Effect of Raloxifene on the Incidence of Invasive Breast Cancer in Postmenopausal Women with Osteoporosis Categorized by Breast Cancer Risk

Identifying women most likely to benefit from breast cancer prevention strategies requires an understanding of any association between risk factors for the disease and benefits of the proposed therapeutic intervention. Women at higher risk for breast cancer are generally older and have a greater lifetime estrogen exposure, as assessed through surrogate indicators of estrogen exposure including age at menopause, body mass index (BMI), estradiol level, and use of estrogen therapy (1–4). A heritable component to disease etiology is also well recognized and women with a first-degree relative with breast cancer are at greater risk of breast cancer (5). The Gail model (6, 7) is the most commonly used tool to estimate breast cancer risk; 5-year predicted risk is calculated based on the presence/absence of recognized breast cancer risk factors. Using this model, women at high risk are generally defined as those with a 5-year predicted risk of invasive breast cancer of ≥1.67%.

Tamoxifen, a selective estrogen receptor modulator, is the only agent currently approved for chemoprevention of breast cancer and is indicated in women considered at high risk for breast cancer based on their 5-year predicted risk assessed using the Gail model.

Raloxifene, a selective estrogen receptor modulator indicated for osteoporosis treatment and prevention, is associated with a reduced incidence of invasive breast cancer in postmenopausal women with osteoporosis based on the results of the Multiple Outcomes of Raloxifene Evaluation (MORE) and the Continuing Outcomes Relevant to Evista (CORE) trials (8–10). MORE was a 4-year double-blind, placebo-controlled osteoporosis treatment trial of 7,705 postmenopausal women (mean age, 67 years), with breast cancer as a secondary end point. CORE was a 4-year follow-up trial to MORE, enrolling 4,011 women of the 7,705 participating in MORE. CORE was conducted to determine the effect of an additional 4 years of raloxifene therapy on invasive breast cancer incidence in postmenopausal women with osteoporosis.

Over the 4 years of the MORE trial, there were 61 confirmed invasive breast cancer cases. Raloxifene was associated with a significant 72% reduction in the incidence of invasive breast cancer compared with placebo (8). In a subgroup analysis of the MORE trial data, the reduction in breast cancer incidence with raloxifene was statistically significant in both women considered at higher and lower risk of breast cancer based on MORE trial baseline characteristics, but was numerically greater in women considered at higher risk (11). Women with a family history of breast cancer had a significantly greater invasive breast cancer risk reduction with raloxifene than those with no family history, as did women with higher bone mineral density versus those with lower bone mineral density.

Over the 8 years of MORE and CORE, there were 98 confirmed invasive breast cancer cases, and raloxifene was associated with a significant 66% decrease in invasive breast cancer incidence (10). In the present analyses, we assessed the effect of raloxifene on invasive breast cancer incidence by the predicted level of breast cancer risk (higher versus lower risk) using data collected in both MORE and CORE. We also sought to confirm risk factors for breast cancer in this population of postmenopausal women with osteoporosis by comparing placebo event rates in these subgroups.

For all subgroup analyses, except in the case of the 5-year predicted risk calculated using the modified Gail model, data from the 7,705 women participating in either the MORE trial only (N = 3,694) or both the MORE and CORE trials (N = 4,011) were considered. For the subgroup analysis by 5-year predicted risk, only women participating in CORE who had not experienced breast cancer in MORE and who underwent a breast cancer risk assessment were considered (N = 3,996).

Study design. The MORE trial was a 4-year (3 years plus 1 year extension) study designed to assess the effect of raloxifene (60 or 120 mg/d) on bone mineral density and vertebral fracture incidence, and to establish the safety profile of raloxifene (12). Secondary objectives included assessment of the possible effect of raloxifene on risk of breast cancer. Eligible women were ≤80 years of age, at least 2 years postmenopausal, and had documented osteoporosis. Exclusion criteria included a history of breast cancer, invasive endometrial cancer, or stroke or venous thromboembolism during the preceding 10 years. At the beginning of MORE, women were randomly assigned to placebo (N = 2,576), raloxifene 60 mg/d (N = 2,557), or raloxifene 120 mg/d (N = 2,572). Participants were discontinued from the MORE trial if they developed cancer of the breast or uterus, or other malignancies considered to be estrogen dependent, or experienced a venous thromboembolism. These women were not followed further unless they chose to continue in CORE, for which they were still eligible.

