Diagnostic Accuracy of Screening Contrast-enhanced Mammography for Women with Extremely Dense Breasts at Increased Risk of Breast Cancer

Abstract

Background

Mammogram interpretation is challenging in female patients with extremely dense breasts (Breast Imaging Reporting and Data System [BI-RADS] category D), who have a higher breast cancer risk. Contrast-enhanced mammography (CEM) has recently emerged as a potential alternative; however, data regarding CEM utility in this subpopulation are limited.

Purpose

To evaluate the diagnostic performance of CEM for breast cancer screening in female patients with extremely dense breasts.

Materials and Methods

This retrospective single-institution study included consecutive CEM examinations in asymptomatic female patients with extremely dense breasts performed from December 2012 to March 2022. From CEM examinations, low-energy (LE) images were the equivalent of a two-dimensional full-field digital mammogram. Recombined images highlighting areas of contrast enhancement were constructed using a postprocessing algorithm. The sensitivity and specificity of LE images and CEM images (ie, including both LE and recombined images) were calculated and compared using the McNemar test.

Results

This study included 1299 screening CEM examinations (609 female patients; mean age, 50 years ± 9 [SD]). Sixteen screen-detected cancers were diagnosed, and two interval cancers occured. Five cancers were depicted at LE imaging and an additional 11 cancers were depicted at CEM (incremental cancer detection rate, 8.7 cancers per 1000 examinations). CEM sensitivity was 88.9% (16 of 18; 95% CI: 65.3, 98.6), which was higher than the LE examination sensitivity of 27.8% (five of 18; 95% CI: 9.7, 53.5) (P = .003). However, there was decreased CEM specificity (88.9%; 1108 of 1246; 95% CI: 87.0, 90.6) compared with LE imaging (specificity, 96.2%; 1199 of 1246; 95% CI: 95.0, 97.2) (P < .001). Compared with specificity at baseline, CEM specificity at follow-up improved to 90.7% (705 of 777; 95% CI: 88.5, 92.7; P = .01).

Conclusion

Compared with LE imaging, CEM showed higher sensitivity but lower specificity in female patients with extremely dense breasts, although specificity improved at follow-up.

© RSNA, 2024

See also the editorial by Lobbes in this issue.

Summary

Compared with low-energy imaging, contrast-enhanced mammography depicted 11 additional cancers in asymptomatic women with extremely dense breasts and showed higher sensitivity but lower specificity, with specificity improving at follow-up examinations.

Key Results

• ■ In this retrospective study involving 609 asymptomatic female patients with extremely dense breasts (Breast Imaging Reporting and Data System category D; 1299 examinations), contrast-enhanced mammography (CEM) depicted an additional 11 cancers compared with low-energy (LE) imaging (incremental cancer detection rate, 8.7 cancers per 1000 screenings).

• ■ CEM showed higher sensitivity than LE imaging (16 of 18 [88.9%] vs five of 18 [27.8%]; P = .003), albeit at the cost of decreased specificity (1108 of 1246 [88.9%] vs 1199 of 1246 [96.2%]; P < .001).

• ■ CEM specificity improved at follow-up examinations compared with baseline (705 of 777 [90.7%] vs 403 of 469 [85.9%]; P = .01).

Introduction

Increased mammographic density is an independent risk factor for breast cancer (1) and is associated with diminished mammographic sensitivity (2–4) and increased risk for interval cancer (5,6). The term dense breasts refers to the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) categories C and D, and the breast density of approximately half of the female population of the United States falls into these two categories (7). Although most women with dense breasts have heterogeneously dense breasts (ie, category C), women with extremely dense breasts (ie, category D) account for approximately 7.4% of the screening population, equating to an estimated 4.7 million women in the United States (8). This population is of particular interest because they present a greater challenge for mammography while carrying a higher risk of breast cancer (9).

