Oral, injected and implanted contraceptives and breast cancer risk among U.S. Hispanic and non-Hispanic white women
Version of Record online: 26 JUL 2007
Copyright © 2007 Wiley-Liss, Inc.
International Journal of Cancer
Volume 121, Issue 11, pages 2517–2523, 1 December 2007
How to Cite
Sweeney, C., Giuliano, A. R., Baumgartner, K. B., Byers, T., Herrick, J. S., Edwards, S. L. and Slattery, M. L. (2007), Oral, injected and implanted contraceptives and breast cancer risk among U.S. Hispanic and non-Hispanic white women. Int. J. Cancer, 121: 2517–2523. doi: 10.1002/ijc.22970
- Issue online: 25 SEP 2007
- Version of Record online: 26 JUL 2007
- Manuscript Accepted: 29 MAY 2007
- Manuscript Received: 9 APR 2007
- U.S. National Institutes of Health. Grant Numbers: CA078682, CA078762, CA078552, CA078802
- National Cancer Institute. Grant Number: N01-PC-67000
- State of Utah Department of Health
- breast neoplasms;
- Hispanic Americans;
- contraceptive agents
Associations between oral contraceptive (OC) use and breast cancer have been reported, but few studies have considered associations in racial and ethnic minorities. Data regarding injected or implanted hormonal contraceptives are limited. In a case–control study of Hispanic (796 cases, 919 controls) and non-Hispanic white (1,522 cases, 1,596 controls) women in the U.S. southwest interviewed in 2000–2005, 49% of Hispanic controls and 66% of non-Hispanic white controls reported having used OC. Breast cancer odds ratios (OR) associated with OC use within the past 5 years were 1.22 (95% confidence interval (CI) 0.80, 1.84) among Hispanics, 1.28 (95% CI 0.93, 1.76) among non-Hispanic whites, 1.27 (95% CI 0.99, 1.63) for both ethnic groups combined and 1.53 (95% CI 0.98, 2.40) for estrogen receptor (ER) negative tumors. OC use for 20 years or longer was associated with ORs of 1.50 (95% CI 1.04, 2.17) for both ethnic groups combined, and 2.23 (95% CI 1.17, 4.25) for ER negative tumors. Hormonal contraceptive injections were used by 3.3% of Hispanic controls and 2.8% of non-Hispanic white controls, OR 1.23 (95% CI 0.88, 1.73). Fifteen cases and 2 controls reported use of a subdermal contraceptive implant, OR 8.59 (95% CI 1.92, 38.39). Associations between OC use and breast cancer in Hispanics are consistent with modestly increased risk among recent users and for ER negative tumors, as observed in other populations. Based on a small number of users of contraceptive implants, a significantly increased breast cancer risk was observed; continued surveillance of implant users may be warranted. © 2007 Wiley-Liss, Inc.
Hispanic women in the southwest U.S., an ethnic group with genetic heritage from American Indian and European populations,1 have a lower incidence of breast cancer than non-Hispanic whites,2 and a higher proportion of breast cancers that express the estrogen receptor (ER).3, 4 Studies in non-Hispanic white populations have reported a modest increase in breast cancer risk associated with recent use of oral contraceptives (OC),5, 6, 7 an association that may differentially affect ER negative tumors.8 Limited data are available to evaluate associations of hormonal contraceptives with breast cancer risk in racial or ethnic groups with different incidence of breast cancer and a different distribution of ER tumor status. One study stated that associations were similar for U.S. blacks and whites.9
OC formulations have changed over time, and it remains relevant to continue to examine the influence of contemporary compounds, which provide a lower dose of hormones, on breast cancer risk.5 Few studies have investigated the potential influence of contraceptive injections and implants, sources of exposure to progestin compounds, on breast cancer risk. We have investigated the associations of hormonal contraceptives with breast cancer in the 4 Corners Breast Cancer Study, a case–control study of Hispanic and non-Hispanic white women residing in the U.S. southwest.