CORE was a 4-year follow-up trial to MORE. The primary objective was to determine the effect of an additional 4 years of raloxifene therapy on invasive breast cancer incidence in postmenopausal women with osteoporosis (10). The only eligibility requirement for CORE was to have been enrolled in the MORE trial. The beginning of the CORE trial did not coincide with the end of the MORE trial; the median time between the end of participation in MORE and subsequent enrollment in CORE was 10.6 months (range, 2.6-62 months) for both treatment groups. CORE enrollees were not rerandomized and continued in their original randomized group established in MORE. Those randomized to raloxifene 60 or 120 mg/d in MORE were assigned to receive raloxifene 60 mg/d (approved dose for osteoporosis treatment and prevention) in CORE (N = 2,725) and those randomized to placebo in MORE were assigned placebo in CORE (N = 1,286). Of the 4,011 CORE enrollees, 21% in the placebo group and 20% in the raloxifene treatment group never resumed study drug in CORE, either because they chose not to or were not allowed to due to diagnosis of any estrogen-dependent malignancy, a history of venous thromboembolism, ongoing treatment with excluded medication, or unblinding to study medication because of a safety concern during the MORE trial. For those women who chose to participate in CORE, the mean time from randomization in MORE to end of participation in CORE, including the period between trials, was 7.8 years for each treatment group.

Breast cancer ascertainment. In the MORE trial, mammograms were required at baseline and at 2, 3, and 4 years, and were optional at year 1. In the CORE trial, mammograms were required within 1 year of entering the trial and at 2 and 4 years. At each visit of MORE (biannually) and CORE (annually), women received a clinical breast examination and were asked if they had been diagnosed with breast cancer or had had a breast biopsy or breast surgery that had occurred since their last visit. If breast cancer was suspected, study medication was stopped and all available information (e.g., mammography reports, histopathology reports, and surgical records) was sent for adjudication by an independent oncology review board, who either rejected or confirmed the diagnosis.

Subgroup analyses. Women were categorized as at higher or lower risk of breast cancer based on age, age at menopause, BMI, estradiol level, bone mass and presence/absence of preexisting vertebral fracture, prior use of oral or transdermal estrogen preparations (yes versus no), a family history of breast cancer (yes versus no), and 5-year predicted risk of invasive breast cancer. All characteristics were collected at baseline of the MORE trial except for 5-year predicted risk, which was calculated using the modified Gail model at baseline of the CORE trial.

Serum estradiol was measured at MORE baseline by a central laboratory using a double-antibody procedure (13). Estradiol levels <5 pmol/L were below the limit of accurate quantification. Bone mineral density was measured at baseline of the MORE trial by dual-energy X-ray absorptiometry. In MORE, all women had osteoporosis defined as a total hip bone mineral density T-score of −2.5 SD or less, calculated using the Hologic reference data, and/or the presence of a radiographically apparent vertebral fracture. For the current analysis, total hip bone mineral density T-scores were recalculated using the Third National Health and Nutrition Examination Survey data set (14). Women were categorized into two groups, consistent with current WHO criteria: those with osteoporosis (T-score ≤2.5 or one or more preexisting vertebral fractures) and those with low bone mass (T-score between −2.5 and −1 and no preexisting vertebral fracture). Family history of breast cancer was defined as a female first-degree relative with breast cancer.

The 5-year predicted risk of invasive breast cancer was calculated using the modified Gail model (7). Criteria considered in the model were age, race, age at menarche, age at first live birth of a child, number of first-degree relatives with breast cancer, number of breast biopsies, and pathologic diagnosis of atypical hyperplasia. Because this information was only collected comprehensively at the start of CORE, the 5-year predicted risk of invasive breast cancer was only calculated at the start of CORE for those women choosing to participate in CORE (N = 3,996).

Statistical analysis. Subgroup factors (age, age at menopause, BMI, estradiol level, prior oral/transdermal estrogen plus progestin or estrogen therapy use, family history of breast cancer, bone mass) were prespecified. Women were categorized as at higher or lower risk of breast cancer based on characteristics collected at baseline of MORE or, in the case of the 5-year predicted risk, the baseline of CORE. For age, age at menopause, and BMI, the study population was dichotomized at the median or using a clinically relevant value close to the median (65 years, 49 years, and 25 kg/m², respectively). For estradiol level, the study population was divided into three groups (<5, 5-10, and ≥10 pmol/L). For 5-year predicted risk for invasive breast cancer, women with a score ≥1.67% were considered as at higher risk. Hazard ratios and 95% confidence intervals were calculated by a proportional hazards model for higher-risk versus lower-risk women receiving placebo, and for raloxifene versus placebo groups in both higher-risk and lower-risk women.

