Increased breast carcinoma risk is a major concern for women. Hormone Replacement Therapy in Relation to Breast Carcinoma Incidence Rate Ratios

Transcription

1 2328 Hormone Replacement Therapy in Relation to Breast Carcinoma Incidence Rate Ratios A Prospective Danish Cohort Study Anne Tjønneland, Ph.D. 1 Jane Christensen, M.Sc. 1 Birthe L. Thomsen, M.Sc. 1 Anja Olsen, M.Sc. 1 Kim Overvad, Ph.D. 2,3,4 Marianne Ewertz, D.M.Sc. 5,6 Lene Mellemkjær, Ph.D. 1 1 Institute of Cancer Epidemiology, Danish Cancer Society, Copenhagen, Denmark. 2 Department of Clinical Epidemiology, Aalborg Hospital, Aalborg, Denmark. 3 Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark. 4 Department of Epidemiology and Social Medicine, University of Aarhus, Aarhus, Denmark. 5 Department of Oncology, Aalborg Hospital, Aalborg, Denmark. 6 Danish Breast Cancer Cooperative Group, Copenhagen, Denmark. BACKGROUND. The goal of the current study was to investigate the relation between hormone replacement therapy (HRT) and breast carcinoma in a prospective study cohort. Particular attention was paid to the type of HRT used and to the association of HRT type with estrogen receptor status and tumor histology. METHODS. Between 1993 and 1997, a total of 29,875 women were enrolled in the Danish Cancer Society s prospective Diet, Cancer and Health study. Among 23,618 women who were assumed to be postmenopausal and for whom information on HRT use was available, we identified 423 cases of breast carcinoma over a median follow-up period of 4.8 years. Statistical analyses were based on the Cox proportional hazards model, with age serving as the time parameter. RESULTS. The breast carcinoma incidence rate ratio (IRR) was 2.22 (95% confidence interval [CI], ) for users of HRT at baseline compared with women who never received HRT. Among HRT users (relative to nonusers), the IRR for estrogen receptor positive tumors (2.38; 95% CI, ) was greater than the IRR for estrogen receptor negative tumors (1.56; 95% CI, ). HRT use at baseline also was analyzed in relation to the incidence of lobular carcinoma and the incidence of ductal carcinoma; the adjusted IRR associated with HRT use was 3.53 (95% CI, ) for lobular carcinoma and 2.10 (95% CI, ) for ductal carcinoma. The likelihood of developing estrogen receptor positive breast carcinoma was found to depend significantly on the type of HRT regimen used (P 0.03), with women receiving continuous therapy having the greatest probability of developing estrogen receptor positive disease. CONCLUSIONS. An increased breast carcinoma IRR was found to be associated with current HRT use. In addition, relative to other types of HRT regimens, continuous estrogen progestin regimens were found to be associated with an increased risk of breast carcinoma, and particularly estrogen receptor positive breast carcinoma. Cancer 2004;100: American Cancer Society. Supported by grants from the Danish Cancer Society and the Europe Against Cancer program (European Prospective Investigation into Cancer and Nutrition [EPIC] project). The authors thank Katja Boll and Jytte Fogh Larsen for their assistance with data collection. Address for reprints: Anne Tjønneland, Ph.D., Danish Cancer Society, Institute of Cancer Epidemiology, Strandboulevarden 49, DK-2100 Copenhagen, Denmark; Fax: (011) ; annet@cancer.dk Received December 18, 2003; revision received February 26, 2004; accepted March 3, KEYWORDS: hormone replacement therapy, estrogen receptor status, postmenopausal breast carcinoma, lobular carcinoma. Increased breast carcinoma risk is a major concern for women receiving long-term hormone replacement therapy (HRT). In the recent Million Women Study, 1 it was reported that current HRT users were more likely to develop breast carcinoma than were women who had never received HRT (risk ratio, 1.66; 95% confidence interval [CI], ). This study also found that current use of estrogen progestin combinations was associated with a breast carcinoma risk that was substantially greater than the risk associated with other HRT regimens. The findings of the Million Women Study are in agreement with 2004 American Cancer Society DOI /cncr Published online 23 April 2004 in Wiley InterScience (

2 HRT and Breast Carcinoma/Tjønneland et al a 1997 metaanalysis of 51 studies involving a combined total of 52,705 women with breast carcinoma; this metaanalysis found that breast carcinoma risk increased by 2.3% with each additional year of HRT use. 2 Nonetheless, during the time period encompassed by the metaanalysis, most women received HRT regimens consisting of estrogens alone, and thus the effect of HRT regimen type could not be assessed. Since 1997, several case control studies 3 10 and a limited number of prospective studies have addressed this issue, with all but one study 7 suggesting that estrogen progestin regimens are associated with an increased breast carcinoma risk. This finding is supported by the recently published results of two large interventional trials. 14,15 Some questions regarding the association between HRT and tumor type remain unanswered. In a study of breast carcinoma prognosis among HRT users, Delgado and Lubian Lopez 16 reported that compared with other women, HRT users were more likely to present with estrogen receptor positive disease. In addition, it has been suggested that HRT use increases the incidence of lobular carcinoma more than it increases the incidence of ductal carcinoma or other histologic disease types. 