ORIGINAL ARTICLE CLINICAL PRACTICE MANAGEMENT

Comparison of Contrast-Enhanced
Mammography With Conventional Digital
Mammography in Breast Cancer Screening:
A Pilot Study
Geunwon Kim, MD, PhD a, Jordana Phillips, MD a , Elodia Cole, MS b, Alexander Brook, PhD a,
Tejas Mehta, MD a, Priscilla Slanetz, MD, MPH a, Michael D. C. Fishman, MD c,
Evguenia Karimova, MD a, Rashmi Mehta, MD a, Parisa Lotfi, MD a, Nancy Resteghini, DO d,
Sean Raj, MD e, Vandana Dialani, MD a

Credits awarded for this enduring activity are designated “SA-CME” by the American Board of Radiology (ABR) and qualify
toward fulfilling requirements for Maintenance of Certification (MOC) Part II: Lifelong Learning and Self-assessment. To
access the SA-CME activity visit https://cortex.acr.org/Presenters/CaseScript/CaseView?CDId=evy9auosYT8%3d.
SA-CME credit for this article expires August 2022.

Abstract
Purpose: To perform a pilot evaluation of contrast-enhanced mammography (CEM) for screening to determine whether it can improve
accuracy and reader confidence in diagnosis.
Methods and Materials: This institutional review board–approved reader study was comprised of 64 de-identified CEM cases acquired
from December 1, 2014, to June 7, 2016, including 48 negative, 5 biopsy-proven benign, and 11 biopsy-proven malignancies. Negative
cases were followed for at least 2 years without evidence of cancer. Ten breast imagers of varying experience first rated the low-energy
(LE) mammogram and then the CEM examination using BI-RADS categories and a 5-point Likert scale for confidence in diagnosis.
Results: There were 635 out a total possible 640 complete reader interpretations included in this analysis. The remaining five
incomplete interpretations were excluded. Median sensitivity and specificity improved with the addition of CEM (sensitivity: 0.86 [95%
confidence interval {CI}: 0.74-0.95] versus 1 [95% CI: 0.83-1.00], specificity: 0.85 [95% CI: 0.64-0.94] versus 0.88 [95% CI: 0.80-
0.92]). Individual receiver operating characteristic curves showed significant improvement with CEM (mean area under the curve
increase ¼ 0.056 [95% CI: 0.015-0.097], P ¼ .002). The addition of CEM significantly improved average confidence in 5 of 10 readers
when compared with LE (P < .0001) and improved pooled confidence across all tissue density categories, except the almost entirely fatty
category. There was a trend toward improved confidence with increasing tissue density with CEM. Degree of background parenchymal
enhancement did not affect readers’ level of improvement in confidence when interpreting CEM.
Summary: CEM improved reader performance and confidence compared with viewing only LE, suggesting a role for CEM in breast
cancer screening for which larger trials are warranted.
Key Words: Contrast enhanced mammography, breast cancer screening, CEM, CESM
J Am Coll Radiol 2019;16:1456-1463. Copyright
2019 American College of Radiology

a
Beth Israel Deaconess Medical Center, Boston, Massachusetts. Dr Phillips reports other from GE Healthcare and personal fees from
b Hologic, during the conduct of the study, and grants from GE Healthcare,
American College of Radiology Center for Research Innovation, Phila-
delphia, Pennsylvania. USA, outside the submitted work. Dr Fishman reports personal fees from
c
Boston Medical Center, Boston, Massachusetts. Zebra Medical Vision and personal fees from Hologic during the conduct of
d the study. The other authors state that they have no conflict of interest
Atrius Health at Harvard Vanguard Kenmore, Boston, Massachusetts.
e related to the material discussed in this article.
American Radiology Associates, Dallas, Texas.
Corresponding author and reprints: Jordana Phillips, MD, Beth Israel
Deaconess Medical Center, 330 Brookline Ave, Boston, MA 02215; e-mail:
jphilli2@bidmc.harvard.edu.

