Breast Cancer Screening in Women With Dense Breasts: Current Status and Future Directions for Appropriate Risk Stratification and Imaging Utilization

Abstract

Breast density continues to be a prevailing topic in the field of breast imaging, with continued complexities contributing to overall confusion and controversy among patients and the medical community. In this article, we explore the current status of breast cancer screening in women with dense breasts including breast density legislation. Risk-based approaches to supplemental screening may be more financially cost-effective. While all advanced imaging modalities detect additional primarily invasive, node-negative cancers, the degree to which this occurs can vary by density category. Future directions include expanding the use of density-inclusive risk models with appropriate risk stratification and imaging utilization. Further research is needed, however, to better understand how to optimize population-based screening programs with knowledge of patients’ individualized risk, including breast density assessment, to improve the benefit-to-harm ratio of breast cancer screening.

Introduction

More than 40% of women in the United States have dense breasts (1). Multifactorial complexities revolve around breast cancer screening in women with dense breasts, contributing to confusion and controversies that can be challenging to navigate for this population and their referring providers. The clinical significance of breast density is two-fold, with implications for risk prediction and supplemental screening related to the risk of dense tissue in breast cancer development and the masking effect of dense tissue on mammography. In this article, we aim to provide an update on breast cancer screening in women with dense breast tissue on mammography. We describe the current status of screening women with dense breasts and provide future directions for appropriate risk stratification and imaging utilization in this group.

Effect of Breast Density on Mammography

While mammography is the only imaging modality proven to decrease breast cancer–specific mortality, interpretative performance characteristics are reduced in women with dense breasts (2). Breast tumors, which share similar x-ray attenuation properties with fibroglandular tissue, may be obscured in areas of high mammographic density, resulting in false negative examinations (3). Overall, the mean sensitivity of screening mammography declines with increasing breast density from 87.0% in women with almost entirely fatty breast tissue to 62.9% in women with extremely dense breast tissue (4,5). The “masking” effect in regions of high mammographic density contributes to increased interval breast cancer rates in women with dense breasts, which tend to have a more unfavorable biologic profile associated with worse prognosis than screen-detected tumors (6,7).

Breast Density Legislation

Since Connecticut became the first state to enact breast density notification legislation more than a decade ago, greater awareness of this topic has catapulted discussions about breast density among patients and providers (8). As of July 2022, strong patient advocacy has led to 38 of the 50 states and the District of Columbia in the United States passing breast density notification laws that mandate women undergoing routine mammography to receive information about their breast density (9). National breast density legislation was passed by Congress in February 2019, directing the Food and Drug Administration (FDA) to establish breast density reporting language. On March 27, 2019, the FDA announced proposed amendments to the Mammography Quality Standards Act (MQSA), including the requirement that facilities inform patients and their health care providers about breast density. The open commentary period ended in March 2020, and an FDA/MQSA Final Rule has yet to be published such that individual state laws continue to guide breast density reporting (9,10). These laws have significantly increased awareness about the implications of dense breast tissue on screening mammography, leading to further discussions between patients and providers about supplemental screening (11).

Despite the well-intentioned aim of improving patient education on breast density to support shared decision-making, gaps exist that may prevent women from benefiting fully from this knowledge. While most women are interested in being informed about their breast density, confusion and misconceptions about breast cancer risk related to breast density remains a challenge, largely due to issues related to health literacy and primary provider discomfort with addressing questions related to this topic (12–14). In addition, despite most states requiring content that includes language that mammograms may be suboptimal in assessment of dense breasts, only a limited number of states mandate insurance coverage for supplemental screening (14).

Breast Density Assessment

The term “breast density” is based on the differential attenuation of mammographic x-ray by the fibrous and glandular components of the breast, which appear whiter, ie, “denser,” on mammography, versus areas of fat in the breast, which appear darker on mammography. The relative amount and distribution of the fibroglandular tissue in the breast represents a normal spectrum and governs the assessment of breast density. The Breast Imaging Reporting and Data System (BI-RADS) 5th edition includes a qualitative scale for assessing breast density divided into four categories: almost entirely fatty, scattered areas of fibroglandular tissue, heterogeneously dense, and extremely dense (15). Approximately 43% of women between the ages of 40 and 74 years of age are considered to have dense breasts (heterogeneously or extremely dense breasts) (16). Breast density is not stagnant, however, and can change over time, demonstrating an inverse association with increasing age. Other factors influencing breast density include body mass index, menopausal hormone therapy, diet, and reproductive factors (17,18).

