|Home | About | Journals | Submit | Contact Us | Français|
Breast density is one of the strongest predictors of breast cancer risk. Women with the densest breasts are 4 to 6 times more likely to develop cancer compared with those with the lowest densities. Breast density is generally assessed using mammographic imaging; however, this approach has limitations. Magnetic resonance imaging and ultrasound tomography are some alternative imaging modalities that can aid mammography in patient screening and the measurement of breast density. As breast density becomes more commonly discussed, knowledge of the advantages and limitations of breast density as a marker of risk will become more critical. This review article discusses the relationship between breast density and breast cancer risk, lists the benefits and drawbacks of using multiple different imaging modalities to measure density and briefly discusses how breast density will be applied to aid in breast cancer prevention and treatment.
Breast cancer is the most common type of cancer among women in North America, accounting for approximately 1 in 3 cancers diagnosed. Breast cancer death rates began to decrease in the United States in 1990. This reduction reflects improvements in breast cancer treatment and possibly early detection via mammographic screening, though the role of screening is controversial. Additional effects that could also help explain the reduction include temporal changes in the frequencies of specific tumors types in response to a shift in prevalence of risk factors.
Available breast cancer risk prediction models have generally good ability to predict the number of breast cancers that will develop in a specific population (calibration), but assessment of individual risk (discrimination) is limited. Expansion of these models to include additional factors could improve discrimination. Of the many risk factors listed by the American Cancer Society (ACS), mammographic breast density is one of the strongest predictors of breast cancer risk. Attempts to incorporate breast density into predictive models have shown modest improvements in risk prediction estimates[4, 6, 7] and efforts to incorporate density in newer risk models are ongoing. Based on age and BMI distribution of US women, approximately 25 million U.S. women have heterogeneously or extremely dense breasts, so the inclusion of breast density is an important risk factor to consider when designing a predictive risk model.
The current gold standard for breast cancer screening is digital mammography. Mammography is a low-dose X-ray procedure that produces high quality images that allow visualization of the internal anatomy of the breast. Digital systems may have higher sensitivity than film based mammography among women under 50 with dense breasts[10, 11] and currently account for approximately 90% of mammographic screens in the US. The ACS suggests that women receive regular and annual screening mammograms beginning at age 40. Screening has been shown to reduce breast cancer mortality[1, 13] and early detection may enable less-aggressive surgeries and treatment options.
However, mammographic screening has come under increased scrutiny recently. Mammography suffers from both false negative and false positive results. High density has been associated with false negative mammographic screens. Screening also tends to favor detection of more indolent tumors that develop slowly and which may not have caused harm leading to “over-diagnosis”. The usefulness of mammographic screening to reduce mortality beyond that of physical self-examination has been called into question, particularly among younger women[15, 16].
For women whose lifetime risk of developing breast cancer is greater than 20%, additional forms of screening are recommended in addition to mammography. Among higher risk women, magnetic resonance imaging (MRI) and ultrasound breast screening are recommended in most states as part of a comprehensive cancer screening program. Since breast density is a strong risk factor for breast cancer, knowledge of breast density can greatly affect the assessment of risk. Therefore, screening methods that can stratify breast density for this higher risk cohort would be very useful in risk assessment.
Systematic studies have consistently demonstrated that increased mammographic density is related to increased breast cancer risk[19–21]. Boyd et al have investigated this relationship thoroughly over many years[22–26]. It was determined that, when compared to women with lower densities, women with the highest densities showed an increased risk of breast cancer of 4 to 6-fold. Other studies carried out by Byrne et al, Byng et al[28, 29], Wolfe et al and Vachon et al further highlight that irrespective of the modality used to measure mammographic breast density, high density is a strong indicator of breast cancer risk. Furthermore, Boyd et al. and others have demonstrated that high mammographic density is strongly associated with elevated risk of breast cancer detected by screening or presenting clinically between screening exams (i.e., interval cancers).
Breast density reflects breast tissue composition and its biophysical properties. Compared with non-dense, adipose-rich tissue, denser tissues contain a higher proportion of fibroglandular elements, including epithelial and stromal cells, connective tissue and intercellular matrix. Dense tissue impedes the passage of X-rays through the breast, resulting in white opaque regions on a mammogram. Dense tissue corresponds to elevated sound speed in ultrasound tomography (UST) and higher water content of fibroglandular tissue in MRI. While measurements obtained using different modalities are correlated, differences between measurements among technologies reflects the underlying physics of the modalities.