3,4,17,18 According to recent studies, the incidence of lobular carcinoma has increased at a greater rate compared with the incidence of ductal carcinoma in many Westernized countries. 19,20 In the Danish Cancer Society s prospective Diet, Cancer and Health study, we had the opportunity to investigate the relation between HRT and breast carcinoma risk. In addition, we were able to analyze the associations of HRT regimen type with estrogen receptor status and tumor histology for breast carcinoma cases diagnosed between 1993 and MATERIALS AND METHODS Between December 1993 and May 1997, 79,729 women were invited to participate in the prospective Diet, Cancer and Health study. We invited all women in the greater Copenhagen and Aarhus areas who were ages years, born in Denmark, and not previously registered in the Danish Cancer Registry as having a diagnosis of malignant disease. Each cohort member was identified by the unique personal identification number (consisting of the member s date of birth together with a four-digit number) allocated by the Central Population Registry of Denmark. A total of 29,875 women, corresponding to 37% of all women invited and to 7% of the entire Danish female population in this age group, were enrolled in the current study. Linkage of records to the Danish Cancer Registry revealed a slightly higher incidence of breast carcinoma among study participants compared with the entire Danish population. 21 This finding reflects the fact that participants were recruited from the two largest cities in Denmark as well as the fact that women with higher incomes and more education had slightly higher participation rates than did other women. 22 Both the Diet, Cancer and Health study and the current substudy were approved by the regional ethics committees on human studies in Copenhagen and Aarhus and by the Danish Data Protection Agency. Informed consent was obtained from all participants. Participants completed a detailed, 192-item food frequency questionnaire, which they received by mail before visiting one of the two study clinics in Copenhagen and Aarhus. Details regarding the development and validation of the questionnaire have been reported elsewhere. 23 On their study clinic visits, participants completed a lifestyle questionnaire, which included questions regarding reproductive characteristics and lifestyle habits. The two self-administered questionnaires were processed via optical scanning and checked for missing information while participants were present at the study clinic, so that direct clarification could be obtained from the participant. A limited number of missing responses were allowable in the background questionnaire, but not in the dietrelated questionnaire. Anthropometric measurements, including participant height and weight, were obtained by professional staff members at the two study clinics. In the lifestyle questionnaire, participants were asked about their use of HRT at baseline (never received HRT, previously received HRT, orcurrently receive HRT), the age at which they began receiving HRT, the duration (in years) of their HRT use, and the brand of HRT product currently used. Participants also were asked whether they previously had received other types of HRT (but not other brands of HRT products). Triers were defined as former HRT users with 6 months of use. Time since HRT cessation was calculated using the participant s age at the start of HRT use and the reported duration of use. Based on the self-reported brand of HRT product used, the HRT regimen was coded either as unopposed estrogen or as an estrogen progestin combination. Estrogen progestin combinations were further classified as either continuously or sequentially administered regimens. The lifestyle questionnaire also provided information regarding duration of schooling (short [ 7 years], medium [8 10 years], or long [ 10 years]), parity, age at first birth, history of hysterectomy or oophorectomy, and history of surgery for benign breast lesions (yes or no). The food frequency questionnaire pro-

3 2330 CANCER June 1, 2004 / Volume 100 / Number 11 vided data on current alcohol consumption. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m 2 ). A total of 326 women (1.1%) who were found to have had a malignancy registered in the Danish Cancer Registry before study entry were excluded from the analysis. In addition, eight women were excluded from the main data set because they chose not to respond to significant portions of the lifestyle questionnaire. Because analyses were restricted to postmenopausal women, 4798 premenopausal women (i.e., women who reported no HRT use and at least 1 instance of menstruation 12 months before study entry) were excluded from the study, as were the 9 women who reported a lifetime history of no menstruation and the 37 women who did not respond to the portion of the questionnaire dealing with current HRT use. Among the remaining 24,697 women who were assumed to have postmenopausal status, we excluded 715 women (2.9%) for whom data on the duration of HRT use or the time since HRT cessation were unavailable, as well as 364 women (1.5%) for whom information on 1 or more of the other confounders was unavailable. These exclusions left 23,618 postmenopausal women (95.6% of all postmenopausal women considered) for analysis in the current study. Cohort members were linked via their personal identification numbers to other registries. Information on