The association between oral contraceptive use and lobular and ductal breast cancer in young women
Recent reports indicate that the incidence of lobular breast cancer is increasing at a faster rate than ductal breast cancer, which may be due to the differential effects of exogenous hormones by histology. To address this issue, we examined whether the relationship between oral contraceptive use and incident breast cancer differs between lobular and ductal subtypes in young women. A population-based sample of in situ and invasive breast cancer cases between ages 20 and 44 were recruited from Atlanta, GA; Seattle-Puget Sound, WA and central New Jersey. Controls were sampled from the same areas by random-digit dialing, and were frequency matched to the expected case age distribution. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated using polytomous logistic regression. Among the 100 lobular cancers, 1,164 ductal cancers, and 1,501 controls, the odds ratios for oral contraceptive ever use were 1.10 (95% CI = 0.68–1.78) for lobular cancers and 1.21 (95% CI = 1.01–1.45) for ductal cancers, adjusted for study site, age at diagnosis, and pap screening history. Our results suggest that the magnitude of the association between ever use of oral contraceptives and breast cancer in young women does not vary strongly by histologic subtype. These results are similar to previous studies that report little difference in the effect of oral contraceptive use on breast cancer by histology. © 2007 Wiley-Liss, Inc.
Recent reports indicate that the incidence rate of invasive lobular breast cancer is increasing faster than invasive ductal breast cancer,1, 2, 3, 4 but few epidemiologic studies have focused on identifying breast cancer risk factors by histologic subtype. Among the studies that have, the association between combination hormone replacement therapy and risk of invasive breast cancer differed for ductal and lobular cancers in most studies,5, 6, 7, 8, 9, 10 but not all.11 These findings suggest that the effects of other exogenous hormonal treatments, such as oral contraceptives, may also differ by histologic type.
It has been consistently shown that oral contraceptive use is associated with an increased risk of breast cancer in young women,12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 but not older women.21, 23, 24, 25 Recently, Newcomer et al.26 reported that current oral contraceptive use, recent use, and older age at first use are associated with an increased risk of invasive lobular breast cancer, but not invasive ductal breast cancer in a population of middle-aged women, suggesting that oral contraceptives may have effects related to specific histologic subtypes. It is unknown whether the associations between oral contraceptive use and breast cancer differs by histologic subtype among younger women—the age group most likely to experience increased risk of breast cancer due to oral contraceptive use.25 To determine this, we explored the relationship of oral contraceptive use and lobular and ductal breast cancer subtypes within a population of young women. Previously, we reported an odds ratio (OR) of 1.2 for the association between oral contraceptive use and all breast cancers in this population,22 which is consistent with other results in the literature for women less than 45 years old.24
Material and methods
Study data were originally collected as part of the Women's Interview Study of Health (WISH), a population-based case–control study designed to investigate the effects of oral contraceptives, alcohol consumption and dietary intake during adolescence on breast cancer risk in young women.22 Cases were recruited from metropolitan Atlanta, GA, Seattle-Puget Sound, WA, and a 5-county area in central New Jersey and were defined as any woman between 20 and 44 years of age who was diagnosed with incident in situ or invasive breast cancer between May 1, 1990 and December 31, 1992. Eligible cases were identified using rapid reporting systems established in each of the 3 geographic areas and checks were made against population-based cancer registries in these same areas. Controls were randomly selected from the same geographic regions as cases using 13 cycles of random-digit dialing, and were matched to the expected age distribution of the cases.
Cases and controls were asked to give written informed consent, and then complete an in-person interview. During the interview, information was collected by a trained interviewer about demographic characteristics, lifestyle, medical history, contraceptive use, diet, reproductive and menstrual histories, physical activity, smoking habits, occupation, exogenous hormone use, alcohol consumption and family history of cancer. Anthropometric measures were also obtained. A book of color photographs and descriptions of oral contraceptive pills organized by brand, name and dose, corresponding packaging, and years the pill was available in the United States were shown to study subjects to help identify specific formulations they used. Additionally, a monthly reproductive history calendar marked with pregnancies, other reproductive events, and contraceptive use since menarche was used to aid in the recall of dates of oral contraceptive use. Interviews were completed for 86.4% of eligible cases and 78.1% of eligible controls.
