Mammographic density, MRI background parenchymal enhancement and breast cancer risk

Abstract

Mammographic density (MD), representing connective and epithelial tissue (fibroglandular tissue, FGT) is a major risk factor for breast cancer. In an analysis of an autopsy series (Bartow SA, Pathak DR, Mettler FA. Radiographic microcalcification and parenchymal patterns as indicators of histologic “high-risk” benign breast disease. Cancer 1990; 66: 1721–1725, Bartow SA, Pathak DR, Mettler FA et al. Breast mammographic pattern: a concatenation of confounding and breast cancer risk factors. Am J Epidemiol 1995; 142: 813–819), MD was found to be strongly correlated with the collagen and epithelial content of the breast (Li T, Sun L, Miller N et al. The association of measured breast tissue characteristics with MD and other risk factors for breast cancer. Cancer Epidemiol Biomarkers Prev 2005; 14: 343–349), and another report showed that breast epithelium was highly concentrated in the areas of collagen concentration (Hawes D, Downey S, Pearce CL et al. Dense breast stromal tissue shows greatly increased concentration of breast epithelium but no increase in its proliferative activity. Breast Cancer Res 2006; 8: R24). Collagen comprises the overwhelming majority of the FGT, occupying an area on the slides obtained from the autopsy series some 15 times the area of glandular tissue. The relationship of MD with breast cancer risk appears likely to be due to a major extent to increasing epithelial cell numbers with increasing MD. FGT is also seen in breast magnetic resonance imaging (breast MRI) and, as expected, it has been shown that this measure of FGT (MRI-FGT) is highly correlated with MD. A contrast-enhanced breast MRI shows that normal FGT ‘enhances’ (background parenchymal enhancement, BPE) after contrast agent is administered(Morris EA. Diagnostic breast MR imaging: current status and future directions. Radiol Clin North Am 2007; 45: 863–880, vii., Kuhl C. The current status of breast MR imaging. Part I. Choice of technique, image interpretation, diagnostic accuracy, and transfer to clinical practice. Radiology 2007; 244: 356–378), and a recent study suggests that BPE is also a major breast cancer risk factor, possibly as important as, and independent of MD (King V, Brooks JD, Bernstein JL et al. BPE at breast MR imaging and breast cancer risk. Radiology 2011; 260: 50–60). BPE is much more sensitive to the effects of menopause and tamoxifen than is FGT (King V, Gu Y, Kaplan JB et al. Impact of menopausal status on BPE and fibroglandular tissue on breast MRI. Eur Radiol 2012; 22: 2641–2647, King V, Kaplan J, Pike MC et al. Impact of tamoxifen on amount of fibroglandular tissue, BPE, and cysts on breast MRI. Breast J 2012; 18: 527–534). Changes in MD and BPE may be most useful in predicting response to chemopreventive agents aimed at blocking breast cell proliferation. More study of the biological basis of the effects of MD and BPE is needed if we are to fully exploit these factors in developing chemopreventive approaches to breast cancer.

introduction

The extent of mammographic density (MD), i.e. the white areas on a mammogram representing connective and epithelial tissue (fibroglandular tissue; FGT) in contrast to fat, is a major breast cancer risk factor with the risk of breast cancer in women of the same age being close to directly proportional to the amount of MD as well as to the percent MD (i.e. MD as a proportion of the area of breast on the mammogram, MD%) [1]. FGT can also be measured using breast magnetic resonance imaging (breast MRI), and it has been shown that this measure of FGT (MRI-FGT) is highly correlated with MD [2]. Breast MRI when used with a contrast agent (contrast breast MRI) is a more sensitive method than mammography in detecting early breast tumors. With contrast breast MRI one compares a baseline MRI with the one taken shortly after the contrast agent is administered; and breast tumors show marked ‘enhancement’ in the post-contrast MRI. Some degree of enhancement is also seen in normal breast tissue—termed background parenchymal enhancement (BPE). The ‘extent’ of BPE varies quite markedly between different women, and in a case–control study completed at the Memorial Sloan-Kettering Cancer Center a woman's risk of breast cancer appeared to be at least as markedly affected by the extent of BPE as it was by the amount of MRI-FGT [3]. Furthermore, this risk from BPE appeared to be close to independent of the risk from MRI-FGT and presumably from MD.

MD

On a mammogram, connective and epithelial tissues (fibroglandular tissue, FGT) appear white (‘dense’), whereas fat appears dark (‘non-dense’). A strong association between the mammographic parenchymal pattern and breast cancer risk was first proposed by Wolfe in 1976 [4–6]. Since that time it has been shown that the extent of dense tissue on the mammogram (MD), i.e. the area of the mammogram considered to be ‘white’, is a reproducible and major risk factor for breast cancer [1]. To permit comparisons between different mammographic methods, MD was often reported as a proportion (percentage) of the area of the breast on the mammogram (MD%).