ª 2019 American College of Radiology

1456 1546-1440/19/$36.00 n https://doi.org/10.1016/j.jacr.2019.04.007
INTRODUCTION
Breast cancer affects one in eight women in an
individual’s lifetime with an annual incidence
over 230,000 [1]. Early detection with screening
mammography has been shown to decrease breast
cancer mortality [2,3]. The existing limitations of
conventional 2-D full-field digital mammography
(FFDM) were improved with the integration of 3-D
digital breast tomosynthesis into clinical practice, lead-
ing to increased sensitivity and specificity and decreased
recall rates from screening [4,5]. However, FFDM and
3-D tomosynthesis both rely on subtle morphologic and
density differences for breast cancer detection, which
can be obscured by overlapping dense glandular tissue.
As a result, women with elevated lifetime risk for breast
cancer as well as with dense breasts may undergo sup-
plemental screening with breast MRI [6–9]. Although
MRI has the highest sensitivity for cancer detection,
MRI is a relatively expensive study that necessitates
approximately 20 to 30 minutes to acquire images
and is not equally accessible to women globally,
even including certain locations in the United States
[3,10–13]. These challenges limit MRI’s universal
applicability as a screening tool, despite the
development of abbreviated protocols to reduce the
overall examination time and cost [14,15].
Contrast-enhanced mammography (CEM) helps
detect cancer by highlighting malignant enhancement
after administration of a contrast agent. CEM has a
similar performance to breast MRI in the diagnostic
setting [16–18]. Unlike MRI, the cost of CEM is similar
to conventional mammography [19], and the time to
image CEM is substantially less than that required for
MRI [20]. Moreover, implementation of CEM is easier
than MRI because it can be implemented with a
software upgrade to some existing digital mammography
equipment rather than requiring purchase of an entirely
new machine [21]. Therefore, based on early studies,
CEM offers a cost-effective alternative to MRI. At pre-
sent, CEM is only approved for diagnostic use, but there
is a potential role for this modality in screening as
well [21–23]. Two studies have begun to evaluate
CEM for screening; both show an overall positive
performance [22,23].
To further explore CEM’s performance for breast
cancer screening and to guide the design of a larger CEM
screening trial, we performed a pilot study comparing
CEM with FFDM to determine differences in diagnostic
performance and reader confidence in diagnosis.
Journal of the American College of Radiology
Clinical Practice Management n Kim et al n Contrast-Enhanced Mammography for Screening
METHODS
Study Design
This was a multireader multicase retrospective pilot study.
Similar to other multireader multicase designs, an
enriched case set was used that included biopsy-proven
malignant cases at a higher proportion than what is pre-
sent in clinical practice to allow for statistical calculations.
It was HIPAA compliant and approved by the insti-
tutional review board at our institution. Informed con-
sent was waived for case inclusion. Reader participation
was voluntary, and reader willingness to participate was
accepted as consent.
Participants
Cases included de-identified CEM images (Senographe
Essential, GE Healthcare, Waukesha, Wisconsin) acquired
from 64 consecutive patients from December 2014 to June
2016 that were obtained for breast cancer screening (as part
of a prospective clinical trial comparing CEM with breast
MRI for breast cancer screening, ClinicalTrials.gov:
NCT02275871, n ¼ 39) or for evaluation of a screen-
detected finding (as part of clinical practice, n ¼ 25).
Although the clinical CEM cases were performed at the
time of diagnostic evaluation, they were performed for a
screen-detected finding before any intervention. As a
result, they still represent a screening population, but the
imaging was performed at a different time in their
screening workup. The images included in the diagnostic
CEM also are the same as those acquired during a screening
CEM (similar to screening and diagnostic breast MRI).
The inclusion criteria for the prospective clinical trial
were patients who were 30 years or older and meeting any
of the following criteria: greater than 20% lifetime risk of
breast cancer, BRCA (BReast CAncer) mutation or other