Subjects and methods
A case–control study of breast cancer was conducted among women in the U.S. states of Arizona, Colorado, New Mexico, and Utah. Methods for selection of subjects and collection of interview data have been described previously.10, 11 Cases diagnosed in October, 1999 to May, 2004 were identified through state-wide cancer registries in each state. All incident breast cancer cases identified as Hispanic (based on cancer registry information or computerized surname search, using the Generally Useful Ethnic Search System algorithm12 and the Census Spanish Surname List13), and an age-matched sample of non-Hispanic white cases, were selected for the study. Control subjects aged 64 years and younger were randomly selected from computerized drivers' license lists in New Mexico and Utah, and from commercially available lists in Arizona and Colorado; subjects aged 65 and older were selected from Center for Medicare Studies lists in all centers. Hispanic controls were initially identified using surname searches. Control groups were frequency-matched to cases on age and ethnicity. Each subject was asked to identify her race or ethnic background; only women who reported non-Hispanic white, Hispanic or American Indian were eligible for the study. Final classification of ethnicity was based on self-report.
An in-person interview was conducted with each subject and the subject's responses were entered directly into laptop computers, using computer assisted personal interview software. Interviews were audio recorded for quality assurance.14 The questionnaire is available at https://zorro.hrc.utah.edu/breast.html. Questionnaire topics included diet, medical history, physical activity, menstrual history and use of hormones, pregnancy history, family history of cancer and tobacco and alcohol exposures. Questions referred to exposures before or during a reference year, 1 year before diagnosis for cases and 1 year before selection for the study for controls. Subjects were asked whether they had ever used hormonal contraceptives, including birth control pills, shots or implants, and if the response was yes, were asked to select the type(s) from among these 3 categories and to report when intervals of use started and stopped. Women were not asked to recall the specific brands, dosages, or formulations of OC used. The interview did not collect information on use of intrauterine devices (IUD), because IUD's available in the U.S. at the time that the interviews began for our study in 2000 did not represent a hormonal exposure. (A hormone-releasing IUD was first approved for use in the U.S. by the Food and Drug Administration (FDA) in December, 2000.15) A few study subjects reported use of a contraceptive “implant,” but when probed for further information, indicated that the product they referred to was an IUD. Because of the potential for confusion about the term implant, for subjects who reported use of a contraceptive implant we confirmed that the product was a subdermal implant by probing during the interview, by later review of the tape-recorded interviews, and in some cases by making call-backs after the interview. Subjects were reclassified as nonusers of implants if the product described on the tape or call-back was not a subdermal implant, or if the product was not described in detail but the date of reported use was before 1990, the date of approval of the product by the FDA.15
Among eligible subjects contacted for the study, cooperation rates for the interview were 63% for Hispanic cases, 36% for Hispanic controls, 71% for non-Hispanic white cases and 47% for non-Hispanic white controls; details of recruitment success and characteristics influencing participation have been reported.10, 11, 16 The study procedures were reviewed and approved by institutional review boards for human subjects' research at each institution, and subjects provided written informed consent.
