Fertility drugs and the risk of breast and gynecologic cancers.

The epidemiology of breast cancer has been extensively studied, with many investigations supporting the notion of an important etiologic role for endogenous as well as exogenous hormones. 75 Surprisingly, few studies have addressed the potential relationships to breast cancer risk of fertility drugs, despite their recognized effects on ovulation and hormone patterns 76 and clinical reports that have suggested an association. 41 , 77 – 81 Despite the biological plausibility of a relationship between the use of these hormonal agents and breast cancer risk, the literature remains quite confusing, with some studies showing a possible increased risk associated with the drugs, 30 , 49 others showing the opposite effect, 82 , 83 and still others showing no substantial relationship. 14 , 29 , 52 , 53 , 84 , 85 Many of these results derive from the same cohort studies that addressed relationships with ovarian cancer, and they are plagued by the same methodological problems previously mentioned. A few, however, merit specific discussion. One of the earliest studies addressing breast cancer relationships was that of Rossing and others, 82 the cohort study that raised substantial concern regarding effects on ovarian cancer risk. For breast cancer ( Table 3 ), the opposite relationship was observed, namely a nonsignificantly reduced risk of invasive and in situ breast cancer associated with clomiphene use (adjusted RR: 0.5; 95% CI, 0.2 to 1.2). This estimate, however, was based on only 12 exposed cases, and there was no indication of any further risk reduction with extended duration of use. A chemopreventive effect of clomiphene would be of interest given that it is a selective estrogen receptor modulator (SERM) and thus could have properties similar to tamoxifen, another SERM. 86 Additional epidemiological support of a reduced risk of breast cancer associated with clomiphene use was provided by results from the Nurses Health Study II, 83 which showed a RR of 0.40 (95% CI, 0.2 to 0.7) associated with the use of clomiphene among women treated for ovulatory infertility. Risk decreased significantly with duration of use of clomiphene, with users of ≥10 months having a RR of 0.21 compared with nonexposed women. The findings were based on self-reports of both drug use as well as causes of infertility. In contrast to these results, a more recent study from Israel, which included 131 breast cancers, found a significantly increased SIR of 1.4 associated with exposure to clomiphene. 30 Given concerns that this might merely reflect confounding by such factors as parity, the authors conducted a nested case-control study within their cohort to obtain additional information on covariates. The excess risk associated with clomiphene persisted and even became stronger after adjustment for a variety of breast cancer risk factors. Women who received more than four cycles of clomiphene were also noted to be at increased breast cancer risk (SIR: 1.90; 95% CI, 1.08 to 3.35) in a Swedish cohort, 57 but this was based on only 12 exposed cases and unadjusted for other breast cancer risk factors. An additional Israeli study, focused on women giving birth after IVF treatment, also noted an increased risk associated with self-reported exposure to ovulation induction drugs (RR: 1.65; 95% CI, 1.15 to 2.36). 49 Additional information on types of drugs used was not available in this investigation. In contrast to these studies showing either reduced or increased risks associated with fertility drug use, studies conducted in both the United States (292 breast cancer cases) 84 and Denmark (331 cases) 31 have noted no substantial relationships. However, in the U.S. study, there was some evidence of an increase in risk of invasive breast cancer with follow-up time, with a significant excess risk seen for clomiphene users with 20 or more years of follow-up (RR: 1.6; 95% CI, 1.0 to 2.5). In the Danish study, 31 no increased risk was seen for clomiphene or gonadotropins, but subjects who were prescribed progestins were found to have more than a twofold increased risk. Another large study, embedded in the French E3N Cohort Study in France, 85 which focused on 92,555 women, 6602 of whom reported infertility problems, found no overall association between self-reports of exposure to fertility drugs and breast cancer risk (with 183 breast cancers occurring among the infertile women), but it did note increased risks associated with treatment among those with a family historyof breast cancer. However, these risks were based on relatively small numbers and require a cautious interpretation. Although most of the investigations of breast cancer have focused on older fertility treatment protocols, a few have assessed IVF effects, including investigations in Australia, 18 the Netherlands, 29 Sweden, 55 , 87 and Israel. 