Proposed explanations for the strong positive relationship between breast density and breast cancer risk include:
It is unclear which, if any, of these hypotheses about density and risk are correct. However, the first two possibilities suggest that breast density might represent a causal intermediate in carcinogenesis, and therefore, lowering density could reduce breast cancer risk. Studies that combine novel density measurements with molecular pathology can inform this issue.
Denser breasts are associated with a higher content of epithelial cells and/or stromal cells which could reflect a greater number of cells at risk of malignant transformation[23, 33, 34]. Higher density is also associated with a greater content of terminal duct lobular units, the specific microanatomic structures from which nearly all breast cancers arise. However, epithelial cell numbers may not fully explain density related risk as most data suggest that percentage density is a stronger marker of breast cancer risk than absolute density. Boyd et al suggested that density is most variable in the population at younger ages and converges to a lower more similar value as women age. Given that cellularity and density both decline with age, while breast cancer incidence increases, higher cell numbers alone is likely too simplistic an explanation for the underpinning of density associated risk.
Higher mammographic density is also associated with greater collagen content. Collagen contributes to increased extracellular matrix stiffness[37, 38], which may increase cancer development[39, 40]. Studies in animal models suggest that tumorigenesis is promoted in the presence of increased collagen. More recently, Keely and colleagues have demonstrated that altered collagen fiber alignment in tissues may be associated with breast cancer, and particularly, with more clinically aggressive tumors[39, 40].
Molecular composition of non-dense tissue may also affect associations of density and breast cancer risk. Pettersson and colleagues found an inverse relationship between non-dense breast area and breast cancer risk, supporting the hypothesis that fat tissue within the breast may be protective against breast cancer development. The molecular relationship between BMI and mammographic breast density is not fully clear. In a recent study, correlations between fibroglandular volume and mammographic dense area were strongest among women that were categorized as lean. However, the strength of the correlations decreased with increasing BMI, suggesting that volumetric and area density measurements do not capture non-dense tissue in a similar manner. Further, data suggest that density is not associated with risk of breast cancer death in analyses adjusted for prognostic pathologic features (i.e. stage, grade, estrogen receptor status); however, a subset analysis has suggested that obese women with low density may be at elevated risk of death.
John Wolfe, at Wayne State University, was the first to propose a relationship between patterns of breast density on a mammogram and breast cancer risk over 35 years ago. Four classifications were used to separate women into groups according to their relative risk of developing breast cancer. In an effort to standardize mammographic reporting, the American College of Radiology developed an alternate visual assessment method known as the Breast Imaging Reporting and Data System (BI-RADS). The BI-RADS is a density estimation technique that typically involves a radiologist’s visual assessment of the mammogram. It is routinely reported for a large proportion of mammograms in the U.S. according to four categories: a (predominately fat), b (scattered densities), c (heterogeneously dense) and d (extremely dense) (Figure 1). However, due to the subjective nature of these classifications, considerable inter- and intra-rater variability exists.
The relationship of Wolfe’s classifications and BI-RADS density score to risk have been affirmed, and recognition that parenchymal patterns are markers of risk has stimulated the development of more quantitative and objective methods of breast density assessment. Quantitative methods to measure breast density involve the use of computer-assisted programs that are based on interactive or automatic thresholding and segmentation. Examples of such programs include Cumulus, Volpara and Quantra. Mammographic percent density (MPD) is determined by finding the ratio of fibroglandular to total breast tissue areas or volumes. MPD can then be treated as either a continuous or categorical variable in subsequent analysis. An example of the CUMULUS software being used to measure MPD is shown in Figure 2.
Despite being the current gold standard for breast imaging, mammography poses the following shortcomings for both cancer detection and breast density measurement:
Breast density calculated by the use of mammography is dependent on the variations in breast tissue composition along with X-ray attenuation properties of those tissues and the methods used to generate the images. The image processing that is applied to digital images to maximize contrast helps detect relevant signs of cancers, but it can also distort the calculation of density. Digital image processing algorithms ultimately function to spread the image more uniformly among all the possible grey levels, while film mammography tends to render most of the image towards the black or white levels with few pixels occupying the intermediate grey levels. The differences between manufacturers in the production of the X-ray hardware and processed mammographic images means that image quality may vary across institutions.
Breast density measurements using mammography are typically based on the 2D projected area of the breast instead of examining the entire 3D volume. For mammographic density measurements, each pixel is simply assumed to represent either completely dense or completely fatty tissue. In reality, each pixel represents the X-ray absorption and attenuation characteristics of both fibroglandular and fatty tissues combined. It is therefore possible that two women may show similar projected areas of dense tissue yet may have different volumes of dense tissue. Efforts have been made to estimate the volume of dense tissue based on these mammographic measurements. These efforts generally involve the use of algorithms or physics models that use the imaging acquisition parameters to convert the mammographic pixel values into estimates of the thickness of the fibroglandular tissue.
