|Home | About | Journals | Submit | Contact Us | Français|
The relationship between hormone therapy (HT) and invasive breast cancer has been extensively investigated, but the relationship between HT and in situ breast cancer has received relatively little attention. We examined the relationship between HT and ductal carcinoma in situ (DCIS) among postmenopausal women who participated in a population-based case control study in Connecticut, USA.
This analysis included 1,179 post-menopausal women (603 controls and 576 cases), who comprised a subset of a population-based case-control study that included all incident cases of breast carcinoma in situ (BCIS) in Connecticut and frequency-matched controls by 5-year age intervals.
We found no association between DCIS and ever use of any HT (adjusted odds ratio (OR) = 0.85, 95 % confidence interval (CI): 0.65-1.11); of estrogen alone (adjusted OR = 0.93; 95 % CI: 0.68-1.29) or of estrogen and progesterone (adjusted OR = 0.75; 95 % CI: 0.52-1.08). There was also no association between DCIS and current use of these hormones. In addition, estimated risk of DCIS did not increase with duration of use of these preparations.
These results add to a small literature that remains inconclusive. To determine whether HT poses risk of in situ breast cancer, larger studies with greater power and precise control of important covariates (e.g., mammography screening) are needed, as are meta-analyses of available data.
Largely attributed to an increase in mammography screening , the incidence of breast carcinoma in situ (BCIS) in the United States has increased from 5.83 cases per 100,000 women in 1975 to 34.59 per 100,000 in 2007 . In situ tumors now account for nearly 22% of all breast cancers . Though we know that women with these lesions are far more likely to develop invasive breast cancer than women without in situ disease , our understanding of the natural history of BCIS remains limited, making it difficult to provide optimum treatment . Epidemiological studies that examine whether established risks of invasive breast cancer are also implicated in non-invasive disease bear on whether these lesions share a common etiology. These studies may also help to elucidate when, in the sequence of events that may culminate in invasive cancer, particular risks factors exert their greatest influence [4, 5].
Evidence is emerging that several risks may contribute to both BCIS and invasive breast lesions including older age , family history of breast cancer [4, 7], and higher breast density . Some endogenous hormonal factors associated with invasive breast cancer have also been associated with in situ disease (e.g., fewer full-term pregnancies, older age at first birth, older age at menopause) , while others have not (e.g., age at menarche, breastfeeding) . For invasive breast cancer, postmenopausal hormone therapy (HT) is a well-established risk. In an analysis of 51 studies of breast cancer and HT conducted in 1997 by the Collaborative Group on Hormonal Factors in Breast Cancer, risk increased with greater duration of use, and current users were at greater risk than past users . More recent studies have refined the relationship between invasive breast cancer and HT and demonstrated greater risk with combined estrogen and progesterone formulations than with estrogen alone . There is also some evidence that among women using combination HT, continuous use of progesterone may put women at higher risk than sequential use . However, with regard to in situ disease, information is much more limited. Two reviews [12, 13] found inconsistent results among studies that examined the relationship between HT and DCIS. The more recent review, commissioned by the National Institutes of Health (NIH) for a state-of-the-science conference on diagnosis and management of ductal carcinoma in situ (DCIS), included only 5 observational studies and one clinical trial . Previous studies included women with both invasive and in situ lesions, with in situ cases comprising a subset of the women in those studies. Given the increase in non-invasive breast cancers and our limited understanding of the risks associated with development and progression of this disease, additional studies that focus on potential risks specifically for in situ breast lesions are needed.
Unlike prior work, our analysis is based on a study of exclusively in situ cases. Our objective was to examine the association between HT and DCIS in a population-based case-control study that included a relatively large number of DCIS cases. In addition to examining ever use of HT, we examined whether: 1) risk increased with duration of use; 2) current users were at greater risk than those who had ever used HT; and 3) continuous users of combined estrogen and progesterone were at higher risk than sequential users.
These analyses are based on data from a population-based case control study that examined a broad range of potential risks of BCIS . The study was approved by the Human Investigations Committees of Yale University School of Medicine and the State of Connecticut Department of Public Health.
The methods have been previously described . Briefly, the study invited to participate all English-speaking incident cases of BCIS among female residents of Connecticut, at least 20 years of age, diagnosed between September 15, 1994 and March 14, 1998. The Rapid Case Ascertainment Shared Resource of the Cancer Center at Yale University identified the cases whose diagnosis was confirmed by review of the pathology report. Additionally, some data used in this study were directly obtained from the CT Tumor Registry in the Connecticut Department of Public Health. The authors assume full responsibility for analyses of these data.