88 These studies for the most part have failed to find an overall difference in risk between exposed and unexposed subjects. However, one of the studies in Israel 88 found higher risks among women who received IVF at age 40 or beyond or those who had four or more cycles of treatment. In the Australian study, 18 a twofold increased risk of breast cancer was observed within 1 year of last treatment. This prompted the suggestion that ovulation-inducing drugs might promote the rapid growth of preexisting tumors, similar to the short-term transient increase in breast cancer following a recent pregnancy. 89 However, several other studies that have assessed the detailed timing effects of last drug use found no support for a promotional effect by either clomiphene or gonadotropins. 29 , 84 The few case-control studies that have addressed the relationship of fertility drug use to breast cancer risk have for the most part not found any substantial associations. 90 – 92 Most of these studies, however, were limited by small numbers of cancers, imprecise information on patterns of or indications for drug use, or incomplete ability to control for other correlates of risk, including well-recognized reproductive risk factors. One large study, involving ≥4500 breast cancer cases, was able to carefully control for potential confounding variables but had to rely on self reports of infertility and had few women exposed to fertility drugs. 93 Although this study found no association of risk related to the use of clomiphene, there was some indication of a risk elevation among women with long-term use of gonadotropins (risks ranging from 2.7 to 3.8 for use of at least ≥6 months or six cycles). The authors speculated that this might be due to increases in serum estrogen and progesterone levels, although whether hormone increases that would result from such exposure could substantially affect subsequent breast cancer risk has been questioned. 94 Only a few studies have focused on whether breast cancers that occur among patients exposed to fertility drugs exhibit unusual clinical characteristics. One study 95 showed that breast cancers developing subsequent to fertility drug use are more advanced and have worse prognostic features (e.g., they are estrogen or progesterone receptor negative). The diagnosis of aggressive tumors predominated within 2 years of drug exposure and appeared similar to pregnancy-associated breast cancers evaluated in the same investigation. However, other investigations have not confirmed poorer prognostic features in patients exposed to fertility drugs. 83 , 96 An evaluation of breast cancer mortality in the Australian study also failed to show any appreciable differences between those who did and did not receive IVF. 97 Apart from the one French study that evaluated effects of fertility drugs among women with a family history of breast cancer, 85 scant attention has focused on familial or genetic modifications. One study evaluated the relationships of fertility drugs among carriers of BRCA1 or BRCA2 mutations but found no unusual effects. 63 An additional study assessed whether fertility drug use might increase mammographic densities, one of the strongest identified risk factors for breast cancer, but found no notable relationships. 98

Numerous clinical reports have expressed concern about a potential link between the use of fertility drugs and ovarian cancer risk. 34 – 42 The association has biological credibility, given that “incessant ovulation” and associated alterations in endogenous hormones during the reproductive years are plausible explanations for several factors that alter disease risk, including nulliparity and oral contraceptive use. 4 , 43 In the 1990s, an epidemiological study raised major concern regarding a potential link between the use of fertility medications and ovarian cancer risk. This study 26 consisted of a meta-analysis of 12 ovarian cancer case-control studies, only 3 of which provided information regarding drug use (focusing on 526 cases and 966 controls). There was scant information about the type of drug or the extent of its use. Self-reported prior use of fertility medications was associated with an odds ratio (OR) of 2.8 [95% confidence interval (CI), 1.3 to 6.1] as compared with women who had no history of infertility. This risk was limited to nulligravid