Menopausal Hormone Therapy and Subsequent Risk of Specific Invasive Breast Cancer Subtypes in the California Teachers Study

There is substantial epidemiologic, experimental, and clinical evidence that the risk of breast cancer is elevated with the use of menopausal hormone therapy (HT), with formulations containing estrogen and progestin (EPT) resulting in a substantially higher risk than formulations containing estrogen only (ET; refs. 1-13). Since the publication of the 2002 report from the Women's Health Initiative (WHI) clinical trial (12), sales of HT have dropped substantially both in the United States (14-17) and elsewhere (18, 19). This decline in hormone use has been accompanied by an unprecedented decline in rates of invasive breast cancers (19), most of which is likely due to changes in HT use (19, 20).

Despite the evidence linking EPT to the risk of developing breast cancer, different formulations of HT are still prescribed by general practitioners and gynecologists who consider them to offer important relief from menopausal symptoms (21). Therefore, to the extent that HT continues to be prescribed, and because individual women need detailed information regarding risks to optimize their own risk-benefit assessments, it remains important to delineate the characteristics of women at the highest risk of HT-associated breast cancer, and how this risk varies with factors such as formulation and duration of HT.

In particular, it remains unclear whether the number of days per month of progestin use affects risk, specifically whether the continuous-combined EPT regimens (e.g., Prempro) differ from sequential regimens. Continuous-combined HT was associated with the highest risk of hormone-sensitive breast cancer in some (2, 8, 22, 23) but not all studies (6, 24). Lower body mass index (BMI) has also been flagged by a number of studies as being an important modifier of HT-associated breast cancer risk (1, 2, 7, 25-28). Some evidence supports a selective increase in risk of particular breast cancer subtypes associated with EPT use, i.e., the estrogen receptor–positive and progesterone receptor–positive subtypes (ER+/PR+; refs. 10, 29, 30). There is also a need to understand how HT affects the development of triple-negative (TPN; ER−/PR−/HER2−) breast cancer, the subtype having the poorest prognosis (31-33).

The California Teachers Study (CTS) has been previously described in detail (34). Briefly, the CTS is a prospective cohort of women who were current, recent, or retired public school teachers and administrators participating in the California State Teachers Retirement System. Cohort members were mailed a detailed baseline questionnaire in 1995 characterizing information on personal medical history, family history of breast cancer, reproductive factors, HT use, and lifestyle factors such as physical activity, diet, alcohol consumption, smoking history, mammographic screening practices, and history of breast biopsy. The CTS has been approved by the Institutional Review Boards of the Northern California Cancer Center, the California Department of Public Health, the University of California, Irvine, the University of Southern California, and the City of Hope National Medical Center in accord with assurances filed with and approved by the U.S. Department of Health and Human Services.

The CTS cohort consists of 133,479 women. For this analysis of HT use, women were excluded, in sequence, if they were not California residents at the time of completing the baseline questionnaire (n = 8,867), had previous or unknown history of breast cancer (n = 6,351), were older than 80 years of age at baseline (n = 4,344), were premenopausal (n = 47,724), were of unknown menopausal status (n = 5,317), or had unknown history of ever using HT (n = 4,010).

Among the remaining 56,867 perimenopausal and postmenopausal women, 2,857 women (5%) were diagnosed with pathologically confirmed invasive breast cancer through December 31, 2006. Follow-up began on the date the baseline questionnaire was completed and ended at the occurrence of invasive cancer in the breast or a censoring event, which included a move outside of California, a diagnosis of in situ breast cancer, death, or December 31, 2006. Incident invasive breast cancer diagnoses (International Classification of Disease Oncology code ICDO-C50) were identified through annual linkages with the California Cancer Registry (CCR), which receives reports of more than 99% of the incident cancer diagnoses occurring among California residents. National Cancer Institute–designated Surveillance, Epidemiology, and End Results registries, including the CCR, routinely collect ER and PR as recorded in the medical records following a breast cancer diagnosis with the additional recording of HER2 status in California beginning in 2000, but HER2 status was only widely collected beginning in 2005. Among 2,857 invasive breast cancers, ER status was available for 2,452 (86%) cases and PR status was available for 2,347 (82%) cases. Joint ER and PR status were available for 2,344 (82%) cases, and ER, PR, and HER2 status were available for 1,276 (46%) cases. A combination of annual mailings within the CTS cohort, linkages with the U.S. Postal Service National Change of Address Database, and participant change-of-address postcards were used to monitor changes in residence.