Although mammographic density estimation using volumetric methods, such as Quantra or Volpara, produce density measurements that correlate strongly with other measures of breast density[47, 48, 52], improvements in risk prediction when using volumetric as opposed to area density are mixed[25, 53–57]. This appears to be counterintuitive as the entire breast volume should logically contain more information than the projected area. Most of the algorithms require knowledge of the breast thickness and the calculated volumes are sensitive to small changes in the measured value. Small errors in the thickness can lead to inaccurate estimations of volume of dense and non-dense tissue that can reduce the performance in risk prediction, thereby off-setting theoretical advantages.
Mammography is currently used in routine clinical practice and is the only screening method that has been shown to reduce mortality from breast cancer[1, 13]. Therefore, it is likely that mammography will remain as the principal method of breast cancer detection. Digital mammography has largely replaced film mammography in the U.S., based on its reported advantage in detecting breast cancer among younger women with dense breasts, yet it is not optimal for the measurement of breast density. Beyond mammography, multiple breast imaging modalities have been developed that could be used to assess breast density, each offering specific advantages and limitations. The major alternatives to mammography are MRI[58–64] and UST[32, 65–74]. Additional imaging modalities that can also measure breast density include double X-ray absorption (DXA)[75–77], digital breast tomosynthesis[78, 79], positron emission mammography (PEM)[80, 81], molecular breast imaging, breast CT[83, 84] and optical imaging[85, 86]. Summaries of the advantages and limitations of these different imaging modalities are discussed in Table 1.
Magnetic resonance imaging (MRI) uses non-ionizing radio waves to produce 3D images of tissue structures with very good soft tissue contrast. For women with a greater than 20% lifetime risk of developing breast cancer, MRI is recommended as an additional form of screening, and MRI has been used to measure breast density as well. Breast density using MRI involves creating images where dense tissue can be segmented from non-dense tissue. Many different sequences have been used to create images. A commonly used protocol involves T1-weighted MRI scans, however correlations between MPD and T2 relaxation time have been measured. Furthermore, sequences known as IDEAL (Iterative Decomposition of water and fat with Echo Asymmetry and Least squares estimation) are believed to directly assess the biochemical features and composition of breast tissue but are still considered a surrogate for the real histological references standard. All of these techniques have shown strong correlations with mammographic density[58–63]. Despite the fact that different sequences have been used to measure breast density on MRI, no definite consensus has been reached about the optimal technical method. In addition, MRI is also associated with false-positive results, has lower specificity compared to mammography and has not yet shown to decrease mortality. Due to the absence of ionizing radiation, MRI may represent a useful screening tool for younger women. However, MRI requires special infrastructure and technical expertise to operate which increases its expense and makes widespread use for density measurement impractical. Current screening sequences also require intravenous contrast injections.
Ultrasound imaging, like MRI, is recommended as an additional form of screening for women with increased risk of developing breast cancer. When using automated breast ultrasound (ABUS) in combination with mammography, cancer detection increases compared with mammography alone. Efforts to measure breast density using ABUS have been attempted, but it is still a relatively new area of research. Using ABUS to measure density is a labor-intensive process that suffers from problems with accurate segmentation of the breast.
In 1976, Greenleaf et al observed that acoustic measurements made with transmission ultrasound (US) could characterize breast tissue and that differentiation between benign masses and cancers was possible based on plots of sound speed and attenuation. This result led to the development of US transmission scanners in an attempt to measure the transmission parameters in vivo. UST is one example of this type of ultrasound scanner. UST does not require ionizing radiation or breast compression. UST examinations are not time intensive and costs of ultrasound technology generally compare favorably with other breast imaging modalities. Therefore, UST may be used for risk assessment in conjunction with mammographic screening for breast cancer.
UST uses whole-breast sound speed as an indicator of breast density. In breast tissue, the speed of sound (v) has the following relationship to the elastic constant (c) and material density (ρ):
Studies have shown that in human tissues, the elastic constant scales proportionally to the cube of density (c ρ3)[90–92]. Substitution into this equation reveals that in human breast tissue, sound speed is directly proportional to density. Therefore, the average density of the breast can be measured by calculating the volume averaged sound speed (VASS) of the breast. This measurement represents an absolute scale that can easily be replicated between current and future machines without the need for phantom calibrations. Examples of the sound speed images produced by UST are shown in Figure 4.