vital status and migration was obtained from the Central Population Register. Information on the occurrence of malignant disease among cohort members was obtained through the linkage of records to the Danish Cancer Registry, which collects information on all Danish inhabitants with malignant disease. Data on histologic disease type also were retrieved from the files of the Danish Cancer Registry. Histologic disease types were distributed as follows: ductal carcinoma (International Classification of Diseases for Oncology [1st edition; ICDO-1] codes and 85213), n 299; lobular carcinoma (ICDO-1 code 85203), n 66; other histologic types, n 26; and unknown histology, n 16. No cases of mixed infiltrating ductal and lobular carcinoma (ICDO-1 code 85223) were reported. Using personal identification numbers, the study cohort also was linked to the database of the Danish Breast Cancer Cooperative Group (DBCG), which has coordinated clinical trials involving patients with breast malignancies since The DBCG holds records on a wide range of characteristics, including tumor size (i.e., largest measured tumor diameter [mm]), lymph node metastases, and estrogen receptor status, for approximately 90% of all breast malignancies diagnosed in Denmark. 24 Malignancies were classified as having positive estrogen receptor status if immunohistochemical staining revealed 10% estrogen receptor positivity. This method of classification suggests that in the current context, negative estrogen receptor status should be considered indicative of an estrogen receptor poor tumor, rather than a tumor in which estrogen receptor expression is completely absent. Data on progesterone receptor status were not available. Each cohort member was followed for breast carcinoma detection from the date of study entry (i.e., the date of the participant s first study center visit) until whichever of the following dates came first: 1) the date of diagnosis of any malignancy (except for nonmelanoma skin malignancy); 2) the date of death; 3) the date of emigration; or 4) December 31, Associations of the age-specific breast carcinoma incidence rate with various exposure variables were analyzed using a Cox proportional hazards model incorporating time-dependent variables, with age serving as the time axis to ensure that all analyses were based on comparisons involving individuals of the same age. Time on study was included as the timedependent variable and was modeled by a linear spline 25 with a boundary at 1 year after entry into the study cohort. All models were adjusted for baseline values of known breast carcinoma risk factors, including parity (entered as a categoric variable [nulliparous or parous] and also as a linear variable [number of births]), age at first birth (linear variable), history of benign breast tumor surgery (yes or no), duration of schooling (short, medium, or long), alcohol consumption (linear variable [g per day]), and BMI (linear variable). In the analyses involving estrogen receptor status, estrogen receptor positive disease and estrogen receptor negative disease were treated as independent, competing causes of failure (i.e., each type of disease was subjected to a separate analysis in which patients whose tumors had the opposite estrogen receptor status were censored at the age of diagnosis). Patients whose tumors had unknown estrogen receptor status were censored at the age of diagnosis in both analyses. Likelihood ratio tests were performed to compare associations between receptor status specific outcomes and various exposure variables. These tests compared the sum of the partial likelihoods obtained in the two separate analyses with the partial likelihood obtained using underlying receptor status specific rates but using only common associations with exposure variables. The two histologic types of breast carcinoma that were analyzed also were treated as independent, competing causes of failure and were compared using the likelihood ratio test statistic. One goal of the current study was to investigate

4 HRT and Breast Carcinoma/Tjønneland et al the effects of different types of HRT on breast carcinoma risk. This investigation included pairwise comparison of the different HRT regimens to estimate risk ratios using a model involving separate, mutually adjusted associations with the three types of HRT. In practice, all of these estimates were obtained by applying the same statistical model but using different reference categories (corresponding to the different types of HRT). The distribution of women undergoing hysterectomy among the three HRT subgroups is highly skewed, with a total of only 186 hysterectomized women (accounting for 10 breast carcinoma cases) receiving progestin as a part of HRT; thus, the inclusion of hysterectomy as a confounder would cause the estimated associations of breast carcinoma risk with HRT type to reflect differences among nonhysterectomized women only and would affect the estrogenbased HRT subgroup almost exclusively. In the comparison of HRT types, we chose to exclude hysterectomized women who received progestin-containing HRT. For the remaining women, history of hysterectomy and HRT type were used to define four mutually exclusive categories: hysterectomized users of unopposed estrogen, nonhysterectomized users of unopposed estrogen, nonhysterectomized users of sequential estrogen progestin, and nonhysterectomized users of continuous estrogen progestin. Our comparison of the