Clinical and histopathologic information for cases was obtained from hospital and medical records and classified according to the International Classification of Diseases.27 Cases with an ICD-O behavior code of 2 were classified as in situ, and cases with a behavior code of 3 were classified as invasive. To facilitate comparisons with a previously published study of the effect of oral contraceptive use on lobular and ductal breast cancer incidence,26 we limited our analysis to cases of ductal carcinoma, NOS (ICD-O-2 code 8500) and lobular carcinoma (ICD-O-2 code 8520). Table I shows the histologic classification of all cases in the study population. There were 1,164 ductal cases with a mean age at diagnosis of 39 years and 100 lobular cases with a mean age at diagnosis of 41 years. The 1,501 controls had a mean age of 38 years old at the time of the telephone screening interview.
Table I. Frequency of Histologic Subtypes Among Breast Cancer Cases
|8001||Tumor cells, malignant||1|
|8020||Carcinoma, undifferentiated type, NOS||1|
|8021||Carcinoma, anaplastic, NOS||2|
|8050||Papillary carcinoma, NOS||1|
|8230||Solid carcinoma NOS||1|
|8500||Duct carcinoma, NOS||1,164|
|8503||Intraductal papillary adenocarcinoma||4|
|8505||Intraductal papillomatosis, NOS||1|
|8510||Medullary carcinoma, NOS||29|
|8511||Medullary carcinoma with amyloid stroma||1|
|8512||Medullary carcinoma with lymph. stroma||2|
|8520||Lobular carcinoma, NOS||100|
|8522||Duct and lobular carcinoma||37|
|8540||Paget disease, mammary||2|
|8541||Paget disease and infiltrating duct carcinoma||2|
|8543||Paget disease and intraductal carcinoma||2|
|8570||Adenocarcinoma with squamous metaplasia||2|
|9680||ML, large cell, diffuse||2|
|Unknown or not classified2||6|
The association between oral contraceptive use and lobular and ductal breast cancers was estimated using polytomous logistic regression.28 All information about oral contraceptive use and other covariates refers to exposures prior to the reference date, which was defined as the age at diagnosis for cases or age at telephone screening interview for controls. Oral contraceptive use variables were created based on categories used in previously published studies.22, 24, 26, 29, 30 Lifetime oral contraceptive use of 6 months or longer was treated as ‘ever use’. Oral contraceptive use of less than 6 months or no use at all was treated as ‘never use’. Among ever users, oral contraceptive use within 24 months of the reference date was classified as recent use. Last use more than 24 months before the reference date was categorized as former use. In addition, we also examined age at first use of oral contraceptives, number of years since last use, duration of use, and the effect of oral contraceptive ever use where cases were further stratified by in situ (LCIS, N = 43; DCIS, N = 152) and invasive (lobular, N = 57; ductal, N = 1012) stage.
A binary variable describing menopausal status was evaluated as an effect measure modifier of oral contraceptive ever use on the multiplicative scale. A woman was considered post-menopausal if she reported of having a hysterectomy, bilateral oopherectomy, or natural cessation of menses; otherwise she was considered premenopausal. The interaction between oral contraceptive ever use and menopausal status failed to improve model fit and was left out of the final model.
Several breast cancer risk factors were assessed for confounding by backwards elimination, using the approach described by Rothman and Greenland.31 A covariate was considered a confounder if elimination of the covariate changed the regression coefficient for oral contraceptive use by more than 0.10 on the natural log scale compared to the fully adjusted regression model. Self-identified race, noncontraceptive exogenous hormone use, history of breast cancer in at least one first degree relative, previous breast biopsy, and menopausal status were coded as binary variables. Age at menarche, total pretax household income for the calendar year prior to interview, highest education level completed, lifetime average alcohol consumption, pap screening history in the past 5 years, and mammographic screening history in past the 5 years were coded as nominal categorical variables. Age at first birth and number of children were coded as categorical variables and examined jointly with a binary variable indicating parity.32 The elimination of pap screening history changed the effect estimate for oral contraceptive use by more than 0.10 on the natural log scale compared to the full model that was adjusted for all potential confounders; thus, pap screening history was retained in the final regression model and adjusted for as a confounder. When we repeated the backwards elimination without including pap screening history in the fully adjusted model, no additional covariates were confounders. All estimates were also adjusted for the frequency matching factors, study site and age at the reference date.
