborative reanalysis of individual data on 53,297 women with breast cancer and 100,239 women without breast cancer from 54 Epidemiological studies. The Lancet. 1996; 347: 1713-1727. [Back]

49 Romieu I, Berlin J, et al. Oral contraceptives and breast cancer. Review and meta-Analysis. Cancer. 1990; 66: 2253-2263. [Back]

50 Olsson H, et al. Her-2/neu and INT2 prot-oncogene amplification in malignant breast tumors in relation to reproductive factors and exposure to exogenous hormones. J Natl Cancer Inst. 1991; 83: 1483-1487. [Back]

51 Olsson H, Ranstam J, et al. Proliferation and DNA ploidy in malignant breast tumors in relation to early contraceptive use and early abortions. Cancer. 1991; 67: 1285-1290. [Back]

52 Ory HW, et al. Long-term oral contraceptive use and the risk of breast cancer (CASH Study). JAMA. 1983; 249: 1591-1595. [Back]

53 Newcomb PA, Storer BE, et al. Pregnancy termination in relation to risk of breast cancer. JAMA. 1996; 275: 283-322. [Back]

54 Pike MC, Henderson BE, et al. Oral contraceptive use and early abortion as risk factors for breast cancer in young women. Br J Cancer. 1981; 43: 72-76. [Back]

1, 2,