Department of Gynecology and Obstetrics, University Breast Center for Franconia, Erlangen University Hospital, Friedrich-Alexander University of Erlangen-Nuremberg, Erlangen-Nuremberg, Germany
Erlangen-Nuremberg Comprehensive Cancer Center, Erlangen-Nuremberg, Germany
Division of Hematology and Oncology, Department of Medicine, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, CA
University Breast Center for Franconia, Department of Gynecology and Obstetrics, Erlangen University Hospital, University Breast Center for Franconia, Friedrich-Alexander University of Erlangen-Nuremberg, Erlangen-Nuremberg, Germany
For many breast cancer (BC) risk factors, there is growing evidence concerning molecular subtypes for which the risk factor is specific. With regard to mammographic density (MD), there are inconsistent data concerning its association with estrogen receptor (ER) and progesterone receptor (PR) expression. The aim of our study was to analyze the association between ER and PR expression and MD. In our case-only study, data on BC risk factors, hormone receptor expression and MD were available for 2,410 patients with incident BC. MD was assessed as percent MD (PMD) using a semiautomated method by two readers for every patient. The association of ER/PR and PMD was studied with multifactorial analyses of covariance with PMD as the target variable and including well-known factors that are also associated with MD, such as age, parity, use of hormone replacement therapy, and body mass index (BMI). In addition to the commonly known associations between PMD and age, parity, BMI and hormone replacement therapy, a significant inverse association was found between PMD and ER expression levels. Patients with ER-negative tumors had an average PMD of 38%, whereas patients with high ER expression had a PMD of 35%. A statistical trend toward a positive association between PMD and PR expression was also seen. PMD appears to be inversely associated with ER expression and may correlate positively with PR expression. These effects were independent of other risk factors such as age, BMI, parity, and hormone replacement therapy, possibly suggesting other pathways that mediate this effect.
Over the last decade, the knowledge about the molecular subtypes of breast cancer (BC) and their specific treatment is increasing.1 Likewise, studies concerning with BC risk are trying to associate risk factors specifically with these specific molecular subtypes.
Nulliparity was associated with a reduced risk specifically for triple-negative BCs (TNBCs) and with an increased risk for estrogen receptor (ER)-positive BC.2 Other reports linked early age at menarche, nulliparity and late age at first full-term pregnancy with an increased risk of ER-positive, but not ER-negative BC.3 With regard to genetic risk factors, some variants could be associated with the risk for ER-positive BC and some others specifically with ER-negative BC.4–11 Some loci could be even associated with the risk for TNBCs.11–14
With an up to fivefold increase in the BC risk, mammographic density (MD) is one of the most important risk factors for BC.15–18 However, data on the association between MD and specific tumor characteristics are limited. Data from several reports are inconclusive and many questions remain unaddressed. A smaller case-only analysis reported an association between MD and PR positivity and between MD and ERβ.19 In another study, MD increased the risk for ER+PR+ tumors, but not for ER–PR− tumors. An increased risk has also been reported for a mixed group consisting of ER+PR− and ER−PR+ patients.20 MD was highest in the ER+PR+ group and in the mixed group.20 Another case–control study found that MD increased the risk for both ER+PR+HER2− and TNBCs.21 A larger case–control study identified MD as being a stronger risk factor for high-grade tumors and ER-negative tumors.22
There are as yet only limited data from larger studies regarding tumor characteristics and MD. The aim of our study was to analyze the association between ER expression, PR expression and MD in a case-only study. We will report these associations in a large case-only study.
Patients and Methods
Patients were selected from the BC database of the University Breast Center for Franconia. A total of 5,110 patients with invasive BC are documented in the database for the period 1995–2008. In the analysis presented here, patients were excluded in the following hierarchical order: no mammogram performed at the University Breast Center at the time of the primary diagnosis (excluding 1,989 patients, 485 of whom were prevalent cases) and unknown ER or unknown PR status (excluding 711 patients). The final study population consisted of 2,410 patients with incident-invasive BC.
Data collection and follow-up
All patient and tumor characteristics were documented as part of the certification processes required by the German Cancer Society (Deutsche Krebsgesellschaft) and by the German Society for Breast Diseases (Deutsche Gesellschaft für Senologie).23 Tumor characteristics, treatment data, some epidemiological data, histopathological characteristics, tumor treatments and follow-up have to be documented and are audited annually. Information about HRT usage in the patient's history was collected from the patients' charts at the time of the primary diagnosis. Body mass index (BMI) at the time of diagnosis was obtained from the measurements in the hospital that were performed for planning of the treatment (i.e., surgery or chemotherapy).
The quantitative computer-based threshold density assessments and breast area measurements were carried out by two different readers with explicit training in the method used. Each mammogram was read by both readers independently of each other. The assessment method has been described and validated previously elsewhere.24
Briefly, the images were digitized using the CAD PRO Advantage® film digitizer (VIDAR®, Herndon, VA) and for assessment of the density fraction; the reader used the Madena software program, Version X (Eye Physics, LLC, Los Alamitos, CA).24 All mammograms were read in random order by two different observers, who were unaware of any previous classifications or pathological findings. The average of both observers' values for percent MD (PMD) was used for analysis.
