Hormone Replacement Therapy in Relation to Breast Cancer

Context Studies of long-term hormone replacement therapy (HRT) suggest an associated increased risk of breast cancer, but whether this association differs according to histologic type of cancer has not been extensively studied.

Objective To determine whether the association between HRT and risk of breast cancer varies by HRT formulation and differs across histologic cancer types.

Design, Setting, and Participants Nested case-control study among 705 postmenopausal women enrolled in the Group Health Cooperative of Puget Sound (GHC) who were aged 50 to 74 years and had primary invasive breast cancer diagnosed between July 1, 1990, and December 31, 1995 (cases), and 692 randomly selected aged-matched female members of GHC (controls).

Main Outcome Measure Incidence and type of breast cancer by duration of HRT use in the 5-year period ending 1 year before diagnosis, which was ascertained from computerized pharmacy records.

Results The incidence of breast cancer, all histologic types combined, was increased by 60% to 85% in recent long-term users of HRT, whether estrogen alone or estrogen plus progestin. Longer use of HRT (odds ratio [OR], 3.07 for 57 months or more; 95% confidence interval [CI], 1.55-6.06) and current use of combination therapy (OR, 3.91; 95% CI, 2.05-7.44) were associated with increased risk of lobular breast cancer. Long-term HRT use was associated with a 50% increase in nonlobular cancer (OR, 1.52 for 57 months or more; 95% CI, 1.01-2.29).

Conclusion Our data add to the growing body of evidence that recent long-term use of HRT is associated with an increased risk of breast cancer and that such use may be related particularly to lobular tumors.

The possible association between use of estrogen replacement therapy or combined estrogen-progestin replacement therapy and the incidence of breast cancer has been assessed in numerous studies.^1-28 The Collaborative Group on Hormonal Factors in Breast Cancer pooled and reanalyzed the data from most of these (52 705 women with breast cancer, 108 411 women without breast cancer).²⁹ They reported a modest increase in the risk of breast cancer associated with ever use of estrogen replacement therapy (relative risk [RR], 1.14; P<.001), with evidence of an increasing RR with increasing duration of use (P = .003). The risk of breast cancer was increased among current users (RR, 1.21; P<.001), but not among past users (RR, 1.07; P = .10). They also found that, among women whose duration of current combination therapy was more than 5 years, the risk appeared to be increased relative to never users, but the estimate was imprecise due to small numbers (RR, 1.52; 95% confidence interval [CI], 0.80-2.92). In addition, there is recent evidence that use of hormone replacement therapy (HRT) may differentially affect the incidence of lobular cancer relative to other types of breast cancer.^30,31

We conducted a nested case-control study to examine the relationship between postmenopausal HRT use and risk of breast cancer by histologic type among female enrollees of the Group Health Cooperative of Puget Sound (GHC).

Study subjects were selected from women enrolled in GHC continuously for at least 2 years before diagnosis date (or comparable date for controls). Cases were women aged 50 to 74 years who were newly diagnosed as having a first primary invasive breast cancer between July 1, 1990, and December 31, 1995, identified through the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry. Controls were randomly selected from the enrollment files of GHC during the years the cases were diagnosed and were frequency matched to cases by year of diagnosis, age at diagnosis (5-year intervals), and years of GHC enrollment (5-year intervals). The reference date for each case was 1 year before the breast cancer diagnosis date and a comparable date (June 30 of the matched year) was assigned to each matched control.

A total of 752 cases and 752 controls were identified as potential subjects. We excluded 1 case with adenosquamous carcinoma, 1 with carcinosarcoma, 6 cases and 18 controls whose menopausal status was unknown, and 39 cases and 42 controls who were premenopausal at their reference date. As a result, 705 case and 692 control women remained in the analysis.

The major source of information regarding the use of HRT was the GHC computerized pharmacy database, which began in March 1977. This database contains detailed information about all prescriptions dispensed from GHC pharmacies, including date, drug name, dosage, formulation, pill quantity, and route of administration. Estrogen and progestin oral pills and topical estrogen vaginal cream prescribed at GHC were considered to be HRT in these analyses. Because of the small proportion of women who received HRT either by patch or injection at GHC or progestin cream (<1%), women who used these only were excluded. Since the cost of HRT medications is minimal for GHC members if they receive the medications through a GHC pharmacy, less than 3% of GHC women use non-GHC pharmacies (Katherine Newton, PhD, GHC, oral communication, 1997). Because some women started using HRT before 1977 or before joining GHC, information regarding estrogen use from the GHC Breast Cancer Screening Program (BCSP) questionnaire was also used to further estimate lifetime exposure to HRT. This questionnaire was sent to all female members of GHC aged 40 years or older with an invitation to participate in the BCSP.^32,33 Eighty-five percent of eligible women completed the questionnaire and the information was updated at the time of each mammogram.³³