Breast density on mammography can be assessed using visual, automated, or semiautomated methods. Visual breast density assessment, the most common method for evaluating breast density, is subjective, which contributes to variable inter- and intraobserver reliability. Redondo et al demonstrated high discordance between adjacent BI-RADS breast density categories: almost entirely fatty and scattered fibroglandular, scattered fibroglandular and heterogeneously dense, and heterogeneously dense and extremely dense, with interobserver discordance of 16.3%, 14.0%, and 6.9%, respectively (19). Sprague et al also demonstrated pervasive and wide variation in qualitative density assessment across radiologists, with 17.2% of discordant dense versus nondense assessments among women with consecutive mammograms interpreted by different radiologists (20).

Increasing awareness about the shortcomings of visual assessment in determining breast density has led to growing discussions about using automated methods to standardize breast density thresholds for supplemental screening (21). Commercially available automated breast density tools are available, which largely determine breast density through computation of dense volume and volumetric percent density using raw images (21–23). While automated and semiautomated methods theoretically offer a promising solution to issues related to reliability and reproducibility associated with visual assessment, technical and woman factors have been shown to play a role in volumetric density classification inaccuracies (24).

Breast Density as a Risk Factor

Dense breast tissue represents a strong independent risk factor for developing breast cancer (25). Cumulative exposure to growth factors and hormones in areas of high mammographic density may stimulate increased cell division in epithelial and nonepithelial cells, which may explain the increased breast cancer risk seen in women with higher ratios of fibroglandular tissue to fatty breast tissue on mammography (26). Women with dense breasts have a 4 to 6 times greater risk of developing breast cancer compared to women with almost entirely fatty breast tissue (2). Wang et al estimate that when comparing women with dense breasts with women with scattered areas of fibroglandular tissue, the relative risk is 1.2 to 1.5 for heterogeneously dense breasts and 2.1 to 2.3 for extremely dense breasts (1).

Twin studies support that mammographic density is heritable. Boyd et al (18) showed the correlation between monozygotic twins was approximately twice as strong compared to that between dizygotic twins. Genetic factors explained the majority of variation in breast density in their study, with heritability estimated to be as high as 75%. Brand et al demonstrated that at least 25% of the variance in volumetric mammographic density is explained by common genetic variants (27). Mechanisms linking genetic variants of mammographic density and breast cancer susceptibility loci were supported in their study, providing more insight into a potential common biological shared pathway.

Risk-based Approach to Breast Cancer Screening in Women With Dense Breasts

Globally, 2.1 million new breast cancer cases are diagnosed each year. Over this same period, 600 000 breast cancer deaths occur, accounting for approximately 15% of all female cancer deaths (28,29). Population-based screening programs established largely during the 1980s, however, have contributed to significant declines in breast cancer mortality (30). The primary aim of population-based screening, which largely determines screening criteria based on age, is to detect breast cancer at an early stage when it is treatable with curative intent (31,32).

Interest in risk-based screening has grown substantially over the past two decades, with increasing awareness of breast density as a risk factor and further discussion about the role of supplemental screening in this group. While data evaluating the efficacy of risk-based screening protocols are limited, simulation modeling suggests screening regimens that incorporate breast cancer risk factors including breast density may lead to a more efficient use of health care resources, leading to greater cost savings and fewer unnecessary recalls and biopsies and increasing the benefit-to-harm ratio of women from different subgroups (33–36).

There are two main considerations for implementing a more tailored approach to breast cancer screening: (1) development of accurate and accessible tools to assess risk that incorporate breast density in individual women and (2) the use of advanced imaging techniques in women who are at increased breast cancer risk or whose mammograms are associated with lower interpretative accuracy (37).