different HRT types involved only the three classes of nonhysterectomized participants, such that the confounding effects of hysterectomy did not influence the analysis. The overall assessment of the effect of HRT type consisted of a likelihood ratio test comparing the aforementioned three groups. Two-sided 95% CIs for the estimated incidence rate ratios (IRRs) were calculated using the Wald test of the Cox regression parameter (i.e., on the log [rate ratio] scale). The PHREG procedure, part of the SAS software package (Version 6.12; SAS Institute, Cary, NC), was used for statistical analyses. We also compared the disease characteristics (namely, tumor size and lymph node involvement) of HRT users at baseline with the corresponding characteristics among participants who had never received HRT. These comparisons were made via the Wilcoxon two-sample test using the NPAR1WAY procedure, which also is contained in the SAS software package. RESULTS Table 1 presents the distribution of known breast carcinoma risk factors among nonusers, triers, past users, and current users of HRT at baseline. Compared with individuals who had never received HRT, current HRT users had more schooling, were more likely to have had a benign tumor removed, had a higher level of alcohol consumption, and were more likely to have undergone hysterectomy. A total of 423 cases of breast carcinoma were identified among 23,618 postmenopausal women over a median follow-up period of 4.8 years. Breast carcinoma IRRs according to HRT use at baseline are summarized in Table 2. Adjustment for known risk factors had little effect on the estimated associations. Although they were not significant, the associations of breast carcinoma risk with duration of HRT and years since cessation exhibited the expected directionality; a risk increase of 7% was associated with every 5 years of HRT use, whereas a risk reduction of 5% was associated with every 5 years that had passed since HRT cessation. When these variables were excluded from the analysis, current users at baseline were found to have a significantly increased breast carcinoma risk (IRR, 2.22; 95% CI, ) compared with participants who had never used HRT. Neither exclusion of cases diagnosed within 1 year of recruitment nor exclusion of cases diagnosed within 2 years of recruitment had a significant effect on the results that were obtained (results not shown). Analysis of HRT use at baseline in relation to estrogen receptor specific disease revealed that relative to participants who had never received HRT, current HRT users had significantly increased risks for both types of breast carcinoma. The IRR associated with current HRT use was higher for estrogen receptor positive disease (IRR, 2.38; 95% CI, ) than for estrogen receptor negative disease (IRR, 1.56; 95% CI, ; P 0.24) (Table 3). When HRT use at baseline was evaluated in relation to lobular carcinoma and ductal carcinoma as competing outcomes, the adjusted IRR associated with current HRT use was found to be somewhat higher for lobular carcinoma (IRR, 3.53; 95% CI, ) than for ductal carcinoma (IRR, 2.10; 95% CI, ), but the difference was not significant (P 0.46). For former HRT users ( 6 months of HRT use), the estimated IRR for lobular breast carcinoma was 2.71 (95% CI, ), compared with 1.16 (95% CI, ) for ductal carcinoma. Finally, although the observed risk elevation was not significant, even former short-term HRT users (i.e., triers) may have an increased risk of developing lobular carcinoma (IRR, 1.66; 95% CI, ). A close relation between hysterectomy and HRT type was noted in the current study; 67% of current unopposed estrogen users had undergone hysterectomy, compared with only 3.8% (n 124) of current sequential HRT users and only 4.6% (n 62) of current continuous HRT users. Among HRT users as a

5 2332 CANCER June 1, 2004 / Volume 100 / Number 11 TABLE 1 Characteristics of 23,618 Postmenopausal Women Enrolled in the Danish Diet, Cancer and Health Study ( ) Lifetime HRT use Characteristic Never used Tried HRT a Previously used b Currently use No. of women (%) 11,801 (50.0) 1929 (8.2) 1641 (6.9) 8247 (34.9) Age at entry (yrs) Mean th 95th percentile range Duration of schooling (%) Short 4203 (35.6) 708 (36.7) 633 (38.6) 2273 (28.8) Medium 5683 (48.2) 939 (48.7) 776 (47.3) 4252 (51.6) Long 1915 (16.2) 282 (14.6) 232 (14.1) 1622 (19.7) Body mass index (kg/m 2 ) Median th 95th percentile range Nulliparity (%) 1466 (12.4) 219 (11.4) 211 (12.9) 1028 (12.5) No. of children Median th 95th percentile range Age at first birth (yrs) Median th 95th percentile range History of benign breast tumor (%) 1407 (11.9) 257 (13.3) 276 (16.8) 1175 (14.25) Alcohol consumption (g/day) Median th 95th percentile range Hysterectomy (%) 1522 (12.9) 320 (16.6) 311 (19.0) 1992 (24.2) Duration of HRT use (yrs) Median th 95th percentile range Current HRT type (%) Estrogen 2447 (35.8) Sequential estrogen/progestin 3119 (45.6) Continuous estrogen/progestin 1278 (18.7) Time since cessation of HRT (yrs) Median th 95th percentile range HRT: hormone replacement therapy. a Women who previously used hormone replacement therapy for 6 months. b Women who previously used hormone replacement therapy for 6 months. whole, hysterectomy was found to be associated with an increased breast carcinoma incidence rate (IRR, 1.20; 95% CI, ) after adjustment for time on study, education, BMI, parity, age at birth of first child, history of benign breast tumor surgery, alcohol