Two separate analyses of HRT use were conducted. First, lifetime use, including use before the establishment of the pharmacy database, was estimated by combining information from the pharmacy database with information on HRT use obtained from the BCSP questionnaire. For women who had used HRT and stopped before 1977, HRT use and age at first use were determined from the BCSP risk factor questionnaire. All women were included in this analysis. Second, recent use examined use during the 5- and 10-year periods before the reference date. For these analyses, exposure was determined solely from the pharmacy database. Only women who were GHC members for a continuous 5 years immediately before their reference dates were included in the 5-year HRT analysis (553 cases and 551 controls). Similarly, only women enrolled for 10 years were included in the 10-year analysis (428 cases and 427 controls). Those who did not have pharmacy records in the 5- or 10-year period were considered nonusers.

Current use was defined as having at least 2 prescriptions for HRT during the 6-month period immediately before the reference date. This definition was used because a woman with only 1 prescription may have taken only a few or no pills before discontinuing use. Those who had 2 or more HRT prescriptions according to GHC pharmacy records but none within the 6 months before the reference date were defined as past users. Women who did not have pharmacy records but answered positively to questions on estrogen use in the BCSP questionnaire were also classified as past users. All others were defined as nonusers.

Oral HRT use was defined as use of estrogen pills with or without progestin pills. Continuous combined therapy was defined as the same number of estrogen and progestin pills in each prescription. Women who had a different number of estrogen and progestin pills in each prescription (usually 25 estrogen pills with 10 progestin pills) were considered sequential therapy users. Pill counts were converted to estimated months of use.

The cumulative dose of estrogen (or progestin) was obtained by multiplying the quantity of pills by dose per pill for each prescription and then summing across all prescriptions in the 5- or 10-year period. Based on the equivalency table in the article by Lobo,³⁴ all oral estrogens were converted to doses equivalent to conjugated estrogens and all progestin doses to equivalents of medroxyprogesterone acetate.

The information about reproductive factors, self-reported height and weight, and family history of breast cancer was taken from the BCSP questionnaire. Mammography use was ascertained from the BCSP database and women not enrolled in the BCSP were coded as having no prior screening mammograms. Since most of our study subjects were enrolled at GHC for at least 6 years, mammograms before enrollment outside GHC would have occurred early enough to have had little impact on risk.

Information about menstrual status at reference date was derived from the BCSP questionnaire completed before a woman's reference date, 1 completed after the reference date, and from medical records when the questionnaires were not sufficient to determine menopausal status. Women were considered premenopausal (and excluded) if they reported menstrual periods after the reference date and postmenopausal if natural cessation of menses had occurred before the reference date. All other women (including women who had undergone hysterectomy without bilateral oophorectomy or with unknown ovary status and women without BCSP information) were classified as postmenopausal if their age at reference was 55 years or older. Otherwise, if they were younger than 55 years at the reference date, their menstrual status was obtained from medical records (216 women). Age at menopause was defined as the age when periods stopped naturally or age of hysterectomy with bilateral oophorectomy. Use of HRT was not used to determine menopausal status, but for women who had undergone a hysterectomy and who retained at least 1 ovary or whose ovarian status was unknown, age at use of HRT was used as age at menopause.

Histologic information, estrogen-progestin receptor status, and stage of the tumors for all cases were ascertained from the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry. Breast cancer cases were also divided into 2 histologic groups: lobular breast cancer, which included lobular carcinoma (not otherwise specified) and infiltrating duct and lobular carcinoma, and nonlobular breast cancer, which included all other histologic types, primarily ductal carcinomas (89%).

Unconditional logistic regression analysis was conducted to estimate, by means of the odds ratio (OR), the relative risk of breast cancer associated with each category of HRT use. Estimates of parameters in the model were computed by maximum likelihood techniques and 95% CIs were based on the SE of the coefficients and the normal approximation. Trends across levels of HRT use were assessed by testing the statistical significance of the category of use (coded 1, 2, and so on), where nonexposed women served as the reference group.

Established and suggested breast cancer risk factors were evaluated as potential confounders, including age at reference, age at menarche, age at menopause, type of menopause, parity and age at first birth, family history of breast cancer, years of oral contraceptive use, and several measures of screening mammography before diagnosis. The matching variables of years of GHC enrollment and year of diagnosis were also evaluated. Those factors that changed the OR estimates to the tenths place of the relationship between measures of HRT use and risk of breast cancer were included in the covariate-adjusted models. Only age at reference, year of breast cancer diagnosis, and number of mammograms before diagnosis were found to be confounders and were adjusted for in the final models. Information on the covariates that were in fact confounding factors were available for all women; therefore, the main analyses were not limited to those who completed the risk factor questionnaire.