To assess the combined effects of the risk factors and their interactions, a multivariate Cox proportional hazards model was also fitted using data from both placebo and raloxifene groups. Variables considered were therapy (raloxifene versus placebo), age, age at menopause, BMI, estradiol level, bone mass and presence/absence of preexisting vertebral fracture (osteoporosis versus low bone mass), prior estrogen therapy, and a family history of breast cancer. For this analysis, data from women with estradiol levels of 5 to 10 pmol/L and ≥10 pmol/L were pooled, so that the study population was divided into approximately equally sized groups (<5 versus ≥5 pmol/L). The model was built by first including therapy and all risk factors without interaction terms (i.e., a main-effects only model). In a second multivariate model, those factors significant at the 0.2 level were retained in the model together with their two-way interactions. The final model included only factors (main-effects and two-way interactions) significant at the 0.2 level. This approach to model building was required because including all factors and two-way interactions in one model would result in an excessive number of terms relative to breast cancer cases (i.e., the total number of terms would exceed one third of the observed number of breast cancers).

There were no significant differences in MORE baseline characteristics between treatment groups in the 7,705 women participating in MORE (Table 1) or in those women who continued in CORE (N = 4,011; data not shown; P > 0.5).

Overall, 3,996 women had a Gail risk assessment at CORE baseline. Of these women, 1,847 had a 5-year predicted risk of invasive breast cancer <1.67% and 2,149 had a 5-year predicted risk ≥1.67%. The mean 5-year predicted risk scores in these two subgroups were 1.34% and 2.45%, respectively. The percentage of women with a 5-year predicted risk ≥1.67% did not differ significantly between the placebo and raloxifene treatment groups (52.7% and 54.3%, respectively; P > 0.05).

For the 7,705 women participating in MORE only or both MORE and CORE and followed for up to 8 years, 98 invasive breast cancer cases were confirmed [placebo, 58 (2.3%); raloxifene, 40 (0.78%); ref. 10]. For the 4,011 women participating in both MORE and CORE, 45 invasive breast cancer cases were confirmed over the 4 years of CORE [placebo, 24 (1.9%); raloxifene, 21 (0.77%); ref. 10].

In the placebo group, women predicted to be at higher breast cancer risk had a numerically greater incidence of breast cancer than those women considered at lower risk (Table 2). Women of ages ≥65 years were at a 1.8-fold greater risk of breast cancer than women of ages <65 years; women with a family history of breast cancer were at 2.3-fold greater risk of breast cancer than women with no family history; and women with an estradiol level of 5 to 10 pmol/L or >10 pmol/L were >2.5-fold greater risk of breast cancer than those with estradiol levels <5 pmol/L (Fig. 1). For age at menopause, BMI, bone mass and presence/absence of preexisting vertebral fracture, and prior use of estrogen preparations, the differences in breast cancer incidence between the higher-risk and lower-risk women were not statistically significant.

During the CORE period, the placebo group breast cancer incidence rate in women with a 5-year predicted risk <1.67% was 24 per 10,000 woman-years versus 64 per 10,000 woman-years in women with a 5-year predicted risk ≥1.67%. Women with a 5-year predicted risk ≥1.67% were at 2.7-fold greater risk of breast cancer than those with a 5-year predicted risk <1.67% (hazard ratio, 2.71; 95% confidence interval, 1.08-6.83).

For subgroups defined by MORE baseline characteristics, raloxifene therapy was associated with a significantly lower incidence of breast cancer over 8 years of follow-up in both women at lower risk and those at higher risk for breast cancer, except in women with estradiol level <5 pmol/L (Fig. 2); hazard ratio point estimates ranged from 0.11 to 0.52, corresponding to a 48% to 89% reduction in breast cancer risk associated with raloxifene therapy versus placebo. The reduction in risk of breast cancer with raloxifene versus placebo was numerically greater in those women considered at higher breast cancer risk, but, within each subgroup, the effect of raloxifene did not differ significantly between women at higher versus lower risk (P > 0.28), except in the case of family history of breast cancer where there was a statistically significant interaction between treatment and reduction in invasive breast cancer risk (P = 0.04). Raloxifene was associated with an 89% and 58% reduction, respectively, in invasive breast cancer risk versus placebo in women with and those without a family history of breast cancer.

