Postmenopausal hormone use and breast cancer associations differ by hormone regimen and histologic subtype

Authors


Abstract

BACKGROUND:

Data from large prospective studies are needed to fully characterize the impact of exogenous hormones on breast cancer incidence by type of hormone preparation and histology of the cancer.

METHODS:

In a prospective cohort of 67,754 postmenopausal women in the US, 1821 cases of invasive ductal cancer and 471 cases of invasive lobular or mixed lobular cancer occurred during 13 years of follow-up. The authors computed age-adjusted rates, as well as age-adjusted and multivariate-adjusted rate ratios (RR) for ductal and lobular breast cancer and for the use of estrogen only (E-only) and estrogen and progesterone (E + P) for current and former hormone users by duration of use and years since last use.

RESULTS:

Current use of E + P was associated with an increased risk of both ductal (RR of 1.75; 95% confidence interval [95% CI], 1.53-2.01) and lobular (RR of 2.12; 95% CI, 1.62-2.77) breast cancer. Risk increased within the first 2 to 3 years of use and attenuated 2 years after cessation. In contrast, current use of E-only was not associated with an overall increased risk of invasive ductal cancer. The only exceptions to this finding were in lean (body mass index <25) women and for ductal cancers diagnosed at the regional/distant stage, where in both cases the use of E-only was associated with an increased risk. E-only use was associated with a 50% increased risk of invasive lobular cancer after ≥10 years of use.

CONCLUSIONS:

Use of E + P is more detrimental to the breast than E-only use, in terms of both ductal and lobular cancer. The findings from the current study suggest a window of 2 to 3 years for the risks of E + P both to become apparent after initial use and to attenuate after cessation. Cancer 2009. © 2009 American Cancer Society.

There is ample evidence from observational studies1, 2 and from a randomized clinical trial3 that hormone replacement therapy after menopause increases breast cancer risk and that use of a regimen that includes both estrogen and progesterone (E + P) is more detrimental for the breast than the use of estrogen alone (E-only).4 Estimates of annual excess relative risk of breast cancer among current hormone users range from 4% to 8% for users of E + P and 0% to 3% for users of E-only.5

Results from 10 recent case-control studies6, 7 and 3 prospective cohorts2, 8, 9 suggest that the increased risk associated with postmenopausal hormone use is greater for lobular carcinoma of the breast than for the more common ductal carcinoma. Two cohort studies10, 11 and a meta-analysis5 did not find lobular cancer to be more strongly associated with hormone use than ductal cancer. Based on data from the 9 oldest cancer registries that participate in the Surveillance, Epidemiology, and End Results (SEER) program, the proportion of all breast cancers with a lobular component increased from 9.5% in 1987 to 16.8% in 2002.12 This trend is consistent with a stronger effect of combination hormone therapy, which gained common use in the mid- to late-1980s, on lobular carcinoma than ductal carcinoma.

We examined prospectively the relation between postmenopausal hormone use and the risk of incident breast cancer in a large cohort of US women. We investigated the use of E-only and E + P in current and former users of varying duration, and the subsequent risk of developing invasive ductal and lobular carcinoma of the breast. We also examined whether the risk for each type of breast cancer and each type of hormone regimen varied by adiposity, as estimated by body mass index (BMI), stage of disease at diagnosis, and estrogen receptor (ER) and progesterone receptor (PR) status.

MATERIALS AND METHODS

Study Population

Women in this study were selected from the 97,786 female participants in the Cancer Prevention Study II (CPS-II) Nutrition Cohort, a prospective study of cancer incidence in approximately 184,000 US adults.13 The Nutrition Cohort is a subgroup of the approximately 1.2 million participants in CPS-II, a prospective mortality study established by the American Cancer Society in 1982.14 Members of the CPS-II cohort who resided in 21 states with population-based cancer registries and who were ages 50 to 74 years in 1992 were invited to participate in the Nutrition Cohort by completing a mailed questionnaire. The recruitment and characteristics of Nutrition Cohort participants are described in detail elsewhere.13 All aspects of the CPS-II Nutrition Cohort study were approved by the Emory University Institutional Review Board.