Several screening strategies have been introduced for women with dense breasts. Tomosynthesis (10,11) was found to have limited added value in a subpopulation of women with extremely dense breasts (12). Adjunct US yielded an increased cancer detection rate of approximately 3.5 per 1000 screenings, although with a substantially increased number of false-positive results (13–15). Recently, the Dense Tissue and Early Breast Neoplasm Screening (DENSE) trial (16) evaluated the use of MRI as a supplementary screening modality in women with extremely dense breasts and found an incremental cancer detection rate of 16.5 of 1000 compared with mammography, alongside a reduction in the interval cancer rate (17,18). Although these findings may have direct implications for screening recommendations (19), from a practical perspective, the implementation of MRI for mass population screening is unlikely because of its cost and limited availability (20).

Contrast-enhanced mammography (CEM) has emerged as a potential alternative to MRI, offering greater availability for vascular imaging at reduced costs (21) while showing similar performance in breast cancer detection (22). CEM combines low-energy (LE) imaging, the equivalent to routine digital mammography, for morphologic evaluation, and recombined imaging, which provides vascular mapping. Promising results have been observed for CEM in the screening setting (23–27). However, scant data exist regarding CEM utility for women with extremely dense breasts. This study aimed to evaluate the diagnostic performance of CEM for breast cancer screening among women with extremely dense breasts.

Materials and Methods

Study Sample

This retrospective Health Insurance Portability and Accountability Act–compliant study was approved by our institutional review board. The need to obtain informed consent was waived.

At our institution, screening CEM must be specifically ordered by the referring physician or performed as part of a research study (23). CEM has been available at this comprehensive cancer center since December 2012 and is used mostly to screen female patients at intermediate lifetime risk (ie, 15%–20% risk for developing breast cancer) (28), including patients with a personal or family history of breast cancer or a history of a high-risk lesion (including atypical ductal hyperplasia, atypical lobular hyperplasia, lobular carcinoma in situ, papillary lesions, radial scar, flat-cell epithelial atypia, mucocele-like lesions, and phyllodes tumor) (29). To a lesser extent, patients at high risk of breast cancer (ie, risk >20%), such as carriers of the BRCA mutation and patients with a history of chest irradiation, may also be screened with CEM instead of digital mammography if their annual mammography and breast MRI are staggered at 6-month intervals or if they are unable to undergo MRI (23). Contraindications for CEM included a history of an allergic reaction to iodine or reduced renal function, in accordance with the American College of Radiology guidelines outlined in the Manual on Contrast Media (30), and large-breasted patients, requiring tiling to image the entire breast.

Inclusion criteria was screening CEM studies with 1-year follow-up in asymptomatic female patients with extremely dense breasts, whereas symptomatic patients and patients without a biopsy reference or at least 1 year of follow-up were excluded (Fig 1). The department database was searched, and screening CEM examinations performed from December 2012 to March 2022 for consecutive asymptomatic patients with extremely dense breasts were identified. The final study sample included 103 female patients who were part of a previous study (23) exploring the efficacy of screening CEM for female patients with elevated breast cancer risk, alongside 51 female patients who were part of an additional study (27) focused on screening CEM in patients after breast-conserving surgery.

Figure 1:

Patient inclusion and exclusion flowchart. Calculated diagnostic performance metrics included sensitivity, specificity, positive predictive value, and negative predictive value. CEM = contrast-enhanced mammography.

CEM Technique

CEM examinations were performed using a dual-energy mammography system (Senographe Essential; GE HealthCare). Intravenous iohexol (Omnipaque 350; GE HealthCare) at a dose of 1.5 mL per kilograms of body weight up to a maximum of 150 mL was automatically injected at a rate of 3 mL per second. Imaging was initiated approximately 2.5 minutes after injection and included almost simultaneous LE (26–30 kVp) and high-energy (45–49 kVp) exposures. For each breast, mediolateral oblique and craniocaudal views were obtained. LE images served as the equivalent of two-dimensional full-field digital mammography. Recombined images that highlighted the areas of contrast enhancement were constructed by a postprocessing algorithm. In women with breast implants (n = 20), LE images were obtained for displaced and nondisplaced implant views and recombined images were acquired in the standard craniocaudal and mediolateral oblique planes for each breast, on the implant displacement views (31).