consumption, and HRT type. The IRR associated with hysterectomy among users of unopposed estrogen was 1.61 (95% CI, ). Table 4 summarizes the relation between breast carcinoma IRR and HRT type among current HRT users who had not undergone hysterectomy. The three HRT types were not significantly different from each other (P 0.08) in terms of IRR, but there was a trend toward increased risk among users of continuous HRT compared with users of sequential HRT, as well as a trend toward decreased risk among users of unopposed estrogen compared with users of each of the other two HRT types. When the analyses were restricted to women who had only ever used a single brand of HRT product, the estimated IRRs were not substantially altered (results not shown). With respect to estrogen receptor positive breast carcinoma risk (Table 4), the three HRT types were significantly different from each other (P 0.03); compared with sequential therapy and unopposed estrogen therapy, continuous therapy was associated with an increased risk of developing estrogen receptor positive disease. No significant difference among HRT regimens in terms of estrogen receptor negative (P 0.66), lobular (P 0.27), or ductal (P 0.27) breast carcinoma risk

6 HRT and Breast Carcinoma/Tjønneland et al TABLE 2 Breast Cancer Incidence Rate Ratios Estimated Using a Cox Proportional Hazards Model with Age Serving as the Time Parameter, with Adjustment for Time on Study HRT use Cases/entire cohort IRR (95% CI) a IRR (95% CI) a,b IRR (95% CI) a,b Never used HRT 144/11, Tried HRT c ( 6 mos) 24/ ( ) 1.08 ( ) 1.01 ( ) Previously used HRT d 28/ ( ) 1.33 ( ) 1.35 ( ) Currently use HRT e 227/ ( ) 2.03 ( ) 2.22 ( ) Risk per 5 yr period of HRT use 1.07 ( ) 1.07 ( ) Risk per 5 yr period since cessation of HRT 0.95 ( ) 0.95 ( ) IRR: incidence rate ratio; CI: confidence interval; HRT: hormone replacement therapy. a Presented estimates are mutually adjusted within each column. b Incidence rate ratios adjusted for duration of schooling (short, medium, or long), body mass index (linear variable), parity/nulliparity, number of births (linear variable), age at birth of first child (linear variable), history of benign breast tumor surgery (yes or no), and alcohol consumption (linear variable). c When adjusted for duration of therapy and time since last use, the presented rate ratios compare triers who stopped receiving hormone replacement therapy at least 1 year before baseline with participants who never received hormone replacement therapy. d When adjusted for duration of therapy and time since last use, the presented rate ratios compare past users who stopped receiving hormone replacement therapy at least 1 year before baseline with participants who never received hormone replacement therapy. e When adjusted for duration of therapy and time since last use, the presented rate ratios compare current users registered at baseline with participants who never received hormone replacement therapy. TABLE 3 Breast Carcinoma Incidence Rate Ratios by Estrogen Receptor Status and Histologic Type for Nonusers, Triers, Past Users, and Current Users of Hormone Replacement Therapy Cases/entire cohort Unadjusted IRR (95% CI) a Adjusted IRR (95% CI) b Estrogen receptor status Positive 301/23,628 Nonusers 98/11, Triers 19/ ( ) 1.17 ( ) Past users 19/ ( ) 1.35 ( ) Current users 165/ ( ) 2.38 ( ) Negative 91/23,618 Nonusers 38/11, Triers 3/ ( ) 0.47 ( ) Past users 8/ ( ) 1.43 ( ) Current users 42/ ( ) 1.56 ( ) Histologic type Lobular carcinoma 66/23,618 Nonusers 15/11, Triers 4/ ( ) 1.66 ( ) Past users 6/ ( ) 2.71 ( ) Current users 41/ ( ) 3.53 ( ) Ductal carcinoma 299/23,618 Nonusers 109/11, Triers 14/ ( ) 0.77 ( ) Past users 18/ ( ) 1.16 ( ) Current users 158/ ( ) 2.10 ( ) IRR: incidence rate ratio; CI: confidence interval; HRT: hormone replacement therapy. a Adjusted for time on study. b Adjusted for duration of schooling (short, medium, or long), body mass index (linear variable), parity/nulliparity, number of births (linear variable), age at birth of first child (linear variable), history of benign breast tumor surgery (yes or no), alcohol consumption (linear variable), age, and time on study. was noted. Further classification of estrogen receptor positive tumors into ductal and lobular subgroups revealed that continuous HRT was associated primarily with estrogen receptor positive ductal carcinoma (P 0.07; data not shown). Unadjusted IRRs are presented in Table 4, as there were no major differences between the unadjusted and adjusted results obtained from these analyses. We also compared the tumor characteristics of current HRT users at baseline with the corresponding

