Women with invasive breast cancer who are treated with breast-conserving surgery and radiotherapy face a cumulative risk of local disease recurrence of approximately 10% at 10 years. To the authors' knowledge, the role of mammographic density as a risk factor for the development of local recurrence has not been thoroughly evaluated to date.
Medical records were reviewed for 335 patients who underwent breast-conserving surgery for invasive breast cancer and for whom a pretreatment mammogram was available. Information was recorded concerning mammographic density as well as tumor features, patient characteristics, and adjuvant treatments received. Patients were categorized for mammographic density based on the Wolfe classification as either low (<25% density), intermediate (25-50% density), or high (>50% density). A multivariate survival analysis was conducted using the Cox proportional hazards model with local disease recurrence as the primary endpoint.
Patients in the high mammographic density group experienced a much greater risk of local disease recurrence compared with women with the least dense breasts (10-year actuarial risks: 21% vs 5%; hazards ratio [HR], 5.7 [95% confidence interval, 1.6-20; P = .006]). The difference in the rates of disease recurrence at 10 years was pronounced for women who did not receive radiotherapy (40% vs 0% for patients with >50% density and <25% density, respectively; P < .0001).
For many women, surgical treatment of invasive breast cancer is comprised of resection of the malignant lesion with preservation of the surrounding breast (breast-conserving surgery). In the majority of cases, breast-conserving surgery is followed by irradiation of the breast to reduce the risk of local (within the breast) disease recurrence. Large, randomized, prospective studies have shown that patients who undergo breast-conserving surgery have a survival experience equivalent to that of women who undergo a total mastectomy.1-4 However, the risk of ipsilateral breast tumor recurrence is approximately 10% at 10 years. Various factors have been shown to influence this risk, including early age at the time of diagnosis, the presence of an extensive intraductal component, and positive surgical margins.5-8 Radiotherapy is usually given after breast-conserving surgery to reduce the risk of local disease recurrence and to improve breast cancer-specific mortality.9 Mammographic density is now recognized as 1 of the most important risk factors for the development of breast cancer.10-13 In the 1970s, Wolfe proposed that breast density plays an important role in carcinogenesis14 and developed a classification scheme based on the appearance of the breast parenchyma in the mammogram. The extremes included N1, a breast with a mostly radiolucent, fatty appearance; and DY, a mammographically dense breast with connective tissue hyperplasia. Intermediate categories were P1, a minimally prominent ductal pattern; and P2, with increasing moderate to severe ductal appearance. Other, quantitative, scales have since been developed. The risk of breast cancer is increased up to 6-fold in women with the highest density scores compared with women with low breast density measurements.15 Boyd et al.10 demonstrated that the increased incidence of breast cancer associated with increased breast density corresponds with the inherent tendency of the breast to form a carcinoma. Levels of circulating growth factors and the quantity and composition of proteins in the breast stromal matrix have been shown to influence breast density16, 17 but, to the best of our knowledge, the biologic basis of the association with cancer risk is largely unknown. High breast density also may reduce the sensitivity of mammography through a masking effect.18
Recently, Park et al19 reported that mammographic density was a risk factor for breast cancer recurrence among 136 women treated for invasive breast cancer with breast-conserving surgery and radiotherapy. Increased breast density also has been demonstrated to increase the risk of invasive breast cancer after surgery for ductal carcinoma in situ (DCIS).20 The purpose of the current study was to evaluate whether mammographic breast density is a risk factor for ipsilateral breast tumor recurrence among women treated with breast-conserving surgery (with or without radiotherapy) for invasive breast cancer.
MATERIALS AND METHODS
Selection of Subjects
Eligible subjects were women who underwent breast-conserving surgery for invasive breast cancer at Women's College Hospital between 1987 and 1998, and for whom a pretreatment mammogram was available for study. Patients were identified from the Henrietta Banting Database. A total of 567 patients underwent breast-conserving surgery and a diagnostic pretreatment mammogram was available for 449 of them (79%). Of these 449 patients, 114 had DCIS and were excluded from the study; thus, 335 women were included in the current analysis.