CEM Assessment

CEM results, background parenchymal enhancement (BPE) grade (ie, grade 0, minimal; grade 1, mild; grade 2, moderate; and grade 3, marked), and density categories were determined on the interpretation of the original interpreting breast radiologist as part of routine clinical care. Radiology reports were reviewed for findings suspicious for cancer and the modality (LE imaging, recombined imaging, or both) at which they were depicted, as well as for any additional imaging performed (ie, recall) and for the final BI-RADS category. Per routine clinical care, the original interpreting breast radiologist evaluated findings on LE images (calcifications, masses, asymmetries, and architectural distortion) with additional mammographic views and/or at US before assigning a final BI-RADS assessment. If lesions were observed on the recombined images but had no correlate on mammographic or US images, then MRI was performed for further evaluation of those lesions. If a correlate suspicious for cancer was shown at MRI, then MRI-guided biopsy was performed. If there was no correlate suspicious for cancer at MRI, then a BI-RADS category 3 was assigned and a 6-month follow-up CEM or MRI examination was recommended.

Medical records were reviewed to obtain information regarding age, risk factors, biopsy results, tumor pathologic characteristics, and reactions to iodine-based contrast material.

Validation Study

The original readers had access to both the LE and recombined images simultaneously, and they were initially assigned a unified BI-RADS score. Therefore, to validate the comparison between LE and CEM, a validation study was conducted and structured as follows: A sample dataset consisting of all cancer cases (both detected and missed interval cancers) was compiled along with randomly selected negative cases (with 1 year of follow-up), at a ratio of 1:2, to create a validation cohort. Two additional readers (M.H. and V.L.M., with 13 and 15 years of experience, respectively) who were unaware of the study outcomes independently reviewed the two datasets. They first examined the LE images and then reviewed the LE images along with recombined images, on separate occasions. They were blinded to clinical history, outcomes, and any additional diagnostic assessments. The sensitivity and specificity of the two separate reading sessions were compared for CEM against LE imaging alone.

Contrast Reaction

Patients who developed a reaction to the contrast agent after iodine injection underwent evaluation by a nurse and radiologist. Contrast agent reactions were recorded and graded as mild, moderate, or severe in accordance with the American College of Radiology guidelines outlined in the Manual on Contrast Media (30).

Statistical Analysis

Descriptive statistics were summarized using frequencies and percentages. Only examinations with an adequate reference standard (biopsy) or at least 1 year of radiologic follow-up were included for the calculation of sensitivity, specificity, positive predictive value, and negative predictive value (23). Any additional imaging performed, including dedicated mammographic views, US, and MRI, was considered to be a recall. The recall rate and diagnostic performance metrics were calculated for LE imaging alone and for CEM collectively (ie, both LE and recombined imaging; referred to as CEM). BI-RADS 1 and 2 scores were classified as negative, whereas BI-RADS 3, 4, and 5 were classified as positive (23). A false-negative finding was defined as a cancer diagnosis within 365 days following a negative CEM result.

To compare the sensitivity and specificity of LE and CEM in the entire data set, the McNemar test was used. The positive predictive value of recall (PPV1), positive predictive value of biopsies recommended (PPV2), and positive predictive value of biopsies performed (PPV3) of LE imaging and CEM were calculated as the rate of cancer detection among recalled lesions, biopsy-recommended lesions, and biopsied lesions, respectively. PPV1, PPV2, and PPV3 were compared between LE imaging and CEM using the generalized estimating equation framework, with pathologic examination (reference standard) as the outcome variable and the modality that helped call for further investigation as the explanatory variable. Negative predictive value comparison was performed using the Leisenring test.

To evaluate the influence of BPE on the performance of CEM, BPE grades were divided into two groups: minimal group versus mild, moderate, and marked combined group, which was previously described (25). Sensitivity and specificity were compared using logistic generalized estimation equations model.

CEM performance at baseline and at follow-up rounds was also estimated in a subgroup analysis. The sensitivity and specificity of the baseline and follow-up CEMs were compared using the generalized estimating equation framework, with the modality-labeled outcome as the dependent variable and the timing of the modality as the independent variable. PPV1, PPV2, and PPV3 were also compared between baseline and follow-up CEM using the generalized estimating equation framework, with cancer diagnosis as the outcome variable and with the modality that helped call for further investigation as the independent variable.