women, who experienced a 27-fold increase in risk associated with drug use (95% CI, 2.3 to 315.6). Although the report caused great medical and lay concern, several editorials and literature reviews disputed the conclusion of markedly increased risk, 44 – 47 pointing out that the risk estimate was based on only 12 exposed cases and 1 exposed control. Moreover, drug use among gravid women was associated with a nonsignificant OR of 1.4 (95% CI, 0.52 to 3.6). Further concern was raised following publication of results from a retrospective cohort study of 3837 women evaluated for infertility in a Seattle practice between 1974 and 1985 who were followed for cancer incidence through a regional cancer registry. 32 Clomiphene use, as documented in medical records, was associated with an adjusted 2.3-fold increased risk (95% CI, 0.5 to 11.4), based on nine ovarian cancers. Five of the nine women with cancer had taken the drug for ≥12 monthly cycles, resulting in a RR of 11.1 (95% CI, 1.5 to 82.3). An enhanced risk associated with long-term treatment was observed in both those with and without ovulatory abnormalities. A large proportion of the observed tumors were borderline (5 of the 11 in the cohort). These two studies prompted several other investigations, including cohort studies in Australia, 18 , 48 Israel, 14 , 16 , 49 , 50 the United States, 28 , 51 the Netherlands, 29 the United Kingdom, 52 , 53 Denmark, 54 and Sweden. 55 – 57 Most of these studies failed to provide confirmation of a link between fertility drug use and ovarian cancers. However, as shown in Table 2 , several of the investigations were limited by extremely small numbers of events. Of studies that have assessed relationships with earlier treatment protocols (in the era before in vitro fertilization [IVF]), the most comprehensive ones were conducted in the United States, 28 Denmark, 54 and the United Kingdom. 52 In the U.S. study, 45 ovarian cancers developed among 12,193 infertile women followed for a median of 18.8 years. This study involved detailed collection of information on drug exposures, causes of infertility, and potential cancer risk factors from both medical records and questionnaires, allowing subjects who had undergone a bilateral oophorectomy to be eliminated from those at risk. 28 The results were largely reassuring, showing no risk increases associated with ever use of either clomiphene or gonadotropins. However, there were nonsignificant risk increases (range of RRs: 1.5 to 2.5) associated with use of either clomiphene or gonadotropins among the subjects followed for the longest periods of time (i.e., ≥15 years). In the Danish study, 54 among 54,362 women diagnosed from 1963 to 1998, there were 156 ovarian cancer cases identified through linkage with the Danish Cancer Registry. In internal analyses, there was no evidence of increased ovarian cancer risk associated with clomiphene, the most commonly used fertility drug during the period of subject accrual (RR: 1.14; 95% 0.79 to 1.64) or with a variety of other fertility drugs documented in medical records. Finally, in a study of 7355 women in the United Kingdom evaluated for infertility between 1963 and 1991 and in whom 21 ovarian cancers were identified, there was no evidence of a significant link between fertility drug use and ovarian cancer, although this study was somewhat more limited than others in not being able to adjust for most other risk factors. 52 Although these previously discussed studies focused on women exposed to ovulation-inducing agents at earlier times, several other studies have concentrated on exposures received during IVF. One study that derived data from 15 California clinics, published to date only in abstract form, 51 in which 50 ovarian cancers developed among 51,371 women, found no evidence for an association of fertility drugs with ovarian cancer risk, even when dose, formulation, and number of treatment cycles were considered. Among 29,666 women referred to 10 Australian IVF clinics, 13 ovarian cancers were observed during follow-up. 18 The investigators had detailed information on indications for drug use but only limited information on patient characteristics. The overall SIR was 0.99, with no higher risk for the women who received at least one IVF treatment cycle (0.88) as compared with those who received no drug treatment (1.16). Women with unexplained infertility were at a significantly increased risk compared with the general population, but within this subgroup there was no difference in risk between treated and untreated women. In a cohort of 25,152 women treated for subfertility in the Netherlands, 17 ovarian cancers developed. 