Abstract

Background: High mammographic density is associated with increased breast cancer risk. Previous studies have shown that estrogens increase breast density on mammograms, but the effect on mammographic density of selective estrogen receptor modulators, such as raloxifene, is unknown. We assessed changes in mammographic density among women receiving placebo, raloxifene, or conjugated equine estrogens in an osteoporosis prevention trial. Methods: In a 5-year multicenter, double-blind, randomized, placebo-controlled osteoporosis prevention trial, healthy postmenopausal women who had undergone hysterectomy less than 15 years before the study and had no history of breast cancer received placebo, raloxifene (at one of two doses), or conjugated estrogens (ERT). Women from English-speaking investigative sites who had baseline and 2-year craniocaudal mammograms with comparable positioning (n = 168) were eligible for this analysis. Changes in mammographic density were determined by digital scanning and computer-assisted segmentation of mammograms and were analyzed with the use of analysis of variance. All statistical tests were two-sided. Results: Among the four treatment groups after 2 years on study, the mean breast density (craniocaudal view) was statistically significantly greater in the ERT group than it was in the other three groups (P<.01 for all three comparisons). Within treatment groups, the mean breast density from baseline to 2 years decreased statistically significantly in women receiving the placebo or either the higher or lower raloxifene dose (P = .003, P = .002, and P<.001, respectively) and showed a nonstatistically significant increase in women receiving ERT. Conclusions: In an osteoporosis prevention trial, raloxifene did not increase breast density after 2 years of treatment. Raloxifene administration should not interfere with, and could even enhance, mammographic detection of new breast cancers.

When mammographic screening programs were found in the early 1980s to decrease the mortality rate due to breast cancer (1), the use of mammograms became standard clinical practice. In addition to detecting existing breast cancer, mammography can also be useful for assessing breast density, which was first proposed as a risk factor for breast cancer in 1976 (2). High breast density is a strong predictor of breast cancer, behind only age and family history of breast cancer (3), and is directly correlated with a fourfold to sixfold increase in breast cancer risk (2,49). Identifying women at high risk for breast cancer is increasingly important as preventive therapies are being studied and becoming available for that population (10,11).

High mammographic density can obscure subtle breast abnormalities, making it not only more difficult to diagnose small-volume breast cancer but also more likely for a woman to have a false-positive mammogram reading (12). Approximately 50% of breast cancers are detected by mammography as masses (13). Such masses are more difficult to detect and to classify than are microcalcifications because surrounding radiographically dense epithelial or glandular tissue can obscure masses, especially in women with radiologically dense breasts (14).

Breast density, as measured by mammography, is determined by the relative amounts of fat and fibroglandular tissue present. Fat is radiolucent and appears dark on radiography. In contrast, fibroglandular tissue, which contains a mixture of fibrous connective tissue (stroma) and glandular tissue (epithelial cells), is more dense and appears white on mammograms. The Wolfe classification (15) uses visual assessments of the relative percentage of fat and fibroglandular tissue to determine breast density. This method identifies four parenchymal patterns (N1, P1, P2, and DY) that represent a spectrum from a low-density, fatty breast (N1) to a significantly dense breast (P2 and DY).

Breast density varies among individuals according to the inherent relative amounts of fat, connective tissue, and epithelial tissue present (16), all of which can be influenced by endogenous estrogen levels (17,18), the use of postmenopausal hormone replacement therapy (HRT) (19,20), and body mass index (21). A recently published subgroup analysis of the prospective, randomized Postmenopausal Estrogen/Progestin Interventions (PEPI) Trial (22) confirmed the results of previous observational studies; i.e., 3 years of estrogen alone or estrogen plus progestin increased mammographic density in 8% and 24% of women, respectively.

In light of the effects of estrogen to increase breast density, the effect of selective estrogen receptor modulators (SERMs) on breast density becomes increasingly important as more SERMs are developed and approved for use in postmenopausal women. SERMs interact with both α and β estrogen receptors, resulting in estrogen agonist or antagonist effects depending on the tissue (23,24). Raloxifene, a SERM with estrogen agonist effects on bone and lipid metabolism and estrogen antagonist effects in the breast and uterus, is approved in the United States for the prevention (25) and treatment (26) of postmenopausal osteoporosis.

A previous study (27) has shown that raloxifene reduces the incidence of invasive breast cancer in postmenopausal women with osteoporosis. However, raloxifene's effects on mammographic breast density have not been studied.

In this study, we evaluated the effects of 2 years of treatment with raloxifene, estrogen, or placebo on breast density using a digitized analysis of mammograms that quantifies the ratio of fibroglandular tissue to fat. We analyzed baseline and 2-year mammograms from a subgroup of postmenopausal women enrolled in an international osteoporosis prevention trial (28).

Subjects and Methods

Study Design and Subjects

The double-blind, randomized, placebo-controlled osteoporosis prevention trial of 619 postmenopausal women was conducted at 38 study centers in 10 countries in North America, Europe, South Africa, and New Zealand (28).

The osteoporosis prevention trial enrolled healthy postmenopausal women, aged 45 through 60 years, who had undergone hysterectomy no more than 15 years before the study. Women in this trial were to have lumbar spine bone mineral density (BMD) measurements in the range of 2 standard deviations (SDs) above to 2.5 SDs below the mean peak lumbar spine BMD for premenopausal women, inclusive. Women were excluded from the osteoporosis study if they had any of the following: breast cancer at any time or any cancer within the previous 5 years, thromboembolic disorders or cerebrovascular accident, diabetes mellitus, chronic liver disease, impaired kidney function, or clinically significant menopausal symptoms.

Systemic HRT was permitted before, but not during, the osteoporosis study and only if it had been discontinued at least 6 months before study entry. Low-dose vaginal estrogens, except estradiol, were permitted up to three times per week during the study. Concomitant use of progestins, androgens, bone-active agents, or other SERMs was not permitted. Women who had ever used systemic fluoride therapy (except for dental prophylaxis) or had taken bisphosphonates were excluded.

In the osteoporosis trial, a randomized block design was used to assign women to receive either raloxifene HCl (Evista®; Eli Lilly and Company, Indianapolis, IN) at a dose of 60 mg/day or 150 mg/day, conjugated equine estrogens (Premarin®; Wyeth Ayerst, Philadelphia, PA [ERT]) at a dose of 0.625 mg/day, or placebo. Randomization was stratified within each investigative site. Because raloxifene and Premarin pills have different appearances, all women took two pills daily—raloxifene or a placebo of identical appearance and Premarin or a placebo of identical appearance—to maintain the double-blind design of the study. Study visits were scheduled every 3 months for 2 years.

The protocol was approved by local ethical review boards, and all women provided written informed consent for participation in accordance with the principles outlined in the Declaration of Helsinki.

Women were selected for participation in this breast density substudy if they were enrolled at English-speaking sites, had both baseline and 2-year mammograms of the craniocaudal (CC) view that had comparable positioning, and met the osteoporosis prevention study criteria described above (Fig. 1). The analysis of this subset of women was planned before any mammograms were digitized and without prior knowledge of which patients at eligible sites would participate. The “Appendix” section lists the investigators who participated in this breast density substudy.

Breast Density Measurements

Mediolateral oblique (MLO) and CC views of each breast were obtained at baseline and at 2 years. Mammograms taken up to 1 year before study entry were considered to be “baseline.” Because the statistical analysis results from the CC and MLO views were similar, all results presented are from the CC view from both breasts. Films were scanned (Lumiscan Model 150; Lumisys, Inc., Sunnyvale CA), and the scanned images were segmented (see description below) by the radiologist with a computer-assisted technique called “interactive thresholding.” The interactive threshold technique uses software that allows the radiologist to draw a boundary around objects that have similar degrees of whiteness on the digitized breast image. To accommodate mammograms with different exposures, a threshold level of whiteness is selected by the radiologist interacting with the computer. Using this method, the radiologist can outline one larger connected area or several separate areas of fibroglandular tissue at the same time in a process called “segmentation.” The computer then calculates the area of tissue that has been interactively outlined. The breast region that projects behind the axillary fold was marked and excluded from further analysis.

With the use of this interactive segmentation software developed at Georgetown University Medical Center (Washington, DC), the edge of the breast was first defined by an automated process. One radiologist who was blinded to therapy (M. Freedman), aided by the software, assigned a threshold to the digitized mammograms to define areas of dense breast tissue. A previous study (29) has indicated that this method results in very low intra-observer and inter-observer variability. The ratio of the area of dense breast tissue to the total breast area is the percent density.

An example of the results of interactive thresholding is shown in Fig. 2. Fig. 2 represents a 2-year digitized mammogram from one subject (CC view, left breast); the right panel in the figure represents the segmentation of the same mammogram, showing fibroglandular tissue (dark area) and surrounding tissue, as defined by this interactive thresholding method. This example is from a subject assigned to receive raloxifene at a dose of 60 mg/day who had an endpoint density of 7.9%.

Each subject's baseline and 2-year films were viewed together (the temporal sequence of the mammograms was known) to ensure that similar threshold criteria were used, to detect any minor differences in positioning that might limit the accuracy of comparison, to ensure that the films had not been mislabeled, and to detect any artifacts (e.g., silicone breast implants) that might interfere with segmentation. Subjects whose pairs of films showed marked differences in the position of the breast on the radiograph, such that glandular tissue observed on one view could not be observed on the other view, were excluded from further analysis.

Statistical Analysis

Films were analyzed for all subjects who had a baseline and a 2-year mammogram where the positions of both breasts in both mammograms were comparable. For subgroup analyses, continuous demographic variables (age, years since menopause, and body mass index) and baseline density were categorized into tertiles; other subgroup variables included previous HRT use, use of alcohol at entry, and smoking at entry. Subgroup analyses were performed with the use of analysis of variance (ANOVA). Subject demographic variables were analyzed for potential treatment-group differences with the use of a one-way ANOVA for continuous variables and a chi-square test for categorical variables. All statistical tests were two-sided.

The endpoint of this study, change in breast density from baseline to 2 years, was analyzed with the use of ANOVA, with treatment and country as fixed effects in the model. Mammographic density is presented as percent density, calculated as described previously, for both breasts combined. This density definition is a weighted average of the left and right breast densities.

Since breast densities measured by this method have not been described previously, a cutoff value for clinically relevant endpoints had not been determined. We, therefore, chose to use the SD of endpoints in the placebo group in this study to define a potentially clinically relevant endpoint. Any change in breast density greater than the mean change in the placebo group plus 1 SD was considered to be potentially clinically relevant; likewise, any decrease greater than the mean placebo change minus 1 SD was considered to be potentially clinically relevant. This definition of clinical relevance was made before the data had been analyzed and is clearly not the only one that could have been selected. However, this cutoff seemed statistically reasonable, as did the use of the placebo group as an internal reference range.

Results

Of the 235 participating patients, 38 were lost to follow-up or had mammograms that were missing or had substantial differences in breast position between baseline and 2 years. Of the remaining 197 women with paired mammograms with comparable positioning, 168 had CC mammograms from both breasts that showed no substantial differences in position between the baseline and the 2-year films (Fig. 1). Data from these 168 patients are presented here. The mean age and years after menopause were 53 years and 6 years, respectively; additional patient characteristics are presented in Table 1. There were no statistically significant therapy-group differences for any patient characteristic at baseline, and the characteristics of the subset of women enrolled in this study were not different from those of the other 451 subjects enrolled in the overall osteoporosis prevention trial (data not shown).

After 2 years of treatment, the change in mean breast density in the ERT group (+1.2%) was statistically significantly different from that in each of the other groups (−1.3%, −1.5%, and −1.7% for the groups receiving placebo, raloxifene at a dose of 60 mg/day, and raloxifene at a dose of 150 mg/day, respectively; P<.01 for all three comparisons). The mean density within the placebo group and both raloxifene groups decreased statistically significantly from baseline (P<.02 for each group), while the mean density in the ERT group increased but not statistically significantly (Table 2). Among all of the variables tested (age, years since menopause, body mass index, previous HRT use, use of alcohol at entry, smoking at entry, and baseline breast density), there were no statistically significant therapy-by-subgroup interactions.

Table 3 lists the percentage of patients in each group who experienced a change in breast density greater than 1 SD above or below the mean change observed in the placebo group; these changes were considered to be “potentially clinically relevant.” Potentially clinically relevant changes, according to our definition, were observed for 11.1% of women on placebo, 6.7% of women on raloxifene at a dose of 60 mg/day, 9.5% of women on raloxifene at a dose of 150 mg/day, and 30.6% of women on ERT (chi-square; P = .010).

Most (98%) of the patients in this substudy reported at least one adverse event during the first 24 months of the trial. Breast pain was reported by 18 patients (10.7%), of which four were assigned to placebo, two were assigned to raloxifene at 60 mg/day, four were assigned to raloxifene at 150 mg/day, and eight were assigned to ERT. (The therapy-group differences were not statistically significant for the proportion of patients reporting this adverse event.) No cases of breast cancer were reported for women participating in this breast density substudy.

Discussion

In this subset of 168 postmenopausal women enrolled in an international osteoporosis prevention trial, the mean breast density decreased by 1.3% in the group receiving placebo, by 1.5% in the group receiving raloxifene at a dose of 60 mg/day, and by 1.7% in the group receiving raloxifene at a dose of 150 mg/day, but it increased in the ERT-treated group by 1.2% at 2 years. The decreases in mean breast density observed with placebo and raloxifene may be attributed partially to age-related decreases in breast density (30). In an analysis of changes either 1 SD above or 1 SD below the mean change in breast density observed in the placebo group, which we considered to be potentially clinically relevant, twice as many women assigned to receive raloxifene had decreases in breast density below the lower 1-SD threshold (13.3% and 19.0% for the groups receiving 60 mg/day and 150 mg/day, respectively) as had increases above the upper 1-SD threshold (6.7% and 9.5%, respectively). In contrast, three times as many women assigned to ERT had a mean change in breast density that was an increase above the upper 1-SD threshold (30.6%) as had a decrease below the lower 1-SD threshold (11.1%).

It has been reported previously that treatment with raloxifene, in addition to its effects on bone density, results in a 76% reduction in the incidence of invasive breast cancer in postmenopausal women with osteoporosis by 40 months of therapy (26,27). This is consistent with the estrogen-antagonist effect of raloxifene in the breast, as is the lack of an effect of raloxifene on breast density observed in the current study. Because raloxifene does not increase breast density, such treatment should also not impede the detection of new breast cancer by mammography.

The absolute mean breast densities measured in this trial were smaller than those observed in another study (22) with similar study populations, a difference that is due, in part, to the various methods used for measuring breast density. In the most commonly used method, radiologists visually inspect mammograms and assign them to one of four subjective categories of breast density (6,8,18,31). Other methods estimate either the percentage of parenchymal tissue present (32) or the changes in density by comparing mammograms taken at different times (19,22,33). All of these methods are visual, subjective, and qualitative. In contrast, quantitative methods that exclude interspersed fat from the analysis of glandular tissue by manual segmentation (6,8,31) or interactive computer analysis (4,34), as was done in this study, measure only glandular tissue and result in overall lower breast density values.

In the recently published PEPI trial (22), 8% of the patients on ERT and 20%–24% of those on HRT (estrogen plus progesterone) had increases in breast density at 12 months. These women had to have at least a “moderate” change in breast density for a change to be recognized in the four-category system used to analyze data in that study. Compared with the women in our trial, PEPI trial participants were about 6 years older, half as likely to have ever used HRT, and were not required to have undergone hysterectomy. Compared with the ERT group in the PEPI trial, a higher percentage (30.6%) of the ERT group in our trial demonstrated a potentially clinically relevant change. This difference is most likely due to the increased sensitivity of the method employed in our study. Despite the differences in density evaluation methods and populations, statistically significantly more patients receiving ERT in both studies developed increased breast density compared with those receiving placebo. The effects of other SERMs, such as tamoxifen or toremifene, on breast density have not been reported in healthy postmenopausal women without breast cancer.

As new therapies, such as SERMs, become available, it is desirable to understand whether specific characteristics predict enhanced response to therapy. For example, some subgroups of women (e.g., those with a higher percentage of glandular tissue) might be more responsive to estrogen or raloxifene, since both therapies work through estrogen receptors that are present in glandular tissue. However, while this study was not sufficiently powered to detect such differences, a trend in that direction was observed.

This study has two primary limitations. First, because this trial used ERT instead of HRT (estrogen plus progesterone), we could not discern the effects of progestin on breast density separately from the effects of estrogen. In other trials (20,22), estrogen plus progestin (given cyclically or in combination) increased breast density beyond the increase observed for estrogen alone. Second, although the digitizing method used here is similar to one published previously (3), to our knowledge, this is the first clinical trial application of this particular method. We are, therefore, unable to directly compare our results with the results of other studies.

Thus, in a study of healthy postmenopausal women enrolled in an international osteoporosis prevention trial, raloxifene given at a dose of 60 mg/day or 150 mg/day did not increase breast density at 2 years. These findings suggest that raloxifene therapy should not impair detection of new breast cancer by mammography. The lack of apparent mammary parenchymal stimulation by raloxifene clearly distinguishes its effects on breast tissue from those of conjugated estrogens, which reproducibly increase breast density. These different effects on breast density may be of benefit in evaluating screening mammograms for women who are receiving raloxifene or HRT to prevent or treat osteoporosis.

Appendix

The following investigators participated in this breast density substudy: Australia—J. A. Eisman, Bone and Mineral Research Program, Garvan Institute of Medical Research, St. Vincent's Hospital, Sydney; A. G. Need, Department of Clinical Biochemistry, Institute of Medical and Veterinary Science, Adelaide; K. W. Ng, Department of Medicine, Saint Vincent's Institute of Medical Research, Fitzroy; R. L. Prince, Department of Medicine, Sir Charles Gairdner Hospital, Nedlands; and E. Seeman, Department of Endocrinology, The Austin Hospital, Heidelberg. Canada—J. D. Adachi, St. Joseph's Hospital, Hamilton; J. P. Brown, Le Centre Hospitalier de L'Universite Laval, Sainte-Foy; D. A. Hanley, Division of Endocrinology and Metabolism, University of Calgary; A. B. Hodsman, Saint Joseph's Health Center, London; R. G. Josse, Division of Endocrinology and Metabolism, Saint Michael's Hospital, Toronto; W. P. Olszynski, Midtown Medical Centre, Saskatoon; L. G. Ste-Marie, Hospital St. Luc, Viallet, Montreal; and J. P. Wade, The Arthritis Centre, Vancouver. United Kingdom—R. Eastall, Northern General Hospital, University of Sheffield; I. Fogelman, Osteoporosis Department, Guys Hospital, London; and I. Smith and S. Young, Coppull Menopause and Osteoporosis Research Unit, Lancashire. New Zealand—I. R. Reid, University of Auckland School of Medicine. South Africa—B. H. Ascott-Evans, Department of Nuclear Medicine, Groote Schuur Hospital, Cape Town; S. Hough, Department of Endocrinology and Metabolism, Tygerberg Hospital; and C. M. Schnitzler, Department of Orthopaedic Surgery, University of the Witwaterstrand, Parktown. United States—J. F. Aloia, Winthrop-University Hospital, Mineola, NY; H. G. Bone III, St. John Hospital Bone and Mineral Clinic, Detroit, MI; C. Cheney, Scripps Clinical Research Foundation, La Jolla, CA; M. H. Davidson, Chicago Center for Clinical Research, IL; S. Greenspan, Beth Israel Hospital, Boston, MA; A. A. Licata, Department of Endocrinology, Osteoporosis/Metabolic Bone Disease Clinic, Cleveland, OH; M. R. McClung, Center for Osteoporosis Research, Portland, OR; M. T. McDermott, Austin Diagnostic Clinic, Center for Clinical Research, TX; P. D. Miller, Colorado Center for Bone Research, Lakewood; R. A. Sachson, Research Institute of Dallas, TX; and N. B. Watts, Division of Endocrinology, Emory Clinic, Atlanta, GA.

Table 1.

Baseline characteristics of study subjects*

 Treatment group 
  Raloxifene  
Characteristic Placebo (n = 45) 60 mg/day (n = 45) 150 mg/day (n = 42) Estrogen, 0.625 mg/day (n =  
*There were no statistically significant therapy-group differences for any baseline patient characteristic, and the subset of women enrolled in this study was not different from the rest of the subjects enrolled in the larger trial (data not shown). kg/m2 = (weight in kilograms)/(height in meters)2; HRT = hormone replacement therapy; CC = craniocaudal. 
†Data are presented as mean±standard deviation. 
‡Patients' self-reported consumption of more than three alcoholic drinks per week. 
Age, y† 52.2±5.1 54.1±4.2 53.3±4.8 52.1±4.6 
Years since menopause † 5.9±4.1 5.5±3.9 6.6±4.1 6.0±4.5 
Body mass index, kg/m2† 28.7±4.7 27.1±4.8 28.1±4.6 27.0±5.2 
Previous HRT use, yes 33% 27% 33% 36% 
Use of alcohol at entry‡ 22% 27% 24% 33% 
Smoking at entry, yes 13% 24% 17% 17% 
Baseline breast density, CC view† 9.8%±9.6% 9.3%±9.1% 8.1%±6.6% 13.5%±11.5% 
 Treatment group 
  Raloxifene  
Characteristic Placebo (n = 45) 60 mg/day (n = 45) 150 mg/day (n = 42) Estrogen, 0.625 mg/day (n =  
*There were no statistically significant therapy-group differences for any baseline patient characteristic, and the subset of women enrolled in this study was not different from the rest of the subjects enrolled in the larger trial (data not shown). kg/m2 = (weight in kilograms)/(height in meters)2; HRT = hormone replacement therapy; CC = craniocaudal. 
†Data are presented as mean±standard deviation. 
‡Patients' self-reported consumption of more than three alcoholic drinks per week. 
Age, y† 52.2±5.1 54.1±4.2 53.3±4.8 52.1±4.6 
Years since menopause † 5.9±4.1 5.5±3.9 6.6±4.1 6.0±4.5 
Body mass index, kg/m2† 28.7±4.7 27.1±4.8 28.1±4.6 27.0±5.2 
Previous HRT use, yes 33% 27% 33% 36% 
Use of alcohol at entry‡ 22% 27% 24% 33% 
Smoking at entry, yes 13% 24% 17% 17% 
Baseline breast density, CC view† 9.8%±9.6% 9.3%±9.1% 8.1%±6.6% 13.5%±11.5% 
Table 2.

Breast density at baseline and after 2 years by treatment group*

 Treatment group 
  Raloxifene  
Breast density Placebo (n = 45) 60 mg/day (n = 45) 150 mg/day (n = 42) Estrogen, 0.625 mg/day (n = 36) 
*Data are presented as mean±standard deviation for craniocaudal views. P values are based on Student's t test. 
†At 2 years, the mean change in breast density in the estrogen group was statistically significantly greater than that in each of the other three groups (P<.01 for all three comparisons). 
‡95% confidence intervals for within-group mean changes. 
Baseline 9.8%±9.6% 9.3%±9.1% 8.1%±6.6% 13.5%±11.5% 
2 y 8.5%±11.4% 7.7%±8.8% 6.4%±5.3% 14.6%±11.4% 
Mean change −1.3%±2.9% −1.5%±4.1% −1.7%±3.4% +1.2%±5.3%† 
95% confidence interval‡ −2.2 to −0.4 −2.7 to −0.3 −2.8 to −0.6 −0.6 to 3.0 
P .003 .002 <.001 .611 
 Treatment group 
  Raloxifene  
Breast density Placebo (n = 45) 60 mg/day (n = 45) 150 mg/day (n = 42) Estrogen, 0.625 mg/day (n = 36) 
*Data are presented as mean±standard deviation for craniocaudal views. P values are based on Student's t test. 
†At 2 years, the mean change in breast density in the estrogen group was statistically significantly greater than that in each of the other three groups (P<.01 for all three comparisons). 
‡95% confidence intervals for within-group mean changes. 
Baseline 9.8%±9.6% 9.3%±9.1% 8.1%±6.6% 13.5%±11.5% 
2 y 8.5%±11.4% 7.7%±8.8% 6.4%±5.3% 14.6%±11.4% 
Mean change −1.3%±2.9% −1.5%±4.1% −1.7%±3.4% +1.2%±5.3%† 
95% confidence interval‡ −2.2 to −0.4 −2.7 to −0.3 −2.8 to −0.6 −0.6 to 3.0 
P .003 .002 <.001 .611 
Table 3.

Percentage of patients with a change in breast density beyond the 1-standard-deviation threshold observed in the placebo group at 2 years*

 Treatment group, % 
  Raloxifene  
Breast density change Placebo (n = 45) 60 mg/day (n = 45) 150 mg/day (n = 42) Estrogen, 0.625 mg/day (n = 36) 
*Changes in breast density were determined for craniocaudal view mammograms. Changes greater than 1 standard deviation (SD) above or below the mean change observed in the placebo group were considered to be potentially clinically relevant. There was a significant therapy difference for the proportion of women whose breast density increased beyond the 1-SD threshold (1.6%) versus those who did not (two-sided chi-square, P = .010). The therapy difference was not significant for the proportion of women whose breast density decreased beyond the 1-SD threshold (4.2%) versus those who did not (two-sided chi-square, P = .735). 
Increase >1 SD above the mean placebo change 11.1 6.7 9.5 30.6 
Within 1 SD of mean placebo change 71.1 80.0 71.4 58.3 
Decrease >1 SD below the mean placebo change 17.8 13.3 19.0 11.1 
 Treatment group, % 
  Raloxifene  
Breast density change Placebo (n = 45) 60 mg/day (n = 45) 150 mg/day (n = 42) Estrogen, 0.625 mg/day (n = 36) 
*Changes in breast density were determined for craniocaudal view mammograms. Changes greater than 1 standard deviation (SD) above or below the mean change observed in the placebo group were considered to be potentially clinically relevant. There was a significant therapy difference for the proportion of women whose breast density increased beyond the 1-SD threshold (1.6%) versus those who did not (two-sided chi-square, P = .010). The therapy difference was not significant for the proportion of women whose breast density decreased beyond the 1-SD threshold (4.2%) versus those who did not (two-sided chi-square, P = .735). 
Increase >1 SD above the mean placebo change 11.1 6.7 9.5 30.6 
Within 1 SD of mean placebo change 71.1 80.0 71.4 58.3 
Decrease >1 SD below the mean placebo change 17.8 13.3 19.0 11.1 
Fig. 1.

Patients recruited, enrolled, and analyzed. *Unable to be evaluated because of the presence of breast implants or substantial weight loss. RLX 60 = raloxifene at a dose of 60 mg/day; RLX 150 = raloxifene at a dose of 150 mg/day; CC = craniocaudal.

Fig. 1.

Patients recruited, enrolled, and analyzed. *Unable to be evaluated because of the presence of breast implants or substantial weight loss. RLX 60 = raloxifene at a dose of 60 mg/day; RLX 150 = raloxifene at a dose of 150 mg/day; CC = craniocaudal.

Fig. 2.

Paired digitized and segmented mammogram. Left panel: digitized mammogram (craniocaudal view, left breast) from a subject treated with raloxifene at a dose of 60 mg/day who had 2-year breast density of 7.9%. Right panel: segmentation of the same mammogram, showing fibroglandular tissue (dark area) and surrounding tissue, as defined by this interactive thresholding method. Areas of glandular tissue intermixed with fat appear as areas of spotty or irregular intermixed black and gray.

Fig. 2.

Paired digitized and segmented mammogram. Left panel: digitized mammogram (craniocaudal view, left breast) from a subject treated with raloxifene at a dose of 60 mg/day who had 2-year breast density of 7.9%. Right panel: segmentation of the same mammogram, showing fibroglandular tissue (dark area) and surrounding tissue, as defined by this interactive thresholding method. Areas of glandular tissue intermixed with fat appear as areas of spotty or irregular intermixed black and gray.

Supported by funds from Eli Lilly and Company, the manufacturer of Evista® (raloxifene). M. Freedmen received an honorarium from Eli Lilly and Company, and M. Lippman served on an advisory board for Eli Lilly and Company.

We thank Joan E. Glusman, M.D., for study design; Teri Scott, M.S.N., for study coordination and data management; Jeffrey Collmann, Ph.D., and Cerone Williams, R.N., for study execution and coordination; and Michele Y. Hill for editorial assistance.

References

1
Tabar L, Fagerberg G, Chen HH, Duffy SW, Smart CR, Gad A, et al. Efficacy of breast cancer screening by age. New results from the Swedish Two-County Trial.
Cancer
 
1995
;
75
:
2507
–17.
2
Wolfe JN. Risk for breast cancer development determined by mammographic parenchymal pattern.
Cancer
 
1976
;
37
:
2486
–92.
3
Byng JW, Yaffe MJ, Jong RA, Shumak RS, Lockwood GA, Tritchler DL, et al. Analysis of mammographic density and breast cancer risk from digitized mammograms.
Radiographics
 
1998
;
18
:
1587
–98.
4
Boyd NF, Byng JW, Jong RA, Fishell EK, Little LE, Miller AB, et al. Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study.
J Natl Cancer Inst
 
1995
;
87
:
670
–5.
5
Boyd NF, O'Sullivan B, Campbell JE, Fishell E, Simor I, Cooke G, et al. Mammographic signs as risk factors for breast cancer.
Br J Cancer
 
1982
;
45
:
185
–93.
6
Wolfe JN, Saftlas AF, Salane M. Mammographic parenchymal patterns and quantitative evaluation of mammographic densities: a case– control study.
AJR Am J Roentgenol
 
1987
;
148
:
1087
–92.
7
Brisson J, Verreault R, Morrison AS, Tennina S, Meyer F. Diet, mammographic features of breast tissue, and breast cancer risk.
Am J Epidemiol
 
1989
;
130
:
14
–24.
8
Saftlas AF, Hoover RN, Brinton LA, Szklo M, Olson DR, Salane M, et al. Mammographic densities and risk of breast cancer.
Cancer
 
1991
;
67
:
2833
–8.
9
Byrne C, Schairer C, Wolfe J, Parekh N, Salane M, Brinton LA, et al. Mammographic features and breast cancer risk: effects with time, age, and menopause status.
J Natl Cancer Inst
 
1995
;
87
:
1622
–9.
10
Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, Cronin WM, et al. Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study.
J Natl Cancer Inst
 
1998
;
90
:
1371
–88.
11
Hartmann LC, Schaid DJ, Woods JE, Crotty TP, Myers JL, Arnold PG, et al. Efficacy of bilateral prophylactic mastectomy in women with a family history of breast cancer.
N Engl J Med
 
1999
;
340
:
77
–84.
12
Kavanagh AM, Mitchell H, Giles GG. Hormone replacement therapy and accuracy of mammographic screening.
Lancet
 
2000
;
355
:
270
–4.
13
Nishikawa RM, Giger ML, Doi K, Metz CE, Yin FF, Vyborny CJ, et al. Effect of case selection on the performance of computer-aided detection schemes.
Med Phys
 
1994
;
21
:
265
–9.
14
Li H, Lo SC, Wang Y, Freedman MT, Mun SK. Mammographic mass detection by stochastic modeling and a multi-nodular neural network.
Proc SPIE
 
1997
;
3034
:
478
–88.1997.
15
Wolfe JN. A study of breast parenchyma by mammography in the normal woman and those with benign and malignant disease.
Radiology
 
1967
;
89
:
201
–5.
16
Feig SA. Breast masses. Mammographic and sonographic evaluation.
Radiol Clin North Am
 