hereditary germ line mutation, lobular carcinoma in situ
(LCIS), history of chest wall radiation, history of breast
cancer at age 40 or earlier, history of breast cancer in first-
or second-degree relative at age 50 or younger, history of
breast cancer that is mammographically occult, or history
of breast cancer for which a medical oncologist feels
breast MRI screening is important. Exclusion criteria for
the prospective trial includes known allergies to contrast
media, severe allergic response to one or more allergens,
severe asthma, renal insufficiency or failure determined by
a point of care renal function blood test defined as
glomerular filtration rate < 60 mL/min/1.73 m2. Eligible
patients were identified at the time they scheduled
MRI examinations. After December 2015, CEM was
1457
implemented as a routine clinical portion of the diag-
nostic examination at our institution. Only asymptomatic
patients who were called back from routine screening
were included. Exclusion criteria were the same as the
prospective trial. All included cases preceded biopsy.
Case Collection
All CEM images were acquired after intravenous admin-
istration of 1.5 mL/kg of Omnipaque 350 (Iohexol, GE
Healthcare) at 3 mL/s. Two minutes after injection, low-
energy (LE) (26-32 kVp) and high-energy (45-49 kVp)
images of both breasts were acquired in craniocaudal (CC)
and mediolateral oblique (MLO) projections. Recom-
bined images (RIs), highlighting contrast enhancement,
were automatically generated by the CEM software.
All CEM cases in the reader study included bilateral LE
and RIs in CC and MLO projections.
Any additional diagnostic imaging that may have
been performed was not included at time of reader
interpretation.
Case Categorization and Determination of
Reference Standard
The enriched case mix for the reader study included 48
negative and 16 positive examinations (total of 64 cases).
All 16 positive cases underwent biopsy, yielding 11 cancers,
resulting in a 17% incidence of cancer status as determined
by histopathology for all biopsied imaging-detected lesions.
For cases with no imaging-detected lesions, at least 2 years
of imaging or clinical follow-up was used to classify the case
as negative. The rationale for building the case set with
cancer incidence greater than in the general population
was to allow for statistical analysis because low incidence
of de novo cancer limits detection of change before and
after addition of recombined images. We simulated the
screening pool using a cancer incidence rate that is between
the stratified random sampling with proportional and
disproportional allocation as described by Zur et al [24].
Information regarding tissue density and background
parenchymal enhancement (BPE) was collected for each case
from the initial CEM imaging report. Tissue density was
recorded as almost entirely fatty, scattered fibroglandular
densities, heterogeneously dense, or extremely dense. BPE
was recorded as minimal, mild, moderate, and marked.
Image Interpretation
Ten radiologists participated in the reader study. Two
were breast imaging fellows and eight were Mammog-
raphy Quality Standards Act (MQSA)-certified breast
1458
imaging attending radiologists with 1 to 20 years of
experience (Table 1). Each radiologist reviewed the same
64 studies in a randomized sequence on a clinical PACS
workstation. No standardized training was provided to
the reader before the reader study. Readers were
blinded to patient history.
For each case, the participating radiologists inter-
preted the LE images first. This interpretation was used as
a surrogate for FFDM [25,26]. Readers were asked
whether they would call back the patient to simulate
screening. If a reader chose to recall a case, they were
asked to provide at most two positive findings per breast
with a forced BI-RADS category per finding. The high-
est BI-RADS category was used as the LE BI-RADS
category for the case. Readers also were asked to provide
a confidence level using a 5-point Likert scale for each
positive finding, or per case score if negative study.
Readers then interpreted the LE in conjunction with
the RI, which was used as the CEM interpretation.
Readers provided new BI-RADS categories and confi-
dence scores for lesions previously identified on the LE. If
LE was negative, the reader was asked whether an inci-