During the 4-year CORE period, in women with a 5-year predicted risk ≥1.67%, the reduction in breast cancer risk with raloxifene was 67% (Fig. 3). In those women with a 5-year predicted risk <1.67%, raloxifene was associated with a 33% reduction in breast cancer risk versus placebo, but this reduction was not statistically significant (Fig. 3). The effect of raloxifene did not differ between those women at higher versus lower risk (P = 0.28). The absolute risk reductions in women with a 5-year predicted risk of ≥1.67% or <1.67% were 43.0 and 8.4 per 10,000 woman-years, respectively. The observed 5-year incidence of invasive breast cancer for higher-risk and lower-risk women, as defined by predicted 5-year risk, is shown by treatment group in Fig. 3.

In the initial multivariate analysis, age, estradiol level, family history of breast cancer (all factors that were significant in the univariate analyses), and therapy were significant at the 0.05 level and were retained for further modeling. In the model with all remaining main effects and their interactions, therapy and family history of breast cancer had a significant interaction (P = 0.024) whereas all other interactions were not significant (P > 0.37). In the final model, higher age and estradiol level were associated with 54% and 72% increase in invasive breast cancer risk, respectively (Table 3). The interaction between family history of breast cancer and therapy remained significant. Raloxifene was associated with a reduction in risk of breast cancer in women with and those without a family history of breast cancer. A family history of breast cancer was a risk factor for breast cancer in the placebo group, but not in the raloxifene group.

In this population of postmenopausal women with osteoporosis followed for up to 8 years, recognized risk factors for breast cancer were confirmed. In the placebo group, women of higher age, with a family history of breast cancer, and with a higher estradiol level or with a 5-year predicted risk ≥1.67% were at greater risk of invasive breast cancer. Higher BMI, lower age at menopause, prior estrogen therapy, and higher bone mass/no preexisting vertebral fracture were also associated with a numerically higher incidence of breast cancer but the difference from those considered at lower risk was not significant. In a multivariate analysis, the only risk factors in the final model were age, estradiol level, and a family history of breast cancer. These results suggest that, in this population and using cutoffs at or close to the median, age, estradiol level, and a family history of breast cancer were strong risk factors whereas age at menopause, BMI, and bone mass and presence/absence of preexisting vertebral fracture were weaker risk factors.

These results are in support of those from the MORE trial (11, 13), although a family history of breast cancer was not associated with a significant increase in invasive breast cancer risk in this MORE trial subgroup analysis (11). This is likely because of too few invasive breast cancer cases during MORE where in the placebo group only 1.9% of women with and 1.4% without a family history of breast cancer had invasive breast cancer (11) compared with 4.2% and 1.9%, respectively, at the conclusion of the CORE trial. In the previous subgroup analysis of MORE trial placebo data, the absence of preexisting vertebral fracture was also associated with a significant increase in breast cancer risk (11). A preexisting vertebral fracture subgroup analysis was not done as part of the current analyses; however, we have previously reported no difference in invasive breast cancer incidence between those with and those without preexisting vertebral fracture (P < 0.1) using the MORE/CORE trials data set (15). Moreover, we found no significant difference in invasive breast cancer incidence in women with osteoporosis (36% with preexisting vertebral fracture) versus those with low bone mass (excluded women with preexisting vertebral fracture) in the present study.

With the exception of women with low estradiol levels (<5 pmol/L), raloxifene therapy was associated with a significant 58% to 89% reduction in invasive breast cancer incidence in all subgroups defined by MORE baseline characteristics. The magnitude of risk reduction was not dependent on level of risk (higher versus lower risk), except in the case of family history of breast cancer. In this case, women with a family history of breast cancer exhibited a significantly greater risk reduction with raloxifene therapy compared to those with no family history. The treatment effect remained in the final multivariate model together with breast cancer risk factors of age, estradiol level, and family history of breast cancer. In this model, an interaction between treatment and family history was evident and, as in the univariate analysis, a greater risk reduction with raloxifene therapy was evident in women with a family history of breast cancer. Moreover, a family history of breast cancer was a risk factor for breast cancer in women receiving placebo but not in women receiving raloxifene. That is, raloxifene therapy seemed to reduce risk in women with a family history to a similar level to women without a family history of breast cancer. These findings are in support of those from the MORE trial alone (11). Interestingly, unlike in the present study, in a further analysis of MORE trial data, a significant interaction between estradiol and breast cancer risk reduction with raloxifene has been reported; i.e., the effect of raloxifene in reducing breast cancer incidence was greater in women with higher estradiol levels (13). In this earlier study, a fuller range of estradiol levels were analyzed than in the present study (0, >0 to <5, 5-10, and >10 pmol/L), but only in a subset of MORE trial participants. In the present analysis, raloxifene was associated with a significant reduction in invasive breast cancer risk in women with estradiol levels of 5 to 10 pmol/L and >10 pmol/L but not in women with the lowest estradiol levels (<5 pmol/L).