Improved predictive accuracy of risk assessment models has been demonstrated by adding mammographic density to classic breast cancer risk models (38). Brentnall et al demonstrated improvements in the area under the receiver-operator curve (AUC) from 0.57 and 0.55 to 0.61 and 0.59 by incorporating breast density into the Gail and Tyrer-Cuzick models, respectively (39). Similar improvements in risk assessment were shown by adding visual and volumetric mammographic density with the Tyrer-Cuzick breast cancer risk model in a case-control study including 474 patient participants and 2243 healthy control participants between the ages of 40 and 79 years, with odds ratios of 1.55 and 1.40, respectively (40). Tyrer-Cuzick version 8 now enables mammographic density to be added using visual assessment and automated methods.

Individualized genomic information is also available to use in risk stratification and management. Over 300 single nucleotide polymorphisms (SNPs) associated with breast cancer have been identified (41,42). A polygenic risk score is a precision-medicine tool that can be used to calculate individualized breast cancer risk using genetic risk factors. The combined effect of individual breast cancer susceptibility variants may lead to a considerable increase in breast cancer risk. Pathogenic variants including CHEK2, ATM, TP53, PTEN, CDH1, STK11, PALB2, BRCA1, and BRCA2 are among numerous genetic mutations that have been associated with moderate to high breast cancer risk (43).

Combining breast density with genetic risk can lead to even greater predictive power. Van Veen et al demonstrated that adding SNPs and mammographic density to classical risk factors improves breast cancer risk prediction and better risk discrimination, separating more women into low and high-risk groups (44). Their study showed that incorporating SNP18 improved discrimination between cases and controls with observed risk from Tyrer-Cuzick demonstrating an AUC of 0.58, observed risk from Tyrer-Cuzick with mammographic density incorporated demonstrating an AUC of 0.64, and observed risk from Tyrer-Cuzick with mammographic density and SNP18 both incorporated demonstrating an AUC of 0.67.

Recent studies have shown that deep learning, an approach to using artificial intelligence based on learning data representations, offers a paradigm of sophisticated techniques that may also aid in providing more accurate risk assessment by incorporating mammographic density directly from mammographic imaging (45). Deep learning models are not limited by missing data points, unlike classical risk models, and have been shown to demonstrate similar sensitivity among specific subgroups. Yala et al developed a deep learning model that demonstrated improved 5-year risk prediction compared to the latest Tyrer-Cuzick risk assessment tool (46). The model accurately identified patients as high-risk in 41.5% of patients who would develop cancer within 5 years compared to the Tyrer-Cuzick model, which identified 22.9% of patients who would develop cancer within the same data set over the same period. Deep learning models may be particularly useful in patients with limited family history of breast or ovarian cancer and offer the ability to provide accurate risk assessment at the time of screening mammography, lessening the burden on primary care providers to identify high-risk women and to provide better guidance for women with dense breast tissue incorporating overall breast cancer risk.

Role of Supplemental Screening

Dense breast tissue represents only one of many breast cancer risk factors. Increasing age and genetic factors continue to represent the strongest risk factors for developing breast cancer. Additional risk factors that place women at high lifetime risk include strong familial history of breast cancer and/or ovarian cancer, prior personal history of chest radiation therapy between the ages 10 and 30 years, prior breast biopsy showing atypical ductal hyperplasia, and familial syndromes including Li-Fraumeni syndrome, Cowden syndrome, and Bannayan-Riley-Ruvalcaba syndrome (47). Major national organizations are largely in support of early or supplemental screening in women at high lifetime risk. Consensus on supplemental screening is highest among women in the highest risk category (>20% lifetime risk) including recommendations for annual high-risk breast MRI in this group, which accounts for nearly one-third of breast MRI examinations performed (48).

Women with dense breasts as their only risk factor have been characterized in the intermediate risk group (lifetime risk: 15%–20%); recommendations for screening strategy in this group are less clear, although advanced imaging may provide benefit. There are no data to support supplemental screening in women with nondense breast tissue at average risk.

A primary aim of supplemental screening is to improve breast cancer detection in women who are at increased risk and who have characteristics that may limit mammographic interpretation. Although long-term mortality benefit of supplemental breast imaging tools has not been confirmed by randomized controlled trials, these techniques have been shown to increase diagnostic yield of small, node-negative tumors (49). Access to these tools, however, may be limited by geographic availability or primary provider familiarity with risk assessment and supplemental screening tools.