Information regarding patient characteristics, tumor features, and adjuvant treatment factors was entered into the database. Patient characteristics included age at diagnosis and menopausal status. Tumor features included size, grade, lymph node status, the presence of lymphovascular invasion (LVI), and estrogen receptor (ER) status. Treatment factors included surgical margin status (negative or positive), requirement for re-resection, adjuvant radiotherapy, and chemotherapy. The main endpoints were local disease recurrence and overall survival.
Two radiologists reviewed the mammograms at the time of diagnosis and assigned a density score. Density was measured on the cranio-caudal and oblique mammographic views. The inter-rater reliability of these density measurements was assessed for the first 100 mammograms. These scores were based on a variation of the Wolfe classification as described by Boyd et al (Table 1).21 For the purposes of the current study, the 9 categories were amalgamated into 3 groups of approximately equal size (low density, intermediate density, and high density). Patients in the low‒density group (<25% density) included those whose mammograms were classified as N, D1, or P1. The intermediate‒density group included patients with dysplasia (D2) or ductal prominence (P2) occupying approximately 25% to 50% of the breast. The high-density group included any patient whose mammogram was designated as D3, P3, D4, or P4, for a density of >50% overall.
Table 1. Categories of Breast Parenchymal Appearance
Fatty breast with <10% dense breast tissue
10-26% Extent of ductal prominence
25-50% Extent of ductal prominence
51-75% Extent of ductal prominence
>75% Extent of ductal prominence
10-25% Extent of dysplasia
26-50% Extent of dysplasia
51-75% Extent of dysplasia
>75% Extent of dysplasia
Correlations between breast density and other patient and tumor factors were examined. F-tests were used to compare means and chi-square tests were used to compare the distributions of these key variables among women in the 3 different density groups. The risk of breast cancer recurrence was compared for women in the 3 groups using the Kaplan-Meier method (univariate) and the Cox proportional hazards model (multivariate). In the multivariate analysis, the hazards ratio (HR) was adjusted for age, menopausal status, and the use of radiotherapy. Kaplan-Meier survival plots were generated and then analyzed using the log-rank test. All statistical tests were performed using SAS statistical software (version 9.1.3; SAS Institute, Inc, Cary, NC).
The 335 patients were distributed into groups with low density (n = 99 patients) intermediate density (n = 107 patients), or high density (n = 129 patients). Patients in the 3 groups were similar with respect to tumor characteristics and adjuvant treatments received (Table 2). Breast density was higher, on average, in younger women; the mean age at the time of diagnosis was 63.5 years for women in the low‒density group versus 54.9 years for women in the high‒density group (P < .01). Menopausal status was also found to be significantly different among the groups, with the highest percentage of postmenopausal patients noted in the low‒density group (82.8%). Only 52.7% of patients in the high‒density group were postmenopausal at the time of diagnosis. ER status was found to be similar in the 3 groups.
Table 2. Comparison of Patient, Tumor, and Treatment Factors for All Patients (N=335)
Low Density n = 99 (29.6%)
Intermediate Density n = 107 (31.9%)
High Density n = 129 (38.5%)
DOB indicates date of birth; LN, lymph node; NA, not available; ER estrogen receptor.
Mean age at diagnosis (range), y
Mean DOB (range)
LN status, no. (%)
No axillary LN staging
Bloom-Richardson grade No. (%)
Lymphovascular invasion, no. (%)
ER status, no. (%)
Menopausal status, no. (%)
Tumor size, mm (%)
Chemotherapy, no. (%)
Tamoxifen treatment, no. (%)
Radiotherapy, no. (%)
Surgical margins, no. (%)
Re-resection performed, no. (%)
The mean periods of follow-up were similar among the 3 groups (Table 3). In total, 34 patients developed a local disease recurrence (9.6%). The actuarial risk of local recurrence at 10 years was 14% (for all patients combined). Of those patients with high breast density, 15.5% (20 of 129 patients) developed a disease recurrence, versus 10.3% (11 of 107 patients) of those in the intermediate‒density group and 3% (3 of 99 patients) of those in the low‒density group (Table 3). The actuarial risk of local disease recurrence at 10 years was 5% for women in the low‒density group, 13% for women in the intermediate‒density group, and 21% for women in the high‒density group (Fig. 1).