7 2334 CANCER June 1, 2004 / Volume 100 / Number 11 TABLE 4 Breast Carcinoma Incidence Rate Ratios a Estimated Using a Cox Proportional Hazards Model with Age Serving as the Time Parameter, with Adjustment for Time on Study, According to Hormone Replacement Therapy Type at Baseline Unadjusted IRR (95% CI) Type of breast carcinoma Continuous vs. sequential Unopposed estrogen vs. sequential Unopposed estrogen vs. continuous P b Any type (n 186) 1.23 ( ) 0.64 ( ) 0.52 ( ) 0.08 Estrogen receptor positive (n 137) 1.42 ( ) 0.57 ( ) 0.40 ( ) 0.03 Estrogen receptor negative (n 34) 0.67 ( ) 0.70 ( ) 1.05 ( ) 0.66 Lobular (n 39) 1.06 ( ) 0.37 ( ) 0.35 ( ) 0.27 Ductal (n 126) 1.32 ( ) 0.79 ( ) 0.60 ( ) 0.27 IRR: incidence rate ratio; CI: 95% confidence interval. a Analysis excludes women who used other unspecified hormone replacement therapy products (n 1217), women for whom no information on hormone replacement therapy type was available (n 64), and women who had undergone hysterectomy (continuous therapy group, n 62; sequential therapy group, n 124). b Probability (likelihood ratio test with two degrees of freedom) that there is no difference in breast carcinoma incidence rates among the three groups analyzed. TABLE 5 Tumor Characteristics among Nonusers and Current Users of Hormone Replacement Therapy Characteristic Never used HRT use Currently use Tumor size No. of women in analysis a Median size (mm) th 95th percentile range (mm) % of women with tumor size 20 mm Lymph node involvement No. of women in analysis b No. of removed lymph nodes Median th 95th percentile range % of women with positive lymph nodes No. of positive lymph nodes Median 3 2 5th 95th percentile range Ratio of positive lymph nodes to removed lymph nodes Median th 95th percentile range HRT: hormone replacement therapy. a Twenty-eight women had missing information on tumor size. b Twenty-four women had missing information on removed lymph nodes or did not have lymph nodes removed. characteristics among participants who had never received HRT (Table 5). Median tumor size was slightly smaller in the former group (15 mm) compared with the latter group (19 mm); however, this difference was not statistically significant (P 0.15). Likewise, the ratio of the number of positive lymph nodes to the total number of removed lymph nodes was lower among current HRT users (0.15) compared with women who had never received HRT (0.18); this difference also was nonsignificant (P 0.09). P DISCUSSION In the current prospective investigation of 423 breast carcinoma cases found among 23,618 postmenopausal women, a significantly increased risk of breast carcinoma was found to be associated with current HRT use at baseline. This risk increase was evident for both estrogen receptor positive and estrogen receptor negative tumors but was greater for estrogen receptor positive tumors. Similar results were noted with respect to lobular and ductal malignancies, with the increase in lobular carcinoma risk being greater than the increase in ductal carcinoma risk. The risk of developing estrogen receptor positive breast carcinoma was significantly greater for current users of continuous estrogen progestin therapy compared with users of sequential therapy and users of unopposed estrogen. The observed differences in terms of tumor size and lymph node involvement between current HRT users and women who had never received HRT were small and nonsignificant, but as expected, current HRT users tended to have smaller tumors and less lymph node involvement. In the current study, data on HRT use were collected prospectively, thereby minimizing the possibility of differential recall bias. To our knowledge, relatively few prospective studies addressing the relation between HRT use and breast carcinoma have been published to date. One potential limitation of the current study was that the accuracy of the data depended on the ability of participants to recall details regarding the timing and duration of their HRT use. In addition, our analysis of the association between breast carcinoma risk and HRT type was based on the participantreported brand of HRT product used at baseline; consequently, the results of this analysis do not reflect previous use of different HRT types. Nonetheless,

8 HRT and Breast Carcinoma/Tjønneland et al analyses restricted to women who had only ever used one brand of HRT product did not yield substantially different results compared with analyses involving all women who had ever received HRT. Another potential weakness of the current study was that the relatively small number of breast carcinoma cases and the imprecision associated with the exposure variable measurements may have combined to limit our ability to detect significant associations in the subgroup analyses. Our results are consistent with the results of other studies, although it is possible that the observed associations are attributable to the characteristics of hormone users rather than to hormone use itself. Nonetheless, our finding of an overall increase in breast carcinoma risk among current HRT users is consistent with the results of a number of case control and cohort studies, 1,3 13 as well as one collaborative analysis. 2 It should be noted that some of these previous studies were unable to distinguish among different HRT regimens. The analysis of estrogen receptor positive breast carcinoma rates revealed an increased risk associated with continuous estrogen progestin therapy relative to other HRT types; this result indicates that continuous administration of progestins may be more hazardous than sequential administration. In support of this hypothesis, it has been shown in experimental studies that progestin has an impact on apoptosis and that discontinuation of progestin triggers the apoptotic mechanism; these findings may explain the increased risk associated with continuous therapy relative to sequential therapy, which more closely mimics the natural hormone cycle. In a recent study involving a population-based Swedish cohort, Olsson et al. 13 reported an IRR of 2.45 (95% CI, ) for women who had ever received continuous HRT, compared with 1.22 (95% CI, ) for women who had ever received sequential therapy; women who had never received HRT represented the referent