At the time of enrollment in 1992, participants completed a self-administered mailed questionnaire that included demographic, medical, behavioral, environmental, occupational, and dietary factors. Follow-up questionnaires were sent to all surviving cohort members in 1997, 1999, 2001, and 2003 to update exposure information and to ascertain newly diagnosed cancers. The response rate for each of the follow-up questionnaires (after multiple mailings) was at least 89%. For the present study, the follow-up period ended on June 30, 2005.

We excluded from this analysis women who were lost to follow-up from baseline through June 30, 2005 (N = 3152). We also excluded women who reported a history of cancer at baseline, except nonmelanoma skin cancer (N = 12,055). Women who developed cancer (other than breast or nonmelanoma skin cancer) during the follow-up period were censored at the time of their report of the diagnosis. Analyses were limited to postmenopausal women. We excluded women who did not become postmenopausal by the end of follow-up or had unknown menopausal status or unknown age at menopause at baseline (N = 2212). Women who became postmenopausal during the follow-up period (N = 2740) contributed to the analysis only after they became postmenopausal. We excluded women with unknown type or duration of hormone use (N = 7255) or with use of oral progesterone alone or vaginal cream alone (N = 1920) at baseline. Cohort members with missing hormone data in 1997, 1999, 2001, or 2003 were censored at the time of their missing data report. Finally, we excluded women whose reported use of hormone therapy was for <1 year (N = 2823). After all exclusions, the final analytic cohort consisted of 68,369 postmenopausal women who were cancer-free at baseline in 1992.

Case Ascertainment

From this analytic cohort, we documented a total of 3199 incident cases of breast cancer diagnosed between enrollment and June 30, 2005. The majority of incident breast cancer cases (N = 3098) were identified initially by self-report on 1 of the follow-up questionnaires; 3043 of these were subsequently verified by obtaining medical records or through linkage with state cancer registries when complete medical records could not be obtained.13 (Previous work linking the Nutrition Cohort members with state cancer registries has indicated that the ability of our respondents to accurately report a past diagnosis of cancer is very high: sensitivity of 0.93, and specificity of >0.99 for report of any cancer).15 Case verification of other cancers through state cancer registries revealed 19 breast cancer cases that were not first reported by the participant. A small number of breast cancer cases (N = 82) were identified initially on a death certificate as the underlying or contributory cause of death among women who died during the follow-up period and who did not report breast cancer at the beginning of the interval. (Ascertainment of all deaths among cohort members is accomplished through biennial linkage of the cohort with the National Death Index.)16 Additional information was subsequently obtained for 62 of these interval deaths through state cancer registries.

Histologic information was obtained from either medical or registry records for 98% (N = 3118) of all breast cancer cases. Cases of in situ breast cancers (N = 534) were excluded from analyses as were the 75 cases without histology and 6 cases whose recorded diagnosis date was ≥6 months after their self-report of cancer. The remaining 2584 cases of invasive breast cancer included 1821 cases of ductal carcinoma not otherwise specified (NOS) (International Classification of Diseases for Oncology [ICD-O] code 8500/3), and 471 cases of lobular (ICD-O code 8520/3, 312 cases) or mixed lobular and ductal (ICD-O code 8522/3, 159 cases) carcinoma. All 471 cases with a lobular component are referred to as “lobular” in this analysis.

Hormone Use

Information regarding postmenopausal hormone use was collected at the time of enrollment and on all follow-up questionnaires. Questions were asked about the type of hormones used in the past and currently, and duration of use. For this analysis, women were categorized at baseline as never hormone users, E-only users, or E + P users. Women in the E-only group had never reported use of a progesterone, whereas women in the E + P group included women who had only used E + P (N = 8158) as well as women who had used both E-only and E + P before baseline (N = 3057).

Never hormone users were defined as women who reported no use of any hormone preparation at the beginning of each follow-up interval. Within the 2 hormone groups, women were defined as current or former users at the beginning of each follow-up interval and their total lifetime years of use were calculated. For women in the E + P group, total lifetime years of use refers only to years of E + P use. Women who changed type of hormone use during the follow-up period were censored at that time. Women who changed from never use to hormone use during the follow-up interval contributed to follow-up in the appropriate hormone use category only after they began use. In current users of E-only, the total duration of use was classified as 1 to <5 years, 5 to <10 years, 10 to <20 years, or ≥20 years. In current users of E + P, the total duration of use was classified as 1 to <5 years, 5 to <10 years, or ≥10 years. Former users were classified as 1 to <5 years or ≥5 years of use. We estimated years since last use at baseline for each former user by subtracting the sum of her age at menopause and total years of use from her age at baseline. For women who changed from current to former use during the follow-up interval, we calculated years since last use based on the reported time of cessation.