Digital Breast Tomosynthesis

Digital breast tomosynthesis (DBT) was approved in 2011 by the U.S. FDA and is now widely available in most U.S. breast centers (50). Digital breast tomosynthesis has largely been incorporated into routine screening regimens at many centers, continuously moving further away from being characterized as a supplemental screening tool over time. Researchers from the Population-based Research Optimizing Screening through Personalized Regimens (PROSPR) consortium observed a trend of decreasing breast density categorization among women screened with DBT versus digital mammography (DM) alone. Data from the PROSPR consortium also found improved specificity of DBT compared to DM alone across all age and breast density groups (51).

Evidence continues to accumulate that combined DBT and 2D DM is superior to 2D DM alone (52). The prospective population-based Oslo Tomosynthesis Screening Trial demonstrated improved sensitivity from 54.1% to 70.5% and improved specificity from 94.2% to 95.0% when DBT was added to screening mammography (53). Nearly two additional cancers per 1000 women screened were identified. Similar improvements in mammography interpretative performance were demonstrated by the Screening with Tomosynthesis Or Mammography (STORM) trial, a prospective comparative study evaluating the effect of integrated 2D and 3D mammography in population breast-cancer screening (52). Cancer detection rates of 5.3 cancers per 1000 screens were shown using 2D only, compared to 8.1 cancers per 1000 screens for combined 2D and 3D screening. In a comparative effectiveness study in women aged 40 to 79 years, Lowry et al demonstrated cancer detection rates increased from 5.9 to 8.8 per 1000 exams and recall rates decreased from 241 of 1000 exams to 204 of 1000 exams in women between the ages of 50 and 59 years (54). Specifically, cancer detection rates were consistently higher and recall rate was lower for DBT than for DM in women with heterogeneously dense breasts across all age groups. However, these rates were not significantly different between DBT and DM for women with extremely dense breasts; thus other supplemental screening methods should be considered for these women.

Improved specificity by using the combined technique offers the advantage of limiting the negative downstream effects associated with false-positive results, including patient anxiety and unnecessary biopsies. Combined gains in improved cancer detection rate and reduction in recall rate have been shown to be the largest in women with heterogeneously dense breasts. However, improved diagnostic interpretative performance in women with extremely dense breast tissue has been modest, similarly supporting the need for additional options for advanced screening techniques (55).

Fully automated software tools for volumetric breast density measures estimated from DBT have been reported, showing stronger associations with breast cancer risk than density measures extrapolated from 2D DM (56). The Tomosynthesis Mammographic Imaging Screening Trial (TMIST) is a large randomized clinical trial currently underway comparing mammographic accuracy between standard digital and tomosynthesis mammography, with the goal of contributing to greater precision in breast cancer screening. Synthetic 2D (s2D) images can be generated from DBT acquisition. Studies have demonstrated equivalent performance of s2D with DBT compared to combined 2D DM with DBT (57). As a result, in an effort to reduce radiation dose, many centers forgo the acquisition of 2D DM in addition to DBT and use s2D with DBT acquisition only. Gastounioti et al has shown increased use of nondense BI-RADS density categories when using s2D by visual assessment, contributing to existing issues related to reliability and reproducibility associated with visual assessment (58).

Whole Breast US

Whole-breast US demonstrates a higher additional breast cancer detection rate in women who have dense breast tissue. Incremental cancer detection rates of 2 to 4 per 1000 women have been reported (59–62). On a single, prevalent screening round for women with heterogeneously or extremely dense breast tissue and elevated risk of breast cancer (at least one other risk factor for breast cancer), supplemental physician-performed screening US detected an additional 4.2 cancers per 1000 women (95% CI: 1.1–7.2) compared to mammography alone in the American College of Radiology Imaging Network (ACRIN) 6666 trial (61). The large majority of the mammographically occult US-detected cancers in this trial were invasive and node-negative, as was shown in other screening US studies. Subsequent supplemental incidence-screening US yielded an additional 3.7 cancers per 1000 screens (95% CI: 2.1–5.8) in the combined second and third years of the ACRIN 6666 trial (63).