group for both of these IRRs. Estrogen receptor status was not considered in this Swedish cohort study. In another recent case control study, Li et al. 29 reported that the use of estrogen progestin was associated with an increase in the risk of developing estrogen receptor positive/progesterone receptor positive breast carcinoma. We are unaware of any study, other than the current one, investigating the association between estrogen progestin HRT administration (continuous vs. sequential) and breast carcinoma estrogen receptor status. The association of increased breast carcinoma risk with current HRT use at baseline was considerably stronger for lobular breast carcinoma than for ductal breast carcinoma; among current HRT users (relative to women who had never received HRT), the IRR for lobular disease (3.53) was almost twice the corresponding IRR for ductal disease (2.10). This finding is in agreement with the findings of a number of recent studies. 3,4,12,17 19,22,29 In the record-linkage study conducted by Chen et al., 12 the estimated odds ratio for lobular carcinoma was found to be 3.1 times the estimated odds ratio for nonlobular disease (3.91 vs. 1.25) when current combination therapy users were compared with women who had never received HRT. Similarly, in a Swedish cohort study, 22 the ratio of the reported IRRs for lobular carcinoma and ductal carcinoma was 3.5 (4.38 vs. 1.25) when current HRT users (all HRT types) were compared with nonusers. In a case control study, Li et al. 19 reported an odds ratio for lobular carcinoma that was 3.7 times the odds ratio for ductal carcinoma (2.6 vs. 0.7) when current users of combination therapy were compared with women who had never received HRT. Newcomb et al., 4 comparing women who had ever used estrogen progestin with women who had never received this type of HRT, found an IRR for lobular carcinoma that was 1.4 times the corresponding IRR for ductal carcinoma (2.01 vs. 1.43). Finally, Daling et al. 3 reported that the ratio of the IRRs for lobular breast carcinoma and ductal breast carcinoma was 1.8 (2.4 vs. 1.3) when current users of continuous HRT were compared with women who had never received HRT. These findings regarding lobular breast carcinoma risk, together with the increasing use of combination therapy, 2 are consistent with reports from a number of Western nations that lobular breast carcinoma incidence is increasing at a greater rate compared with ductal breast carcinoma incidence among women age 50 years. Taken together, these observations suggest that this more rapidly increasing lobular carcinoma incidence rate may be attributable to HRT use. 19,20 Compared with ductal tumors, lobular tumors are more likely to have positive estrogen receptor and progesterone receptor status and therefore may have a stronger association with HRT use. 12 Biologically, there is reason to suspect that HRT may have different effects with respect to the risks of developing different histologic types of breast carcinoma. Breast carcinomas originate from the cells that line the duct lobular units. The lobulus can be classified based on development and differentiation. Type 1 lobuli, which are found predominantly in postmenopausal women, are undifferentiated and represent the site of origin for ductal carcinomas. 30 Hormonal stimulation with combination HRT may cause type 1 lobuli to become more differentiated type 2 lobuli, with this

9 2336 CANCER June 1, 2004 / Volume 100 / Number 11 change leading to an increased risk of lobular disease, which develops from type 2 lobuli. 3 We also identified hysterectomy as a risk factor (after adjustment for HRT type) for current HRT users. The majority of previous studies did not identify hysterectomy as a risk factor, although one Swedish study did find that hysterectomy was associated with a slight increase in breast carcinoma risk when the operation was performed for myoma, abnormal bleeding, or (possibly) severe forms of endometriosis. 34 The risk factors associated with myoma (e.g., nulliparity, obesity, and earlier age at menarche) suggest that myoma and breast carcinoma may have related hormonal etiologies. It has been reported that among women with breast carcinoma, previous HRT users have lower mortality rates than do women who have never received HRT 35 ; in addition, HRT users have been reported to have smaller, better-differentiated tumors at diagnosis compared with women who have never received HRT. 16,18 In the current study, we were unable to identify a significant difference in tumor size or number of positive lymph nodes between these two groups; however, as expected, HRT users tended to have smaller tumors and fewer positive lymph nodes. In conclusion, the current prospective study confirms the finding of an increased postmenopausal breast carcinoma risk associated with current HRT use, with more pronounced increases in estrogen receptor positive breast carcinoma risk and lobular breast carcinoma risk. Our findings suggest that continuous estrogen progestin regimens are associated with a greater risk than are unopposed estrogen and sequential estrogen progestin regimens, especially with respect to estrogen receptor positive disease. Additional prospective data are needed to confirm these findings. REFERENCES 1. Beral V. Breast cancer and hormone-replacement therapy in the Million Women Study. Lancet. 