Covariate Information

Confounders included in final multivariate models were age at menopause (single year), type of menopause (natural, surgical/radiation), hysterectomy (yes, no), bilateral oophorectomy (yes, no), history of mammography as a time-dependent variable (never, within the past 2 years, not within the past 2 years), BMI (<25, 25-<30, ≥ 30), and education (high school or less, some college or vocational school, college graduate). We ran models that also included oral contraceptive use, diethylstilbestrol (DES) use, age at menarche, number of live births, age at first live birth, history of breast cysts, family history of breast cancer in a mother or sister, alcohol use, and exercise. Inclusion of these additional covariates did not change the relative risks and so they were dropped from the final models.

Statistical Analysis

Age-adjusted breast cancer incidence rates were calculated for categories of hormone use by type and duration and separately for ductal and lobular cancers. All rates were standardized to the age distribution of women in the CPS-II Nutrition Cohort at baseline.

We used Cox proportional hazards modeling to examine the association of hormone use and incident breast cancer while adjusting for other potential risk factors. All Cox models were stratified on single year of age at enrollment by including age in the strata statement of the Cox model. Results from both age-adjusted and multivariate-adjusted models are presented. Relative risks reported in the text are hazard rate ratios from the multivariate models, unless otherwise noted. Hormone use was modeled as a time-dependent variable in this analysis. Women who never used hormones comprised the referent group for all comparisons.

For current users of hormone therapy, we examined the potential modifying effect of BMI, stage of disease at diagnosis, and receptor status on the exposure-disease relation. For this purpose, BMI was categorized as <25, 25 to <30, or ≥30; stage as localized or regional/distant; and receptor status as ER+PR+, ER−PR−, or mixed receptor status (ER+PR− or ER−PR−). ER and PR status were missing for 720 breast cancer cases; these cases were excluded from the receptor analysis. Statistical interaction between hormone use and BMI was assessed in multivariate models using the likelihood ratio test.

RESULTS

In this cohort of postmenopausal women, 47.6% were never hormone users at baseline, 21.2% were current E-only users, 13.0% were current E + P users, 14.6% were former E-only users, and 3.6% were former E + P users. Among former E-only users of at least 1 year, the average total years of use was 6.7, and average years since last use was 15.7. Among former E + P users of at least 1 year, the average total years of use was 6.0, and average years since last use was 6.4.

Hormone users, especially current users, were much leaner, had higher levels of education, and were more likely to have had a mammogram within the past year than never users (Table 1). E-only users were far more likely to have had a hysterectomy and younger age at menopause than never users. The average age at study entry in 1992 varied across the hormone use groups (lowest in current and former users of E + P, highest in former users of E-only), likely reflecting temporal trends in the use of these different preparations.

Table 1. Characteristics of Women According to Hormone Use at Baseline
 Hormone Regimen
  EstrogenEstrogen Plus Progesterone
CharacteristicNever (n=32,309)Current Use (n=14,339)Former Use (n=9891)Current Use (n=8785)Former Use (n=2430)
  • BMI indicates body mass index.

  • *

    Adjusted to the age distribution of the Nutrition Cohort.

  • Columns may not add to 100% due to missing data.

Average age in 1992, y6361665961
Average age at menopause, y*4945465149
Mean BMI*26.125.126.124.125.1
 Age-adjusted Percents*
BMI     
 <2547.955.046.266.456.4
 25-<3032.231.333.624.929.0
 ≥3018.212.318.97.713.3
 Unknown1.71.51.41.01.3
Education     
 High school or less41.833.641.520.729.2
 Some college28.834.933.332.532.5
 College graduate28.630.924.646.237.6
 Unknown0.70.70.60.70.6
Hysterectomy and oophorectomy     
 No organs or 1 ovary only removed79.09.441.296.770.7
 Uterus with no, 1, or unknown no. of ovaries12.336.822.70.010.3
 Both ovaries (with or without uterus)6.751.633.90.317.0
 Unknown if organs removed2.02.22.23.02.0
Mammogram within last y     
 Yes55.677.060.084.472.4
 No44.022.839.615.427.5
 Unknown0.40.20.40.20.1
Type of menopause     
 Natural84.124.749.296.277.6
 Surgical or radiation15.274.850.40.621.4
 Unknown0.70.50.43.20.9

There was no suggestion of any increased risk of invasive ductal carcinoma with current use of E-only, even after ≥20 years of use (Table 2). In contrast, a significantly increased risk of approximately 50% for invasive lobular carcinoma was observed for E-only use of ≥10 years duration. Relative risks for former users of E-only were less than unity for both ductal and lobular cancer, although not significantly so.

Table 2. Relative Risk of Ductal and Lobular Breast Cancer According to Type and Duration of Hormone Use
   DuctalLobular
Hormone UseYears of UseSince Last UseNo.Rate*Age-adjusted RR (95% CI)Multivariate- adjusted RR (95% CI)No.Rate*Age-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)
  • RR indicates rate ratio; 95% CI, 95% confidence interval; E, estrogen; P, progesterone.

  • *

    Rate per 100,000 person-years standardized to the age distribution of the Nutrition Cohort.

  • Adjusted for age at 1992 survey.

  • Adjusted for age at 1992 survey, hysterectomy and oophorectomy, time-dependent mammography, education, body mass index, type of menopause, and age at menopause.

Never  780279111856611
Current E15.7 3652780.99 (0.87-1.12)0.99 (0.84-1.17)104831.23 (0.97-1.57)1.30 (0.94-1.78)
 l-<5 y2.8 522180.91 (0.68-1.21)0.89 (0.67-1.20)7320.58 (0.27-1.24)0.57 (0.26-1.24)
 5-9 y7.2 823671.18 (0.94-1.49)1.18 (0.92-1.51)15621.04 (0.61-1.77)1.08 (0.61-1.90)
 10-19 y14.4 1402600.94 (0.78-1.12)0.95 (0.77-1.17)49991.48 (1.08-2.03)1.59 (1.07-2.35)
 >20 y26.3 912230.97 (0.78-1.20)1.03 (0.80-1.33)33761.32 (0.91-1.92)1.58 (1.01-2.47)
 Per y of use    1.00 (0.99-1.01)1.00 (0.99-1.01)  1.01 (1.00-1.02)1.02 (1.00-1.04)
Current E+P9.0 3644981.75 (1.54-1.99)1.75 (1.53-2.01)1011572.29 (1.78-2.95)2.12 (1.62-2.77)
 l-<5 y2.9 1114571.49 (1.21-1.82)1.46 (1.19-1.80)321402.13 (1.45-3.14)1.91 (1.29-2.84)
 5-9 y7.1 1234541.76 (1.45-2.14)1.77 (1.45-2.16)341572.39 (1.64-3.47)2.17 (1.48-3.19)
 >10 y13.7 1305382.02 (1.67-2.45)2.07 (1.70-2.52)351572.34 (1.61-3.39)2.19 (1.50-3.22)
 Per y of use    1.06 (1.05-1.08)1.06 (1.05-1.08)  1.07 (1.05-1.10)1.07 (1.04-1.09)
Former E6.715.72272440.88 (0.75-1.02)0.88 (0.75-1.04)58550.84 (0.63-1.13)0.88 (0.65-1.21)
 l-<5 y2.020.01312410.85 (0.71-1.03)0.86 (0.71-1.04)36610.90 (0.63-1.29)0.94 (0.65-1.35)
 >5 y12.110.9962490.91 (0.73-1.13)0.94 (0.75-1.18)22440.77 (0.49-1.20)0.86 (0.54-1.37)
 Per y of use    0.99 (0.97-1.01)0.99 (0.98-1.01)  0.97 (0.93-1.01)0.98 (0.94-1.02)
Former E+P6.06.4852891.08 (0.86-1.35)1.08 (0.86-1.35)23821.30 (0.84-2.01)1.24 (0.80-1.92)
 l-<5 y2.58.2372210.79 (0.57-1.10)0.78 (0.56-1.09)15921.44 (0.85-2.44)1.35 (0.80-2.30)
 >5 y9.14.8483991.52 (1.14-2.05)1.53 (1.14-2.06)8651.10 (0.54-2.24)1.06 (0.52-2.17)
 Per y of use    1.05 (1.01-1.08)1.05 (1.01-1.08)  1.01 (0.94-1.09)1.01 (0.93-1.08)

Incidence rates were significantly higher among current users of E + P compared with never users for both ductal (rate ratio [RR] of 1.75; 95% confidence interval [95% CI], 1.53-2.01) and lobular cancer (RR of 2.12; 95% CI, 1.62-2.77) (Table 2). A positive dose-response relation with increasing duration of use was observed for ductal cancer (for 1-<5 years of use: RR of 1.46 [95% CI, 1.19-1.80]; and for ≥10 years of use: RR of 2.07 [95% CI, 1.70-2.52]). However, for lobular cancer, the approximate doubling of risk with use of E + P appeared within the first 5 years of use and did not increase appreciably with longer durations of use.

We further examined current E + P users of 1 to <5 years duration by single years of use to determine whether the risk associated with hormone use varied within the first 5 years of use (data shown in text only). No increase in risk was observed with E + P use of <2 years for ductal (RR of 0.72; 95% CI, 0.40-1.29) or lobular (RR of 0.79; 95% CI, 0.25-2.52) breast cancer. However, we observed a significant increase in risk for both types of breast cancer at 2 to 3 years of use (for ductal cancer: RR of 1.91 [95% CI, 1.39-2.61]; for lobular cancer: RR of 1.95 [95% CI, 1.04-3.65]) and all years thereafter.

Former users of E + P for ≥5 years were at significantly increased risk of ductal carcinoma (RR of 1.53; 95% CI. 1.14-2.06) (Table 2). On further analysis, this risk was limited to women who had ceased hormone use within the past 2 years (RR of 2.38; 95% CI, 1.50-3.78) and was not observed in women with ≥2 years since last hormone use (RR of 1.20; 95% CI, 0.80-1.81) (data shown in text only). These risk estimates did not change appreciably when we restricted the analysis to women who reported having had a mammogram during the past 2 years. The RRs for lobular cancer in former E + P users (based on a small number of cases) were consistent with a null finding (Table 2).

Body mass modified the effect of E-only use on ductal cancer (P for interaction = .003) (Table 3). There was a significantly increased risk in women with BMI <25 (RR of 1.28; 95% CI, 1.03-1.60) and a clear trend toward decreasing risk with increasing BMI. In the heaviest women (BMI ≥30), RRs were below unity, although not significantly so. There was a suggestion that the effect of E + P on ductal cancer was higher in lean women than in heavy women, but the interaction was not significant (P = .09) (Table 3). BMI did not significantly modify the effect of E-only or E + P on lobular cancer.

Table 3. Relative Risk of Ductal and Lobular Breast Cancer in Current Versus Never Users of Hormones According to Body Mass Index
BMIHormone UseDuctalLobular
No.Rate*Age-adjusted RR (95% CI)Multivariate- adjusted RR (95% CI)No.Rate*Age-adjusted RR (95% CI)Multivariate- adjusted RR (95% CI)
  • BMI indicates body mass index; RR, rate ratio; 95% CI, 95% confidence interval; E, estrogen; P, progesterone.

  • *

    Rate per 100,000 person-years standardized to the age distribution of the Nutrition Cohort.

  • Adjusted for age at 1992 survey.

  • Adjusted for age at 1992 survey, hysterectomy and oophorectomy, time-dependent mammography, education, type of menopause, and age at menopause.

 Estrogen
BMI <25Never30222711836211
Current E2042831.26 (1.05-1.50)1.28 (1.03-1.60)55791.31 (0.93-1.85)1.2.5 (0.81-1.94)
BMI 25-<30Never27730211555911
Current E1122800.89 (0.71-1.11)0.90 (0.70-1.16)34921.44 (0.94-2.21)1.38 (0.83-2.27)
BMI 30+Never18236611408311
Current E452840.72 (0.52-1.00)0.74 (0.52-1.05)15851.16 (0.64-2.11)1.13 (0.59-2.16)
     Pinteraction =.003   Pinteraction = .86
 Estrogen plus Progesterone
BMI <25Never30222711836211
Current E+P2314782.15 (1.80-2.56)2.10 (1.75-2.52)601402.25 (1.59-3.18)2.04 (1.43-2.92)
BMI 25-<30Never27730211555911
Current E+P905021.58 (1.24-2.01)1.54 (1.20-1.97)291812.82 (1.78-4.47)2.56 (1.59-4.10)
BMI 30+Never18236611408311
Current E+P427041.73 (1.23-2.42)1.69 (1.19-2.38)101792.00 (0.99-4.03)1.79 (0.88-3.65)
     Pinteraction = .09   Pinteraction = .63

The effect of E-only use on ductal cancer varied by stage at diagnosis (data shown in text only). There was a significantly increased risk in women with regional or distant disease (RR of 1.59; 95% CI, 1.10-2.30) but not for women with localized disease (RR of 0.89; 95% CI, 0.72-1.09). E + P was associated with significantly increased risks for both localized lobular cancer and regional/distant lobular cancer; however, the effect was larger for regional/distant disease (RR of 2.79; 95% CI, 1.65-4.73) than for localized disease (RR of 1.92; 95% CI, 1.37-2.69). The effect of E-only on lobular cancer and E + P on ductal cancer did not vary appreciably by stage at diagnosis.

ER and PR status did not appear to play an important role in the association between E-only and ductal breast cancer; no association was observed regardless of receptor status. In contrast, the association between E + P and ductal cancer was limited to ER+PR+ tumors (RR of 2.28; 95% CI, 1.86-2.81) and, to a lesser extent, to mixed receptor status tumors (RR of 1.54; 95% CI, 0.97-2.42) (data shown in text only). E + P was not associated with ER−PR− ductal breast cancer (RR of 1.14; 95% CI, 0.73-1.80). For both types of hormone use and lobular cancer, associations were statistically significant only for ER+PR+ cancers (E-only: RR of 2.19 [95% CI, 1.33-3.61]; E + P: RR of 2.84 [95% CI, 1.95-4.15]). The numbers of lobular cases with ER−PR− status were too small to be informative.

DISCUSSION

In this prospective cohort study, we found use of E + P to be more detrimental to the breast than use of E-only, in terms of both ductal and lobular cancer, with risk significantly and substantially increased within 3 years of beginning hormone use. Combination E + P was associated with a doubling of the risk of lobular cancer with 3 years of use and a doubling of the risk of ductal cancer with 10 years of use. In contrast, current use of E-only resulted in a 50% increased risk of lobular cancer after 10 years of use and was not associated with an overall increased risk of ductal cancer.

To our knowledge, this is the fifth prospective cohort study to examine separately the association between the 2 major types of hormone replacement therapy, E-only and E + P, and the 2 major histologic types of invasive breast cancer, ductal and lobular.2, 9, 10, 11 The 4 previous cohort studies, the Million Women Study in the UK, the Multiethnic Cohort, the E3N cohort in France, and NIH-AARP cohort, also observed stronger associations between breast cancer and E + P use than E-only use, confirming results of past case-control studies. Our study, the Million Women Study, and the E3N cohort also found stronger associations for both types of hormone therapy with invasive lobular cancer than with invasive ductal breast cancer, again supporting results of past case-control studies.

To our knowledge, few studies to date have examined duration of hormone use in increments <5 years to determine when risk begins to rise and potentially identify a “safe” period of time for hormone use. In the Women's Health Initiative (WHI) randomized trial of E + P,3 the risk for all invasive breast cancer in the treatment group began to diverge from the control group at year 4 of the trial. However, in a sensitivity analysis of adherent participants, risk between the 2 groups began to diverge at year 3. In the current study, we found that current E + P users of 2 to 3 years' duration were at significantly increased risk of both ductal and lobular breast cancer. No increase in risk was noted in women who used E + P for <2 years. In addition, the increased risk for ductal cancer observed in long-term past users of E + P was greatly attenuated 2 years after the cessation of hormone use. In combination, these findings suggest a window of 2 to 3 years for the risks of E + P both to become apparent after initial use and to attenuate following cessation.

A significant interaction between hormone replacement therapy and BMI for postmenopausal breast cancer risk was first demonstrated over a decade ago.17 Most, but not all, subsequent studies supported these findings. In the current study, we noted a significant interaction for E-only use and ductal breast cancer, and the suggestion of an interaction for E + P use and ductal breast cancer. Importantly, the increased risk of ductal breast cancer with E-only use was entirely restricted to lean women. Results from the Million Women Study also indicated no adverse effect of E-only therapy on risk of ductal cancer in overweight or obese women.2 In our study, the risk of lobular cancer associated with hormone use of either type was not modified by BMI, a result that does not agree with the Million Women Study. In that study, they found a gradient of risk for lobular with E-only and E + P use, with the highest risks noted in the leanest women.2

The distribution of BMI in the study population likely impacts the overall association between E-only therapy and ductal breast cancer. Because this association is limited to lean women, it is more likely to be observed in a study population with a high proportion of lean women and unlikely to be observed in study populations with higher average BMI. Similar to the results from the current study, many other studies conducted in the US have demonstrated no increased risk overall of ductal breast cancer with E-only use.4 The WHI randomized trial of equine estrogen alone conducted in the US, with an average of 6.2 years of follow-up, did not report an increased risk for all invasive breast cancer (RR of 0.77; 95% CI, 0.59-1.01).18 It is important to note that WHI participants in the equine estrogen trial were, on average, an obese population, with a mean BMI of 30 and 45% of all women having a BMI >30.

The relative risks associated with 5 years of E + P use reported in the current study (1.46 for ductal and 1.91 for lobular cancer) and in the Million Women Study (1.71 for ductal and 2.46 for lobular cancer)2 were considerably greater than that reported in the WHI randomized trial of E + P, in which the RR was 1.26 for all invasive breast cancer with an average of 5.2 years of follow-up.19 These differences could be due to chance; the WHI had only 290 total cases of invasive breast cancer and the 95% CI around the RR was wide (1.00-1.59). They might also be explained by the BMI distribution of the WHI study population. The mean BMI of participants in the E + P trial was 28.5, which is considerably higher than in the current study or the Million Women Study; this difference in BMI would reasonably result in lower RRs, especially for ductal cancer, than would be observed in leaner populations.

Results of previous studies have been inconsistent as to whether hormone replacement therapy is more strongly associated with hormone receptor-positive disease, with some studies demonstrating stronger associations for receptor-positive than receptor-negative breast cancer, and some reporting no differences.20 Our results supported, in general, a stronger association between hormone replacement and ER+PR+ breast cancer than other receptor types. Our finding of a significantly increased risk of regional/distant ductal carcinoma with use of E-only is not supported by other studies and may be due to chance.

The primary strength of the current study is its prospective design, which eliminates the possibility of biased recall of hormone use, and its large size. We believe this is the largest prospective study in the US published to date to examine separately the impact of E-only and E + P use on breast cancer by histologic subtype. In addition, we had repeated measures of hormone use throughout the follow-up period, so that hormone use in the interval before the diagnosis of breast cancer could be assessed. We also controlled for mammography screening in a time-dependent manner, minimizing the potential for screening bias. Finally, because we had lifetime history of hormone use, we could define a group of E-only users who had never received progesterone therapy.

A limitation of the current study is that hormone use was self-reported. However, studies of self-reported hormone therapy have found reports to be in good agreement with more objective sources.21 In addition, these results are from a cohort of predominantly white, middle-aged or elderly, educated women, and they may not be generalizable to other populations of interest, especially to populations with a different BMI distribution.

In summary, we found the use of E + P to be associated with much higher risks of breast cancer than the use of E-only for both ductal and lobular carcinoma, with risk significantly increased within 3 years of beginning hormone use. Our results suggest a window of 2 to 3 years for the risks of E + P both to become apparent after initial use and to attenuate after cessation of use. We found stronger associations for lobular cancer than for ductal cancer with use of both types of hormone therapy. E-only use increased the risk of ductal carcinoma only in lean women. The estimate of risk of total invasive breast cancer associated with the use of E + P from the WHI randomized trial3 likely underestimates the risk for US women due to the high BMI distribution in that study population.

Conflict of Interest Disclosures

The authors made no disclosures.

Ancillary