Table 1 shows the relationship between MD% and breast cancer risk in a classic study [7]. The risk increases steadily with increasing MD% and some 4.6-fold between women with an MD% ≥75% compared with a woman with an MD% of <10%.

Table 1.

MD % and RR of breast cancer

MD%	RR	Population (%)
None	1.0	7
<10	1.2	17
10−	2.2	21
25−	2.4	27
50−	3.4	19
75+	5.3	9

MD%, mammographic density %; RR, relative risk.

Data from Boyd et al. [7].

MD and MD% are highly correlated with each other, and Table 2 shows that they are almost equally connected with the risk of breast cancer [8]. For a given MD, MD% is reduced with increasing body mass index (BMI) and studies of the effect of MD% are almost invariably improved by adjustment for BMI. In a study where the effects of MD and MD% were examined in detail, the fit of MD% with breast cancer risk was somewhat improved with adjustment for the non-dense area of the mammogram, whereas the fit of MD was unaffected and the authors concluded that MD was to be preferred on grounds of simplicity [8].

Table 2.

MD and MD% and RR of breast cancer

MD		MD%
Quintile	RR	Quintile	RR
Q1	1.00	Q1	1.00
Q2	1.41	Q2	1.62
Q3	1.50	Q3	1.46
Q4	1.94	Q4	1.89
Q5	2.85	Q5	2.71

MD, mammographic density; MD%, mammographic density percent; RR, relative risk.

Data from Stone et al. [8].

Similar to MD, FGT can be seen on breast MRI (MRI-FGT). Studies have shown that the extent of MRI-FGT correlates strongly with MD, and, since breast MRI is three-dimensional, MRI-FGT or MRI-FGT% (based on volume of the breast) may provide a better measure of breast cancer risk than MD, but this has not been demonstrated [9–12].

MD has a large genetic component [13, 14], but is also clearly related to various breast cancer risk factors. This is most clearly seen by considering the effect of menopause on breast cancer risk. Figure 1 shows the age incidence of breast cancer in US white females before the introduction of extensive screening programs, which distorted the curve by the inclusion of very early lesions [15, 16]. For most non-hormone-dependent cancers incidence increases exponentially with increasing age and a straight line is obtained if we plot the logarithm of incidence against the logarithm of age, but as is clear from Figure 1 there is a clear slowing down of the rate of increase of breast cancer around age 50; this has been demonstrated to be an effect of menopause [17]. The lower estrogen and particularly progesterone levels in postmenopausal women are less stimulating to the breast than the hormonal profile of premenopausal women [18]. This change in the rate of increase in breast cancer and the lower level of proliferation of the breast tissue is reflected in a decrease of MD after the decline in ovarian function and the onset of menopause and this decline is reversed by use of menopausal estrogen-progestin therapy [1, 19]. MD density is also less in parous women and declines further with increasing numbers of live births [20].

Figure 1.

Age-specific incidence rates for breast cancer in US White females (Third National Cancer Survey).

association of MD with breast tissue components

Using breast tissue samples from a large non-selected forensic autopsy series compiled by Bartow et al. [21, 22], Boyd et al. [23] showed that MD%—measured in this case by Bartow et al. in an overall assessment of the Faxitron images of the breast slices made of the whole breast and named Faxitron density %—was strongly correlated with the proportion of tissue area on the associated ‘random’ microscope slide occupied by collagen. Faxitron density % was also strongly correlated with the proportion of tissue area occupied by the nuclei of breast epithelia and by glandular tissue. Collagen comprised the greatest quantity of fibroglandular tissue in the breast [20]; collagen occupied an area some 15 times the area occupied by glandular tissue. The latter result is not generally recognized.

We initially reported that very early DCIS lesions in the breast were invariably in areas of the breast containing significant densities [24]; and moreover found in a study of breast tissue from women having a reduction mammoplasty that normal breast epithelium was highly concentrated in the areas of the breast containing a high concentration of inter-lobular collagen [25] and subsequently confirmed this in large studies of breast tissue from normal weight women [26, 27]. Taken together with the results reported by Boyd et al. [23], these results show that the number of epithelial cells varies many fold between different women mirroring the wide variation in MD, so that the relationship of MD with breast cancer risk may be due to a significant extent simply to increased epithelial cell numbers. We found that normal breast epithelial cell proliferation was higher in those areas where there was little associated collagen [25]; this is evidence against the densities as such being an important modifier of breast cancer risk, but other evidence does suggest a direct effect and this is an area of intense research [28, 29].