The 5-year predicted risk for invasive breast cancer calculated using the modified Gail model considers age and family history of breast cancer, as well as other risk factors for breast cancer. In the present study, women with a 5-year predicted risk ≥1.67% were at higher risk for invasive breast cancer, and in these women raloxifene therapy was associated with a significant 66% reduction in invasive breast cancer incidence. Indeed, treatment with raloxifene reduced the calculated 5-year incidence to <1.67%. In women with a 5-year predicted risk <1.67%, the reduction in breast cancer risk with raloxifene was not significant, most likely because of the too few breast cancer cases in this subgroup.

The benefits of any therapy must outweigh any risks associated with drug use, and both the benefits and risks are dependent on the characteristics of the individual receiving therapy. Raloxifene reduces vertebral fracture risk and increases bone mineral density in women with or at increased risk for osteoporosis (12), and is indicated for osteoporosis treatment and prevention. It is also associated with a reduced risk for invasive breast cancer (10) but is not approved by the U.S. Food and Drug Administration for breast cancer prevention. Based on the results of the present study, this effect was numerically greater in women at increased breast cancer risk. Raloxifene is associated with an increased risk for venous thromboembolism (16); it should not be prescribed in women with a history of venous thromboembolism and should be temporarily discontinued before and during prolonged immobilization (e.g., postsurgical recovery, prolonged bed rest). The raloxifene risk-benefit profile has previously been described with a global index (17).

There are a number of limitations to this analysis. Whereas invasive breast cancer was a primary end point of CORE, it was a secondary end point of MORE. Women were considered at higher or lower risk of breast cancer using a value at or close to the median as a cutoff for age, age at menopause, and BMI. These cutoffs were thus specific to this population of postmenopausal women of osteoporosis participating in the MORE trial. For example, whereas women were categorized as having low bone mass or osteoporosis based, in part, on their bone mineral density, the MORE population overall had relatively low bone mineral density. Only limited information on prior use of oral/transdermal estrogen-containing preparations was collected, so that consideration of type and duration of therapy was not possible in this present analysis. The data comprised those from the MORE and CORE trials, and not all women from MORE chose to participate in CORE. Thus, there was an opportunity for selection bias at the beginning of CORE, and women experiencing breast cancer or other adverse events in MORE would be expected to be less likely to participate in CORE. However, treatment groups remained balanced in CORE according to MORE baseline characteristics. Five-year predicted risk was calculated at the beginning of CORE and only for women who participated in CORE. Because raloxifene acts through the estrogen receptor, its effect in reducing breast cancer incidence seems to be limited to estrogen receptor–positive invasive breast cancer (10). The present analyses considered all invasive breast cancers, irrespective of estrogen receptor status, because the estrogen receptor status was not known for all of the 98 invasive breast cancers and, of those invasive breast cancers for which estrogen receptor status was known (n = 88), the majority (75%) were estrogen receptor positive. Identification of risk factors for estrogen receptor–positive breast cancer, as recently described by Cummings et al. (18), may be useful in identifying women who would more likely benefit from selective estrogen receptor modulator therapy.

The effect of raloxifene on invasive breast cancer incidence was specifically addressed in the recently completed Study of Tamoxifen and Raloxifene (STAR) and Raloxifene Use for the Heart (RUTH) trials. In the STAR trial, enrolling more than 19,000 women at high risk of breast cancer, raloxifene was as effective as tamoxifen in reducing the risk of invasive breast cancer (19). In the RUTH trial, enrolling more than 10,000 postmenopausal women with or at increased risk for coronary events, the incidence of invasive breast cancer was a significant 44% lower in women receiving raloxifene versus placebo (20).

In conclusion, raloxifene therapy was associated with a lower incidence of invasive breast cancer versus placebo in this population of postmenopausal women with osteoporosis (mean age, 67 years) irrespective of the presence or absence of risk factors for breast cancer. The effect of raloxifene was greater in women with versus those without a family history of breast cancer.

We thank Steve Zheng for statistical programming and analysis support, Melinda Rance for editorial assistance, and all the MORE and CORE investigators who made this study possible; a complete list has previously been published (10, 12).