29 Strengths of this study included detailed information on causes of sub-fertility and drug exposures from medical records, as well as on cancer risk predictors obtained through completed questionnaires from many of the study subjects. Results showed no difference in risk between treated and untreated subjects, even when the number of cycles or ampoules received were considered. The latest results regarding IVF and ovarian cancer risk derive from several studies in Sweden, 55 – 57 all of which focused on cancer developing among women who gave birth following IVF treatment. One of these observed only one ovarian cancer developing after IVF 57 ; and the other two studies had more power to evaluate effects. Although seemingly focused on the same populations of women, the studies derived discrepant results, possibly due to differences in analytical approaches. One investigation 55 found a significantly elevated risk of ovarian cancer following IVF treatments (OR: 2.09; 95% CI, 1.39 to 3.12); the other 56 found no overall effect of IVF but a significantly increased risk of invasive ovarian cancer associated with exposure to gonadotropins (RR: 5.28; 1.70 to 16.47). Although reassuring results regarding the effects of fertility drugs on ovarian cancer risk have emerged from several casecontrol studies, most have been limited by the small number of subjects reporting prior drug use. 23 , 58 – 62 For example, in the largest study, 61 based on 1031 cases and 2411 hospital controls, only 1.1 to 1.5% of the subjects reported fertility drug use, resulting in only 15 cases and 26 controls with relevant exposures. To derive larger numbers, Ness and others 23 undertook a meta-analysis of eight studies involving data on 1060 cases and 1337 controls. In this study, after adjustment for types of infertility, the risk associated with drug use was somewhat higher among nulligravid women (1.8) and among those who had ≥4 months of exposure (RRs: 1.5–1.7), but none of these risks was statistically significant. Although the results of the most recent studies are much more reassuring than early studies, several observations indicate a need for further monitoring. These include the findings in two studies 23 , 28 of modest risk increases with either extended follow-up or increased drug exposure. Given that these medications became available in the United States beginning in the early 1960s, women who were subsequently exposed to them are now entering the age range of highest ovarian cancer incidence. Thus additional follow-up data are needed to evaluate their effects. Several investigations 23 , 26 , 28 have noted higher risks associated with fertility drug use among nulligravid women, suggesting the possibility of an enhanced effect of the medications among women with certain indications for use. However, other investigations have not confirmed this relationship. 62 A possible additional subgroup of interest with respect to effects of fertility drugs are women at high genetic risk for ovarian cancer. Women with BRCA mutations exposed to fertility drugs have been evaluated for breast but not ovarian cancer risk. 63 It has also been questioned whether fertility medications might have a preferential effect on certain ovarian tumor types. Clear cell, 64 , 65 malignant germ cell, 66 and granulosa cell 42 tumors have been linked by case reports to the use of ovulation-inducing drugs. Gonadotropins have been shown to induce granulosa cell tumors in rodents 67 and stimulate cells in human in vitro models, 68 and clomiphene has been shown to increase granulosa, theca, and luteal cells in a rat model. 69 Epidemiological studies of these rare tumor types are not available, but several descriptive analyses in Finland 70 , 71 fail to provide support for an effect of fertility drug use on granulosa cell tumors. That fertility drugs might preferentially affect the risk of borderline ovarian tumors is suggested by both cohort 32 , 56 , 72 and case-control 23 , 73 investigations, with associated risk ratios in the range of 3 to 5. The relationship has been found to predominate among nulligravid women, 23 users of clomiphene, 56 and users of gonadotropins. 23 , 72 , 73 These findings, in conjunction with case reports of ovarian cancer developing in women during or shortly after treatment with ovulation-inducing agents, 34 – 38 , 40 , 42 , 74 have led to speculations that ovarian stimulation may induce growth in existing highly differentiated indolent tumors. Alternatively, the findings simply could reflect more intensive medical surveillance among infertile women.