dental finding was present on the RI, for which BI-RADS
categories and confidence scores were given. The most
concerning BI-RADS category was used as the CEM BI-
RADS category for the case.
The BI-RADS categories of 1 through 5 were used for
interpretation per the BI-RADS lexicon, fifth edition
[27]. The 5-point Likert scale used the following cate-
gories for interpretation: very confident (5), confident (4),
neutral (3), not confident (2), and not at all confident (1).
At the beginning of the study, readers were surveyed
for years of experience in breast imaging starting with
Table 1. Readers’ overall years of experience in breast
imaging and number of CEM cases that were clinically
interpreted before participating in the study
Experience in breast Experience in CEM
Reader imaging (years) (cases)
1 0 10-24
2 0 0-9
3 1 10-24
4 2 0-9
5 3 10-24
6 5 0-9
7 8 10-24
8 12 10-24
9 19 24-100
10 20 24-100
CEM ¼ contrast-enhanced mammography.
Journal of the American College of Radiology
Volume 16 n Number 10 n October 2019
fellowship, and the approximate number of prior CEM
cases interpreted.
Statistical Analysis
After excluding five responses due to incomplete entries,
635 responses were included from 10 independent
readers who each interpreted the 64 cases. Data analysis
was performed using Matlab (MathWorks, Natick,
Massachusetts.
For each case, readers’ BI-RADS categories of 1, 2,
and 3 were classified as negative and BI-RADS 4(A, B, C)
and 5 were classified as positive. Readers’ positive answers
were recorded as true-positive if a reader marked a case
with known malignant pathology as BI-RADS 4A or
higher and false-negative if a reader marked the case as
BI-RADS 3 or lower. For cases with biopsy-proven
benign pathology or no abnormality, the interpretation
was recorded as true-negative if the reader marked the
case as BI-RADS 3 or lower and as false-positive if the
reader marked the case as BI-RADS 4A or higher.
Almost entirely fatty and scattered fibroglandular
tissue densities were classified as nondense, whereas het-
erogeneously dense and extremely dense were classified as
dense. Minimal and mild BPE combined were classified
as “low” enhancement, and moderate and marked BPE
were classified as “high” enhancement.
Receiver operating characteristic curves for each
reader comparing LE and LE þ RI were generated.
Interobserver variability was calculated using Fleiss’ k.
Individual sensitivity and specificity was compared using
McNemar test. Median pooled sensitivity and specificity,
areas under the curve (AUCs), and individual confidence
levels were compared using Wilcoxon test. Analysis of
variance was used to assess change in confidence, overall
change, and change by density level. Overall comparisons
for confidence, sensitivity, and specificity versus density
and BPE levels were performed using paired t test. As-
sociation of density and BPE was evaluated using c2 test.
RESULTS
Patient Characteristics
All 64 patients were women, with an average age of 52.2
years (range 31-71 years). Thirty-five women (55%) were
postmenopausal. One patient experienced a rash on her
face 4 days after contrast administration; however, she
also had permanent eye makeup applied 1 day before the
CEM, and it was inconclusive whether the reaction was
related to contrast administration. The patient’s symp-
toms resolved with a course of oral steroids.
Journal of the American College of Radiology
Clinical Practice Management n Kim et al n Contrast-Enhanced Mammography for Screening
Table 2. Histopathologic findings of 16 patients who
underwent biopsy
Findings Number of lesions
Malignant (n ¼ 11)
IDC 7
ILC 1
DCIS 3
Benign (n ¼ 5)
Complex sclerosing lesion 1
PASH 1
Benign fibrous tissue 1
Apocrine metaplasia 2
Three patients underwent two different biopsies. Only the highest-
grade lesion was counted toward the analysis. DCIS ¼ ductal
carcinomas in situ; IDC ¼ invasive ductal carcinomas; ILC ¼ invasive
lobular carcinoma; PASH ¼ Pseudoangiomatous stromal hyperplasia.
Sixteen of the 64 patients (25%) with at least BI-
RADS 4A underwent biopsy or surgical excision of the
lesion identified on CEM (Table 2), including 11
patients (17%) with malignant pathology (Table 2).
Thirteen of those 16 patients (81.3%) had a single site
biopsy, yielding 5 (31.3%) invasive ductal carcinomas
(IDCs), 3 (18.8%) ductal carcinomas in situ (DCISs),