Controls were English-speaking female residents of Connecticut, frequency matched to the cases by 5-year age group. The consulting firm Northeast Research of Oreno, Maine selected the controls by random digit dialing. Cases and controls with a history of breast cancer or of breast biopsy with unknown outcome were excluded from the study, as were cases that could not be identified as either DCIS or LCIS by pathology report review. Only cases with physician permission to participate were included. Potential subjects received an introductory letter, followed by a telephone call from a trained interviewer, requesting participation. Subjects verbally consented to be interviewed by telephone and they gave written consent to release medical records. The structured telephone interview obtained information on demographics, family medical history, menstrual and pregnancy histories, exogenous hormone use, personal medical and screening histories, and alcohol and smoking behaviors. Risks were assessed to date of diagnosis for cases and to date of interview for controls. The original study population included 999 controls and 998 cases (n = 875 with DCIS and n = 123 with lobular carcinoma in situ (LCIS)).
As the effect of HT on ductal and lobular breast cancer may differ , we included only DCIS cases in these analyses and excluded the 123 LCIS cases. We also restricted our sample to postmenopausal women, defined as those who: 1) reported natural menopause; 2) had bilateral oophorectomy; 3) had a hysterectomy without bilateral oophorectomy or had periods stopped by chemotherapy, radiation, or hormones and were at least age 55 years; and 4) started HT before their periods stopped. We excluded 396 controls and 299 cases from the original study population who were: 1) self-identified as pre- or perimenopausal; 2) had a hysterectomy without bilateral oophorectomy or had periods stopped by chemotherapy, radiation, or hormones and were less than age 55 years, as these women might be premenopausal; and[L1] 3) were missing information on menopausal status. The subjects for these analyses, a subset of the original study participants, thus included 1,179 postmenopausal women (603 controls and 576 cases with DCIS).
In keeping with most other studies, summarized by the Collaborative Group on Hormonal Factors in Breast Cancer , we defined age at menopause as age at last menstrual period (LMP) for women with natural menopause and age at bilateral oophorectomy for women who had undergone that procedure. Where age at menopause was not known, we used common conventions to impute age. First, among women who began HT before their periods stopped, we estimated age at menopause as the age they began HT, in keeping with the method employed by Ross . Second, for the few women who reported natural menopause but did not recall their age at menopause, we use age 55 years. Third, there[L2] were several groups of women who reported that their periods stopped as a result of medical treatments, including: 1) hysterectomy without bilateral oophorectomy; 2) radiation or chemotherapy; and 3) hormonal treatments (whether HT or another hormone treatment could not be determined). Because it is not known when ovarian functioning ceases among women with simple hysterectomy, and because menstruation may resume once women discontinue chemotherapy [16, 17] and other treatments, we only included women in these categories if they were ages 55 years and older and we used age 55 years as their age at menopause. We selected age 55 years because 90% of our sample had reached menopause by that age. We did not assign different ages at menopause to smokers and non smokers as have some others ; in our sample, age at menopause did not differ between these groups (p =0.250).
To examine HT use, we asked participants if they had used estrogen, progesterone, or other female hormones for purposes other than family planning or birth control. We then asked women who had used these hormones to provide the name(s)/brands(s) and dose of the medication(s) if known, the dates of use, the form in which they took the medications (e.g., oral, topical), and the pattern of use (continuous or sequential). In these analyses, we focused on oral preparations used during the postmenopausal period. We grouped hormone use as: 1) any HT (estrogen only or estrogen and progesterone); 2) estrogen only; and 3) estrogen and progesterone (continuous or sequential). If a woman used estrogen only and estrogen plus progesterone at different points in time, she was included in the estrogen plus progesterone group. We examined: 1) ever versus never use of these hormones; 2) current versus never use, where current use was defined as HT use within 12 months of diagnosis (cases) or interview (controls); and 3) duration of use.
We examined a number of covariates. These included sociodemographic factors (age, race, marital status, and education); potential risks of invasive and/or in situ breast cancer (first degree relative with the disease, younger age at first menstrual period, older age at first birth of a child, and fewer pregnancies); potential confounders of the relationship between HT and risk of breast cancer (time since menopause [9, 18], body mass index , frequency of mammography screening , and history of breast biopsy ); and oral contraceptive use.