In the course of routine patient care, dedicated breast pathologists at the University Breast Center for Franconia examined pathologic specimens from all of the patients included in our study. The histological type, grade, resection status and TNM stage were determined, and the expression of ER, PR and HER2/neu was analyzed immunohistochemically in accordance with the standard practice in certified breast centers in Europe.25 ER and PR status are based on estimates of the percentage of clearly positive-stained cancer cells (from 0 to 100%). In clinical practice, ER or PR negativity was defined as less than 10% of the cells staining negative. Information about tumor characteristics was subsequently transferred from the histopathological reports into our clinical database.
Statistical methods and considerations
The association between MD and various patient characteristics and predictors was analyzed using linear models. Simple linear regression analyses were used for ordinal predictors (age at diagnosis, BMI, parity, pT and grading), and one-way analyses of variance were used for categorical predictors (menopausal and HRT status, nodal status, HER2 status, ER and PR; each one categorized as in Table 1). Mean PMDs with 95% confidence intervals and the p-value of the F-test are shown.
Table 1. Percentage mammographic density (PMD) relative to patient characteristics
The association between ER/PR and MD, taking into account the well-known predictors mentioned above, was studied using multifactorial analyses of covariance (ANCOVA) with PMD as target variable. Initially, an ANCOVA with all predictors except for ER and PR was fitted. Then, backward stepwise variable selection was carried out to obtain the best model in accordance with the Akaike information criterion (final model without ER/PR). Next, another ANCOVA was fitted containing ER, PR, the interaction between ER and PR, the predictors from the final model without ER/PR and the interactions of ER and PR with these predictors. The variable selection procedure described above was carried out again, but with the condition that the selected predictors from the final model without ER/PR were retained. The p-values for the F-tests (Type III analysis) of the resulting model (final model with ER/PR) and adjusted mean PMD values with 95% confidence intervals were shown. If the F-test for ER, PR or an interaction term produced a significant p-value, then corresponding pairwise post hoc tests were performed.
The model selection procedure described above was evaluated by tenfold crossvalidation with 20 replications. The mean R2 statistic of the final models with ER/PR was shown.
All of the tests were two-sided, and a p-value of <0.05 was regarded as statistically significant. Calculations were carried out using the SAS software package (version 9.2, SAS Institute, Cary, NC) and the R system for statistical computation (version 2.13.1; R Development Core Team, Vienna, Austria, 2011).
At least one risk factor and both ER expression and PR expression were available for a total of 2,410 patients. Mean age was 58.9 years (±12.8 years) and mean BMI was 26.1 kg/m2 (±4.8 kg/m2). PMD was measured on average as a percentage of 33.0 (±19.9). The study included 593 premenopausal or perimenopausal women (25%), 1,779 postmenopausal patients (75%) and 59 women with unknown menopausal status. The patients' characteristics are shown in Table 1.
In the univariate analysis, PMD was strongly associated with age, BMI, number of live births and menopausal status, as well as with the use of HRT (all p < 0.00001). With regard to tumor characteristics, there were no associations with tumor size, nodal status, grading, PR or HER2 status. Tumors with a high level of ER expression appeared to have a lower PMD in this univariate analysis (p < 0.00001).
For multifactorial analysis, HER2 status was excluded, as a high percentage of data values were lacking (20.4%). The initial ANCOVA, which did not include ER or PR, identified age, BMI, parity, menopausal and HRT status, pT and grading as relevant predictors of PMD. Nodal status was dropped as a predictor during the variable selection process, that is, the predictive value of nodal status appeared to be irrelevant or it was already explained by the other selected variables.
The ultimate ANCOVA, taking into account the relevant predictors described above and also ER and PR and their interactions with those predictors, showed that the prediction of PMD can be improved overall by including ER (p < 0.0001, F-test) and PR (p = 0.06, F-test). All interaction terms were dropped during the variable selection process, that is, the association of ER, PR and PMD does not essentially differ between patient groups. The adjusted mean PMD for each predictor and the p-values from the F-tests are shown in Table 1. Commonly known influencing factors continued to be strongly associated with MD. Although not significantly associated, tumor size (pT) appeared to correlate positively with MD (p = 0.07, F-test). Grading had no correlation with the adjusted PMD.
Women with a low PMD were more likely to have higher ER expression. The adjusted mean PMD decreased significantly from 38% for ER-negative tumors to 35% for tumors within the highest ER category (p < 0.00001). More precisely, the mean PMD did not differ significantly between patients with ER-negative tumors and tumors with low ER expression (p = 0.21) or between patients with ER expression in the second and third tertiles (p = 0.83). However, there were substantial differences between the first and second tertiles (p < 0.00001; Tables 1 and 2).