Data describing tumor histology and ER expression were obtained from information recorded and coded by the cancer registries.17
We compared frequency distributions of characteristics of hormonal contraceptive users and nonusers in the control population, using χ2 tests with Cochran-Mantel-Haenszel adjustment for age. We evaluated associations between use of OC, contraceptive injections or subdermal implanted contraceptives and breast cancer by calculating odds ratios (OR) and 95% confidence intervals (CI) from logistic regression models. Associations according to timing and duration of exposure were examined, including categories of years since most recent use, duration of use and year of first use. Our primary analysis considered Hispanic and non-Hispanic white women separately, as well as the groups combined with adjustment for ethnicity. There were too few American Indian participants (n = 127) for separate analysis, so American Indian women were grouped with Hispanics (with whom they share genetic heritage18) for analysis. All analyses were adjusted for age and study center, which were frequency-matching variables, and for education to account for the any possible influence of education on study participation.16 ORs for oral and injected contraceptive use were calculated from a model adjusting for the other exposure as a covariate; contraceptive implant use could not be adjusted for as a covariate because of thesmall number of users, and therefore users of contraceptive implants were excluded from models when estimating ORs for the other 2 types. We evaluated other potential confounders to detect a confounding effect, i.e., a change of 10% or more in the coefficient for the exposure of interest; variables that met this criterion for any exposure of interest were included in all final models. Fisher's exact p-values were calculated for case–control comparisons, in which the number of exposed cases or controls was very small or zero. We estimated ORs for associations between hormonal contraceptive use and subgroups of breast cancers according to ER status of tumors, comparing each group to controls in separate logistic regression models. SAS 9.1 (SAS Institute, Cary, NC) was used for statistical analysis.
Two-thirds of non-Hispanic white controls (1,064 of 1,596) reported that they had used hormonal contraceptives, whereas only 51% (466 of 919) Hispanic controls has used these preparations (Table I; p < 0.0001 for difference in prevalence of use by ethnicity, age-adjusted). Users of hormonal contraceptives had lower parity than never-users in both ethnic groups. Hormonal contraceptive users among Hispanics had more education than never-users. Among non-Hispanic whites, hormonal contraceptive users had older ages at first birth.
|Total (n = 919)||Hormonal contraceptive use||p1||Total (n = 1,596)||Hormonal contraceptive use||p1|
|No (n = 453)||Yes (n = 466)||No (n = 532)||Yes (n = 1,064)|
|Less than high school graduate||267||29.1||41.9||16.7||<0.0001||73||4.6||6.6||3.6||0.14|
|High school graduate||241||26.3||27.1||25.5||343||21.5||26.7||18.9|
|Bachelor's degree, or higher||154||16.8||11.5||21.9||583||36.6||30.1||39.8|
|Family history of breast cancer|
|Age at menarche|
|Age at first birth|
|Number of births2|
|5 or more||177||21.3||29.8||12.9||172||12.5||18.9||9.4|
|Body mass index (kg/m2)|
Ever-use of OC was not associated with an overall increased in risk of breast cancer (Table II). A modest elevation of risk, ORs of 1.22 (95% CI 0.80, 1.84) for Hispanics and 1.28 (95% CI 0.93, 1.76) for non-Hispanic whites, was observed for recent (within 5 years) use of OC. The elevated risk for recent-users compared to never-users approached statistical significance, OR 1.27 (95% CI 0.99, 1.63; p = 0.06) when both ethnic groups were combined. OC use for 20 years or longer was also associated with elevated risk of breast cancer, with ORs 1.43 (95% CI 0.69, 2.95) among Hispanics, 1.49 (95% CI 0.96, 2.30) for non-Hispanic whites and 1.50 (95% CI 1.04, 2.17) for both ethnic groups combined. We further examined associations among recent users according to year of first use of OC (data not shown in table). A large proportion of women who had used OC within 5 years of the reference year, 81 controls and 107 cases, reported that their first use was before 1980 and the OR for breast cancer for this group compared to never-users was 1.43 (95% CI 1.02, 2.02). Recent users with first use in the 1980's had an OR of 1.14 (0.75, 1.72), and recent users with first use in 1990 or later had an OR of 1.26 (0.76, 2.09).