The distribution of breast cancer risk factors that were evaluated as confounders is shown in Table II. Although estimates are imprecise, the associations between lobular cancer and late age at first birth, high household income, postgraduate education, frequency of pap screening, and ever having mammographic screening are stronger than the associations between those same risk factors and ductal cancer (Table II).
Table II. Demographic and Reproductive Characteristics of Cases and Controls
|Age at diagnosis (years)|
| 20–24||9||1||0||0|| ||4||0|| |
| 25–29||57||4||2||2|| ||34||3|| |
| 30–34||225||15||6||6|| ||141||12|| |
| 35–39||474||32||19||19|| ||362||31|| |
| 40–44||736||49||73||73|| ||623||54|| |
|Age at menarche (years)|
| 11 or younger||348||23||23||23||1.31 (0.67–2.56)||278||24||1.16 (0.92–1.47)|
| 12||402||27||32||32||1.52 (0.81–2.87)||364||31||1.31 (1.05–1.65)|
| 13||444||30||30||30||1.35 (0.71–2.57)||313||27||1.04 (0.82–1.30)|
| 14 or older||305||20||15||15||Referent||207||18||Referent|
| Missing||2|| || || || || || || |
| Parous||1,165||78||80||80||0.83 (0.50–1.40)||866||74||0.77 (0.64–0.93)|
| Missing||1|| || || || ||1|| || |
|Number of children3|
| 2–3||742||64||58||73||1.11 (0.65–1.92)||563||65||0.92 (0.75–1.13)|
| 4 or more||126||11||3||4||0.33 (0.10–1.16)||61||7||0.58 (0.41–0.82)|
|Age at first birth (years)3|
| 20–24.9||370||32||18||23||0.89 (0.42–1.91)||260||30||1.17 (0.90–1.51)|
| 25–29.9||324||28||24||30||1.52 (0.73–3.13)||246||28||1.30 (1.00–1.69)|
| ≥30||215||18||26||33||2.47 (1.21–5.07)||207||24||1.60 (1.21–2.12)|
|Non-contraceptive hormone use|
| Yes||139||9||9||9||0.83 (0.40–1.71)||68||6||0.56 (0.41–0.75)|
| Missing||4|| || || || ||2|| || |
| Post-menopausal||186||12||11||11||0.69 (0.35–1.34)||119||10||0.72 (0.56–0.93)|
| Missing||5|| ||1|| || ||1|| || |
| $25K–$49K||522||36||28||29||1.61 (0.72–3.59)||359||31||0.85 (0.68–1.06)|
| $50K–$89K||473||32||42||43||2.45 (1.13–5.33)||402||35||1.04 (0.83–1.30)|
| ≥$90K||203||14||20||20||2.42 (1.03–5.66)||171||15||1.00 (0.76–1.32)|
| Missing||36|| ||2|| || ||20|| || |
| Non-white||319||21||15||15||0.71 (0.40–1.26)||236||20||0.92 (0.75–1.12)|
| Missing||2|| || || || ||6|| || |
| High School or less||402||27||24||24||Referent||315||27||Referent|
| Technical School-College grad||898||60||53||53||1.10 (0.67–1.81)||675||58||0.98 (0.82–1.17)|
| Postgraduate work||201||13||23||23||1.87 (1.02–3.41)||174||15||1.09 (0.85–1.40)|
|Family history of breast cancer4|
| 1 or more relatives||95||6||15||15||2.41 (1.33–4.36)||164||14||2.42 (1.86–3.16)|
|Ever had a pap smear|
| No||34||2||0||0|| ||11||1||Referent|
| Yes||1,465||98||100||100|| ||1,153||99||2.38 (1.20–4.73)|
| Missing||2|| || || || || || || |
|Number of pap smears in the past 5 years5|
| 4–5||755||52||71||71||2.53 (1.54–4.15)||682||59||1.33 (1.13–1.58)|
| ≥6||169||12||7||7||1.36 (0.56–3.27)||97||8||0.87 (0.66–1.17)|
|Ever had a mammogram|
| Yes||729||49||74||74||2.17 (1.34–3.53)||646||55||1.23 (1.04–1.45)|
| Missing||1|| || || || || || || |
|Number of mammograms in the past 5 years6|
| 0–1||424||58||23||31||0.23 (0.12–0.45)||290||45||0.37 (0.27–0.52)|
| 2–3||231||32||31||42||0.50 (0.27–0.94)||225||35||0.55 (0.39–0.77)|
| Missing||2|| || || || ||2|| || |
|Alcohol use (average drinks per week)|
| <1–6.9||754||50||63||63||1.67 (1.06–2.63)||587||51||1.14 (0.96–1.35)|
| ≥7.0||167||11||7||7||0.94 (0.40–2.19)||175||15||1.54 (1.20–1.98)|