We began with descriptive analyses to characterize the study population, by case control status, using frequency distributions. Next, using chi square tests, we examined bivariate associations between case control status and: 1) ever use of HT; and 2) each covariate. Then, we conducted multivariate analyses that examined the relationship of HT use to DCIS, adjusted by covariates that were associated with the outcome (p<.05, 2-sided) at the bivariate level. In separate models, we examined: 1) ever use of HT; 2) current use; and 3) duration of use. We conducted the multivariate analyses with unconditional logistic regression to generate odds ratios (ORs) and 95% confidence intervals (CIs) as estimates of risk. To test for linear trend (i.e., whether the risk of DCIS increased with more years of HT use), we constructed a model with duration of use in its ordinal form, adjusted for the same group of significant covariates included in the above multivariate models. In addition, because BMI has been shown to modify the association between HT and breast cancer, and HT has been shown to modify the association between BMI and breast cancer , we examined interaction terms for use of HT (ever/never) and each category of BMI to ensure that we were not overlooking potential modifying effects of these variables on DCIS.
We conducted these analyses on all menopausal women; i.e., those with known and unknown (imputed) age at menopause (Analysis 1, n=1,179). Then, we repeated the analyses on the subset of menopausal women with known age at menopause; i.e., 1) women with known age at natural menopause; and 2) women with bilateral oophorectomy (Analysis 2, n=926). We conducted these analyses with SAS software, version 9.1 (The SAS System for Windows. Cary, NC: SAS Institute, Inc., Copyright 2003).
Among subjects invited to participate in the original study, 76% of cases and 70% of controls agreed. Among the 1,179 postmenopausal women included in these analyses, as shown in Table 1, there was a significant difference in age between the 576 cases and 603 controls. Cases and controls did not, however, differ on race, marital status or education. Similar to what would be expected among women with invasive breast cancer, fewer pregnancies, no births or older age at birth of first child, breast cancer in a first degree relative, history of breast biopsy, and a greater number of mammography screenings were associated with increased risk of DCIS. All covariates in Table 1 that were significantly associated with the outcome were included in the multivariable models, coded as shown in that table.
Table 2 presents the adjusted ORs and CIs for associations between HT and DCIS. Among all postmenopausal women (Analysis 1), in relation to women who had never used HT, we found no evidence of increased risk of DCIS with ever use of any HT preparations (adjusted OR: 0.85, 95% CI: 0.65-1.11). For estrogen alone, the adjusted OR was 0.93 (95 % CI: 0.68-1.29), while for those who had used estrogen and progesterone, the adjusted OR was 0.75 (95% CI: 0.52-1.08). Moreover, neither continuous nor sequential use of combined estrogen and progesterone had a significant impact. In addition, among current users of these HT preparations, we found no increased risk of DCIS: adjusted OR = 0.87; 95 % CI: 0.64-1.18 (any HT); adjusted OR = 0.98; 95% CI: 0.66-1.41 (estrogen only); adjusted OR=0.78; 95 % CI: 0.52-1.16 (estrogen and progesterone). In addition, we found no significant associations between DCIS and any category of duration of use of: 1) any HT; 2) estrogen; and 3) combined estrogen and progesterone (Table 2). Tests for trend by duration of any HT, estrogen only, and estrogen and progesterone use were also not significant. In addition, the interactions of HT and each category of BMI were non-significant.
When we repeated these analyses on the subset of women with know age at menopause (Table 2, Analysis 2), we found no association between any form of HT use and DCIS, whether examined as ever use, as current use, or by duration of use.
We found that use of any HT, estrogen alone, and progesterone and estrogen combined were not associated with DCIS among ever users or current users. In addition, none of the categories of duration of use of these hormones were significantly associated with DCIS, nor was there evidence of increased risk with longer duration of use. This held for the analysis that included all women, as well as that restricted to women with known age at menopause.
Notwithstanding the relatively large sample size of this case control study, power was limited for some of the analyses. In particular, though current use of HT has been more strongly associated with breast cancer than ever use , we had less power to examine current than ever use in our study. An additional limitation is that we gathered hormone data by self-report. One[L3] study found good agreement between self reported HT data (obtained from a mailed questionnaire) and prescription information, though the questionnaire under-estimated short-term use . However, non-differential misclassification and recall bias are always potential problems in studies where data is collected retrospectively and by self-report. Our first analysis included women with unknown age at menopause for whom we imputed age values, based on widely used conventions [18, 21]. However, some have raised concerns that imputing age may produce biased estimates [15, 22]. Thus, a strength of this study is the conduct of a second analysis that included only those women with known age at menopause. Further strengths of this work include the use of population-based cases and controls, relatively high participation among invited subjects, and detailed assessment of hormone use. In addition, we were able to obtain information on a range of covariables that could impact the association between HT and DCIS, notably time since menopause and use of mammography screening.