|All subjects||Hispanics||non-Hispanic whites|
|Controls n||Cases n||OR1 (95% CI)||Controls n||Cases n||OR1 (95% CI)||Controls n||Cases n||OR1 (95% CI)|
|No use of oral contraceptives||1,011||809||1.00 Referent||464||365||1.00 Referent||547||444||1.00 Referent|
|Oral contraceptives, any use||1,502||1,494||1.08 (0.94, 1.24)||453||423||1.10 (0.88, 1.37)||1,049||1,071||1.08 (0.90, 1.29)|
|Oral contraceptive use, by recency and duration|
|When last used|
|<5 years ago||200||223||1.27 (0.99, 1.63)||71||75||1.22 (0.80, 1.84)||129||148||1.28 (0.93, 1.76)|
|5–9 years ago||114||99||0.99 (0.72, 1.35)||38||39||1.10 (0.66, 1.85)||76||60||0.93 (0.62, 1.39)|
|10–19 years ago||311||336||1.14 (0.93, 1.40)||92||111||1.35 (0.96, 1.91)||219||225||1.08 (0.83, 1.41)|
|≥20 years ago||873||831||1.04 (0.89, 1.21)||249||196||1.00 (0.77, 1.30)||624||635||1.05 (0.87, 1.27)|
|Duration of use|
|<5 years||688||685||1.13 (0.97, 1.33)||203||186||1.14 (0.87, 1.49)||485||499||1.14 (0.93, 1.40)|
|5–9 years||405||392||1.02 (0.84, 1.22)||120||115||1.06 (0.77, 1.46)||285||277||0.99 (0.78, 1.25)|
|10–19 years||351||330||0.97 (0.80, 1.18)||113||102||1.03 (0.74, 1.43)||238||228||0.96 (0.75, 1.23)|
|≥20 years||54||84||1.50 (1.04, 2.17)||15||19||1.43 (0.69, 2.95)||39||65||1.49 (0.96, 2.30)|
|Decade of first use|
|1960's||643||641||1.13 (0.96, 1.34)||160||131||1.17 (0.86, 1.58)||483||510||1.11 (0.91, 1.37)|
|1970's||569||609||1.07 (0.90, 1.27)||186||197||1.15 (0.87, 1.53)||383||412||1.04 (0.82, 1.31)|
|1980's||218||185||0.93 (0.71, 1.22)||79||73||0.92 (0.60, 1.40)||139||112||0.98 (0.68, 1.40)|
|1990 or later||69||59||1.03 (0.68, 1.57)||27||22||0.80 (0.41, 1.57)||42||37||1.20 (0.69, 2.09)|
|Timing of first use2|
|Before first birth||513||464||1.12 (0.95, 1.33)||193||150||0.99 (0.75, 1.31)||320||314||1.20 (0.96, 1.50)|
|Not before first birth||796||820||1.03 (0.86, 1.23)||213||230||1.17 (0.88, 1.56)||583||590||0.97 (0.77, 1.23)|
We evaluated associations between OC exposures and subgroups of breast cancers by ER status and tumor histology. Cases with missing ER status (32.5% of Hispanic cases and 32.5% of non-Hispanic white cases) were similar to those with known ER status for associations between contraceptive use and breast cancer risk (data not shown). The association between OC use and breast cancer was limited to ER negative tumors (Table III), and this pattern was very similar for Hispanics and NHW (data not shown). OR for ductal and lobular tumors in association with recent use of OC were almost identical.