| Missing||4|| || || || ||2|| || |
|Previous breast biopsy|
| Yes||95||6||15||15||2.27 (1.25–4.12)||108||9||1.44 (1.08–1.92)|
The ORs for ever use of oral contraceptives, adjusted for age and study site only (without adjustment for pap screening history), are 1.27 (95% CI = 0.79, 2.05) for lobular breast cancer and 1.27 (95% CI = 1.07, 1.52) for ductal breast cancer. The estimated effects of different patterns of oral contraceptive use on the relative odds of ductal and lobular cancer, adjusted for age, study site, and history of pap screening, are shown in Table III. Ever use of oral contraceptives is associated with a slightly elevated OR for ductal breast cancer (OR = 1.21, 95% CI = 1.01–1.45). Recent oral contraceptive use is more strongly associated with ductal cancer than former use (ductal, recent use: OR = 1.45, 95% CI = 1.08–1.96 vs. ductal, former use: OR = 1.17, 95% CI = 0.97–1.41). A shorter time interval since last use of oral contraceptives, extended use of oral contraceptives, and later age at first use were also associated with increased odds of ductal cancer (Table III). The association between oral contraceptive ever use and lobular cancer (OR = 1.10, 95% CI = 0.68–1.78) is similar to the association for ductal cancer but is not statistically significant. The association between lobular cancer and patterns of oral contraceptive use are shown in Table III, however, because of the small number of lobular cases in this analysis, many of the estimates are imprecise and no conclusions can be made about trends by recency, duration or age of first oral contraceptive use.
Table III. Odds Ratios for the Association Between Breast Cancer and Oral Contraceptive use, by Histological Type
|Ever||1,076||72||75||75||1.10 (0.68, 1.78)||888||76||1.21 (1.01, 1.45)|
| Former4||911||85||73||97||1.16 (0.72, 1.89)||750||84||1.17 (0.97, 1.41)|
| Recent5||165||15||2||3||0.33 (0.08, 1.47)||138||16||1.45 (1.08, 1.96)|
|Age at first use (years)|
| <20||557||52||30||40||1.01 (0.57, 1.78)||433||49||1.17 (0.95, 1.44)|
| 20–24||448||42||36||48||1.06 (0.61, 1.81)||380||43||1.19 (0.97, 1.48)|
| ≥25||71||7||9||12||1.96 (0.87, 4.44)||75||8||1.56 (1.08, 2.24)|
| p-trend|| || || || ||0.31|| || ||0.02|
|Time since last use (years)|
| 0–4||219||20||4||5||0.45 (0.15, 1.36)||194||22||1.55 (1.18, 2.03)|
| 5–9||165||15||9||12||0.99 (0.45, 2.22)||134||15||1.27 (0.95, 1.68)|
| ≥10||692||64||62||83||1.21 (0.74, 1.99)||560||63||1.11 (0.91, 1.36)|
| p-trend|| || || || ||0.29|| || ||0.63|
|Duration of use (years)|
| <1||91||8||9||12||1.63 (0.72, 3.65)||70||8||1.13 (0.80, 1.61)|
| 1–3||395||37||31||41||1.23 (0.70, 2.14)||295||33||1.11 (0.89, 1.38)|
| ≥4||590||55||35||47||0.92 (0.53, 1.59)||523||59||1.30 (1.06, 1.59)|
| p-trend|| || || || ||0.69|| || ||0.01|
We explored whether the association between oral contraceptive ever use and breast cancer differed by stage among lobular and ductal subtypes, although our estimates are based on small numbers. Among the lobular cases, the effect of ever using oral contraceptives was higher for the in situ cases (LCIS, ever use: OR = 1.60, 95% CI = 0.72–3.54; invasive lobular, ever use: OR = 0.86, 95% CI = 0.47–1.56) (Table IV). Among the ductal cases, the effect of ever using oral contraceptives was higher for the invasive cases (DCIS, ever use: OR = 0.75, 95% CI = 0.51–1.10; invasive ductal, ever use: OR = 1.30, 95% CI = 1.07–1.58) (Table IV).