and 5 (31.3%) with benign pathology (Table 2). All 5
patients (100%) with benign biopsy had at least 2 years
of clinical or imaging follow-up with no interval devel-
opment of cancer. Three of sixteen patients (18.8%)
underwent two different site biopsies on the same day;
one patient had a single site biopsy in each breast, which
yielded invasive lobular carcinoma in one and fibroade-
nomatous change in the other. Two patients had biopsies
of two different sites on the same breast, which yielded
IDC and DCIS in one patient, and IDC and invasive
lobular carcinoma in the other patient. One of these
patients had a lesion detected only on RI, which subse-
quently led to a targeted ultrasound and biopsy yielding
IDC.
All of the 48 patients (100%) with BI-RADS 3 or
lower on the initial evaluation had at least 2 years of
follow-up without interval development of cancer.
In the cohort of 64 patients, 41 (64%) had dense
breast tissue (31 had heterogeneously dense tissue and 10
had extremely dense tissue); 23 (36%) had nondense
tissue (21 scattered fibroglandular densities and 2 almost
entirely fatty).
Forty-nine of the sixty-four patients (77%) demon-
strated low-level background enhancement (minimal or
mild) with the following overall breakdown: minimal
47% (n ¼ 30), mild 30% (n ¼ 19), moderate 19%
(n ¼ 12), and marked 5% (n ¼ 3).
1459
 There was no association between background paren-
chymal enhancement (BPE) and tissue density (P ¼ .37).
In evaluating the malignant cases at the per-finding
level, two cases highlight CEM’s ability to identify ma-
lignancy in the setting of multiple otherwise indetermi-
nate imaging findings. In one case, the patient had
multiple benign biopsies but could not tolerate MRI due
to claustrophobia. CEM detected an enhancing region
near a previously biopsied complex sclerosing lesion,
which was subsequently identified and biopsied with ul-
trasound guidance, demonstrating high-grade DCIS. In
another patient, calcifications were identified in one
quadrant yielding DCIS, but an incidental area of
enhancement in separate quadrant only seen on RI
resulted in multicentric disease after IDC diagnosis on
ultrasound core biopsy.
Reader Accuracy
Addition of RI to LE images increased sensitivity in 6 of
10 readers, though not reaching statistical significance.
Similarly, specificity increased upon addition of RI in 7 of
10 readers with two readers reaching statistical signifi-
cance (Table 3). Individual receiver operating
characteristic curves and AUCs demonstrated
statistically significant improvement in performance
with CEM compared with LE alone (mean AUC
increase ¼ 0.056; 95% confidence interval [CI]: 0.015-
0.097; P ¼ .002) (Fig. 1). Interobserver variability in
identifying cases as positive was k ¼ 0.39  0.02
(95% CI: 0.38-0.40) for LE alone and k ¼ 0.62 
0.03 (95% CI: 0.59-0.61) for CEM. Intraclass
correlation for BI-RADS categories was r ¼ 0.46
(95% CI: 0.35-0.58) for LE alone and r ¼ 0.62 (95%
CI: 0.53-0.71) for CEM.
There was no statistically significant improvement in
sensitivity between cases with low (combined category of
minimal and mild) versus high (moderate and marked)
BPE (P ¼ .27).
There was no significant influence of tissue density
(minimal or mild versus moderate or marked) on speci-
ficity (P ¼ .6).
Reader Confidence
CEM interpretation demonstrated significant difference
in the confidence level for some readers when compared
with using only LE images (Table 4). The addition of RI
significantly improved average confidence in 5 of 10
readers when compared with LE alone (Table 4). The
remaining readers showed no significant change in
average confidence, though there was overall significant
improved average confidence (Table 4). Although the
change in confidence did not correlate with overall
years of training or with the amount of experience with
CEM cases, four of the six readers with less than or
equal to 5 years of experience demonstrated
significantly improved confidence with CEM.
Furthermore, change in pooled reader confidence did
not differ between low (minimal or mild) versus high
(moderate or marked) BPE (P ¼ .9).
There was an improvement in the pooled average
reader confidence level with CEM compared with LE
alone for all density categories, reaching statistical sig-
nificance in all except the almost entirely fatty category;
however, this group only had two cases (Table 5).