The evaluation of cancer risk among patients treated for infertility is complex given the need to consider indications for use, details of the treatment protocols, and effects of other factors that could have independent effects on risk (including whether the fertility treatment is successful and produced a pregnancy, a major protective factor for breast and gynecologic cancers). Many of the available studies to address these issues have also had small numbers and/or short follow-up times. Although most of the available studies have focused on effects of older treatment protocols (most notably clomiphene exposures), more recent studies are beginning to focus on the long-term effects of drugs used in IVF. Because the women who have received such treatments in the past are still relatively young, it may be some time before we have sufficient follow-up to resolve relationships fully. Interpretation of effects will depend on the availability of detailed information on a wide array of fertility treatments (including clomiphene, gonadotropins, gonadotropin-releasing hormone agonists and antagonists, and progestational agents). These studies will clearly need to be large for the evaluation of rare tumors and should incorporate special efforts to evaluate some of the subgroups that have been suggested to possibly be more susceptible (e.g., nulligravidas) and focus on the possibility of heterogeneous effects within broader disease categories (including whether relationships vary by tumor histology and markers). Fortunately, there are several ongoing studies in the United States, Israel, and several Scandinavian countries, whose emerging results in the near future should bring additional clarity to the issue of how use of a broad spectrum of fertility drugs may have an impact on the subsequent occurrence of breast, gynecologic, and other cancers.

Endometrial cancers are well recognized as hormonally sensitive. 99 Surprisingly, there has not been as much focus on this cancer site as on ovarian and breast cancers, but several recent reports indicate that of the three cancer sites that endometrial cancer may be the one most strongly influenced by fertility drug use. However, as with the other cancers, results are not entirely consistent, with a number of studies showing no associations. 16 , 18 , 29 , 48 , 53 However, these studies involved relatively short follow-up times, as well as small numbers, as did the one previous case-control study that addressed the topic. 100 There is, however, some consistency in several of the larger cohort studies of increased risks of endometrial cancer related to either fertility drug use in general or more specifically to the use of clomiphene ( Table 4 ). The possibility of a relationship was first raised in one of the earlier Israeli cohorts in which a significant twofold increased risk was noted for fertility drug use. 11 An increased risk of self-reported exposure to ovulation induction drugs was also noted in the Jerusalem Perinatal Study, although interpretation of the effect was hindered by the fact that the study was restricted to parous women and that the increased risk (3.32; 95% CI: 1.31 to 8.42) was based on only five exposed women. 49 Several studies that have addressed relationships with clomiphene use have noted some risk elevation. In a U.S. cohort, ever use of clomiphene was associated with a RR of 1.8 (95% CI, 0.9 to 3.3), with significant risks seen for subjects with higher dosages and longer follow-up periods. 101 In a study in the United Kingdom, a slight increase was observed for any use of ovarian stimulants, with the highest risk seen in those with a total cumulative dose of clomiphene of ≥2250 mg (RR: 2.62; 95 CI, 0.94 to 6.82). 52 Although not all studies have found an increased risk of uterine cancers associated with clomiphene, 102 it is of interest that the SERM clomiphene is structurally similar to tamoxifen, 6 a drug that has been repeatedly linked with increases in endometrial cancer risk. 103 The biological plausibility of a relationship is also supported by a clinical report of three cases of adenomatous hyperplasia of the endometrium, a precursor condition, occurring among women exposed to ovulation-inducing agents. 104 Less information is available on effects of drugs other than clomiphene. However, a recent cohort study in Denmark noted a significantly increased risk of uterine cancer associated with gonadotropin exposure. 102 Venn and others 18 reported an increased incidence of uterine sarcomas in their IVF-exposed population, based on four observed cases, resulting in an SIR of 8.56 (95% CI, 3.21 to 22.8). This finding requires further follow-up, although this will be difficult given the rarity of the tumor. Along these lines, it would be of particular interest to evaluate whether there are differences between type 1 (endometrioid) and type 2 (nonendometrioid) uterine tumors, given the evidence that the former are especially hormonally responsive. 105