|Controls (n)||Estrogen receptor+||Estrogen receptor−||Ductal||Lobular|
|n||OR1 (95% CI)||n||OR1 (95% CI)||n||OR1 (95% CI)||n||OR1 (95% CI)|
|No use of oral contraceptives||1,011||442||1.00 Referent||98||1.00 Referent||575||1.00 Referent||53||1.00 Referent|
|Oral contraceptives, any use||1,502||772||1.02 (0.87, 1.21)||241||1.38 (1.04, 1.84)||1,044||1.05 (0.90, 1.22)||121||1.20 (0.82, 1.75)|
|Oral contraceptive use, by recency and duration|
|<5 years||688||356||1.08 (0.89, 1.31)||96||1.28 (0.92, 1.78)||487||1.13 (0.95, 1.34)||47||1.09 (0.70, 1.71)|
|5–9 years||405||199||0.94 (0.75, 1.18)||68||1.33 (0.92, 1.91)||263||0.94 (0.77, 1.16)||39||1.38 (0.86, 2.23)|
|10–19 years||351||171||0.93 (0.73, 1.18)||61||1.48 (1.01, 2.16)||235||0.95 (0.76, 1.18)||26||1.05 (0.62, 1.79)|
|≥20 years||54||44||1.39 (0.90, 2.14)||15||2.23 (1.17, 4.25)||56||1.36 (0.91, 2.04)||9||2.08 (0.93, 4.62)|
|<5 years ago||200||111||1.25 (0.93, 1.70)||47||1.53 (0.98, 2.40)||164||1.23 (0.94, 1.62)||13||1.21 (0.59, 2.45)|
|5–9 years ago||114||43||0.83 (0.55, 1.24)||26||1.64 (0.97, 2.77)||73||0.98 (0.69, 1.38)||10||1.67 (0.77, 3.65)|
|10–19 years ago||311||167||1.07 (0.83, 1.37)||57||1.47 (0.99, 2.18)||236||1.11 (0.89, 1.39)||30||1.38 (0.81, 2.33)|
|≥20 years ago||873||447||0.98 (0.82, 1.18)||110||1.28 (0.93, 1.76)||566||1.00 (0.85, 1.18)||68||1.12 (0.74, 1.69)|
Overall 79 cases and 73 controls reported use of contraceptive shots, OR 1.23 (95% CI 0.88, 1.73). There were no apparent trends in breast cancer risk with recency or duration of use of contraceptive shots (Table IV). The highest risk estimate was for women who first used contraceptive shots after age 35. ORs for ever use of contraceptive injections and subgroups of ER positive and ER negative tumors were 1.32 (95% CI 0.88, 1.99) and 0.63 (95% CI 0.29, 1.35), respectively (data not shown in table).
|All subjects||Hispanics||non-Hispanic whites|
|Controls (n)||Cases (n)||OR1 (95% CI)||Controls (n)||Cases (n)||OR1 (95% CI)||Controls (n)||Cases (n)||OR1 (95% CI)|
|Never used||2,440||2,224||1.00 (Referent)||888||756||1.00 (Referent)||1,552||1,468||1.00 (Referent)|
|Ever used||73||79||1.23 (0.88, 1.73)||29||32||1.08 (0.62, 1.86)||44||47||1.22 (0.79, 1.89)|
|When last used|
|<5 years ago||29||29||1.14 (0.66, 1.95)||14||9||0.63 (0.26, 1.52)||15||20||1.53 (0.75, 3.09)|
|5–9 years ago||10||18||1.96 (0.89, 4.33)||6||8||1.10 (0.37, 3.34)||4||10||3.06 (0.91, 10.25)|
|10–19 years ago||15||12||0.86 (0.39, 1.88)||7||8||0.97 (0.33, 2.84)||8||4||0.58 (0.17, 1.97)|
|≥20 years ago||18||18||1.23 (0.62, 2.41)||2||6||3.91 (0.76, 20.13)||16||12||0.85 (0.39, 1.85)|
|Duration of use|
|<5 years||60||65||1.22 (0.84, 1.76)||25||26||1.03 (0.57, 1.86)||35||39||1.23 (0.76, 1.99)|
|5–9 years||11||10||1.17 (0.48, 2.83)||4||4||0.77 (0.18, 3.29)||7||6||1.24 (0.40, 3.85)|
|≥10 years||1||2||1.65 (0.14, 19.68)||0||1||1||1||0.81 (0.05, 14.27)|
|Age at first use|
|<25||22||17||0.98 (0.51, 1.90)||9||7||0.69 (0.24, 1.99)||13||10||0.98 (0.42, 2.31)|
|25–29||13||13||1.26 (0.57, 2.80)||6||6||1.02 (0.31, 3.30)||7||7||1.49 (0.50, 4.42)|
|30–34||18||13||0.71 (0.34, 1.49)||8||7||0.72 (0.25, 2.08)||10||6||0.62 (0.22, 1.79)|
|35+||19||35||1.98 (1.12, 3.52)||6||12||2.33 (0.83, 6.54)||13||23||1.75 (0.87, 3.51)|
|Timing of first use2|
|Before first birth||50||56||1.29 (0.86, 1.92)||22||23||1.10 (0.59, 2.06)||28||33||1.33 (0.78, 2.26)|
|Not before first birth||12||14||1.37 (0.61, 3.04)||6||6||0.72 (0.21, 2.39)||6||8||1.78 (0.60, 5.31)|
|Never used||2,513||2,303||1.00 Referent||917||788||1.00 (Referent)||1,596||1,515||1.00 (Referent)|
|Ever used||2||15||8.59 (1.92, 38.39)||2||8||4.16 (0.83, 20.70)||0||7||Undefined34|
|<5 years ago||1||1||1.20 (0.07, 19.50)||1||0||0||1|
|5–9 year ago||1||12||13.70 (1.74, 108.06)||1||8||0||4|
|≥10 years ago||0||2||Undefined3||0||0||0||2|
|Duration of use5|
|<5 years||2||8||4.74 (0.98, 23.01)||2||4||0||4|