We had fairly detailed information on screening behavior: i.e., whether women were not receiving screening mammograms or were receiving mammograms: 1) less often than once a year; 2) once a year; or 3) more often than once per year. We found that this mammography screening variable had more impact on the relationship between HT and DCIS than any covariable in the dataset, reducing the estimated risk by nearly 11%. Thus, studies that are unable to control for mammography screening, or do so with sufficient precision (e.g., ever had a screening mammogram) may inflate the impact of postmenopausal hormone use. Because in situ cancers that might otherwise go undetected can be found on screening , controlling for this variable may be even more important in studies of in situ than invasive breast cancer.
We offer these results as part of a small, but developing, literature on HT and non-invasive breast cancer. We located fifteen studies [4, 11, 15, 24-35] that examined this association (Table 3). These include one randomized trial , seven prospective studies [11, 26, 28, 30-33], and seven case control studies [4, 15, 24, 27, 29, 34, 35]. The one clinical trial, an analysis of Women’s Health Initiative data, did not find an association between combined estrogen and progesterone use and DCIS, but few DCIS cases were included (n=84) . The larger prospective studies by Reeves and colleagues with 1,913 BCIS cases ; Kerlikowske and colleagues with 583 DCIS cases ; and Lyytinen et al. with 404 in situ cases , all found that HT was associated with BCIS. Reeves found a stronger risk of DCIS, and especially of LCIS, with current than past use. The Lyyntinen group found increasing risk with greater duration of use, with a standardized incidence ratio (SIR) of 2.28 after 10 or more years. However, the latter study was not able to control for important potential confounders, including mammography screening. The Kerlikowske study was based on mammography registry data and accounted for time between screenings, while the Reeves study, based on data from the Million Women Study, recruited participants at the time they[L4] attended a routine screening.
The case control studies that assessed the relationship between HT use and BCIS did adjust for multiple covariables, though they varied in which covariables were included, and in the precision of covariable measurement. Four of the seven case control studies showed modestly elevated risk with estrogen and/or progesterone use in some analyses [4, 15, 24, 29], usually stronger with current use or longer duration, though one of these studies found decreasing risk with longer use . Of these four studies, one adjusted for number of mammograms , one included a less precise measure of screening (ever use) , one did not appear to have adjusted for screening , and one included participants in the Breast Cancer Detection Demonstration Project and did account for time in the program . Three case control studies had negative findings, though two of these studies included very few subjects with in situ disease [27, 34]. However, the present analysis, which included more in situ cases (n=573) than the prior case control studies we identified, also yielded a null result.
Differences in exposure measures may have contributed to inconsistent findings across studies. For example, risk may differ for estrogen only and estrogen-progesterone regimens . Combining all HT medications in one analysis, as did some investigators [4, 26, 31, 32], may have obscured differences in the effect of these preparations on BCIS. Our study did examine use of any HT, but we also examined estrogen alone and combined estrogen and progesterone use. Although not significant, the impact of type of HT on DCIS did differ in our study. However, across studies, even when examined by type of hormone, results varied. Some investigators found increased risk with estrogen only [15, 24, 29, 30] whereas other did not [28, 34]. Results of studies that examined combined estrogen and progesterone use were similarly inconsistent. This suggests that it may not be enough to examine specific hormones; more refined analyses may be needed. For example, studies of breast cancer and HT conducted in Finland  and Germany  suggest that different types of progestins used in combined regimens may impact risk differently. Among subjects in our study, medroxyprogesterone acetate was most frequently prescribed, but others were also used. With regard to estrogens, subjects in our study reported use of conjugated equine estrogens and esterified estrogens, among others. Not distinguishing among these forms of estrogen and progesterone may have obscured some differences in risk of DCIS by HT regimen within our study and may have also contributed to differences across studies, especially when comparing findings conducted across countries where commonly prescribed regimens may differ. For example, though medroxyprogesterone acetate was the most common progestin reported in our study conducted in the United States, norethisterone acetate was more common in the Finnish study noted above. Large studies of in situ breast cancer with sufficient power will be needed to under take analyses that can meaningfully assesses risk of BCIS associated with specific forms of estrogen and progesterone.