All analyses were performed (V.S.) using software (R, version 4.2; R Foundation for Statistical Computing). For the primary analyses comparing LE imaging and CEM, the permissible type I error was adjusted to 0.0083 after multiple comparison corrections were made. For all other statistical comparisons, P < .05 was considered to indicate statistical significance. This study was powered at 80% to detect a difference of 60% in sensitivity to detect breast cancer, given that the eventual total sample size of biopsy-confirmed cancers in this study is 18. This calculation was performed using the biostatUZH package in R, version 4.2.

Results

Patient Characteristics

After the exclusion of patients without an adequate reference (n = 35), the study consisted of 1264 CEM screenings in 609 patients with extremely dense breasts (mean, 2.1 examinations per patient; range, one to eight examinations) (Fig 1), including 476 baseline examinations and 788 follow-up examinations (in addition to the 476 baseline CEM examinations, 133 patients had undergone CEM previously with reported mammographic density of heterogeneously dense breast tissue, which did not meet the inclusion criteria). The mean patient age was 49.8 years ± 9.3 (SD) (age range, 25–79 years). Most patients had an increased lifetime risk for developing breast cancer, including 233 of 609 (38.3%) patients with a personal history of breast cancer, 183 of 609 (30.0%) patients with a family history of breast cancer, and 171 of 609 (28.1%) patients with a history of a high-risk lesion. A minority of patients had two risk factors (46 of 609; 7.6%). The characteristics of the sample are summarized in Table 1.

Table 1:

Patient Characteristics

Screen-detected Cancers

Sixteen screen-detected cancers were diagnosed in 16 patients. Patient, pathologic, and imaging characteristics of the detected cancers are summarized in Table 2. Fourteen of the 16 cancers were invasive, with a median size of 9 mm; 12 of 16 (75%) cancers were hormone receptor positive, and 15 of 16 cancers were node negative. One ductal carcinoma in situ was observed on LE images only and manifested with calcifications. Four cancers manifested as enhancing masses and were detected on both LE and recombined images. Finally, 11 of 16 (69%) cancers were detected only on the basis of enhancement of the recombined images, which increased the number of screen-detected cancers by 220% (from five to 16) (Figs 2–4).

Table 2:

Patient, Tumor, and Imaging Characteristics Among Women With Detected Breast Cancer

Figure 2:

Images in a 56-year-old female patient with a history of atypical ductal hyperplasia show cancer depicted on contrast-enhanced mammograms only; representative case 1. Screening (A, B) mediolateral oblique views and (C, D) craniocaudal views on (A, C) low-energy (LE) images and on (B, D) recombined images in the right breast. At the central slightly outer area in the right breast, mid-depth, there is a new enhancing mass (arrows) with no correlate on LE images. Subsequent targeted US showed a 0.5-cm irregular hypoechoic mass (not shown) for which US-guided biopsy showed invasive carcinoma, with a 1.2-mm invasive component at surgical pathologic examination and with negative nodes.

Figure 4:

Images in a 45-year-old female patient with a personal history of breast cancer (representative case 3) show cancer depicted only on (A–D) contrast-enhanced mammograms with (E, F) subsequent MRI images. Screening (A, B) craniocaudal and (C, D) mediolateral oblique views on (A, C) low-energy (LE) images and (B, D) recombined images, and (E, F) subsequent MRI images, including (E) a diagnostic axial contrast-enhanced first subtraction image and (F) a sagittal contrast-enhanced first subtraction image, served for MRI-guided biopsy in the right breast. In the upper outer breast, mid-depth, there is a new linear enhancement (arrows) on the recombined images with correlates at neither LE imaging nor targeted sonography. Subsequent MRI and MRI-guided biopsy showed ductal carcinoma in situ at surgical pathologic examination and negative nodes.

Figure 3:

Images in a 64-year-old female patient with a family history of breast cancer show cancer depicted on contrast-enhanced mammograms only; representative case 2. Screening (A, B) mediolateral oblique views and (C, D) craniocaudal views on (A, C) low-energy (LE) images and (B, D) recombined images in the left breast. At the lower slightly inner area in the left breast, mid-depth, there is a new enhancing mass (arrows) on the recombined images (B, D) with correlates at neither LE imaging nor targeted sonography. Subsequent MRI and MRI-guided biopsy (not shown) showed invasive carcinoma, with 4-mm invasive component at surgical pathologic examination and negative nodes.