Fifteen cases and 2 controls, representing less than 1% of the study population, reported having used contraceptive implants (Table IV). This case–control difference is consistent with a more than 8-fold increase in breast cancer risk associated with use of this hormonal preparation (OR 8.59; 95% CI 1.92, 38.39, exact p = 0.0009). The 17 subjects who remain classified as users after data review include 15 whose tape or call-back provided specific information describing a subdermal implant, e.g., stated “Norplant” or “implant in the arm,” and 2 subjects with a plausible reported date of use and insufficient detail on the tape to classify them as non-users. The case–control difference remained significant if the latter 2 were also excluded (exact p = 0.003). Most users of implants reported first use between 5 and 9 years before the reference year. The majority of implant users (13 of 17, 76%) had also used contraceptive pills. All of the implant users reported that the product was prescribed in the U.S. Users of contraceptive implants differed from all subjects in that 65% (11 of 17) resided in Colorado, whereas only 20% of the study population was drawn from this state, and the median age of implant users in the reference year was 37, somewhat younger than the study population median of 54. In other respects, users of contraceptive implants were similar to other study subjects: 3 (18%) reported a family history of breast cancer, they had first pregnancies at a median age of 21, and the median number of live births was two. Breast cancer characteristics of cases who had used contraceptive implants did not differ markedly from all cases: 67% of cases with known ER status were ER positive, compared to 78% of all cases and 80% had ductal histology, compared to 74% of all cases.
The influence of environmental and lifestyle risk factors on breast cancer may differ for U.S. Hispanic populations compared to non-Hispanic whites,11, 19, 20 but for most exposures, few data have been reported for the Hispanic population specifically. Our data do not indicate that breast cancer in Hispanics is affected by use of OC to any greater or lesser extent than the transient, modest increase in risk that has been reported from past studies that included primarily non-Hispanic white subjects.7
When exposures to exogenous estrogen and progestin in the form of OC became widespread, concerns were raised about potential effect health effects, including breast cancer risk.21 The possible association between use of OC and breast cancer has been examined in a large number of epidemiologic studies, reviewed in detail elsewhere.5, 6, 22, 23 Many studies found no overall association between OC use and breast cancer. However, increased breast cancer risk has been reported for certain subgroups, specifically women who are young and/or who are recent users of OC. The association with recent use has been supported by analyses of aggregate data from multiple studies.7, 23
The hormone exposures from OC have decreased since these products were first introduced. Typical OC formulations prescribed in the U.S. in the 1960's contained 1, 2 or 10 mg of progestin and 50–100 μg of estrogen,24 and thus women reporting use during this time period would have been exposed to what are now considered high doses. OC use during the 1970's represents a mixed set of exposures, as products containing ≤0.5 mg of progestin and <50 μg of estrogen became available, but a large proportion of prescriptions continued to be for the higher dose formulations.24 During the 1980's, most prescriptions contained ≤1 mg of progestin and ≤50 μg of estrogen.24 Since that time, OC containing 0.15–1 mg of progestin and 35 or 30 μg of estrogen have remained prevalent; compounds containing 20 μg of estrogen or less were introduced but represented only 8% of prescriptions in 1998.25 It has been suggested that increased breast cancer risk may have been associated with exposures to older, high-dose formulations of OC, but that modern, low-dose OC will not be associated with breast cancer.5, 9, 26