Table IV. Odds Ratios for the Association Between Breast Cancer and Ever use of Oral Contraceptives, by Histology and Stage
| Ever||1,076||72||35||81||1.60 (0.72, 3.54)||108||71||0.75 (0.51, 1.10)|
| Ever||1,076||72||40||70||0.86 (0.47, 1.56)||780||77||1.30 (1.07, 1.58)|
The aim of this analysis was to investigate the effect of oral contraceptive use on specific breast cancer histologies in a population of young women. We first reported an association between oral contraceptive use and breast cancer in the WISH population in 1995.22 Our current results suggest that the effect of oral contraceptive ever use is similar for ductal and lobular breast cancer. However, the estimated effect for lobular cancer is imprecise because of the small number of lobular cases in this population. Among ductal cancers, we were also able to estimate increased odds of breast cancer with recent use, shorter time since last use, and long duration of oral contraceptive use. As expected, our estimated association between oral contraceptive use and ductal breast cancer is similar to previously reported positive associations between breast cancer and current use, more recent use, and long duration of use of oral contraceptives13, 14, 16, 17, 18, 20, 21, 23, 24, 33, 34, 35; this is likely because ductal cancers make up approximately 75% of all invasive breast tumors diagnosed in women under age 50.36
Previous studies have addressed the effect of oral contraceptive use by breast cancer histology, but among study subjects with a wide age range. Analyzing younger and older women together could mask heterogeneity of effects due to age, since previous age-specific analyses of the association between oral contraceptive use and breast cancer (regardless of histologic type) have observed variation by age.17, 21, 23, 24, 25 Romieu et al.18 reported that there was little difference in the effect of oral contraceptive use among different breast cancer histologies, but the authors did not report the effect estimates for the different histologies. Newcomer et al.26 reported ORs of 1.2 (95% CI = 0.9–1.6) for the association between oral contraceptive ever use and invasive lobular cancer and 1.0 (95% CI = 0.9–1.1) for the association between oral contraceptive ever use and invasive ductal cancer. Despite the younger age of our study subjects, our results are compatible with those of Romieu et al.18 and Newcomer et al.26, in that the association with oral contraceptive use did not differ strongly for ductal and lobular cancer.
Previous studies that have considered differences in the effect of oral contraceptive use on histologic subtypes of breast carcinoma in situ have found a negative association with LCIS among women 20–79,37 no association with DCIS among premenopausal women,38 and no association with any type of breast carcinoma in situ among recently screened women 35–64 years old.39 Although we determined that mammographic screening was not a confounder in our data, it is possible that the confounding effect of mammography differs in populations where screening is more prevalent. This is especially relevant to the study of in situ lesions, which are detected primarily as a consequence of mammographic screening. Given that many of the women in the WISH study population had not yet reached the recommended age for mammographic screening, our positive association for LCIS and negative association for DCIS may not be comparable to results from populations where more women are of screening age. Our findings by stage may be attributed to random variation due to small sample sizes, and thus additional studies of oral contraceptive use and DCIS and LCIS are needed. A more precise estimation of the effect of oral contraceptive use on in situ breast cancer may clarify whether there is heterogeneity of effect among in situ subtypes.