Data from the J-START study found that in women aged 40 to 49, mammography with the addition of US increased sensitivity and decreased the number of interval cancers compared to mammography alone (64). The Adjunct Screening with Tomosynthesis or Ultrasound in Mammography-Negative Dense Breast (ASTOUND) trial showed that adjunct screening with US demonstrates statistically significantly better breast cancer detection than DBT in women with dense breasts (65). Multiple studies have demonstrated a reduction in interval cancer rates to below 10% by combining screening US with mammography (49,63). While increased false-positives rates associated with screening US exams are not insignificant, improvements in recall rates from the first prevalence screen compared to subsequent incidence screens have been demonstrated when prior US studies are available (66). Automated whole breast US (ABUS), approved by the FDA in 2012, addresses many of the technical disadvantages associated with whole-breast handheld US, including requirements for technical expertise to perform the exam (67). However, further evaluation of the role of ABUS in the clinical care setting is needed related to operational workflow and best practices for provider and patient guidance for use (68).

Breast MRI

Breast MRI is the most sensitive supplemental screening technique, demonstrating an additional cancer detection rate of around 15 per 1000 women screened (69–73). The improved cancer detection performance by MRI is particularly relevant for women with dense breasts, who would be considered at intermediate breast cancer risk while subject to limited sensitivity of mammography (72). A single screening MRI increased cancer detection yield by 14.7 per 1000 women (95% CI: 3.5–25.9) beyond that of combined screening mammography plus US for women with dense breasts and elevated breast cancer risk in the ACRIN 6666 substudy (63). All participants of the ACRIN 6666 trial had dense breasts and at least one other breast cancer risk factor, a substantial majority of whom were at intermediate breast cancer risk.

In the randomized, controlled Dense Tissue and Early Breast Neoplasm Screening (DENSE) trial in the Netherlands, which included women solely with extremely dense breasts and no other breast cancer risk factors, the first or prevalent supplemental screening MRI detected 16.5 additional cancers per 1000 screening examinations (95% CI: 13.3–20.5) and resulted in a significantly lower interval-cancer rate than in the mammography-only group (74). In the second screening round of the DENSE trial, MRI after a negative mammogram demonstrated an incremental cancer detection rate of 5.8 per 1000 screening examinations (95% CI: 3.8–9.0), alongside a concurrent decrease in false-positive rate compared to the first round of MRI screening among women with extremely dense breasts (75).

Several studies have shown that abbreviated breast MRI (AbMRI) may achieve a similar cancer detection rate as full multiparametric MRI protocols without sacrificing diagnostic accuracy (76,77). Moreover, reduced costs and magnet time for patients may increase availability of MRI screening to additional subgroups at increased risk who may derive benefit, including women with dense breasts and other women at intermediate risk for breast cancer. Comstock et al conducted a multicenter study that demonstrated a significantly higher rate of invasive breast cancer detection among women with dense breasts undergoing screening with AbMRI compared with DBT (11.8 per 1000 women vs 4.8 per 1000 women, respectively) (78). Only one cancer (high-grade Ductal Carcinoma In Situ) was not seen on AbMRI and detected on DBT alone. The results of this trial suggested that in women with dense breasts undergoing AbMRI screening, the contribution of mammography and/or DBT is limited (78).

Contrast-enhanced Mammography

Contrast-enhanced mammography (CEM) is an additional vascular-based imaging modality that improves cancer detection by identifying the angiogenic effects associated with tumor growth. There is accumulating evidence that CEM demonstrates improved diagnostic performance characteristics compared to digital mammography alone with better lesion detection and improved sensitivity and specificity (79,80). Pooled sensitivity and specificity estimates for CEM in the diagnosis of breast cancer are 0.89 and 0.84, respectively (81,82). Sung et al demonstrated improvements in sensitivity, specificity, and negative predictive value in the screening setting in 858 women with at least 1 year of follow-up. Six of 14 cancers were detected only due to contrast enhancement (83).