Table 3. Comparison of Local and Distant Recurrence and Overall Survival (N = 335)
Low Density n = 99
Intermediate Density n = 107
High Density n = 129
Follow-up (range), y
Local recurrence (%)
Mean y to local recurrence (range)
Distant recurrence (%)
Both univariate and multivariate analyses demonstrated statistically significant differences between the groups with regard to the risk of local disease recurrence (Table 4). After controlling for age, menopausal status, and the receipt of adjuvant radiotherapy, patients in the high‒density group were found to have an HR of 5.7 for the development of local disease recurrence (95% confidence interval [95% CI], 1.6-20.0; P = .006) on multivariate analysis, compared with patients in the low‒density group. Using multivariate analysis again, patients in the intermediate‒density group were found to have an elevated HR compared with those in the low-density group (HR, 3.6; 95% CI, 1.0-13.2 [P = .05]). There was no difference in the rate of distant disease recurrence noted among women in the 3 mammographic density groups (data not shown).
Table 4. Hazards Ratios for Local Disease Recurrence in Univariate and Multivariate Analyses
As expected, the risk of local disease recurrence was found to be higher for women who did not receive radiotherapy (22% at 10 years) than for women who received radiotherapy (10% at 10 years). Among women who received radiotherapy, the prognostic effect of mammographic density was found to be minimal and nonsignificant (Fig. 2). However, among women who did not receive radiotherapy, the difference was marked and was found to be statistically significant (Fig. 3). Among women who did not receive radiotherapy, the actuarial risk of local disease recurrence at 10 years was 40% for women in the high‒density group. None of the 34 women with low breast density who did not receive radiotherapy experienced a local disease recurrence (P < .01 for difference).
We found mammographic breast density to be a significant risk factor for the development of local disease recurrence after breast-conserving surgery. To our knowledge, the current study is 1 of the few studies published to date to evaluate breast density as a predictor for breast cancer recurrence; in general, these studies have demonstrated consistent results. Recently, Park et al19 conducted a nested case-control study of 136 patients with invasive breast cancer who were treated with breast‒conserving surgery and radiotherapy. They examined the influences of mammographic density and obesity on the risk of local disease recurrence. Patients with the highest breast density (≥75%) had a greater than 4-fold incidence of local, but not distant, disease recurrence compared with women with low breast density. We observed a (nonsignificant) difference in the risk of local disease recurrence associated with mammographic density in women after radiotherapy, but the effect was strong and was statistically significant in women who did not receive radiotherapy. Breast density is also a risk factor for the development of invasive breast cancer among women treated for DCIS. Habel et al20 assessed the risk of subsequent breast cancer (DCIS or invasive disease) for 504 patients who had been treated for DCIS. After adjusting for age, body mass index, and adjuvant radiotherapy, patients with highly dense breasts (>75% of the breast occupied by dense tissue) were found to have approximately 3 times the risk of a subsequent ipsilateral breast cancer than did patients with <25% density.20
Mammographic density appears to be 1 of the few risk factors for local disease recurrence after breast-conserving surgery. Arriagada et al5 found that among 717 women treated with breast-conserving surgery, the strongest risk factors for local disease recurrence were age <40 years and tumor location within an inner quadrant of the breast. Neither tumor size, tumor grade, nor lymph node involvement were found to be predictive of disease recurrence. However, breast density was not measured in their study, and it may be that the observed age effect was a reflection of the high mean breast density in young women.