The estimated 22% increase in breast cancer risk for Hispanic women and 28% increase for non-Hispanic whites associated with use of OC within 5 years of the reference year in the 4 Corners study are consistent with the 24% increased risk associated with recent use reported from a pooled analysis of a number of earlier studies.7 Thus the 4 Corners study data, describing a population for whom recent use refers to the years 1994–2003, estimate a modestly increased breast cancer risk for recent users that is similar in magnitude to studies from earlier decades. A limitation of the present data is that specific formulations of OC exposures can not be assessed; however, decade of use can be used to infer general categories of high- or low-dose. Our analysis by duration of use indicates that women who used OC for 2 decades or longer, whose use history would include older, high-dose formulations, are at highest risk. Among the recent users, women who had first used OC after 1980 had somewhat lower ORs than those with an earlier date of first use, but the ORs for these subgroups had wide CIs.
Some studies have suggested that OC use affects risk of specific subtypes of breast cancer. Associations with ER negative tumors8 and with lobular histologies27, 28 have been reported. The present data concur with prior reports that the influence of OC appears to be stronger for ER-negative tumors, but do not support a model in which the association of OC with breast cancer is specific to lobular histology. The number of subjects with lobular tumors was small however, and CIs were wide.
In the 4 Corners study population, 73 controls and 79 cases reported a history of use of injected hormones for birth control. This represented an estimated 23% higher risk of breast cancer, but with this small number of users, the association could be the result of chance. Other studies have generally reported no overall association between contraceptive injections and breast cancer29 but some studies have noted some evidence of an association with recent use,30, 31 with recently starting use,32 or with first use before age 25.33 A pooled analysis of prior studies found an increased risk of breast cancer associated with use of injected progestin-only contraceptives that was of similar magnitude to the increased risk associated with estrogen–progestin OC.7 In the 4 Corners study data, no clear patterns of association with timing or duration of use were observed except for an increased risk with first use after the age of 35.
The significant increase in breast cancer risk associated with use of contraceptive implants is a novel finding. Data on a possible association between use of this product and breast cancer are quite limited. A case–control study reported that among more than 9,000 women interviewed in the U.S., 5 cases and 7 controls reported implant use,29 with no increased breast cancer risk associated with use. The time since first exposure for women in that study would have been relatively short, between the approval of contraceptive implants use in the U.S. in December, 199015 and diagnosis of cases eligible for the study in 1994 through 1998. In the 4 Corners study, most cases reported first use 5 or more years before the reference year. A postmarketing surveillance study conducted in 8 developing countries reported 4 invasive and 2 in situ breast cancers during 5 years of follow-up among almost 8,000 users of subdermal implants (mean age of 28.5 years), compared to 1 invasive and zero in situ breast cancers among a comparable number of women, using IUD's or sterilization for birth control, representing a nonsignificant incidence rate ratio of 4.1 for invasive breast cancer.34 The prospective study had very limited power to detect an association because of the low breast cancer incidence among young women in developing countries, but the magnitude of elevated incidence of breast cancer in the cohort study is consistent with the association in the present case–control study.