The baseline questionnaire contained detailed questions regarding the use of HT. The questionnaire included questions on ever use as well as duration of use of various HT regimens including ET, progestin alone (PT), and combined EPT. This information was used to create progressively detailed HT variables. The least detailed HT variable compares never users to women with any history of HT. Each of the three classes of HT regimen, ET alone, PT alone, or EPT, was further characterized into current use at the time of baseline questionnaire or past use at the time of baseline questionnaire. Durations of ET and EPT use were calculated separately using questions on ages at start and stop of hormone use as well as total years of use. These separate durations of use were then categorized finely (never use, <2, 3-5, 6-9, 10-14, 15-19, and 20+ years) as well as in collapsed categories (never use, ≤5, 6-14, and 15+ years).

To determine how many days a month progestin was taken in EPT regimens, we used the responses to the question “When using a progestin or progesterone, for how many days each month would you usually take it?” EPT users reporting usually taking progestin for 10 or fewer days per month were classified as short-sequential EPT users, those usually taking the progestin component for longer than 10 days but fewer than 28 days were classified as long-sequential EPT users, whereas women with 28 or more days of progestin in their regimen were classified as continuous-combined EPT users. In analyses of duration of use, duration was calculated to include all regimens taken, thus current users of continuous-combined regimens who had used EPT for 15 or more years may have used sequential regimens for part of this period.

We obtained information regarding other established risk factors for breast cancer from the self-administered baseline questionnaire, including race/ethnicity (White, Black, Hispanic, Asian/Pacific Islander, other/mixed/unknown), first-degree family history of breast cancer (no, yes, unknown/adopted), BMI (<25.0, 25.0-30.0, >30.0 kg/m², unknown), smoking status (never, former, current, unknown), alcohol intake in the year prior to completing the questionnaire (none, <20, ≥20 g/d, unknown), age at first full-term pregnancy (15-24, 25-29, 30+ years old, unknown), parity (nulliparous, one to three full-term pregnancies, four or more full-term pregnancies, unknown), age at menarche (≤10, 11-12, 13-14, 15-16, ≥17 years old, unknown), age at menopause (<35, 35-39, 40-43, 44-46, 47-49, 50-52, 53-55, ≥56 years old, unknown), and history of breast biopsy (no, yes). We combined age at first full-term pregnancy and parity into a single variable. We also considered whether the woman had a screening mammogram within the past 2 years (yes, no, unknown).

Menopausal status was classified using information regarding age at baseline, age at last menstrual period with corresponding reason for cessation, and surgical oophorectomy and/or hysterectomy status. Women reporting their last menstrual period within 6 months of baseline were considered to be perimenopausal. Women reporting that their last menses was more than 6 months prior to completing the baseline questionnaire, that they had had both ovaries removed, or who were older than 56 years and who did not specify that they were premenopausal were considered postmenopausal. Women who reported initiating HT use before experiencing menopause were considered menopausal but were assigned an unknown age at menopause.

We used multivariable Cox proportional hazards regression methods (35) to estimate the association between HT and incident invasive breast cancer risk. Hazard rate ratios are presented as relative risks (RR) with associated 95% confidence intervals (CI) and were estimated using age in days at the start and end of follow-up as the time metric. Associations between each of the HT use variables with breast cancer risk were evaluated with two proportional hazard models, one simple model stratified by age at baseline and a multivariable adjusted model, which estimated the risk for invasive breast cancer associated with HT use stratified by age at baseline questionnaire and adjusted for a number of potentially confounding variables selected a priori, including baseline reports of race/ethnicity, first-degree family history of breast cancer, BMI, smoking history, alcohol consumption during the year prior to baseline, mammographic screening over the past 2 years, parity/age at first full-term pregnancy, age at menarche, age at menopause, and history of breast biopsy. Women who could not be classified into one of the HT exposure categories for each analysis were included in the model as a separate category (other/unknown HT). These categories are presented in Tables 1 and 2, but are not presented for the purpose of clarity in the remainder of the results.