Studies to clarify the relationships are difficult to undertake, with the interpretation of results of many of the existing investigations hindered by a variety of methodological limitations. Although some clinical studies have suggested a link between fertility drugs and the risk of various cancers, the absence of comparison groups in such studies precludes definitive conclusions. Case-control studies have been undertaken, but such studies rely on patient reports of exposure to infertility treatment, which can be imprecise or reflect the influence of selective recall. Cohort studies that provide more definitive results have relied on both retrospective and prospective approaches to evaluate relationships of medication use to subsequent cancers. These studies also are usually hindered by imprecise information on indications for drug use (i.e., causes of infertility) 7 that can independently affect cancer risk. For example, anovulation has been linked to increased risks of endometrial 8 – 11 and possibly breast 8 , 12 – 18 cancers, endometriosis has been linked to ovarian 19 – 23 and breast 20 , 22 cancers, and tubal factors to ovarian cancers. 23 – 26 Further, case-control studies usually focus on an unselected sample of cases diagnosed in the general population (rather than on, for example, a group of infertile women), resulting in low rates of use of fertility medications. Thus, even though case-control studies may start with a large number of subjects, their ability to evaluate specific associations may be limited. For example, in a population-based case-control study of breast cancer, 27 which included 4,566 cases and 4,676 controls, <5% of the study subjects reported prior fertility drug use (184 cases and 200 controls). An additional complexity of case-control studies is that some have used hospitalized women as the control group. Because hospitalized women are likely to have better access to and more utilization of medical services compared with controls selected from the general population, studies using hospital controls may derive different risk estimates than those using controls selected from the general population. As a result of the limitations of case-control studies, more credibility regarding the relationship of fertility treatments to cancer risk has derived from prospective or cohort studies, reviewed in Table 1 , that define exposures prior to the onset of disease. Most cohort studies, however, are limited by small numbers of cancers and lack of information on other cancer predictors. Many cohort studies have had short follow-up periods; thus effects that require long latency intervals may remain undetectable. Participants in these studies are often still young and have not yet reached the age of peak cancer incidence. The unavailability of appropriate comparison groups is also problematic for cohort studies. In many of these studies, the disease experience of cohorts of infertile women is compared with the experience of the general population through the calculation of standardized incidence ratios (SIRs). SIRs compare the number of observed cancers in the cohort to the number expected based on incidence rates in the general population. These comparisons take into account age, race, and calendar time, but usually there is no information regarding other cancer predictors. Of notable concern is the inability to adjust for parity, a recognized risk factor for breast, endometrial, and ovarian cancers. 1 Thus comparisons of cancer rates among infertile women (with or without ovulation induction) to cancer rates in the general population can be difficult to interpret. Cohort studies are most informative if they allow internal comparisons that enable adjustment for a variety of potential cancer risk factors. Through the calculation of relative risks (RRs) rather than SIRs, disease risks can be compared between treated and untreated women while holding constant (or controlling for) other cancer predictors. Few cohort studies, however, have had access to data on all pertinent risk factors. In most cases, when covariate data are available, they are limited to data abstracted from medical records, with only a few studies having attempted direct administration of questionnaires to study subjects. 28 , 29 Because of the large numbers of women usually required for evaluation of longterm effects of fertility treatments, several studies have collected extensive risk factor information only from cancer cases and a sample of nondiseased subjects, either through nested case-control studies within the cohort 30 or from case-cohort investigations. 31 , 32 Although cohort studies are generally preferred for the evaluation of effects of fertility drugs, they are not without their inherent limitations. Losses to follow-up are common in prospective studies. Although poor rates of response to questionnaires are a particular concern for case-control studies, they are also problematic for cohort studies if information on potential confounding variables is attempted through questionnaires. Surveillance bias can also come into play in both case-control and cohort studies if women who receive fertility drugs are under close medical scrutiny. 33 When interpreting reported disease associations, we must give particular consideration to the strengths and limitations of the individual studies, as discussed next.