Table 2. p-values for pairwise post hoc tests for difference in mammographic density as defined by groups of different estrogen receptor (ER) expression levels (ANCOVA, final model with ER/PR)
In contrast, women with a high PMD were more likely to have high PR expression. The adjusted mean PMD increased from 35% for PR-negative tumors to 38% for tumors in the highest PR category. However, the difference was not statistically significant (p = 0.06; Table 1).
Crossvalidation of the model selection process yielded an average R2 = 0.36 for the regression models finally selected.
The association between MD and ER expression was inverse, and the association between PMD and PR expression was positive. To visualize this association, the final ANCOVA model was applied to calculate PMD values for each of the possible combinations of ER and PR expression, adjusted for age, BMI, menopausal and HRT status, parity, grading and ER or PR expression (Fig. 1).
This case-only study showed that PMD is inversely associated with the ER expression level and has a statistical trend toward a positive association with the PR expression level in invasive breast tumors. This association was independent of other risk factors such as age at diagnosis, BMI, HRT use and menopausal status. The association was also independent of other tumor characteristics such as tumor stage and grading.
Most previous studies, either with case-only designs or with case–control designs, have not found consistent associations between MD and ER or PR status.20, 26–30 Most of these studies were underpowered for detecting differences. However, the largest study so far showed that in postmenopausal patients, the risk for ER-negative BC is clearly higher in women with a high MD, whereas this effect was much smaller for ER-positive tumors.22 The OR for women with a PMD > 50% in comparison with those with a PMD < 10% was 4.78 (95% CI = 2.42–9.42) for ER-negative tumors and only 2.94 (95% CI = 2.02–4.27) for ER-positive tumors. The difference was statistically significant (p = 0.04). In our case-only study, women with a higher PMD were more likely to have lower ER expression levels. The direction of this effect is therefore consistent with the finding in the case–control study by Yaghjyan et al.22
With regard to PR status in the study by Yaghjyan et al., the risks for PR-negative tumors and for PR-positive tumors were approximately the same. When women with a MD > 50% were compared to women with a PMD < 10%, the ORs were 3.21 (95% CI = 2.17–4.77) for PR-positive tumors and 3.68 (95% CI = 2.12–6.37) for PR-negative tumors.22 In our case-only study, women with higher PR expression were more likely to have a higher PMD, and the combination of high PR expression and low ER expression resulted in the highest mean values for PMD in the respective subgroup (Fig. 1). Our findings concerning the PR status are consistent with a smaller previous case-only analysis.19
However, comparison with other studies is generally difficult, as the subgroups for the two hormone receptors are not shown in detail in most of the studies. Expression levels of the hormone receptors are usually summarized. This assigns a variety of tumors to a group that may still show different biological behaviors (no expression, low expression levels and high expression levels, as well as ER positives and/or PR positives).
With regard to tumor biology, there are some data from animal models concerning the progesterone receptor (PR)-driven development and proliferation of the mammal breast. In mouse models, mammary stem cells proliferate under the influence of progesterone. It is thought that progesterone drives a series of events, including WNT4 and RANKL (TNFSF11) expression, to induce the proliferation of basal mammary gland cells and the branching of the mouse mammary gland during pregnancy.31, 32 Although there are no experimental data available for humans, there are indications that progesterone and not estrogen only has a specific effect on MD.33
This study has several strengths and limitations. The BC cases included were hospital based and were not recruited from a population-based screening program. Breast density is thought to contribute to a higher likelihood of tumors being missed during early detection methods for BC. It might therefore be possible for patients with a higher MD to have larger tumors, either caused by detection problems or by a different tumor biology. A statistical trend toward such an association was seen in our study, after adjustment for other BC risk factor. Patients with larger tumors had higher PMD values (p = 0.07). The association with ER status followed the same direction. Patients with ER-negative tumors, which are thought to have greater proliferation, also had higher PMD values. Another weakness of the study might be the use of prospectively assessed immunohistochemistry for the expression analysis of the hormone receptors. Although the assessment of biomarkers was performed at a central institution using the same method, several pathologists were involved in the assessment of hormone receptor expression, which might lead to heterogeneity in the assessment of these biomarkers.
One strength of our study is the semiautomated method of quantifying MD, with two readers for all images and a mean value for PMD being used. This may reduce measurement inaccuracies. Another strength is the use of only incident BC cases, avoiding the risk of bias in the selection of patients due to effects of tumor biology (e.g., hormone receptor status) on survival in patients with BC.
We could show a significant negative association between breast density and ER expression and a statistical trend toward a positive association between PR expression and PMD. This sheds some light on the possible involvement of the PR and related pathways specifically in mammographic dense breasts. Future research concerning with BC prevention should specifically address this pathway.
The association between PR status and MD needs further evaluation in other studies. The association between ER status and MD appears to be consistent in larger studies, and further research is warranted to elucidate the biological mechanisms behind these findings.
Katharina Heusinger was funded by the ELAN Program at Erlangen University Hospital, Friedrich-Alexander University of Erlangen-Nuremberg.