Previous work done by Glide et al.[70, 71] was the first to examine the relationship between UST density measurements and mammography density measurements. It was found that sound speed correlated strongly with the known mass densities and CT numbers of the different regions in an anthropomorphic breast phantom. In vivo comparisons of the breast volume averaged sound speed (from UST) to film MPD measurements also gave strong correlations. A volumetric estimation of US percent density (USPD) also showed strong correlations with MPD.
More recent studies carried out by Duric et al. using a larger data set support the results by Glide et al. They showed that the average UST breast density correlates strongly with both film and digital mammographic density measurements. UST density measurements are positively associated with the amount of dense tissue on a mammogram and negatively associated with the amount of non-dense tissue. UST density is higher in premenopausal than in postmenopausal women and shows expected inverse associations with age and weight. Since sound speed is more directly linked to the physical density of the breast tissue than mammography, it has the potential to be a more accurate marker of breast tissue density than mammographic density measures. However, these results do not link UST density measurements with breast cancer risk.
Computer-aided detection (CADe) and computer-aided diagnosis (CADx) are techniques that are designed to assist in the interpretation of medical images[93, 94]. CAD systems are designed to improve interpretation of the abundant visual data presented in radiologic images and highlight specific questionable areas for greater scrutiny. CAD techniques have been applied to mammography to aid radiologists in the detection of masses, microcalcifications and architectural distortions. CAD may be viewed as a second reader that assists the radiologist by identifying areas that could have been missed. The impact of breast density on CAD systems has been investigated. Microcalcification sensitivity is independent of mammographic breast density but density decreases mass sensitivity. The false positive rate of mammographic screening is lower in non-dense breasts versus dense breasts. When using CAD, breast cancer detection was not affected by breast density. This suggests that CAD may be advantageous for patients with dense breasts. As mammographic CAD techniques continue to improve, women with dense breasts may see improvements in early detection and decreased masking bias that currently disproportionally affects them. In the future, extension of CADe and/or CADx to UST-based screening in women with dense breasts appears feasible, as well as helping to automate the monitoring of subtle density changes during chemoprevention and treatment monitoring.
The link between mammographic breast density and breast cancer risk is well documented. However, the link between other measures of breast density, such as UST sound speed measurements, and breast cancer risk has yet to be investigated. Since the other measures of breast density correlate strongly with mammographic density, it is expected that breast cancer risk would also correlate with the non-mammographic density measures. However, this must be formally validated in future studies.
Breast density can also potentially be used as a marker to observe changes in breast tissue in response to preventive and therapeutic strategies. Several studies have shown that approximately one-third to one-half of women receiving tamoxifen experience a decrease in mammographic density within about a year and that these density declines could predict favorable responses in both the treatment and preventive settings[98–100]. Preliminary work using UST density measurements shows that, among some women, changes in density among those treated with tamoxifen are detectable within a few months of initiating treatment. These results show potential to rapidly provide encouragement to responders to adhere to therapy, while re-directing non-responders to alternative treatment options. UST density has also been used to assess response to neoadjuvant chemotherapy.
In future studies, integrating imaging characteristics with tissue analysis of cellular composition, biophysical properties of tissues and molecular markers in tissues and blood, may reveal mechanisms that account for density associated risk. Understanding such mechanisms would provide directions for future research aimed at refining risk estimates and developing means to lower density, which could improve radiologic screening, and perhaps, prevent breast cancer.
Breast density is becoming an important topic in the discussion of women’s health. Although breast density is one of the strongest risk factors in determining breast cancer risk, work by Manning et al.  shows that women in Detroit, Michigan were generally unaware that increased breast density is associated with increased risk. Furthermore, they found that approximately one third of women in the study knew their own breast density. In an effort to increase awareness of breast density among women undergoing screening, a growing number of states have passed legislation requiring the reporting of breast density to patients. As more states begin mandating the reporting of breast density information to patients, general knowledge of the link between breast density and breast cancer risk will increase. These interests will converge to place greater demands for improved methods of accurately and reproducibly measuring breast density, assessing its role in risk prediction and guiding screening recommendations. Given the potentially substantial cost implications of these concerns, future research on breast density is needed.
This research was supported in part by the Intramural Research Program of the National Institutes of Health, National Cancer Institute. The authors acknowledge the support of the National Institutes of Health through contract No HHSN261201400038P and grant No R44CA165320 from the National Cancer Institute. Neb Duric and Peter Littrup have financial interests in Delphinus Medical Technologies Inc. whose SoftVue scanner was used in our previous studies involving sound speed measurements of the breast. Financial conflicts of interest are being managed by Wayne State University.