Availability of CEM also provides an alternative screening method for women who may meet guidelines to receive breast MRI screening from high-risk groups but who have contraindications including metallic implants, weight limitations, or claustrophobia. Current-generation mammography systems are often delivered with CEM capabilities, alleviating equipment acquisition costs and space allocation needs. A primary issue related to implementation, however, includes creating efficient systems and operational workflow to administer intravenous iodinated contrast safely.

Molecular Breast Imaging

Molecular breast imaging (MBI) is a technique that uses small semiconductor-based cameras to provide high-resolution functional images of the breast. Molecular breast imaging has a high sensitivity for the detection of breast lesions with an overall sensitivity of 90%, and sensitivity of 82% for lesions less than 10 mm in size (84). Shermis et al demonstrated an incremental cancer detection rate of 7.7 per 1000 with the use of MBI in 1696 women with dense breast tissue and negative mammography results (85). Rhodes et al demonstrated an improved invasive cancer detection rate in their assessment of 1585 asymptomatic women with mammographically dense breasts from 1.9 per 1000 with mammography use alone to 8.8 per 1000 women with combined mammography and adjuvant MBI. However, there was a statistically significant decrease in specificity from 89% with mammography alone to 83% for the combination, an increase in recall rate from 11.0% with mammography alone to 17.6% for the combination, and an increase in biopsy rate from 1.3% for mammography alone to 4.2% for the combination (86). Breast-specific gamma imaging (BSGI) uses a different detector than MBI, but similarly it has been shown to detect an additional 14 mammographically occult breast cancers in a sample of 849 women (1.65%) with increased risk of breast cancer, or 16.5 cancers per 1000 women screened (87). Examples of successful implementation of MBI in large community-based practices exist that use MBI as a supplemental tool in women with dense breasts, evaluating positive findings on MBI with US and US-guided biopsy when suspicious findings are identified (85). Increased whole-body radiation dose and prolonged compression represent primary challenges of MBI.

Challenges of Risk-based Screening

There are numerous challenges associated with screening women with dense breasts. Primary issues that need to be addressed include improving reliability and reproducibility of breast density assessment and developing clearer recommendations guiding the role of supplemental screening in this group. Policy aimed at (1) improving access to breast cancer risk assessment, (2) ensuring insurance coverage for advanced imaging services beyond mammography, and (3) better educating patients at all literacy levels about the benefits of multimodality screening regimens in women at increased risk of breast cancer, including women with dense breasts, are needed.

Future Directions

With mandatory breast density reporting leading to greater patient and physician awareness about the implications of dense breasts, opportunities exist to support discussions between patients and primary providers about the implications of dense breast tissue. Breast radiologists, who have expertise in multimodality screening regimens and who are well-informed about national guidelines, can leverage their knowledge to help fill this gap, providing both physician and patient education on this topic. There are numerous forums in which these discussions can take place, including the breast imaging suite, educational conferences, community events, and on social media.

In addition, greater efforts to discuss breast density in the context of overall breast cancer risk should be made to improve risk stratification and screening recommendations (26). To achieve this goal, more accurate and readily accessible risk assessment tools that incorporate mammographic density are needed to inform both patients and providers about their risk. Both approaches, including available genetic information into risk assessment tools and leveraging advances in artificial intelligence to perform more accurate and readily accessible risk assessment, offer promise in furthering these efforts.

Conclusion

Overall, broadening the conversation from breast density to the primary issue of breast cancer risk can help reduce patient and provider confusion and misinformation and support multimodal screening strategies that are based on a patient’s overall breast cancer risk, comorbidities, preferences, and values. Further research is needed, however, to better understand how to optimize population-based screening programs with knowledge of patients’ individualized risk, including breast density assessment, to improve the benefit-to-harm ratio of breast cancer screening, without leading to adverse results in women in lower risk categories (88).

Funding

None declared.

Conflict of Interest Statement

R.C.M. has an early career award funded by General Electric (GE) through the AUR GE Radiology Research Academic Fellowship (GERRAF) Award (https://www.aur.org/en/awards/aur-gerraf-award). He is also a consultant at Hologic. A.P. is associate editor of the Journal of the American College of Radiology; consultant at Hologic; and medical advisor for Kheiron Medical. All other authors declare no conflict of interest.

References