Pattern of use may also impact risk and should be considered. As reviewed by Lyytien , several, but not all, investigators found that continuous use of progesterone in combined regimens put women at higher risk of invasive breast cancer than sequential use. Only our study and a study by Ross et al.  report on continuous and sequential hormone use in relation to in situ cancers, but small numbers prohibit drawing definitive conclusions. An additional factor to consider is that HT dosage and preparation may impact BCIS risk. Since the Women’s Health Initiative finding were made public in 2002, there has been a decrease in overall HT use and in standard dose oral HT use, but there has been an increase in lower dose oral medications and an increase in use of vaginal preparations . These products may impact BCIS differently, but the studies reviewed here did not report on, and presumably did not have the data and/or power to conduct, such analyses, nor did we. Future studies will need to account for these changes in prescribing patterns over time.
Across studies, there were also differences in the outcomes investigated. The majority reported on non-specific BCIS [11, 15, 24, 25, 27, 29, 30, 32-34], whereas others reported on DCIS [26, 28, 31, 35] or LCIS [4, 31] separately. Because there is some evidence that HT may be more strongly associated with lobular than ductal lesions in invasive disease [14, 38, 39], and in situ cancers , reporting on BCIS without examining specific histology may have missed important distinctions. Our study’s focus specifically on DCIS is an improvement over the more generic BCIS studies. Yet, it may be necessary to even more precisely examine outcome; for example, by DCIS subtype . A recent study by Phillips and colleagues  is, to our knowledge, the first to examine HT use in relation to comedo and non-comedo DCIS, in addition to examining the effect on DCIS overall (shown in Table 3) and on invasive breast cancer. Among the subset of postmenopausal women in that study, the impact of HT on DCIS did not differ by DCIS subtype, but numbers of women in each group were small. Additional factors to consider in the assessment of BCIS outcomes include the expression of hormone receptors and other biomarkers. We do know that in situ tumors express receptors for estrogen and progesterone , but to our knowledge, the assessment of HT in relation to BCIS by hormone receptor status and other common markers has yet to be undertaken. Breast cancers may include distinct entities that can be differentiated based on specific tumor characteristics, including hormone receptor status, and HT may differentially affect the development of these tumors . Combining, for example, estrogen receptor (ER) positive and ER negative DCIS tumors as one outcome could, potentially, obscure a significant association if HT contributes primarily to the development of DCIS that is ER positive.
In this analysis, we examined HT use during the postmenopausal period, in keeping with prior work and allowing for comparison across studies. However, examining the potential risk of HT at different points in a woman’s life cycle may be indicated. An intriguing study by Shantakumar and colleagues found that premenopausal HT, which may be prescribed to treat a range of gynecological problems, had a greater impact on breast cancer risk (invasive and in situ combined) than postmenopausal use . In a post hoc analysis, compared with the postmenopausal women in this analysis, we, too, found a higher, albeit non-significant, risk of DCIS among women in the original study who had used HT and who were pre- or perimenopausal, or whose menopausal status was unknown. Thus, to fully assess risk, future studies should account for all HT use over a woman’s lifetime.
Future studies that include large numbers of subjects will be needed to examine lifetime hormone use by more precise BCIS outcomes and exposures. In the meanwhile, a careful meta-analysis that thoughtfully examines and parses out the different exposures and outcomes captured by current studies may yield analyses with adequate power to draw more definitive conclusions regarding postmenopausal HT and in situ breast cancers. To ensure that the meta-analysis is free of publication bias, it is critical that both positive and null study results, such as these, are published and made available.
This study was supported by the National Institutes of Health RO1-CA81393 and United States Army Medical Research and Material Command under DAMD-17-95-5006.
The authors greatly appreciate the time and effort of the women who participated in this study and would like to thank Sheila Griffin and Marjorie Jasmin for their work on this project.
We also acknowledge the cooperation of the following Connecticut Hospitals in allowing patient access: Bridgeport Hospital, Bristol Hospital, Charlotte Hungerford Hospital, Danbury Hospital, Day-Kimball Hospital, Eastern Connecticut Health Network, Greenwich Hospital, Griffin Hospital, Hartford Hospital, John Dempsey Hospital, Johnson Memorial Hospital, Lawrence Memorial Hospital, Middlesex Hospital, MidState Medical Center, Milford Hospital, Hospital of Central Connecticut, New Milford Hospital, Norwalk Hospital, Sharon Hospital, St. Francis Hospital and Medical Center, St. Mary’s Hospital, Hospital of St. Raphael, St. Vincent’s Medical Center, Stamford Hospital, Waterbury Hospital, William Backus Hospital, Windham Hospital, Yale-New Haven Hospital.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest statement None of the authors have any competing financial or personal interests that could potentially influence or compromise the work included in this manuscript.