Because a combined EPT pill was marketed just as our baseline questionnaire was being distributed, there is a possibility that some women who used this pill reported it as ET. Considering that this misclassification of ET use should be less likely in hysterectomized women, we conducted one analysis in which we examined the association between ET and breast cancer risk among hysterectomized women only. Similarly, we conducted one analysis in which we examined the association between EPT use and breast cancer risk among nonhysterectomized women, and finally one analysis in which we only included women reporting natural menopause. These sensitivity analyses indicated that the results were similar across types of menopause (results not shown). Additionally, we repeated the analyses halting follow-up at the end of 2002 to explore the possible effect of changes in HT use associated with the release of the WHI findings, when large numbers of women in the United States reported stopping HT use (12, 14, 36).

Tests for trend within each of the HT variables were done by fitting the median value for each HT duration category as a continuous variable in the multivariable age-stratified model (35), and the statistical significance for the β slope parameter estimate (Wald P value) is reported. We examined several potential modifiers of the HT-breast cancer association, including baseline BMI, alcohol consumption, parity, and mammographic screening. We also determined whether risk differed by tumor size or tumor subtype defined by ER, PR, and HER2 status. Differences in the risk association between strata were examined by conducting a test for effect modification using a likelihood ratio test for heterogeneity of trends (35).

The validity of the proportional hazards assumption inherent to Cox proportional hazards modeling (37) was confirmed by visual inspection of Kaplan-Meier survival curves and plotted scaled Schoenfeld residuals to assess the correlation of residuals to time on study. All reported P values were two-sided. The SAS software program (SAS version 9.1, SAS Institute, Inc.) was used to perform all statistical analyses.

The mean (±SD) age at diagnosis of women with incident invasive breast cancer was 67.1 (±9.0) years. Compared with women with no history of HT use, women who reported ever having used ET only, EPT only, or any combination of the two, or could not indicate the form of HT they had used were younger, more likely to self-identify as non-Hispanic white, report no family history of breast cancer, be parous, have a younger age at menarche and lower baseline BMI, have ever smoked or consumed alcohol, have had a screening mammogram within the 2 years prior to baseline, be younger at menopause, and report a history of breast biopsy (Table 1).

Compared with women with no history of HT use, women who had ever used HT were at a 40% elevated risk of breast cancer (95% CI, 1.26-1.55; Table 2). Women who had used ET or EPT were at an increased risk of breast cancer, and EPT users were at increased risk regardless of the pattern of use, i.e., regardless of whether they had used EPT only, or also had used ET or PT. Additionally, current use of either ET or EPT was associated with higher risk than past use of the corresponding hormone regimen. The greatest risk elevation was associated with current EPT use in which EPT was the only formulation ever used (RR, 1.69; 95% CI, 1.50-1.90).

In general, former HT users were more likely to be short-duration users than current HT users. Although 74% of former ET users had used ET ≤5 years and only 8% had used it ≥15 years, the numbers were 26% and 43% among current ET users. Similarly, among former EPT users, 85% had used EPT ≤5 years and 1% had used EPT ≥15 years, whereas the numbers were 60% and 6% among current EPT users. The associations of increasing duration of both ET and EPT use with increasing breast cancer risk were statistically significant. Compared with women who had never taken HT, women with 20 or more cumulative years of ET use at baseline were at a 28% increased risk (95% CI, 1.09-1.51; P_trend = 0.01), whereas EPT users for 20 or more years were at 92% increased risk (95% CI, 1.36-2.72; P_trend < 0.0001; Table 3). Compared with never users, ET use for more than 15 years among current users was associated with a RR of 1.35 (95% CI, 1.15-1.58), whereas 15 or more years of EPT use among current users was associated with a RR of 1.94 (95% CI, 1.53-2.44).