2003;362: Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormone replacement therapy: collaborative reanalysis of data from 51 epidemiological studies of 52,705 women with breast cancer and 108,411 women without breast cancer. Lancet. 1997;350: Daling JR, Malone KE, Doody DR, et al. Relation of regimens of combined hormone replacement therapy to lobular, ductal, and other histologic types of breast carcinoma. Cancer. 2002;95: Newcomb PA, Titus-Ernstoff L, Egan KM, et al. Postmenopausal estrogen and progestin use in relation to breast cancer risk. Cancer Epidemiol Biomarkers Prev. 2002;11: Weiss LK, Burkman RT, Cushing-Haugen KL, et al. Hormone replacement therapy regimens and breast cancer risk. Obstet Gynecol. 2002;100: Kirsh V, Kreiger N. Estrogen and estrogen-progestin replacement therapy and risk of postmenopausal breast cancer in Canada. Cancer Causes Control. 2002;13: Li R, Gilliland FD, Baumgartner K, Samet J. Hormone replacement therapy and breast carcinoma risk in Hispanic and non-hispanic women. Cancer. 2002;95: Ross RK, Paganini-Hill A, Wan PC, Pike MC. Effect of hormone replacement therapy on breast cancer risk: estrogen versus estrogen plus progestin. J Natl Cancer Inst. 2000;92: Magnusson C, Persson I, Adami HO. More about: effect of hormone replacement therapy on breast cancer risk: estrogen versus estrogen plus progestin. J Natl Cancer Inst. 2000; 92: Magnusson C, Baron JA, Correia N, Bergstrom R, Adami HO, Persson I. Breast-cancer risk following long-term oestrogenand oestrogen-progestin-replacement therapy. Int J Cancer. 1999;81: Schairer C, Lubin J, Troisi R, Sturgeon S, Brinton L, Hoover R. Menopausal estrogen and estrogen-progestin replacement therapy and breast cancer risk. JAMA. 2000;283: Chen CL, Weiss NS, Newcomb P, Barlow W, White E. Hormone replacement therapy in relation to breast cancer. JAMA. 2002;287: Olsson HL, Ingvar C, Bladstrom A. Hormone replacement therapy containing progestins and given continuously increases breast carcinoma risk in Sweden. Cancer. 2003;97: Hulley S, Furberg C, Barrett-Connor E, et al. Noncardiovascular disease outcomes during 6.8 years of hormone therapy: Heart and Estrogen/Progestin Replacement Study follow-up (HERS II). JAMA. 2002;288: Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women s Health Initiative randomized controlled trial. JAMA. 2002;288: Delgado RC, Lubian Lopez DM. Prognosis of breast cancers detected in women receiving hormone replacement therapy. Maturitas. 2001;38: Li CI, Weiss NS, Stanford JL, Daling JR. Hormone replacement therapy in relation to risk of lobular and ductal breast carcinoma in middle-aged women. Cancer. 2000;88: Manjer J, Malina J, Berglund G, Bondeson L, Garne JP, Janzon L. Increased incidence of small and well-differentiated breast tumours in post-menopausal women following hormone-replacement therapy. Int J Cancer. 2001;92: Li CI, Anderson BO, Daling JR, Moe RE. Trends in incidence rates of invasive lobular and ductal breast carcinoma. JAMA. 2003;289: Verkooijen HM, Fioretta G, Vlastos G, et al. Important increase of invasive lobular breast cancer incidence in Geneva, Switzerland. Int J Cancer. 2003;104: Tjonneland AM, Overvad OK. [Diet, cancer and health a population study and establishment of a biological bank in Denmark]. Ugeskr Laeger. 2000;162: Olsen A, Tjonneland A, Engholm G, Overvad K. Socio-economic determinants for participation in the Danish EPIC Diet, Cancer and Health cohort. IARC Sci Publ. 2002;156:55 57.

10 HRT and Breast Carcinoma/Tjønneland et al Tjonneland A, Overvad K, Haraldsdottir J, Bang S, Ewertz M, Jensen OM. Validation of a semiquantitative food frequency questionnaire developed in Denmark. Int J Epidemiol. 1991; 20: Fischerman K, Mouridsen HT. Danish Breast Cancer Cooperative Group (DBCG). Structure and results of the organization. Acta Oncol. 1988;27: Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology. 1995;6: Sutherland RL, Prall OW, Watts CK, Musgrove EA. Estrogen and progestin regulation of cell cycle progression. J Mammary Gland Biol Neoplasia. 1998;3: Foidart JM, Colin C, Denoo X, et al. Estradiol and progesterone regulate the proliferation of human breast epithelial cells. Fertil Steril. 1998;69: Gompel A, Somai S, Chaouat M, et al. Hormonal regulation of apoptosis in breast cells and tissues. Steroids. 2000;65: Li CI, Malone KE, Porter PL, et al. Relationship between long durations and different regimens of hormone therapy and risk of breast cancer. JAMA. 2003;289: Russo J, Ao X, Grill C, Russo IH. Pattern of distribution of cells positive for estrogen receptor alpha and progesterone receptor in relation to proliferating cells in the mammary gland. Breast Cancer Res Treat. 1999;53: Kreiger N, Sloan M, Cotterchio M, Kirsh V. The risk of breast cancer following reproductive surgery. Eur J Cancer. 1999; 35: Parazzini F, Braga C, La Vecchia C, Negri E, Acerboni S, Franceschi S. Hysterectomy, oophorectomy in premenopause, and risk of breast cancer. Obstet Gynecol. 1997;90: Schairer C, Persson I, Falkeborn M, Naessen T, Troisi R, Brinton LA. Breast cancer risk associated with gynecologic surgery and indications for such surgery. Int J Cancer. 1997; 70: Lindegard B. Breast cancer among women from Gothenburg with regard to age, mortality and coexisting benign breast disease or leiomyoma uteri. Oncology. 1990;47: Jernstrom H, Frenander J, Ferno M, Olsson H. Hormone replacement therapy before breast cancer diagnosis significantly reduces the overall death rate compared with neveruse among 984 breast cancer patients. Br J Cancer. 1999;80: