© The Author 2007. Published by Oxford University Press.
ARTICLES |
Longitudinal Measurement of Clinical Mammographic Breast Density to Improve Estimation of Breast Cancer Risk
For the National Institutes of Health Breast Cancer Surveillance Consortium
Affiliations of authors: Departments of Epidemiology and Biostatistics (KK, RSB) and Radiology (RSB), and General Internal Medicine Section, Department of Veterans Affairs (KK), University of California, San Francisco, CA; Group Health Center for Health Studies, Seattle, WA (LI, DLM, DSMB); Department of Biostatistics, University of Washington, Seattle, WA (DLM); Departments of Medical Biostatistics and Pathology, University of Vermont, College of Medicine, Burlington, VT (PMV); Department of Radiology, University of North Carolina, Chapel Hill, NC (BY); Departments of Family Medicine and Public Health and Preventive Medicine, Oregon Health and Science University, Portland, OR (PAC); Applied Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD (RBB)
Correspondence to: Karla Kerlikowske, MD, General Internal Medicine Section, San Francisco Veterans Affairs Medical Center, 111A1, 4150 Clement St, San Francisco, CA 94121 (e-mail: karla.kerlikowske{at}ucsf.edu).
 |
ABSTRACT |
|---|
 Top Abstract Context and CaveatsMethodsResultsDiscussionReferencesNotes |
|---|
Background: Whether a change over time in clinically measured mammographic breast density influences breast cancer risk is unknown.
Methods: From January 1993 to December 2003, data that included American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) breast density categories (1–4 in order of increasing density) were collected prospectively on 301 955 women aged 30 and older who were not using postmenopausal hormone replacement therapy and underwent at least two screening mammography examinations; 2639 of the women were diagnosed with breast cancer within 1 year of the last examination. Women's first and last BI-RADS breast density (average 3.2 years apart) and logistic regression were used to model the odds of having invasive breast cancer or ductal carcinoma in situ diagnosed within 12 months of the last examination by change in BI-RADS category. Rates of breast cancer adjusted for age, mammography registry, and time between screening examinations were estimated from this model. All statistical tests were two-sided.
Results: The rate (breast cancers per 1000 women) of breast cancer was higher if BI-RADS breast density category increased from 1 to 2 (5.6, 95% confidence interval [CI] = 4.7 to 6.9) or 1 to 3 (9.9, 95% CI = 6.4 to 15.5) compared to when it remained at BI-RADS density of 1 (3.0, 95% CI = 2.3 to 3.9; P<.001 for trend). Similar and statistically significant trends between increased or decreased density and increased or decreased risk of breast cancer, respectively, were observed for women whose breast density category was initially 2 or 3 and changed categories. BI-RADS density of 4 on the first examination was associated with a high rate of breast cancer (range 9.1–13.4) that remained high even if breast density decreased.
Conclusion: An increase in BI-RADS breast density category within 3 years may be associated with an increase in breast cancer risk and a decrease in density category with a decrease in risk compared to breast cancer risk in women in whom breast density category remains unchanged. Two longitudinal measures of BI-RADS breast density may better predict a woman's risk of breast cancer than a single measure.
Prior knowledge High mammographic breast density as measured by the radiologist in the clinic is strongly associated with an increased risk of breast cancer. It was not known whether temporal increases or decreases in a woman's breast density affects her risk of developing breast cancer. Study design The risk of breast cancer was estimated from clinical breast density data that was collected prospectively from registries linked to state tumor registries or regional SEER program data on breast cancer incidence. Contribution This study showed that two breast density measurements separated by an average of 3 years predicted the odds that a women would develop breast cancer more accurately than one measure. Implications Both current and previous breast density measurements should be used by clinicians when evaluating risk with patients. Limitations Further study will be required to determine the relative contributions of the increased accuracy provided by two measures on the one hand and real changes in breast density on the other to the apparent association of temporal changes in breast density and breast cancer risk.
|
Breast density is one of the strongest predictors of breast cancer risk. Women who have dense tissue occupying greater than 50% of the area of a mammographic breast image have high breast density and are at three- to five-fold greater risk of breast cancer than women with less than 25% dense area (1–5). About 30% of postmenopausal women have high breast density, a frequency that is greater than the frequency of most recognized risk factors; for example, a family history of breast cancer occurs in only 10% of women (6). Both quantitative (1–3,7,8) and qualitative (4,5,7–11) measures of breast density have been used to demonstrate the association of increased breast density with increased risk of breast cancer. Quantitative measurements of breast density are taken mainly in research settings, and due to practical challenges, they are not routinely used in clinical care. Instead, the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) ordinal breast density categories are routinely collected in clinical practice in the United States (12).
The extent of breast density can be modified by several factors. Increasing age and menopause (13–15) are independent contributors to a decrease in breast density (16). Elevated body mass index has been found to be associated with low breast density, whereas increased age at first birth has been associated with high breast density (17–20). Pregnancy at an early age decreases breast density, and this beneficial effect appears to be permanent (18,21). Postmenopausal hormone therapies that include both estrogen and progesterone are associated with an increase in breast density that decreases upon discontinuation of therapy (14,22,23). Intervention trials have shown that decreases in breast density are associated with tamoxifen treatment (24–26), a therapy proven to decrease breast cancer risk (27,28).
These studies suggest that breast density may decrease over time and that it may be modified by risk factors and interventions, but they do not provide information on the association of changes in breast density with breast cancer risk. A small study used a computerized method to quantify breast density from digitized films collected for a 10-year period from 108 postmenopausal breast cancer patients and 400 control subjects (29). It observed a trend of greater risk of breast cancer among women who had low density initially that increased over time compared to women with low breast density that remained low. It also found that, among women with high density that decreased over time, there was no evidence of decreased breast cancer risk relative to that in women whose high density persisted. A recent longitudinal investigation of patients who were preponderantly Hawaiian and Japanese did not find a difference in the rate of change in percent density between women with and without breast cancer (30).
The goal of this study was to determine whether a change in breast density as reflected in a change in BI-RADS category was associated with breast cancer risk in a population of 301 955 women who underwent screening mammography, of whom 2639 were diagnosed with breast cancer. The underlying hypothesis was that decreases in breast density over time are associated with a lower risk of breast cancer and that increases are associated with a higher risk of breast cancer relative to women with unchanging density.
|
Methods |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionReferencesNotes |
|---|
Data Sources
Data were pooled from seven mammography registries that participated in the Breast Cancer Surveillance Consortium (31) (http://breastscreening.cancer.gov) supported by the National Cancer Institute (NCI): San Francisco Mammography Registry, Group Health's Breast Cancer Surveillance, Colorado Mammography Advocacy Project, Vermont Breast Cancer Surveillance System, New Hampshire Mammography Network, Carolina Mammography Registry, and New Mexico Mammography Registry. These registries collect information on screening and diagnostic mammography examinations performed in their defined catchment areas. Each mammography registry annually links women in their registry to a state tumor registry or regional Surveillance, Epidemiology, and End Results (SEER) program to obtain population-based cancer data. Some registries additionally link to pathology databases. Each registry obtains annual approval from their Institutional Review Board for consenting processes or a waiver of consent, enrollment of participants, and ongoing data linkages for research purposes. All registries have received a Federal Certificate of Confidentiality that protects the identities of research participants.
Subjects
We defined screening mammography as a bilateral examination indicated by the radiologist as being performed for screening among women without a history of breast cancer and without breast implants. The accessible study sample (N = 518 739) included women aged 30 years and older who underwent at least two screening mammography examinations between January 1, 1993, and December 31, 2003, that were more than 9 months apart. Both examinations had to include a BI-RADS breast density measure (12) that was assigned in clinical practice by a radiologist at the time of screening mammography. Screening examinations that occurred after December 2003 were not included to ensure that there would be at least 12 months for reporting cancers to tumor registries after the most recent screening examination. Women who indicated they were using postmenopausal hormone therapy at the time of any mammography examination during the study period were excluded (N = 216 504). With the exclusion of women of invalid age or incomplete cancer diagnosis information (N = 280), the final study sample consisted of 301 955 women.
Measurements and Definitions
Demographic information and a self-reported breast health history were obtained at the time of each screening examination by having women complete a questionnaire that requested information on menopausal status, history of breast cancer in first-degree relatives (mother, sister, or daughter), history of oophorectomy, age at first live birth, and height and weight. Women were considered to have a family history of breast cancer if they reported having at least one first-degree relative with breast cancer. Women were considered to be postmenopausal if both ovaries had been removed, if they reported that their menstrual periods had stopped permanently, or if they were aged 55 years or older.
If a woman had more than two screening mammography examinations with BI-RADS density measures during the study period, we used breast density measures from the earliest screening examination (first) and most recent examination (last) and demographic and clinical characteristics obtained at the time of the last examination. If data on demographic and/or clinical characteristics were missing on the last screening examination, we used the last known value obtained before the last screening examination. Time between mammography examinations was determined using dates of the first and last screening examination.
The association of breast density with breast cancer risk was assessed using four breast density measures that were based on the BI-RADS breast density categories assigned at the first and/or last screening examination for each woman. The BI-RADS categories were 1 = almost entirely fat (also referred to as low density), 2 = scattered fibroglandular densities (average density), 3 = heterogeneously dense (high density), 4 = extremely dense (very high density) (12). The first measure was the BI-RADS category assigned to the first screening examination for each woman. The second measure was the BI-RADS category assigned to the last screening examination for each woman. The third measure was the average of the BI-RADS categories from the first and last screening examination. The fourth measure was a summary measure that consisted of classifying women into 1 of 16 possible combinations of BI-RADS categories assigned to the first and last screenings.
Women were considered to have breast cancer if reports from a breast pathology database, SEER program, or state tumor registry showed any invasive carcinoma or ductal carcinoma in situ (DCIS) within one year of their last screening examination in the study period. Women with lobular carcinoma in situ only were not considered to have breast cancer.
Statistical Analysis
All analyses were performed using women as the unit of analysis. Frequency distributions of various risk factors—age, ethnicity, family history, age at first birth, menopausal status, body mass index, and time between first and last screening examination—were determined for women with and without breast cancer. We also determined the frequency distributions of BI-RADS density scores (from both the first and last screening examination), according to decade of attained age and menopausal status for women with and without breast cancer.
Multivariable logistic regression was used to assess the association between breast density measures and breast cancer risk. All models were adjusted for mammography registry, time between the two screening examinations (as a categorical variable [<2, 2 to <3, 3 to <4, or 
4 years]), and age (as a continuous variable with both a linear and a quadratic term because breast cancer incidence increases with age in a nonlinear fashion). All covariates included in the models were statistically significant. We used BI-RADS category 2 (scattered fibroglandular densities) as the referent group when evaluating the association of first, last, and average density measures with breast cancer risk since this group was the most common. When evaluating the association of change in BI-RADS category from first to last examination with cancer risk, we stratified by BI-RADS category on the first examination and used women whose density category did not change as the referent group within each BI-RADS category. Because BI-RADS breast density is a categorical measure, women with an observed change in clinically rated breast density may have had breast density values between BI-RADS breast density categories. Therefore, we also stratified women by the same average breast density of the first and last screening examination, but different BI-RADS values at the first and last examination to try to account for the effect of variation in assigning BI-RADS breast density categories by radiologists on observed associations between changes in breast density and breast cancer risk. We used the Wald test to test for statistically significant linear trends within breast density groups. We also tested for an interaction between age and change in breast density and breast cancer risk and time between screening examinations and change in breast density and breast cancer risk using a linear regression model with age as a linear term and change in breast density and time between screening examinations as categorical variables.
Adjusted breast cancer rates per 1000 women were calculated from the logistic regression model using predicted margins, also known as marginal standardization (32,33). Based on parameter estimates from the logistic regression model, we estimated the probability of breast cancer for each combination of age, mammography registry, time between screens, and breast density. The adjusted rates for each breast density category were then estimated by calculating a weighted average of these probabilities based on the proportion of women in the corresponding age group, mammography registry, and time between screening examination strata observed in the data. Confidence intervals (CIs) for the adjusted rate of breast cancer were computed based on simulations. Each simulation sampled values for each of the parameter estimates from their estimated joint multivariable normal distribution. We ran 100 000 simulations, and the 2.5th and 97.5th percentiles from the simulations were used to determine a 95% confidence interval.
We were not able to adjust for body mass index or age at first birth due to missing values. In the final analysis, we also did not adjust for race, family history of breast cancer, menopausal status at last examination, or change in menopausal status over time because including these variables in the models did not change the overall results. We adjusted for misclassification of breast density in each logistic regression model using an exact approach based on the method of Reade-Christopher and Kupper (34). The misclassification matrix used for adjustment was based on a study of the rating of breast density of 324 women by 19 different radiologists. Since the adjusted and unadjusted results were similar, we did not adjust for misclassification in the final models.
All statistical calculations were performed using SAS (version 9.1; SAS Institute, Cary, NC). Results of two-sided statistical tests in which P values were less than .05 were considered to be statistically significant.
|
Results |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionReferencesNotes |
|---|
Among 301 955 women aged 30 years and older who underwent at least two screening mammography examinations, 2639 developed breast cancer within 12 months of the last examination. Of these, 2089 were diagnosed with invasive breast cancer and 550 with DCIS. In our sample, women with breast cancer, as compared to women without breast cancer, were older and more likely to be white and have a family history of breast cancer and less likely to have given birth before the age of 30 years (Table 1). For about half of all women, 3 or more years had elapsed between the first and last screen.
|
The frequency distribution of the four BI-RADS breast density categories at the first and last screen was calculated for different age categories and according to menopausal status (premenopausal or postmenopausal) among women with and without breast cancer (Table 2). Among women without breast cancer, women aged 70 years and older were four times more likely to be assigned to the lowest breast density category (category 1) and six times less likely to be assigned to the highest breast density category (category 4) than were women aged 30–39 years. Thus, breast density decreased with age. Women who developed breast cancer were more likely to have BI-RADS breast density categories of 3 or 4 on the first and last screen than women without breast cancer of similar age and menopausal status (Table 2).
|
The frequency distribution of breast density measures (values at the first screening, last screening, or the average of the values obtained at first and last screenings) among women of all ages who did or did not have breast cancer is shown in Table 3, along with the adjusted rate of breast cancer calculated from the logistic regression model based on these data. A total of 51.3% of women with breast cancer had a BI-RADS breast density of 3 or 4 on the last screening examination compared with 46.4% of those without breast cancer. The odds ratios (ORs) for experiencing breast cancer for women with a given value of breast density were calculated relative to women with a breast density of 2 on first or last screen or after averaging first and last screens. Compared with women assigned to a BI-RADS breast density category of 2 on the last examination, women with a BI-RADS category of 1 on the last examination had a lower rate of breast cancer (3.5 versus 7.2 cases per 1000 women, OR = 0.49. 95% CI = 0.40 to 0.59), while women with a BI-RADS 4 had a higher rate of breast cancer (11.5 cases per 1000 women, OR = 1.61, 95% CI = 1.38 to 1.87). The association between breast cancer risk and BI-RADS breast density category on the first screen was similar to that between breast cancer risk and BI-RADS density on the last screen.
|
When BI-RADS measurements from the first and last screening examination were averaged, 61.4% of women who developed breast cancer had an average breast density greater than 2.0 compared with 55.9% among those who did not (Table 3). The rate of breast cancer increased in a linear fashion with increases in the average of the BI-RADS categories on the first and last examination, ranging from 3.0, 95% CI = 2.3 to 3.9, for women with an average of 1—to 12.6, 95% CI = 10.6 to 15.2, for those with an average of 4.
The frequency distribution of the 16 possible combinations of first and last breast density values for women with and without breast cancer and the adjusted rate of breast cancer for women in each of these categories were calculated (Table 4). Odds ratios of developing breast cancer were calculated for women who started in a given breast density category compared with women who experienced no change serving as the referent group (Table 4). A total of 19.6% of all women had an increase in breast density category and 18.5% had a decrease, with a median time of 3.2 years between screens. Of women with an initial BI-RADS breast density category of 1, 2, or 3, 22% had an increase in breast density category upon the last screening, and of women with an initial BI-RADS breast density category of 2, 3, or 4, 22% had a decrease. The majority of women had a BI-RADS breast density category of 2 or 3 on the first and last examination, and this was true of women with and without breast cancer: 30.0% of women without breast cancer and 29.4% of women with breast cancer had BI-RADS scores of 2 on the first and last screens, and 22.9% of women without breast cancer and 24.3% women with breast cancer had scores of 3 on both screens (Table 4).
|
The rate (per 1000 women) of breast cancer was higher if BI-RADS breast density category increased from 1 to 2 (5.6, 95% CI = 4.7 to 6.9) or 1 to 3 (9.9, 95% CI = 6.4 to 15.5) compared to when it remained at BI-RADS density of 1 (3.0, 95% CI = 2.3 to 3.9; P<.001 for trend, Table 4). Women were at higher risk of breast cancer if their BI-RADS breast density category increased from 2 to 3 (9.7 breast cancers per 1000 women, 95% CI = 8.7 to 10.9) and lower risk if their BI-RADS category decreased from 2 to 1 (4.0, 95% CI = 3.1 to 5.2) compared to women who remained at BI-RADS 2 (6.9, 95% CI = 6.4 to 7.5; P = .004 for trend). The rate of breast cancer was higher if BI-RADS density category increased from 3 to 4 (10.8 per 1000 women, 95% CI = 8.6 to 13.5) and lower if density category decreased from 3 to 1 (4.5, 95% CI = 1.9 to 10.8), compared with when it remained at BI-RADS density 3 (9.8, 95% CI = 9.0 to 10.7; P = .049 for trend). Women with BI-RADS breast density category of 4 on the first examination were at similar risk of breast cancer regardless of the density category on the last examination (P = 0.87 for trend).
Most women remained at BI-RADS breast density category of 2 and were designated as average-risk women with a rate of breast cancer of 6.9 per 1000 women (Table 4). Rate of breast cancer was low among women with a BI-RADS breast density category of 1, 2, or 3 on the first examination and a 1 on the last examination (Table 4). Compared with average-risk women, women had a higher rate of breast cancer if they changed from a BI-RADS breast density category of 1 to 3, or 2 to 3, or 2 to 4, or remained at category 3, or changed from 3 to 2 or 3 to 4 (Table 4). Women with a BI-RADS breast density of 4 on the first examination had the highest rate of breast cancer. In this category, even women who were assigned to a lower breast density category on the last examination remained at high risk (Table 4).
Risk of breast cancer associated with change in BI-RADS breast density categories was different by age (test for interaction, P = .03) and similar by time between screening examinations (test for interaction, P = .74). Increases or decreases in risk of breast cancer associated with change in BI-RADS breast density categories were more prominent for women aged 50 years and older than for women younger than age 50 years (Table 5).
|
We stratified women by the same average breast density of the first and last screening examination, but different BI-RADS values at the first and last examination to try to account for variation in assigning BI-RADS breast density categories by radiologists (data not shown). Among women who had an average BI-RADS breast density of 1.5, those whose BI-RADS density decreased from a 2 to 1 were at lower risk of breast cancer than women whose BI-RADS increased from 1 to 2 (P = .03). Among women whose BI-RADS density decreased from a 3 to 1 there was a trend of lower risk of breast cancer than women whose BI-RADS density increased from a 1 to 3 (P = .069). There were no additional differences among women with the same average density.
|
Discussion |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionReferencesNotes |
|---|
In this analysis of data from more than 300 000 women, we found that an increase in BI-RADS breast density category over a relatively short period of time was associated with an increase in breast cancer risk within 12 months after the last screening examination and that a decrease in breast density category was associated with a decrease in risk compared with women in whom breast density category remains unchanged. This trend was seen for women initially assigned to BI-RADS breast density categories 1, 2, and 3 but not for women in category 4, the highest density category.
Consistent with other studies (8–11), we found that a single measure of breast density, measured by BI-RADS category, was associated with breast cancer risk. In our study, women with a BI-RADS breast density category of 1 were at decreased risk of breast cancer and women with BI-RADS category of 3 and 4 were at increased risk of breast cancer when compared to women with BI-RADS breast density of 2, the most common density group. The choice of those with BI-RADS breast density 2 as the referent group led to lower estimates of increased relative risk associated with high breast density than have been reported in prior studies that used the lowest density group as the reference (1,3,9). We also found that both recent and prior BI-RADS breast density measures were associated with breast cancer risk, similar to what has been observed when more quantitative measures of breast density were used (1).
A trend of greater risk of breast cancer among women who started at low breast density and were found to have higher breast density over time was reported by van Gils et al. (29). For women who started at high density and whose measured breast density decreased during the study period, these authors found no evidence of decreased risk with decreased density, possibly because of the small study sample size (108 postmenopausal breast cancer patients and 400 control subjects) (29). We found that two BI-RADS breast density measures may predict a women's risk of breast cancer more reliably than a single measure. Our results are consistent with those of van Gils et al. in that we found a higher rate of breast cancer among women who started at low breast density and in whom density increased over time relative to women who remained at low density. Also, consistent with the van Gils et al. study we found that presence of the highest breast density category was associated with the highest rate of breast cancer even if breast density decreased to a lower category over time. Contrary to the results of van Gils et al., and possibly due to our much larger sample size, we found that women with average or high density (BI-RADS density category of 2 or 3) that decreased to a lower BI-RADS density category had a lower rate of breast cancer than women who stayed at average or high density.
A recent case–control study did not observe any association between breast cancer and rate of change in percent density (30). Compared to the results of Maskarinec et al. (30), our study had a relatively large proportion of premenopausal women who have been reported to have a higher rate of decline of breast density over time, this and the large sample size may account for the observed changes in density and associated breast cancer risk that we report here (30). It is not known why some women may have an increase in breast density over time with increasing age. Decreasing weight, increased alcohol intake, or changes in diet or medication may contribute to increasing breast density (35), and their possible effects on changes in breast density over time warrant further study.
Mammographic density is highly influenced by genetic factors. In a study of monozygotic and dizygotic twins from Australia, Canada, and the United States, Boyd et al. (36) determined that the majority (60%–75%) of the variation in the fraction of dense tissue measured on mammograms in women between the ages of 40 and 70 years is controlled by genetic factors and that environmental factors account for 20%–30% of the age-adjusted variation in the percentage of dense tissue (2). Environmental factors that influence breast density include menopausal status, weight, parity, and exogenous and endogenous levels of hormones (18,21,23).
Women who become menopausal have an absolute decrease in mean breast density of 5% within 1.5 years and 8% within 5 years (16). Hormone therapy that includes estrogen plus progesterone has been shown to increase BI-RADS breast density in approximately 16%–28% of menopausal women within the first year of starting hormone therapy (14,22). Women at high risk for breast cancer taking standard tamoxifen therapy for prevention have been shown to have an absolute mean decrease in breast density of 5% at 18 months and 7% at 52 months (26). We used a categorical variable to assess breast density and found that a modest proportion of women had an increase (19.6%) or decrease (18.5%) in BI-RADS category within an average of 3 years. Thus, absolute changes in breast density over time due to environmental factors appear moderate, influence a modest proportion of women, and can occur in a relatively short period of time. The proportion of women who were observed in our study to change in BI-RADS category was somewhat higher than in a study of women aged 67 years and older (22) that reported that 11.6% of women had an increase and 6.5% a decrease, in BI-RADS category within an average of 2 years. Our longer average time between screening examinations and younger study population may account for the differences between studies.
Our results are consistent with the hypothesis that women with sustained levels of high breast density are at highest risk of breast cancer and those with low levels of breast density over time are at lowest risk. This is not surprising in that the extent of breast density reflects morphological differences that are related to breast cancer risk. For example, high mammographic density is associated with greater total nuclear area of both epithelial and nonepithelial cells (37). Furthermore, a greater percentage of epithelium in benign tissue biopsies has been associated with an increased risk of hyperplasia (with or without atypia) and/or carcinoma in situ, and these histology findings are associated with increased risk of breast cancer (38–41). Cumulative exposure to growth factors, including higher plasma concentrations of insulin-like growth factor 1 in premenopausal women and higher prolactin levels in postmenopausal women (42,43), may influence proliferation of epithelial and stromal cells in the breast. Thus, the long-standing presence of extensive or high breast densities may reflect exposure to hormones and growth factors that stimulate cell division in the breast and may influence breast cancer risk. This suggests that cumulative exposure to high breast densities could have a corresponding effect of increasing the incidence of breast cancer (30).
Women with a change in BI-RADS category from 1 to 2 or from 2 to 1 may have a breast density value between these two measures. If the majority of women with a change in BI-RADS category from 1 to 2 or from 2 to 1 had a similar average breast density value of about 1.5, these two groups of women should have similar breast cancer risk. In contrast, if the underlying breast density in the majority of these women represents an actual change in breast density, we would expect women who change from 1 to 2 to have a higher breast cancer risk than women who go from 2 to 1, as we observed in these groups of women. We also demonstrated a trend of different breast cancer risk, albeit one that was not statistically significant, among women with an average BI-RADS breast density of 2.0 who changed from 1 to 3 or 3 to 1. The fact that the direction of breast density change is predictive of risk in categories of women for whom the average of the first and last measurements is the same suggests that the observed changes in BI-RADS breast density categories reflect actual changes whose direction influences breast cancer risk.
Our study has both strengths and limitations. A major strength of our study is the large number of women with two breast density measures and the large number of breast cancers. A limitation of our study is that the interrater agreement of the BI-RADS breast density measure is moderate (44,45). Therefore, misclassification of BI-RADS categories may have influenced our results, such that some of the differences we observed could result in an under- or overestimation of associations (46). In addition, some groups that changed BI-RADS breast density category were very small and the numbers of cancers in these groups were few, limiting our ability to identify important associations. It is also not possible to determine from our data whether the observed risk of breast cancer is due to changes in breast density, the combined effect of two density measurements, or both.
In this study, values for some variables associated with breast density and breast cancer risk were missing. Body mass index, age at first live birth, and race were the variables that were most often lacking because some participating facilities did not consistently collect this information. Adjusting for body mass index, age at first live birth, family history of breast cancer, menopausal status, change in menopausal status over time, and race did not substantially alter the breast cancer risk estimates in this study for the subset of women with covariate data available, consistent with other studies (9,47). It is possible, however, that adjustment for body mass index in the full cohort would have strengthened the association of breast density and breast cancer risk, given its negative association with breast density and positive association with breast cancer risk in postmenopausal women (48). In this study, we excluded women receiving postmenopausal hormone therapy because of a limited number of breast cancers among women who were on hormone therapy for 5 or more years and had BI-RADS density measures. Future analyses might examine the association of change in density and breast cancer risk among women taking postmenopausal hormone therapy.
BI-RADS density measures are routinely included in mammography reports to practitioners, but they are not currently used to estimate the risk of breast cancer at the time of screening mammography. Risk models have shown that BI-RADS breast density is a statistically significant addition to prediction of breast cancer in premenopausal and postmenopausal women (8,49). A model based on breast density alone and adjusted for age and ethnicity is as accurate as the Gail model in predicting breast cancer (8). Postmenopausal women are rarely assessed for risk of developing breast cancer and rarely receive preventive therapy, even if they are at high risk for breast cancer (50,51). The assessment, concurrently with screening mammography, of breast cancer risk based on breast density and standard risk factors creates an opportunity to discuss with the patient the risk of breast cancer in the next 5 years and, for those at high risk, prevention strategies. Our study suggests that physicians should take into account both current and previous breast density measurements when evaluating risk with their patients.
In summary, we found that with average follow-up of three years, a modest proportion of women change BI-RADS breast density category. An increase in BI-RADS breast density category may be associated with increased breast cancer risk, and a decrease may be associated with decreased breast cancer risk relative to the level of risk when the breast density does not change. Thus, two longitudinal measures of BI-RADS breast density may better predict a woman's risk of breast cancer than a single measure.
|
NOTES |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionReferencesNotes |
|---|
This work was supported by the NCI-supported Breast Cancer Surveillance Consortium cooperative agreement (U01CA63740, U01CA86076, U01CA86082, U01CA63736, U01CA70013, U01CA69976, U01CA63731, U01CA70040) and Dr K. Kerlikowske is also supported by PO1 CA107584.
The study sponsor had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript, or the decision to publish.
|
REFERENCES |
|---|
|
TopAbstractContext and CaveatsMethodsResultsDiscussionReferencesNotes |
|---|
(1) Byrne C, Schairer C, Wolfe J, Parekh N, Salane M, Brinton LA, et al. (1995) Mammographic features and breast cancer risk: effects with time, age, and menopause status. J Natl Cancer Inst 87:1622–9.
(2) Boyd NF, Lockwood GA, Byng JW, Tritchler DL, Yaffe MJ. (1998) Mammographic densities and breast cancer risk. Cancer Epidemiol Biomarkers Prev 7:1133–44.[Abstract]
(3) Boyd NF, Byng JW, Jong RA, Fishell EK, Little LE, Miller AB, et al. (1995) Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study. J Natl Cancer Inst 87:670–5.
(4) Saftlas AF, Wolfe JN, Hoover RN, Brinton LA, Schairer C, Salane M, et al. (1989) Mammographic parenchymal patterns as indicators of breast cancer risk. Am J Epidemiol 129:518–26.
(5) Wolfe JN, Saftlas AF, Salane M. (1987) Mammographic parenchymal patterns and quantitative evaluation of mammographic densities: a case-control study. AJR Am J Roentgenol 148:1087–92.
(6) Colditz GA, Willett WC, Hunter DJ, Stampfer MJ, Manson JE, Hennekens CH, et al. (1993) Family history, age, and risk of breast cancer. JAMA 270:338–43.[Abstract]
(7) Prevrhal S, Shepherd JA, Smith-Bindman R, Cummings SR, Kerlikowske K. (2002) Accuracy of mammographic breast density analysis: formal training of operators with varying background. Cancer Epidemiol Biomarkers Prev 11:1389–93.
(8) Tice JA, Cummings SR, Ziv E, Kerlikowske K. (2005) Mammographic breast density and the Gail model for breast cancer risk prediction in a screening population. Breast Cancer Res Treat 94:115–22.[CrossRef][ISI][Medline]
(9) Vacek PM and Geller B. (2004) A prospective study of breast cancer risk using routine mammographic breast density measurements. Cancer Epidemiol Biomarkers Prev 13:715–22.
(10) Ziv E, Tice J, Smith-Bindman R, Shepherd J, Cummings S, Kerlikowske K. (2004) Mammographic density and estrogen receptor status of breast cancer. Cancer Epidemiol Biomarkers Prev 13:2090–5.
(11) Ziv E, Shepherd J, Smith-Bindman R, Kerlikowske K. (2003) Mammographic breast density and family history of breast cancer. J Natl Cancer Inst 95:556–8.
(12) American College of Radiology. (2003) The American College of Radiology Breast Imaging Reporting and Data System (BI-RADS®) 3rd ed (American College of Radiology, Reston (VA)).
(13) Kerlikowske K, Grady D, Barclay J, Sickles EA, Ernster V. (1996) Effect of age, breast density, and family history on the sensitivity of first screening mammography? JAMA 276:33–8.[Abstract]
(14) Greendale GA, Reboussin BA, Sie A, Singh HR, Olson LK, Gatewood O, et al. (1999) Effects of estrogen and estrogen-progestin on mammographic parenchymal density. Ann Intern Med 130:262–9.
(15) Kerlikowske K, Miglioretti DL, Barbash R, Weaver D, Buist DSM, Barlow W, et al. (2003) Prognostic characteristics of breast cancer among postmenopausal hormone users in a screened population. J Clin Oncol 21:4314–21.
(16) Boyd NF, Martin L, Stone J, Little L, Minkin S, Yaffe M. (2002) A longitudinal study of the effects of menopause on mammographic features. Cancer Epidemiol Biomarkers Prev 11:1048–53.
(17) White E, Velentgas P, Mandelson MT, Lehman CD, Elmore JG, Porter P, et al. (1998) Variation in mammographic breast density by time in menstrual cycle among women aged 40 to 49 years. J Natl Cancer Inst 90:906–10.
(18) Vachon CM, Kuni CC, Anderson K, Anderson E, Sellers TA. (2000) Association of mammographically defined percent breast density with epidemiologic risk factors for breast cancer (United States). Cancer Causes Control 11:653–62.[CrossRef][ISI][Medline]
(19) Salminen TM, Hakama M, Heikkila MM, Saarenmaa IE. (1998) Favorable change in mammographic parenchmal patterns and breast cancer risk factors. Int J Cancer 78:410–4.[CrossRef][ISI][Medline]
(20) Warwick J, Pinney E, Warren RML, Duffy SW, Howell A, Wilson M, et al. (2003) Breast density and breast cancer risk factors in a high-risk population. Breast 12:10–6.[CrossRef][ISI][Medline]
(21) El-Bastawissi AY, White E, Mandelson MT, Taplin SH. (2000) Reproductive and hormonal factors associated with mammographic density by age (United States). Cancer Causes Control 11:955–63.[CrossRef][ISI][Medline]
(22) Rutter CM, Mandelson MT, Laya MB, Seger DJ, Taplin S. (2001) Changes in breast density associated with initiation, discontinuation, and continuing use of hormone replacement therapy. JAMA 285:171–6.
(23) Greendale GA, Reboussin BA, Slone S, Wasilauskas C, Pike C, Ursin G. (2003) Postmenopausal hormone therapy and change in mammographic density. J Natl Cancer Inst 95:30–7.
(24) Atkinson C, Warren R, Bingham SA, Day NE. (1999) Mammographic patterns as a predictive biomarker of breast cancer risk: effect of tamoxifen. Cancer Epidemiol Biomarkers Prev 8:863–6.
(25) Chow CK, Venzon D, Jones EC, Premkumar A, O’Shaughnessy J, Zujewski J. (2000) Effect of tamoxifen on mamographic density. Cancer Epidemiol Biomarkers Prev 9:917–21.
(26) Cuzick J, Warwick J, Pinney E, Warren RML, Duffy SW. (2004) Effect of Tamoxifen on breast density in women at increased risk of breast cancer. J Natl Cancer Inst 96:621–8.
(27) Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, Cronin WM, et al. (1998) Tamoxifen for prevention of breast cancer: report of the national surgical adjuvant breast and bowel project P-1 study. J Natl Cancer Inst 90:1371–88.
(28) Cuzick J, Powles T, Veronesi U, Forbes J, Edwards R, Ashley S, et al. (2003) Overview of the main outcomes in breast cancer prevention trials. Lancet 361:296–300.[CrossRef][ISI][Medline]
(29) van Gils CH, Hendriks JHCL, Holland R, Karssemeijer N, Otten JDM, Straatman H, et al. (1999) Changes in mammographic breast density and concomitant changes in breast cancer risk. Eur J Cancer Prev 8:509–15.[ISI][Medline]
(30) Maskarinec G, Pagano I, Lurie G, Kolonel LN. (2006) A longitudinal investigation of mammographic density: the multiethnic cohort. Cancer Epidemiol Biomarkers Prev 15:732–9.
(31) Ballard-Barbash R, Taplin SH, Yankaskas BC, Ernster VL, Rosenberg RD, Carney PA, et al. (1997) Breast Cancer Surveillance Consortium: a national mammography screening and outcomes database. AJR Am J Roentgenol 169:1001–8.
(32) Lane P and Nelder J. (1982) Analysis of covariance and standardization as instances of prediction. Biometrics 38:613–21.[CrossRef][ISI][Medline]
(33) Graubard B and Korn E. (1999) Predictive margins with survey data. Biometrics 55:652–9.[CrossRef][ISI][Medline]
(34) Reade-Christopher SJ and Kupper LL. (1991) Effects of exposure misclassification on regression analyses of epidemiologic follow-up study data. Biometrics 47:535–48.[CrossRef][ISI][Medline]
(35) Boyd NF, Rommens JM, Vogt K, Lee V, Hopper JL, Yaffe MJ, et al. (2005) Mammographic breast density as an intermediate phenotype for breast cancer. Lancet Oncol 6:798–808.[Medline]
(36) Boyd NF, Dite GS, Stone J, Gunasekara A, English DR, McCredie MR, et al. (2002) Heritability of mammographic density, a risk factor for breast cancer. N Engl J Med 347:886–94.
(37) Li T, Sun L, Miller N, Nicklee T, Woo J, Hulse-Smith L, et al. (2005) The association of measured breast tissue characteristics with mammographic density and other risk factors for breast cancer. Cancer Epidemiol Biomarkers Prev 14:343–9.
(38) Warner E, Lockwood GA, Tritchler DL, Boyd NF. (1992) The risk of breast cancer associated with mammographic parenchymal patterns: a meta-analysis of the published literature to examine the effect of method of classification. Cancer Detect Prev 16:67–72.[ISI][Medline]
(39) Gertig DM, Stillman IE, Byrne C, Spiegelman D, Schnitt SJ, Connolly JL, et al. (1999) Association of age and reproductive factors with benign breast tissue composition. Cancer Epidemiol Biomarkers Prev 8:873–9.
(40) Boyd NF, Jensen HM, Cooke G, Lee Han H, Lockwood GA, Miller AB. (2000) Mammographic densities and the prevalence and incidence of histological types of benign breast disease. Eur J Cancer Prev 9:15–24.[CrossRef][ISI][Medline]
(41) Hartmann L, Sellers TA, Frost MH, Lingle WL, Degnim AC, Ghosh K, et al. (2005) Benign breast disease and the risk of breast cancer. N Engl J Med 353:229–37.
(42) Byrne C, Colditz GA, Willett WC, Speizer FE, Pollak M, Hankinson SE. (2000) Plasma insulin-like growth factor (IGF) I, IGF-binding protein 3, and mammographic density. Cancer Res 60:3744–8.
(43) Byrne C, Hankinson SE, Colditz GA, Willett W, Speizer FE. (2001) Plasma prolactin and mammographic breast density in postmenopausal women. Proc Am Assoc Cancer Res 42:153.
(44) Kerlikowske K, Grady D, Barclay J, Frankel S, Ominsky S, Sickles EA, et al. (1998) Variability and accuracy in mammographic interpretation using the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS). J Natl Cancer Inst 90:1801–9.
(45) Ciatto S, Houssami N, Apruzzese A, Bassetti E, Brancato B, Carozzi F, et al. (2005) Categorizing breast mammographic density: intra- and interobserver reproducibility of BI-RADS density categories. Breast 14:269–75.[CrossRef][ISI][Medline]
(46) Gustafson P. (2004) Measurement error and misclassification in statistics and epidemiology: impacts and Bayesian adjustments(CRC Press, Boca Raton (FL)).
(47) McCormack VA and dos Santos Silva I. (2006) Breast density and parenchymal patterns as markers of breast cancer risk: a meta-analysis. Cancer Epidemiol Biomarkers Prev 15:1159–69.
(48) Boyd NF, Martin LJ, Sun L, Gun H, Chiarelli A, Hislop G, et al. (2006) Body size, mammographic density, and breast cancer risk. Cancer Epidemiol Biomarkers Prev 15:2086–92.
(49) Barlow W, White E, Ballard-Barbash R, Vacek PM, Titus-Ernstoff LT, Carney PA, et al. (2006) A prospective breast cancer risk prediction model among women undergoing screening mammography. J Natl Cancer Inst 98:1204–14.
(50) Kaplan C, Haas JS, Perez-Stable EJ, Somkin C, Gregorich S, Des Jarlais G, et al. (2006) Breast cancer risk reduction options: awareness, discussion, and use among women from four ethnic groups. Cancer Epidemiol Biomarkers Prev 15:162–6.
(51) Armstrong K, Quistberg A, Micco E, Domchek S, Guerra C. (2006) Prescription of tamoxifen for breast cancer prevention by primary care physicians. Arch Intern Med 166:2260–5.
Manuscript received June 16, 2006; revised December 19, 2006; accepted January 6, 2007.
Related Article in JNCI
CiteULike 
Connotea 
Del.icio.us What's this?
J Natl Cancer Inst 2007 99: 337.
This article has been cited by other articles:
|
|
 |
|
  M. B. Terry, D. S.M. Buist, A. Trentham-Dietz, T. M. James-Todd, and Y. Liao Nonsteroidal Anti-inflammatory Drugs and Change in Mammographic Density: A Cohort Study Using Pharmacy Records on Over 29,000 Postmenopausal Women Cancer Epidemiol. Biomarkers Prev., May 1, 2008; 17(5): 1088 - 1095. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Byrne Invited Commentary: Assessing Breast Density Change--Lessons for Future Studies Am. J. Epidemiol., May 1, 2008; 167(9): 1037 - 1040. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. A. Tice, S. R. Cummings, R. Smith-Bindman, L. Ichikawa, W. E. Barlow, and K. Kerlikowske Using Clinical Factors and Mammographic Breast Density to Estimate Breast Cancer Risk: Development and Validation of a New Predictive Model Ann Intern Med, March 4, 2008; 148(5): 337 - 347. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. S. Hwang, D. L. Miglioretti, R. Ballard-Barbash, D. L. Weaver, K. Kerlikowske, and for the National Cancer Institute Breast Cancer Su Association between Breast Density and Subsequent Breast Cancer Following Treatment for Ductal Carcinoma In situ Cancer Epidemiol. Biomarkers Prev., December 1, 2007; 16(12): 2587 - 2593. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||