To address the concern that the elevated risks associated with current ET might be confounded by past EPT use, we conducted a separate analysis restricted to current ET users without any history of EPT. The effect estimates for each of the ET duration categories remained stable and showed a clear dose-response trend, with RRs for ≤5, 6 to 14, and 15+ years among current users being 1.23 (95% CI, 0.99-1.54), 1.38 (95% CI, 1.14-1.67), and 1.44 (95% CI, 1.21-1.71), respectively (results not shown). Furthermore, when restricting the analysis to women with a self-reported hysterectomy, the risk estimates for ET use were similar (results not shown).

When we restricted analyses of EPT use to current users who reported having only used EPT, the RRs for ≤5, 6 to 14, and 15+ years of use were 1.60 (95% CI, 1.39-1.86), 1.74 (95% CI, 1.51-2.02), and 1.78 (95% CI, 1.36-2.33), respectively (results not shown).

Because of concern that many women might have ceased HT use after the publication of the WHI trial results (12), we ran one analysis excluding follow-up after December 31, 2002. The resulting risk estimates became slightly stronger; the RR for more than 15 years of current ET use was 1.44 (95% CI, 1.19-1.75) and the RR for long-term current EPT use was 2.25 (95% CI, 1.71-2.96).

We further characterized the risk associated with current EPT use by classifying the HT regimen according to the usual length of use per month of the progestin component. We found that risk increased with the usual number of days per month of progestin in the current EPT regimen, with the greatest risk among current continuous-combined EPT users (RR, 1.83; 95% CI, 1.61-2.08; P trend = 0.0113; Table 4).

The dose-response relationships between duration of EPT and of ET use and risk for invasive breast cancer seemed to be modified by BMI. Both the magnitude and trend in risk estimates diminished as BMI increased from <25.0 to >30.0 kg/m² (Table 5). However, these apparent effect modifications by BMI were not statistically significant (P values for effect modification by BMI of the ET and EPT effects were both >0.20). We observed no statistically significant evidence to indicate effect modification of any associations by level of alcohol intake, parity, or mammographic screening within the last 2 years (P > 0.25; results not shown).

We examined risks separately by ER/PR receptor status of tumor as well as by a combined ER/PR/HER2 classification scheme (Table 6). The associations with current ET use and current EPT use were statistically significant only for tumors that were ER+PR+. Compared with non-HT users, women with 15+ years of ET who were current users were at a 33% increased risk (95% CI, 1.08-1.66; P_trend = 0.0036) whereas women with 15+ years of EPT use had an 84% increased risk of breast cancer (95% CI, 1.32-2.56; P_trend < 0.0001). The risk associated with current EPT was even stronger for ER+PR+ tumors that overexpressed HER2, but the numbers for this subtype-specific analysis were limited. Specifically, long duration of EPT use among current users was most strongly associated with breast cancer tumors that were either ER+ or PR+ and HER2+, in which women who were current users with 15+ years of EPT had a >2.5-fold risk elevation compared with hormone never users (RR, 2.64; 95% CI, 1.07-6.53; P_trend = 0.02). There was no evidence that long duration of current EPT use was associated with triple-negative (ER−PR−HER2−) tumors.

When considering breast tumor size, we found that current EPT use was associated with increased risk of tumors irrespective of size, whereas current ET use was only associated with small (<1 cm) or large (>5 cm) breast tumors; although the number of large breast tumors among current ET users was small (n = 21; Supplementary Table S1). Furthermore, when categorizing breast tumors by histology, we found that the associations with ET and EPT use were similar for ductal (ICDO code 8500) and lobular tumors (ICDO code 8520), and somewhat stronger for tumors with mixed ductal and lobular histology (ICDO code 8522), although the numbers for this subtype-specific analysis were limited (Supplementary Table S2).

In this prospective observational study of women from California, we found that cumulative measures of ET and EPT use assessed at study entry were associated with the subsequent development of invasive breast cancer. Elevated risk was limited to women who were current HT users at baseline. The greatest risk elevations were seen among current long duration EPT users, but increased risk was also seen for ET users in most analyses.

In general, our findings were consistent with those from previous studies describing an increased RR for breast cancer related to HT. We found a higher risk associated with ever use of menopausal hormones than some (1, 4, 12, 24) but not all (38, 39) studies. This likely reflects the fact that in the CTS, ∼75% of ever HT users were current users at baseline; this is a higher fraction than seen in several other studies (1, 24). The fact that current users at baseline have higher risk than past users is consistent with previous studies reporting that increased risk due to HT diminishes with time since last use (1, 24, 36, 40).

The baseline questionnaire used to characterize HT use was completed in 1995 to 1996 whereas follow-up for breast cancer diagnoses continued through 2006. With the early termination of the WHI randomized controlled clinical trial in 2002, national trends in the prescription of EPT decreased markedly (14, 36). As a result of this change in practice, substantial misclassification of current HT use, as measured at study baseline, may have occurred. It is also possible that some of the perimenopausal women in our study commenced HT use after completing the baseline questionnaire. However, both types of misclassification are likely to have been nondifferential with respect to the outcome, and hence, would most likely have resulted in a bias towards the null in our findings for current HT use. Consistent with this, when we truncated follow-up after December 2002, we observed that the measured associations with current ET and EPT became stronger. On the other hand, there is an alternative explanation of the latter result. It could be that EPT users who decided to discontinue their HT use in 2002 might have had a higher risk of developing breast cancer compared with women who used EPT at baseline and continued EPT after 2002. Because we did not have information on reason(s) for stopping EPT in 2002, we cannot address this issue.

We found evidence of a dose-response pattern with the cumulative duration of ET and EPT use measured at baseline. A number of previous studies have found a dose-response pattern in the duration of EPT use, although our findings were somewhat stronger than some (1) but not all previous studies (24, 38, 39). It is possible that we might have underestimated the duration women used EPT or ET as a result of continued use after baseline at least until 2002. Consequently, some of the women contributing to the hazard estimate for current short-duration EPT or ET use may in fact be longer-duration current users.

Our findings are consistent with previous studies suggesting a stronger effect of EPT than ET on the risk for breast cancer (1, 6, 7, 24). This can be explained by an increased mitogenic effect of progesterone in the breast tissue (4, 6). In our study, when EPT use was categorized by the usual number of days per month progestin was taken, we saw that the risk of breast cancer increased with the number of days per month that the progestin component was used. These findings would support the growing evidence that continuous-combined EPT elevates breast cancer risk more than short- or long-sequential regimens of EPT (2, 8, 23, 24, 40). However, some observational studies, including one from Los Angeles (6), have shown a lower RR associated with use of continuous-combined EPT regimens than with sequential regimens and this remains a point for further investigation.

Although the risk estimates associated with overall ET use in this study are similar to those of some previous studies (1, 41), our estimates for the risk associated with current ET use, particularly for the short duration of use, were somewhat higher than some studies (7, 42). This observation was not due to past EPT use, as seen in the analysis restricted to ET-only users. Furthermore, to better understand this finding, we confirmed that women who reported changing from EPT to ET use had largely (84%) done so after hysterectomy. This proportion of women who had undergone hysterectomy was substantially higher than those observed among never users of HT (13%) and women who changed from ET to EPT (17%). Another possible explanation for the risk estimates we observed for long duration of ET use could involve longer true duration of ET use among the current users than indicated in the baseline questionnaire resulting in an upward bias in the RR estimates.

