From the 1Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle, WA, and 2Division of Public Health Sciences, The Fred Hutchinson Cancer Research Center, Seattle, WA.
Address correspondence to: Noel S. Weiss, Department of Epidemiology, School of Public Health and Community Medicine, Box 357236, University of Washington, Seattle, WA 98195-7236.
The study of the relation of exogenous hormones to cancer incidence has turned up several associations that are regarded as causal. One of these is the low risk of ovarian cancer among women who have used oral contraceptives (OCs), an association widely interpreted as reflecting genuine protection. To what extent should the findings of Ness et al., 1 reported in this issue and which seem to indicate that use of OCs has no more favorable an influence on ovarian cancer risk than the use of other, non-hormonal forms of contraception, modify this interpretation?
Our feeling is that not much modification is called for. That feeling is based in part on some methodologic issues present in the study by Ness et al., and in part on the manner in which these authors chose to analyze their data. We are concerned, first, that the inability to interview a high percentage of potential subjects could on its own account for the relatively modest reductions in risk (generally odds ratios of around 0.8) seen in association with “ever” use of most non-hormonal forms of contraception. Ness et al. interviewed approximately 60% of potentially eligible cases and approximately 70% of controls. It is not that these levels of participation are so different than those the rest of us are able to obtain when we conduct population-based case-control studies of cancer that seek exposure information by means of an interview. Rather, it is that all of us have to be cognizant that this large degree of non-participation, combined with the fact that the reasons for non-participation between cases and controls are likely to differ, have the potential to produce an odds ratio at least as low as 0.8 (or as high as 1.2) in the absence of any true association. A possible indication of the presence of selection bias of this sort (or of information bias, or of both) can be seen in the far higher proportion of cases than controls (about 6%vs less than 1%) who reported never having had sexual intercourse with a man. This strong association would appear to be well beyond the magnitude attributable to nulliparity or its correlates, suggesting that at least some of the association may be artifact.
The authors’ focus on “ever” use, especially in the crucial data in Table 2 in which the types of contraception are compared directly with one another, is also of concern. Since there is evidence 2,3 that it is particularly a long duration of OC use that is associated with a low risk of ovarian cancer, an analysis that mixes short-term and long-term contraceptive users will not obtain a result that accurately characterizes the possible influence of either of these. Therefore, in our reading of the article by Ness et al., we give relatively greater weight to the data presented in Table 4 pertaining to specific durations of contraceptive use. In those data, only for OCs is there consistent evidence of a reduced risk of ovarian cancer associated with long-term use. Similarly, when examining the potential altered risk associated with other forms of contraception, controlling for use of OCs by simply creating the dichotomy of “ever”vs “never” may not be adequate to remove all of the confounding influence of this variable.
Does use of OCs protect against the development of ovarian cancer? The answer is almost certainly “yes”: (1) Epidemiologic studies are virtually unanimous (for once!) in finding a reduced risk associated with OC use, one that is greatest when duration of use has exceeded several years; and (2) There are some plausible means (suppression of ovulation or suppression of gonadotropin production, among others 4) by which this protection could be accomplished. Does use of a non-hormonal form of contraception (other than tubal ligation 5) also protect against ovarian cancer? Do the results of Ness et al. “imply mechanisms other than hormonal or ovulatory by which ovarian cancer is reduced”? As we’ve tried to indicate, unless considerably stronger evidence were to become available than is available now, the answer to these latter two questions is “no.”
1. Ness RB, Grisso JA, Verzona R, Klapper J, Morgan M, Wheeler JE, for the Study of Health and Reproduction (SHARE) Study Group. Oral contraceptives, other methods of contraception and risk reduction for ovarian cancer. Epidemiology 2001; 12: 307–312.
2. Whittemore AS, Harris R, Itnyre JL. Characteristics relating to ovarian cancer risk: Collaborative analysis of 12 US case-control studies. II. Invasive epithelial ovarian cancers in white women. Collaborative Ovarian Cancer Group. Am J Epidemiol 1992; 136: 1184–1203.
3. Hankinson SE, Colditz GA, Hunter DJ, Spencer TL, Rosner B, Stampfer MJ. A quantitative assessment of oral contraceptive use and risk of ovarian cancer. Obstet Gynecol 1992; 80: 708–14.
4. Risch HA. Hormonal etiology of epithelial ovarian cancer, with a hypothesis concerning the role of androgens and progesterone. J Natl Cancer Inst 1998; 90: 1774–1786.
5. Weiss NS, Cook LS, Farrow DC, Rosenblatt KA. Ovarian cancer. In: Schottenfeld D, Fraumeni JF, ed. Cancer Epidemiology and Prevention, 2nd ed. New York: Oxford University Press, 1996.