CEM Assessment Results

The final BI-RADS assessments were as follows: Of 1299 examinations, BI-RADS 1 or 2 was assessed in 1137 (87.5%) examinations, BI-RADS 3 in 97 (7.5%) examinations, and BI-RADS 4 or 5 in 65 (5.0%) examinations. Across the entire study sample, an additional imaging evaluation (recall) was performed for 240 of 1299 (18.5%) examinations, most frequently in the form of dedicated mammographic views (ie, spot compression and/or magnification views, n = 58), targeted US (n = 34), or both dedicated mammographic views and US (n = 53). For the remaining 95 of 1299 (7.3%) examinations, following equivocal or negative findings at US and/or on mammographic views, MRI was performed. After MRI, the BI-RADS scores in 17 patients were downgraded to BI-RADS 2. In addition, 50 patients were classified as BI-RADS 3, and these patients were recommended to undergo a 6-month CEM follow-up. Finally, the BI-RADS scores in 28 patients were upgraded to BI-RADS 4, and these patients were recommended to undergo biopsy.

Recall was performed for 78 of 1136 (6.9%) BI-RADS 1 or 2 examinations; of these, findings suspicious for cancer were observed on LE images for 46 examinations, on recombined images for 27 examinations, and on both modalities for five examinations. Recall was performed for all 97 BI-RADS 3 examinations, with the finding suspicious for cancer observed on LE images for 17 examinations, on recombined images for 75 examinations, and on both for five examinations. Recall was also performed for all 65 BI-RADS 4 or 5 examinations, with the finding suspicious for cancer observed on LE images for 23 examinations, on recombined images for 33 examinations, and on both for nine examinations. Lesions where biopsy was recommended were sampled, except for one that was canceled because of nonvisualization; biopsies yielded 13 high-risk and 37 benign lesions, in addition to the 16 screen-detected cancers.

Diagnostic Performance Metrics

In addition to the 16 screen-detected cancers, two women developed interval breast cancer. One female patient presented with a palpable concern 10 months after undergoing CEM with negative results and underwent targeted US, which depicted a 0.7-cm mass. A subsequent US-guided biopsy yielded low-grade ductal carcinoma in situ. The second female patient presented with a palpable concern 6 months after CEM and underwent targeted US, which depicted a 1.8-cm mass; invasive carcinoma was found at a subsequent US-guided biopsy. Retrospective evaluation of the CEM images showed focal enhancement in the area of the interval cancer, which was misinterpreted as a moderate asymmetric BPE. Consequently, the sensitivity of LE imaging was 27.8% (five of 18), which improved to 88.9% (16 of 18) for CEM (P = .003). Overall, the cancer detection rate of LE imaging was 4.0 per 1000 screenings, which increased to 12.7 per 1000 screenings for CEM.

Based on histopathologic and/or 1-year follow-up breast imaging data, which were available for 1264 CEM examinations, there was no evidence of difference between PPV1, PPV2, and PPV3 in the CEM and LE groups. The specificity for CEM was 88.9% (1108 of 1246), which was lower than the specificity of 96.2% (1199 of 1246) for LE imaging (P < .001). The negative predictive value for CEM was 99.8% (1108 of 1110), which was higher than the negative predictive value of 98.9% (1199 of 1212) for LE imaging (P = .001). The diagnostic performance metrics are summarized in Table 3.

Table 3:

Summary of Diagnostic Performance Metrics

Validation Study

Overall, the validation study encompassed 54 cases, including 18 cancers (16 detected and two interval) and 36 examinations with negative findings. Each modality was independently evaluated by two blinded readers, leading to a total of 216 readings. In accordance with the original reading, an additional 11 detected cancers were found at CEM compared with LE imaging alone for both of the blinded readings. The first reader’s sensitivity improved from 33.3% (six of 18) to 94.4% (17 of 18) (P = .003), including the detection of one of the interval cancers. The second reader’s sensitivity improved from 22.2% (four of 18) at LE imaging alone, to 83.3% (15 of 18) using CEM (P = .003). For both readers, no change in specificity was found (Table 4); however, the assessment of specificity in this validation study was constrained by the small number of examinations negative for cancer.