A collaborative reanalysis of the effect of HT on breast cancer risk reported increased HT-related breast cancer risk in postmenopausal women with lower BMI (1). This effect modification by BMI has been corroborated by subsequent studies (2, 7, 25-28). Our findings, although not statistically significant, are consistent with these previous studies in that the RRs for ET and EPT by duration were stronger in women with lower BMI. In the absence of endogenous ovarian hormones, peripheral aromatase activity in the fat tissue is the primary source of circulating estrogens (43). Therefore, in the absence of significant adipose-derived estrogen in women with low BMI, we would expect to see stronger breast cancer risk related to HT use.

Previous studies have shown the strongest associations between HT and hormone receptor–positive breast tumors (10, 29, 44). Accordingly, we found elevated risk associated with current ET and current EPT only in breast tumors that were ER+ and PR+. Interestingly, the point estimates were slightly higher for the subset of tumors that were ER+ or PR+ and HER2+, but this may be a statistical fluke as the numbers for this subtype-specific analysis were very limited because information on HER2 status was not recorded in the cancer registry records until the year 2000. Risk was not elevated in triple-negative breast tumors and this is consistent with other studies (45) and is clinically relevant, given the poor prognosis associated with these tumors (32, 46).

In this study, the EPT association was of similar magnitude and trend across all tumor sizes, whereas the association with long duration ET seemed to be confined to tumors between 1 and 5 cm. This is consistent with the reported findings of some (47) but not all previous studies (10). The persistence of association in larger tumor sizes suggests that the increased risk associated with both ET and EPT use cannot only be due to increased screening in HT users. The findings with respect to histology are similar to what has been reported previously, the association with EPT seems to be strongest for cancers with a lobular component (10, 47, 48).

An important limitation of our analysis includes the characterization of HT use at cohort entry only, as it is clear that the most correct assessment of a woman's HT use during this time period (1995-2006) would have necessitated frequent assessments of HT use and analyses taking into account repeated measures over time. We were also limited by the number of women with complete data needed for some subtype analyses, as there were substantial numbers of women with missing information regarding HER2, which was not collected widely for breast cancers of all stages until the year 2005.

The strengths of the present findings include the CTS prospective cohort design and the large number of participants. Detailed information on reproductive variables, including the use of HT, provided an opportunity to find associations by duration and formulation. Breast cancer case ascertainment was through linkage with the highly complete California Cancer Registry. The relative uniformity of the cohort with respect to education is another strength, as is the high rate of medical insurance coverage among women comprising the CTS, ensuring similar access to health care and mammographic screening. The relatively high rate of HT use resulted in substantial power to detect the effects of long-term HT use.

In our observational study, we found an overall increase in breast cancer risk associated with both ET and EPT use after an average of 9.8 years of follow-up, although the risk was stronger for EPT use. Both HT formulations show increasing risk associated with recent as compared with prior use, and with increasing duration of total use. The individual characteristics of a woman taking HT, such as her BMI, as well as the specific regimen of HT a woman uses, might modify the extent to which a woman's risk for breast cancer is elevated as well as the receptor profile for the malignancy. These findings, taken in context of the larger literature on this topic, continue to underscore the need to personalize risk-benefit discussions for women contemplating the use of HT.

C. Clarke served as an expert witness for the plaintiff lawyers pursuing Prempro litigation (Williams, Love, O'Leary, and Powers).

We acknowledge the late Dr. Ronald Ross for his major contribution in establishing the California Teachers Study cohort. We also acknowledge the entire CTS project team for their efforts.

Grant Support: R01 CA77398, K05 CA136967, and P01 CA017054 from the National Cancer Institute and contract 97-10500 from the California Breast Cancer Research Fund. The funding sources did not contribute to the design or conduct of the study, nor to the writing or submission of this manuscript. The collection of cancer incidence data used in this study was supported by the California Department of Health Services as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885, the National Cancer Institute's Surveillance, Epidemiology and End Results Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention's National Program of Cancer Registries, under agreement no. U55/CCR921930-02 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the authors and endorsement by the State of California, Department of Health Services, the National Cancer Institute, and the Centers for Disease Control and Prevention or their contractors and subcontractors is not intended nor should be inferred.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.