The biological basis of the relationship of collagen in the human breast with epithelium is unclear. Blocking ovarian function with a gonadotropin-releasing hormone (GnRH) analog reduces MD by approximately one-third [30–32], which then returns when the GnRH analog is withdrawn [33]. Some densities appear, therefore, to be closely related to current hormone exposure with its associated greater breast cell proliferation, while others change only slowly with increasing age after the menopause. Differences between these types of density have not been elucidated.

breast BPE

Breast MRI when used with a contrast agent (contrast breast MRI) is a more sensitive method than mammography in detecting early breast tumors as a consequence of the different behavior of the contrast agent in tumors when compared with normal breast tissue. BPE refers to the volume and intensity that normal fibroglandular tissue ‘enhances’ after intravenous contrast agent administration [34, 35]. BPE is categorized in the BIRADS system as minimal, mild, moderate or marked. BPE is thought to be a measure of the amount of blood flow in the dense tissue and may represent breast activity. As we noted earlier, baseline breast MRI before the administration of the contrast agent can also be used to measure the extent of FGT in the breast. MRI-FGT is categorized in the BIRADS system as fatty (<25% of breast comprises fibroglandular tissue), scattered (25%–50% of breast comprises fibroglandular tissue), heterogeneously dense (HD, 51%–75% of breast comprises fibroglandular tissue) or dense (>75% of breast comprises fibroglandular tissue).

Table 3 shows the relationship between BPE and MRI-FGT and breast cancer risk in a case–control study we recently carried out at Memorial Sloan-Kettering Cancer Center [3]. BPE varies markedly between different women, and risk increases steadily with increasing BPE. The risks associated with BPE are large, of the same order of magnitude as are found with MD. As we would expect, the risk also increases steadily with increasing MRI-FGT but possibly not to quite the same extent as with BPE. Adjustment for MRI-FGT resulted in only a slight decrease in the estimated effects of BPE. BPE is therefore a new significant breast cancer risk factor, essentially independent of MRI-FGT. Inspection of the results for the control women in Table 3 shows that there is a marked decline in the level of BPE between premenopausal and postmenopausal women: the percentage of controls showing moderate or marked BPE declined from 38.5% to 12.2%. Table 3 shows that there is a much smaller decline in MRI-FGT between premenopausal and postmenopausal women: the percentage of controls showing HD or dense MRI-FGT declined from 69.2% to 63.4%.

Table 3.

BPE level and RR of breast cancer

BPE	Cases N (%)	Controls N (%)	RR	MRI-FGT	Cases N (%)	Controls N (%)	RR
Minimal/mild	16 (41.0)	63 (80.8)	1.0	Fatty/scattered	7 (18.0)	26 (33.3)	1.0
Moderate	15 (38.5)	12 (15.4)	8.2	HD	19 (48.7)	35 (44.9)	2.0
Marked	8 (20.5)	3 (3.8)	18.2	Dense	13 (33.3)	17 (21.8)	3.2
Minimal/mild	16 (41.0)	63 (80.8)	1.0	Fatty/scattered	7 (18.0)	26 (33.3)	1.0
Moderate/marked	23 (59.0)	15 (19.2)	10.1	HD/dense	32 (82.0)	52 (66.7)	2.3
Premenopausal:				Premenopausal:
Minimal/mild	4 (23.5)	16 (61.5)	1.0	Fatty/scattered	1 (5.9)	8 (30.8)	1.0
Moderate/marked	13 (76.5)	10 (38.5)	5.1	HD/dense	16 (94.1)	18 (69.2)	5.2
Postmenopausal:				Postmenopausal:
Minimal/mild	12 (54.6)	36 (87.8)	1.0	Fatty/scattered	6 (27.3)	15 (36.6)	1.0
Moderate/Marked	10 (45.4)	5 (12.2)	14.2	HD/dense	16 (72.7)	26 (63.4)	1.6

CI, confidence interval; HD, heterogeneously dense; RR, relative risk.

Data from King et al. [3].

We confirmed these lower postmenopausal levels of BPE in a subsequent study that measured BPE in individual women before and after menopause [36]. Table 4 shows that in this study of individual women, the percentage of moderate and marked BPE declined from 53.6% to 14.3%. The percentage of HD or dense MRI-FGT also declined, from 85.7% to 64.2%, but to a lesser extent than BPE.

Table 4.

BPE and MRI-FGT change with menopause

BPE	Pre N (%)	Post N (%)
Minimal	4 (14.3)	15 (53.6)
Mild	9 (32.1)	9 (32.1)
Moderate	11 (39.3)	3 (10.7)
Marked	4 (14.3)	1 (3.6)
MRI-FGT	Pre N (%)	Post N (%)
Fatty	0 (0.0)	2 (7.1)
Scattered	4 (14.3)	8 (28.6)
HD	19 (67.9)	16 (57.1)
Dense	5 (17.9)	2 (7.1)

HD, heterogeneously dense; Pre, premenopausal; Post, postmenopausal.