We also evaluated the relationship between HRT use and risk of breast cancer within subgroups of women according to tumor characteristics and certain risk factors. Analyses by tumor characteristics were performed by polytomous logistic regression for each group of cases to all controls. The statistical significance of the difference in the ORs between the cancer cases defined by tumor characteristics was examined by only including the case groups in logistic regression models, because these comparisons use the same (full) control group. For the other factors, the presence of effect modification was tested on the entire sample by use of an interaction term between the HRT trend variable and the group variable in logistic regression analyses. This P value for interaction examines a departure from a multiplicative relation. Analyses were conducted using SAS version 6.12 (SAS Institute Inc, Cary, NC) and Stata version 6.0 (Stata Corp, College Station, Tex).

The distributions of selected characteristics among the 705 breast cancer cases and 692 matched controls are shown in Table 1. Women with breast cancer were more likely than controls to have had a family history of breast cancer and screening mammograms at GHC.

Compared with never use, current use of estrogen and progestin pills (combination therapy) was associated with an increased risk of breast cancer, whereas past use of any type of preparation was not (Table 2). Time since first and last use of HRT were evaluated and no significant trends were found (data not shown). In the analysis of HRT use during the 5 years ending 1 year before diagnosis, women with longer duration of oral HRT (measured as months of estrogen pill use with or without progestin pills) were at increased risk of breast cancer compared with women with no HRT use during the 5-year period (OR, 1.70 for 57 months or more of estrogen tablets [1425 estrogen pills, assuming 25 pills a month]; 95% CI, 1.15-2.50; P for trend = .002). The results were similar (60%-85% increased risk) for use of oral estrogens alone and for combination therapy, and for sequential and continuous combined HRT. The results were also similar when cumulative estrogen dose was computed as milligram equivalents of conjugated equine estrogen and cumulative progestin dose was computed in equivalency as medroxyprogesterone (data not shown).

We also examined whether the association between breast cancer risk and oral HRT use (the most significant factor in Table 2) varied within subgroups of women (Table 3). The association appeared stronger among women with estrogen or progestin receptor–positive cancer and among women who were leaner or with surgical menopause. However, none of the differences between subgroups was statistically significant.

All measures of HRT appeared more strongly associated with risk of lobular breast carcinoma than nonlobular cancer (Table 4). Current use of either oral estrogen alone or in combination with progestin pills was associated with increased risk of lobular cancer, with particularly high risk associated with current combination therapy (OR, 3.91; 95% CI, 2.05-7.44). Recent oral HRT use of 57 or more months was associated with a 3-fold increase in risk of lobular cancer (OR, 3.07; 95% CI, 1.55-6.06) and a 50% increase in nonlobular tumors (OR, 1.52; 95% CI, 1.01-2.29). This difference between risk of lobular and nonlobular tumors was of borderline statistical significance (P = .06). When we classified HRT by type of formulation, the results for estrogen alone and for combination therapy were similar to those for total oral HRT use for both lobular and nonlobular cancers. Continuous combined therapy was associated with high risk of lobular breast cancer (OR, 6.07; 95% CI, 2.13-17.3 for ≥11 months), whereas it was not associated with nonlobular cancer (P for difference between tumor types = .03), but this was based on very small numbers.

In an effort to disentangle the effects of current vs long-term use and estrogen vs progestin use, we included duration of oral HRT use (measured by use of estrogen pills), duration of progestin use, current estrogen use (yes vs no), and current progestin use (yes vs no) in the same mutivariate model (with the adjustment variables) among all cases, lobular breast cancer, and nonlobular breast cancer (data not shown). Duration of oral HRT use accounted for most of the elevated risk associated with total breast cancer and nonlobular breast cancer; other aspects of hormone use did not contribute to the model. In addition to duration of oral HRT use (P for trend = .03), current use of progestin was associated with increased lobular breast cancer risk (OR, 2.06; 95% CI, 1.05-4.04).

The relationship of breast cancer to HRT use in the recent 10-year period was similar to the results for the recent 5-year analyses (data not shown).

In this nested case-control study, we found an elevated risk of invasive breast cancer among postmenopausal women who were long-term, recent users of oral estrogen, either alone or in combination with progestin. These results are generally consistent with the results from other case-control^9,10,13,28 and cohort studies^1-5 and the recent collaborative analysis.²⁹

Among users of combined HRT, the size of the increase in risk was similar whether the women had followed a sequential or a continuous regimen. Similar results were found in 1 cohort² and 3 case-control studies,^9,10,15 in which the elevated risks ranged from 38% to 50% for sequential use and 9% to 41% for continuous use.