Table 4:

Diagnostic Performance of Two Blinded Readers in the Validation Study

Baseline versus Follow-up CEM Performance Subgroup Analysis

For all but one of the 476 baseline CEM examinations, a previous digital mammographic examination was available for comparison with the LE images. Of the 476 patients with baseline CEM examinations, six had screen-detected cancers and one had interval cancer; however, of the 788 patients with follow-up CEM examinations, 10 had cancers and one had interval cancer. Overall, at follow-up rounds, CEM specificity improved to 90.7% (705 of 777) compared with 85.9% (403 of 469) at baseline (P = .01), whereas there was no evidence of difference between sensitivity, PPV1, PPV2, PPV3, and negative predictive value. The comparison is summarized in Table 5.

Table 5:

Diagnostic Performance of Contrast-enhanced Mammography at Baseline and Follow-up

BPE Effect on CEM Performance

BPE grades were distributed as follows: minimal, 260 of 1264 (20.5%); mild, 508 of 1264 (40.2%); moderate, 346 of 1264 (27.4%); and marked, 150 of 1264 (11.9%). When divided into two BPE groups (minimal vs mild, moderate, and marked combined), the minimal BPE group exhibited increased sensitivity (100%; three of three) and specificity (96.1%; 247 of 257) compared with the combined group of higher BPE scores, which exhibited a sensitivity of 86.7% (13 of 15) (P < .001) and a specificity of 87.1% (861 of 989) (P < .001) (Table 6).

Table 6:

Sensitivity and Specificity of Contrast-enhanced Mammography as a Function of Background Parenchymal Enhancement

Contrast Agent Reactions

Contrast agent reactions occurred in eight of 1316 (0.6%) examinations, and these patients experienced mild to moderate reactions (hives). All patients were discharged after the symptoms resolved within 30 minutes of observation without intervention (n = 6) or after a single dose of diphenhydramine (n = 2).

Discussion

In recent years, several investigators reported on the utility of contrast-enhanced mammography (CEM) in the screening setting, primarily in women with breast tissue that is dense at mammography (23,25). However, only a small fraction of the women in these studies had extremely dense breasts. Thus, the value of screening CEM solely in patients with extremely dense breasts has not been well established in the literature. In our study, the diagnostic performance of screening CEM in patients with extremely dense breasts was compared with that of screening low-energy (LE) imaging. Our results showed that the addition of recombined images enabled the detection of an additional 11 cancers (with an incremental cancer detection rate of 8.7 cancers per 1000 screenings), increasing the sensitivity to 88.9%, compared with a sensitivity of 27.8% achieved by LE imaging (P = .003). On the other hand, CEM showed an increased recall rate and decreased specificity (P < .001). Nonetheless, the specificity of follow-up CEM was greater than that of baseline CEM (P = .01).

The increased sensitivity of CEM compared with LE imaging, also reproduced in the validation study, is in line with the MRI results of the DENSE trial (17). This finding highlights the additive diagnostic value of contrast-enhanced over anatomic imaging among women with extremely dense breasts. Nevertheless, unlike the DENSE trial, our study was not composed of patients with extremely dense tissue at average risk but rather mostly included patients at intermediate risk and few at high risk; thus, the extrapolation of our results to a more general population is limited. In accordance with the DENSE trial, most of the cancers screen-detected with CEM were T-stage 1, node-negative, invasive carcinomas, thus demonstrating the potential for early detection of possibly aggressive breast cancer. Our CEM recall rate of 18.5% was increased compared with the MRI recall rate of 9.5% with MRI as reported in the DENSE trial. However, a recent report (32) demonstrated a lower recall rate for CEM compared with MRI when these modalities were compared head-to-head in women screened with both examinations. Screening CEM compares favorably with US when the two are performed together, with the latter reaching only 8% specificity (33).