Data from King et al. [36].

In a further study, we showed that both BPE and MRI-FGT decrease in the contralateral breast of breast cancer patients treated with tamoxifen (Table 5) [37]. With tamoxifen treatment, the percentage of moderate and marked BPE declined from 59.0% to 5.7%. The percentage of HD or dense MRI-FGT also declined, from 85.2% to 73.9%, but again to a much lesser extent than BPE. The decreases in BPE were not associated with the duration of tamoxifen treatment; decreases occurred ‘even in patients who were on tamoxifen for a short period, as little as 39 days’ [37]. In contrast, the proportion of patients showing a decrease in MRI-FGT became larger with a longer duration of tamoxifen treatment [37].

Table 5.

BPE and MRI-FGT at baseline and on tamoxifen

BPE	Baseline N (%)	On tamoxifen N (%)
Minimal	14 (15.9)	46 (52.3)
Mild	29 (33.0)	37 (42.0)
Moderate	27 (38.5)	5 (5.7)
Marked	18 (20.5)	0 (0.0)
MRI-FGT	Baseline N (%)	On tamoxifen N (%)
Fatty	2 (2.3)	6 (6.8)
Scattered	11 (12.5)	17 (19.3)
HD	34 (38.6)	41 (46.6)
Dense	41 (46.6)	24 (27.3)

HD, heterogeneously dense.

Data from King et al. [37].

The same decreases occurred, but to a lesser extent, in patients treated with the aromatase inhibitor, anastrazole (Table 6) [38]. With anastrozole treatment, the percentage of moderate and marked BPE declined from 25.7% to 8.3%. The percentage of HD or dense MRI-FGT also declined but to a very small extent, from 45.9% to 43.2%. The tamoxifen treated patients were, of course, much younger than the anastrozole-treated patients, who were all postmenopausal. This is the reason for the much higher baseline levels of both BPE and MRI-FGT in the tamoxifen-treated patients.

Table 6.

BPE and MRI-FGT at baseline and on anastrozole

BPE	Baseline N (%)	On AI N (%)
Minimal	17 (15.6)	34 (31.2)
Mild	64 (58.7)	66 (60.6)
Moderate	27 (24.8)	9 (8.3)
Marked	1 (0.9)	0 (0.0)
MRI-FGT	Baseline N (%)	On AI N (%)
Fatty	15 (13.8)	18 (16.5)
Scattered	44 (40.4)	44 (40.4)
HD	35 (32.1)	32 (29.4)
Dense	15 (13.8)	15 (13.8)

AI, anastrozole; HD, heterogeneously dense.

Data from King et al. [38].

discussion

Tamoxifen has been repeatedly shown to reduce MD [39–41]. The reduction in MD with tamoxifen use is much more pronounced in younger women. In the report from Brisson et al. [40], the reduction in MD% due to tamoxifen use was 8.5% in women <50 years of age and 2.2% in women ≥50 years of age. In the report from Cuzick et al. [41], the reduction in MD% was 13.4% in women ≤45 years of age and was 1.1% in women ≥56 years of age. These results are in close agreement with the results we found with MRI-FGT. In their analysis of the IBIS-I randomized prevention trial of tamoxifen use, Cuzick et al. [42] reported that the benefit in terms of reduced breast cancer incidence in the tamoxifen arm of the trial was restricted to women whose MD% was reduced by at least 10%. It is difficult to reconcile this result with their previous finding that MD% was only reduced by 1.1% in women ≥56 years of age [42], when the NSABP P1 tamoxifen prevention study [43] found that the benefit of tamoxifen use was equally strong in such older women. As Cuzick et al. [42] state in their paper, their results need ‘confirmation in a separate study’.

The results presented in Table 5 suggest that BPE may be a better marker of the effect of tamoxifen treatment. Validating this in a prevention setting will be difficult if not impossible to do, but it may be possible to show that the extent of the benefit of tamoxifen treatment of breast cancer patients was associated with the degree to which BPE and MRI-FGT (or MD) was changed. The same applies to validating the use of BPE and possibly MRI-FGT with the use of aromatase inhibitors.

Much more study of the biological basis of these major risk factors in normal human breast is needed if we are to exploit them in some way to help in developing chemopreventive approaches to breast cancer.

funding

This work was supported by a Department of Defense Congressionally Directed Breast Cancer Research Program Grant BC044808 and by the USC/Norris Comprehensive Cancer Center Core Grant P30 CA14089 from the National Cancer Institute. The funding sources had no role in this report.

disclosure

The authors have declared no conflicts of interest.