When we divided breast cancer cases into lobular and nonlobular (primarily ductal) cancers, we found somewhat divergent patterns associated with HRT. The association with HRT was considerably stronger for lobular breast cancer, with an approximately 3-fold increased risk associated with longer duration of HRT and a 4-fold risk for current use of combination therapy. Two prior studies, one a population-based case-control study conducted from 1988 to 1990 in Seattle, Wash,³⁰ and the other a multicenter case-control study³¹ conducted from 1989 to 1991, also found a positive association between current use of combination therapy and lobular breast cancer (Seattle study: for current use of at least 6 months' duration, OR, 2.6; 95% CI, 1.1-5.8; multicenter study: for current use defined as within 2 years of diagnosis, OR, 3.1; 95% CI, 1.8-5.3). These findings are consistent with the results of our study, although the number of cases of lobular breast cancer in this study was small (n = 91) and the follow-up time since combination therapy has become popular in the United States (late 1980s) is limited. Nevertheless, the magnitude of risk, the fact that an association is biologically plausible, and a similar finding in other studies all argue in support of the causal nature of association.

Our finding that recent, longer HRT use is associated with a 50% increase in risk of nonlobular (primarily ductal) cancer also deserves mention, because ductal breast cancers are much more common than lobular. If our results are correct, then nonusers of HRT would have an incidence rate of ductal cancer of about 230 per 100 000 women per year, whereas women with 5 years of recent HRT use would have a rate of 349 per 100 000 women per year (using the 1993-1995 rates of US breast cancer by histologic type for women aged 60 to 69 years³⁵ to represent the incidence of postmenopausal breast cancer and the distribution and ORs for oral HRT groups as in Table 4). Similarly, a 3-fold risk of lobular cancer associated with HRT use would translate into an incidence of lobular cancer among non-HRT users of 23 per 100 000 women per year and 70 per 100 000 women per year among women with 5 years of recent HRT use. Thus, a woman with long-term HRT use would still be 5 times more likely to develop ductal rather than lobular breast cancer.

Our study had several strengths. First, since no direct subject participation was required, the possibility of selection bias was minimal. Selection bias is a particular concern in the prior case-control studies of HRT use and breast cancer, because control women who participate may differ in health behaviors, including HRT use, from nonparticipants. Second, the ascertainment of exposure status for the recent 5- and 10-year periods was based solely on a computerized pharmacy record database and thus was not vulnerable to errors in recall on the part of study subjects. On the other hand, the pharmacy records provide no data on medications actually consumed. Because of this, we chose to define women receiving only 1 prescription of HRT as nonusers, since a single HRT prescription dispensed might not have been taken.

Another limitation of this study is that we had no information on some potential confounders. We were able to consider some breast cancer risk factors from questionnaires, including reproductive history, family history, and body mass index, and several measures of screening from a GHC database on history of screening mammograms, including screening mammograms within the 2-year period before diagnosis, number of prior mammograms at GHC, and the interval between mammograms. Only number of prior mammograms was a confounder of the HRT–breast cancer relationship. However, we did not have information on history of previous breast biopsies, breastfeeding, physical activity, alcohol use, and educational level. However, most of the prior studies did not find substantial confounding effects by physical activity, alcohol consumption, and educational level when examining the association between HRT and breast cancer.

It has long been hypothesized that the development of breast cancer is hormonally influenced, based on studies in rodents³⁶ and on the observations that early age at menarche and late age at menopause are associated with higher risk of breast cancer.³⁷ It is well established that estrogens are mitogenic in the breast in eliciting ductal hyperplasia.^38,39 Supported by the evidence that the mitotic activity in the breast reaches its peak during the luteal phase of the menstrual cycle when progesterone concentrations are highest, some argue that progesterone also may influence the risk of breast cancer.^40,41 Recent studies,^42,43 both in vivo and in vitro, have demonstrated both stimulating and inhibitory properties of progestins on breast epithelium cell proliferation. Hofseth et al⁴⁴ analyzed breast tissues from 86 postmenopausal women and found that the breast epithelium of women who had received either estrogen plus progestin or estrogen alone had significantly higher proliferating cell nuclear antigen indices than epithelium not exposed to hormones. Also, treatment with estrogen plus progestin was associated with a significantly higher index (proliferating cell nuclear antigen and Ki-67) than treatment with estrogen alone. Proliferation associated with estrogen plus progestin was localized to the terminal duct-lobular unit of the breast, which is the site of origin of most breast cancers. Lobular tumors are more frequently estrogen and progesterone receptor positive than ductal cancers⁴⁵ and thus might be particularly influenced by HRT use.

In summary, 2 prior studies have observed a 2- to 3-fold increased risk of lobular breast cancer associated with current combination therapy, and we found similarly large risks of lobular cancer associated with current combination therapy and longer duration of all formulations of HRT. A true increase in the risk of lobular breast cancer could have implications for screening, because lobular carcinomas are relatively more difficult to palpate⁴⁶ and more difficult to diagnose by mammography.⁴⁷ However, until more is known about the costs and benefits of different screening modalities for women using HRT, it would be premature to use our results as a basis for modifying early detection activity in them.