The effect of oral contraceptives on incident breast cancer is small (i.e., close to the null), and so it is possible that even small amounts of residual confounding or other biases could distort the true effect. We used backwards elimination to select confounders to identify joint confounding by 2 or more covariates. Using this method, only pap screening history was a confounder in this dataset. However, to be a confounder pap screening must be a risk factor for oral contraceptive use and not affected by use.40 Because it is possible that some women receive pap screening for the purpose of obtaining an oral contraceptive prescription, making pap screening the result of oral contraceptive use, pap screening may or may not meet the criteria of a confounder. To ensure that our results would not be affected by adjusting for a variable whose role as a confounder is unclear, we repeated the backwards elimination procedure without pap screening in the model. No other covariates met our statistical criteria for confounding, suggesting that the inclusion of pap screening in our initial model did not mask confounding by other covariates. Furthermore, there was no substantial difference in the effect of oral contraceptives on lobular and ductal breast cancer when we did not adjust for pap screening.
The small number of lobular cases in our study limits our power to detect effects associated with detailed patterns of use. Although the primary aim of the WISH study was to examine breast cancer risk in younger women in relation to patterns of oral contraceptive use,22 it was not designed specifically to evaluate hypotheses about lobular breast cancer. Lobular cancers are a small proportion of all breast cancer cases diagnosed, and the mean age at diagnosis for lobular breast cancer is older than the mean age at diagnosis for ductal cancer,36 and so it was expected that there would be a small number of lobular cases in the WISH population. Because of the small number of lobular cases, it is possible that some effect estimates do not reflect the true association of oral contraceptive use, and instead may be the result of random error. A further limitation was our inability to estimate the association separately in white and African American women, because of both the small number of lobular cases and small proportion of African American participants. Therefore, additional research in younger women is needed to confirm our results. In particular, future studies should be designed to have adequate numbers of women with lobular subtype so that the effect of different patterns of oral contraceptive use on in situ and invasive breast cancer can be analyzed in addition to ever use of oral contraceptives.
Despite the imprecision of the lobular results, by focusing specifically on young women our analysis provides information that has not yet been reported by previously published studies. Additionally, the WISH study design limits the chance that the results are biased due to selection bias or disease misclassification. Study subjects were accrued from areas with high quality, population-based cancer registries that operated with a standard protocol for case ascertainment and data reporting, so case identification procedures were consistent across the study centers. Furthermore, previous studies have shown that women are able to recall past oral contraceptive use with a high degree of accuracy,41, 42 and that recall is not differential when exposure information is collected after a breast cancer diagnosis.43 Misclassification of oral contraceptive use was further minimized by using memory aids that have been shown to enhance recall of oral contraceptive use.44 Oral contraceptive information was collected after informed consent, and the interviewer was blind to case status. Thus, any potential misclassification of oral contraceptive use should be independent of case status. Histology and stage data were abstracted directly from patient records, so there is potential for misclassification of histological data or inconsistency of diagnosis between study sites. However, studies of diagnostic inter-observer reproducibility have shown good agreement among pathologists in distinguishing breast lesions' stage and histologic subtype.45, 46, 47
In conclusion, our study adds to the available data characterizing breast cancer in young women. Although our estimates are imprecise, our data suggest that the relationship between oral contraceptive use and lobular and ductal cancers is similar in young women, as was reported previously for older women by Romieu et al.18 and Newcomer et al.26 Several approaches, including histological and molecular subtyping, have been used to define the etiologic heterogeneity of breast cancer. However, the knowledge that lobular cancer is increasing emphasizes the promise of continued research on the relationship between risk factors and specific breast cancer histologic subtypes. Future research could aid in determining whether different breast cancer histologic subtypes have different etiologies, and may identify factors that contribute to the increasing incidence rates of lobular cancer.