Our results, which indicate increased specificity for follow-up CEM compared with baseline CEM, align with the results of the second screening round of the DENSE trial (18). This could be attributed to the availability of previous images at follow-up rounds of imaging, which facilitates the comparison of contrast enhancement and therefore allows for the consideration of enhancement as unchanged, and therefore benign. Furthermore, as clinical experience with CEM increases, it may be possible to avoid additional evaluation and/or biopsies for nonenhancing masses and asymmetries (33). Increased experience with this relatively recently developed technology may also decrease errors when background parenchymal enhancement is differentiated from cancer, which occurred in one interval cancer in our study.

Our results also highlight the effect of BPE grade on the diagnostic performance of CEM, with increased sensitivity and specificity for examinations with minimal BPE compared with higher levels of BPE. BPE, the normal fibroglandular tissue enhancement, may change according to hormonal regulation and is known to have both diagnostic and prognostic implications (33,34). Our findings agree with previous work by Sorin et al (25), who reported decreased specificity of screening CEM examinations with higher than minimal BPE. Our additional finding regarding increased sensitivity for examinations with minimal BPE should be interpretated with caution considering our small sample size and would be better evaluated with a meta-analysis.

The increased risk of breast cancer among women with dense breasts and the diagnostic challenge of screening mammography in this population has prompted legislation mandating that health care providers inform patients about their breast density (35), as well as recommendations for supplemental screening (36). Investigators of several studies have reported on the incremental cancer detection rate of adjunct screening US in patients with dense breasts (13–15), albeit at the cost of substantially increased false-positive results (13). Supplemental US screening in women with dense breasts increases costs considerably while providing relatively small benefits in terms of the number of breast cancer deaths averted and quality-adjusted life-years gained (37).

The use of MRI or abbreviated MRI in addition to tomosynthesis has also been studied in women with dense breasts undergoing screening, demonstrating the superior cancer detection rate of MRI compared with that of tomosynthesis (38). In the multicenter, randomized controlled DENSE trial, MRI screening was offered to women with extremely dense breasts who had negative results at mammography (16–18). Remarkably, in the first screening round, MRI achieved a cancer detection rate of 16.5 cancers per 1000 screenings and a reduced interval cancer rate of 0.8 per 1000 screenings, compared with 5.0 interval cancers per 1000 screenings in the mammography-only group (P < .001) (17). Additionally, whereas the second round of biannual MRI screening achieved a reduced incremental cancer detection rate of 5.8 cancer lesions per 1000 screenings, this change was accompanied by a marked reduction in the number of false-positive results (18).

Despite the success of the DENSE trial, its authors were fully aware of the necessity of finding a more feasible and cost-effective option (18). Abbreviated MRI, which offers shorter acquisition and reading times than does the full protocol (39), could be an option. Another alternative is CEM, which offers noninferior sensitivity and specificity compared with MRI (40). In terms of cost-effectiveness, CEM is an affordable alternative for patients and is less costly for implementation and operation than MRI (41). In terms of availability, existing mammographic equipment can be modified with relative ease to enable CEM, possibly increasing patient access to contrast-enhanced breast imaging at minimal incremental cost (41).

Our study had limitations. First, this was a nonrandomized, single-institution, retrospective study. The relatively high number of patients lost to follow-up (n = 35) may be because the study was retrospective. Second, our study sample included female patients with other risk factors in addition to having extremely dense breast tissue. Third, the breast density category, and therefore this inclusion criterion, was not determined by automatic software, as in the DENSE trial, but was determined subjectively by the interpreting radiologist; thus, it was prone to reader variability. Finally, the original interpreting radiologists read images from each of the two modalities, LE and the recombined imaging, but was not blinded to the other. However, this was addressed in the validation study.

In conclusion, contrast-enhanced mammography (CEM) was found to increase the sensitivity of standard mammography in women with extremely dense breasts, albeit at the cost of additional diagnostic evaluation and biopsies. The increased specificity of follow-up CEM examinations compared with baseline CEM examinations suggests that the diagnostic performance of CEM may be improved even further in the future. Although validation with further large-scale studies is needed, this study shows the potential of CEM to address the unmet diagnostic challenge of breast cancer screening in women with extremely dense breasts with potentially more cost-effectiveness and availability than MRI.