1992
;
30
:
67
–92.
17
Rosenberg RD, Hunt WC, Williamson MR, Gilliland FD, Wiest PW, Kelsey CA, et al. Effects of age, breast density, ethnicity, and estrogen replacement therapy on screening mammographic sensitivity and cancer stage at diagnosis: review of 183,134 screening mammograms in Albuquerque, New Mexico.
Radiology
 
1998
;
209
:
511
–8.
18
White E, Velentgas P, Mandelson MT, Lehman CD, Elmore JG, Porter P, et al. Variation in mammographic breast density by time in menstrual cycle among women aged 40–49 years.
J Natl Cancer Inst
 
1998
;
90
:
906
–10.
19
Persson I, Thurfjell E, Holmberg L. Effect of estrogen and estrogen–progestin replacement regimens on mammographic breast parenchymal density.
J Clin Oncol
 
1997
;
15
:
3201
–7.
20
Lundstrom E, Wilczek B, von Palffy Z, Soderqvist G, von Schoultz B. Mammographic breast density during hormone replacement therapy: differences according to treatment.
Am J Obstet Gynecol
 
1999
;
181
:
348
–52.
21
Boyd NF, Lockwood GA, Byng JW, Little LE, Yaffe MJ, Tritchler DL. The relationship of anthropometric measures to radiological features of the breast in premenopausal women.
Br J Cancer
 
1998
;
78
:
1233
–8.
22
Greendale GA, Reboussin BA, Sie A, Singh HR, Olson LK, Gatewood O, et al. Effects of estrogen and estrogen–progestin on mammographic parenchymal density. Postmenopausal Estrogen/Progestin Interventions (PEPI) Investigators.
Ann Intern Med
 
1999
;
130
:
262
–9.
23
Yang NN, Venugopalan M, Hardikar S, Glasebrook A. Identification of an estrogen response element activated by metabolites of 17beta-estradiol and raloxifene [published erratum appears in Science 1997;275:1249].
Science
 
1996
;
273
:
1222
–5.
24
Bryant HU, Dere WH. Selective estrogen receptor modulators: an alternative to hormone replacement therapy.
Proc Soc Exp Biol Med
 
1998
;
217
:
45
–52.
25
Delmas PD, Bjarnason NH, Mitlak BH, Ravoux AC, Shah AS, Huster WJ, et al. Effects of raloxifene on bone mineral density, serum cholesterol concentrations, and uterine endometrium in postmenopausal women.
N Engl J Med
 
1997
;
337
:
1641
–7.
26
Ettinger B, Black DM, Mitlak BH, Knickerbocker RK, Nickelsen T, Genant HK, et al. Reduction of vertebral fracture risk in postmenopausal women with osteoporosis treated with raloxifene: results from a 3-year randomized clinical trial. Multiple Outcomes of Raloxifene Evaluation (MORE) Investigators.
JAMA
 
1999
;
282
:
637
–45.
27
Cummings SR, Eckert S, Krueger KA, Grady D, Powles TJ, Cauley JA, et al. The effect of raloxifene on risk of breast cancer in postmenopausal women: results from the MORE randomized trial. Multiple Outcomes of Raloxifene Evaluation.
JAMA
 
1999
;
281
:
2189
–97.
28
Cohen FJ, Eckert S, Mitlak BH. Geographic differences in bone turnover: data from a multinational study in healthy postmenopausal women.
Calcif Tissue Int
 
1998
;
63
:
277
–82.
29
Byng JW, Boyd NF, Fishell E, Jong RA, Yaffe MJ. The quantitative analysis of mammographic densities.
Phys Med Biol
 
1994
;
39
:
1629
–38.
30
Stomper PC, D'Souza DJ, DiNitto PA, Arredondo MA. Analysis of parenchymal density on mammograms in 1353 women 25–79 years old.
AJR Am J Roentgenol
 
1996
;
167
:
1261
–5.
31
Laya MB, Gallagher JC, Schreiman JS, Larson EB, Watson P, Weinstein L. Effect of postmenopausal hormonal replacement therapy on mammographic density and parenchymal pattern.
Radiology
 
1995
;
196
:
433
–7.
32
Leung W, Goldberg F, Zee B, Sterns E. Mammographic density in women on postmenopausal hormone replacement therapy.
Surgery
 
1997
;
122
:
669
–73; discussion 673–4.
33
Berkowitz JE, Gatewood OM, Goldblum LE, Gayler BW. Hormonal replacement therapy: mammographic manifestations.
Radiology
 
1990
;
174
:
199
–201.
34
Byng JW, Yaffe MJ, Lockwood GA, Little LE, Tritchler DL, Boyd NF. Automated analysis of mammographic densities and breast carcinoma risk.
Cancer
 
1997
;
80
:
66
–74.