Differential error in recall between cases and controls may introduce bias in case–control studies. In the present study, respondents who reported use of contraceptive implants were probed to confirm that the product described was a subdermal implant. It seems unlikely that differential error in recall alone could be the explanation for 15 cases, but only 2 controls, reporting use of contraceptive implants. Study subjects who initially reported implant use but were reclassified as nonusers based on data review or call-backs included 5 cases and 6 controls, indicating that recall errors that we could detect occurred in approximately equal proportions in cases and controls. The influence of selection bias on estimates of exposure-disease associations is another concern in case–control studies. In the present study the use of hormonal contraceptives by nonparticipants can not be evaluated, but for characteristics that we were able to assess we did not detect differential effects on study participation.16 Estimated exposure-disease relationships for oral and injective contraceptives in the present study were similar to associations reported in other studies, providing some confidence that the data is not unduly influenced by selection bias with respect to contraceptive use. We note that the one other study that observed a comparable, although statistically nonsignificant, increased risk of breast cancer among contraceptive implant users was prospective,34 so that the association in that setting could not be explained by biased recall or selection.
The active ingredient of subdermal implants available in the U.S. during the 1990's was levonorgestrel,15 a progestin compound. A breast cancer-promoting effect of exogenous progestins is plausible given evidence from other research. Mechanistic and laboratory studies indicate that progestins act synergistically with growth factors to stimulate breast cell proliferation.35, 36, 37 Other observational epidemiologic studies are suggestive of a role for progestins in breast cancer.38 A randomized trial indicated that hormone replacement therapy increased breast cancer risk only when the preparation included a progestin.39, 40 Our data are consistent with a stronger positive association of subdermal implants with breast cancer than is observed for injectable progestin contraceptives.7, 29, 30, 31, 32, 33 Injectable contraceptives contain a different progestin compound, medroxyprogesterone acetate. The biological actions of progestins differ depending on the specific compound,41 including different effects on serum estrogens42 and on expression of estrogen metabolizing enzymes in breast cells.43 The levonorgestrel-releasing IUD is a technology which provides the same progestin compound as the subdermal implant, but at a lower systemic dose44, 45; breast cancer incidence in a cohort of users of that product (171 cases in ∼150,000 person-years) did not exceed the number expected for most age groups.46 Limitations of the study of levonorgestrel-releasing IUDs were that the cohort of users was compared to national incidence rather than a defined comparison group, that no information on other risk factors for breast cancer was taken into account in the analysis, and that follow-up from time of first use was relatively short, through the year 2000 for women who initiated use in 1990–1993.
In summary, the direction and magnitude of associations between oral and injected hormonal contraceptives and breast cancer in Hispanic women are concordant with associations characterized in other populations. Subdermal implant contraceptives have not been widely used in the U.S., and the disproportionate use of implanted contraceptives reported by cases compared to controls in the present study is based on a small number of users, but indicates a possible strong association. Implantable contraceptives have been more widely used on other parts of the world, and the current data suggest that continued research on cancer risk among users of these preparations may be warranted.
The authors acknowledge the contributions of Ms. Leslie Palmer, Mr. Roger Edwards, Ms. Karen Curtin, Ms. Betsy Risendal, Ms. Tara Patton, Mr. Jason Witter and Ms. Kelly May to data collection and management.
- 2National Cancer Institute DCCPS Surveillance Research Program Cancer Statistics Branch. Surveillance, epidemiology, and end results (SEER) program, SEER*Stat Databases: incidence—SEER 13 Registries, public use data set, November 2004, submission vol. 2006. Available at www.seer.cancer.gov.
- 7Collaborative 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. Lancet 1996; 347: 1713–27.
- 15Drugs@FDA; FDA approved drug products, vol. 2007. Washington, DC: U.S. Food and Drug Administration, Center for Drug Evaluation and Research. Available at http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfm?fuseaction=Search.DrugDetails.
- 17Division of Cancer Control and Population Sciences. The SEER program code manual, 3rd ed. Bethesda, MD: National Cancer Institute, U.S. Department of Health and Human Services, 1998.