To the best of our knowledge, the biologic bases for the associations between mammographic density, breast cancer risk, and breast cancer recurrence are not known. The mammographic appearance of the breast is correlated with the relative proportions of stroma, epithelium, and fat. Guo et al22 demonstrated that tissue from dense breasts had relatively high quantities of collagen and high concentrations of insulin-like growth factor (IGF-1) and tissue inhibitor of metalloproteinase 3 (TIMP-3). IGF‒1 and TIMP-3 have been identified as paracrine regulators within the breast,17, 22, 23 and circulating levels of IGF-1 have been reported to be predictive of breast cancer risk in premenopausal women.22
The gene expression pattern of the breast stromal tissue has been shown to affect distant recurrence rates in breast cancer patients. Finak et al24 were able to predict the risk of distant metastases based on the gene expression profiles of the tumor stroma within and adjacent to a breast cancer, independent of other traditional prognostic variables. To our knowledge, it is not yet clear whether the observed stromal expression profile is a stable host factor that is characteristic of the breast itself or is dynamic and under the influence of the adjacent cancer. It will be of interest to learn whether this prognostic signature also predicts the rate of local disease recurrence and of breast cancer per se.
We are hopeful that similar gene expression experiments will identify a signature characteristic of high mammographic density (and ultimately that the modification of the expression of the relevant genes will be therapeutic). There has been considerable attention paid to the identification of modifiable risk factors for the prevention of primary and recurrent disease and to the potential use of breast density as an intermediate biomarker of risk. Breast density has been found to be modifiable to some extent by physical activity25 and hormone therapy.26 However, it is not clear whether a reduction in breast density through dietary or lifestyle modifications is associated with a reduction in the rate of local disease recurrence.
In the current study, mammographic density was assigned to patients based on the categories originally proposed by Wolfe.14 Recently, quantitative classification schemes have been proposed that assign an index of percentage breast density between 0% and 100%.17 These indicators of total breast density are highly predictive of breast cancer risk; it will be of interest to establish which indicator best predicts local disease recurrence. In the study of Habel et al20 the percentage breast density score was found to be a better predictor of disease recurrence than the Wolfe classification.
We observed a large effect of mammographic density on local disease recurrence, but not on distant recurrence or death. The data from the current study support the observation of Arriagada et al5 that risk factors for local disease recurrence are distinct from those that predict distant recurrence (which is consistent with clinical data indicating that a reduction in the risk of local disease recurrence [eg, through radiotherapy] is not necessarily followed by a commensurate reduction in the rate of distant recurrence) and raises the question of whether radiotherapy reduces the risk of local disease recurrence through a cytotoxic effect on residual cancer cells or by rendering the stromal matrix less likely to support local tumor growth.
One of the strengths of the current study is that patients were diagnosed as long as 20 years previously, and breast density was measured and recorded at the time of diagnosis. In addition, because of the clinical standards in place at our hospital during that period, a substantial proportion of patients (29%) did not receive radiotherapy. In studies in which the majority of (or all) women receive radiotherapy, the association may be attenuated. It is important that the findings of the current study be confirmed in similar cohorts, although this may be difficult using patients treated with current protocols because the majority of women who undergo breast‒conserving surgery are currently treated with radiotherapy.
Studies are currently underway that have been designed to determine whether partial breast irradiation is as effective as whole‒breast radiotherapy in terms of local control (eg, the National Surgical Adjuvant Breast and Bowel Project [NSABP] B-39 protocol). Given the results of the current study, it is important that the baseline distributions of breast densities should be similar among women in the 2 treatment arms. It is also possible that the relative effectiveness of the 2 regimens might vary according to breast density; it will be of interest to compare the results of these studies in the subgroups of women with high and low breast densities.
In the current study, we observed no local disease recurrence among the 34 women who had low breast density and who did not receive radiotherapy. Unlike the women in the other 2 categories, we were unable to demonstrate that these women were compromised by the decision to forego radiotherapy. However, this is a small sample of women and our results are not definitive. If it could be confirmed that women with low breast density (comprising approximately 30% of the current study population) do not benefit from radiotherapy, this could then result in considerable savings, reduced morbidity, and improved quality of life. Similarly, we observed a local recurrence rate of 40% at 7 years among women in the high‒density group, compared with 8% for women with low breast density who received radiotherapy. We believe these data confirm the benefit of radiotherapy in women with dense breasts.
We thank the Canadian Breast Cancer Foundation (Ontario Chapter) for their support of the Banting Breast Cancer Database at Women's College Hospital.