Table 3. Pooled sensitivity and specificity of each reader

Sensitivity Specificity
(n ¼ 11) (n ¼ 53)
Reader LE LE þ CEM P value LE LE þ CEM P Value
1 1 0.91 .99 0.34 0.72 .0002
2 0.82 1 .5 0.83 0.87 .75
3 1 1 - 0.85 0.87 .99
4 1 1 - 0.49 0.74 .007
5 0.91 1 .99 0.85 0.81 .63
6 0.64 0.64 - 0.85 0.89 .5
7 0.73 0.82 .99 0.92 0.92 .99
8 0.82 1 .5 0.92 0.96 .5
9 0.91 1 .99 0.91 0.91 -
10 0.64 0.82 .5 0.94 0.91 .63
Average 0.85  0.14 0.92  0.12 0.79  0.20 0.86  0.08
Median 0.86 1 .08 0.85 0.88 .2
P value shown for comparison between LE and LE þ CEM diagnostic performance. CEM ¼ contrast-enhanced mammography; LE ¼ low energy.

1460 Journal of the American College of Radiology

Volume 16 n Number 10 n October 2019
Fig 1. Receiver operating characteristic curves for each reader for low-energy and contrast-enhanced mammography.
CESM ¼ contrast-enhanced spectral mammography; MG ¼ low energy mammography.

Although there was no significant difference in mean extremely dense breast tissue. This suggests that CEM
improvement across tissue densities, there was a trend may have a role in cancer detection in the screening
toward improved confidence with increasing tissue setting for women with dense breast tissue.
density. There was no significant difference in the Other studies have reported increased confidence
improvement of confidence for low versus high BPE with CEM interpretation for lesion detection in diag-
(P ¼ .9). nostic settings [29]. This has clinical implications because
a high level of confidence in interpretation of screening
DISCUSSION mammography was associated with the detection of a
CEM has been shown to improve reader performance in true lesion [30] and shorter viewing times [31].
diagnostic settings and its use has been approved for Furthermore, the improved confidence with CEM in
diagnostic evaluation by the FDA [16,28]. There is less experienced breast imagers in our study is a notable
growing interest in offering CEM for breast cancer trend because it could potentially reduce false-positive
screening, particularly in women with dense breast rates without sacrificing sensitivity.
tissue or those at high risk for breast cancer, as an Although not statistically significant at a group level,
alternative to breast MRI. However, prospective our study did demonstrate increased sensitivity with
evaluation of CEM in the screening setting requires a
large number of patients, due to a relatively low breast Table 4. Confidence level of each reader (mean  SD)
cancer incidence in the screening population. This All Breast Density (n ¼ 64)
challenge was highlighted in a recent screening trial Reader LE LE þ CEM P Value
comparing CEM with breast MRI with only eight 1 3.6  0.8 4.0  0.7 .0008
cancers detected in a cohort of 307 patients [23] and 2 3.2  0.8 3.9  1.0 <.0001
another recent prospective study with only 21 cancers 3 3.8  0.6 3.8  0.6 .99
detected [22]. As a result, a retrospective reader study is 4 3.3  0.7 4.4  0.7 <.0001
a useful approach for studying screening CEM, because 5 3.2  0.9 3.3  0.9 .18
a large number of cases with breast cancer can be 6 3.8  0.8 4.1  0.6 <.0001
compared across modalities. 7 3.8  0.6 3.9  0.7 .83
Our study is the first to systematically show that the 8 3.5  1.0 3.7  1.0 0.99
9 3.3  1.1 4.4  0.7 <.0001
addition of CEM can lead to improved confidence in the
10 4.1  0.4 4.2  0.5 .016
interpreting radiologist compared with LE (surrogate for
Average 3.6  0.3 4.0  0.3 <.0001
FFDM) in a screening setting. Moreover, there was a
Confidence ranges 1-5. P values shown comparing confidence inter-
trend toward increasing confidence with increasing tissue preting LE versus LE þ CEM images. CEM ¼ contrast-enhanced
density, with the largest improvement in women with mammography; LE ¼ low energy.

Journal of the American College of Radiology 1461

Clinical Practice Management n Kim et al n Contrast-Enhanced Mammography for Screening
Table 5. Pooled average reader confidence by density exploratory analysis, we noted that CEM demonstrated
category (mean  SD) borderline significant improvement in specificity when
LE þ Mean P comparing cases with minimal BPE against increased
Density n LE CEM Increase Value BPE. This information will help determine which pop-
Almost entirely fat 2 3.9  4.0  0.2 .34 ulation will benefit the most from the added information
0.7 0.8 of contrast enhancement during screening.
Scattered 21 3.7  4.1  0.39 <.0001 In conclusion, our pilot reader study suggests CEM’s
fibroglandular 0.8 0.8 potential to improve cancer detection while minimizing
Heterogenous 31 3.5  3.9  0.37 <.0001
false-positives in breast cancer screening and potentially
fibroglandular 0.8 0.8
improving radiologist confidence in interpretation. Given
Extreme 10 3.3  3.9  0.58 <.0001
fibroglandular 0.8 0.8
these promising results, further investigation of CEM for
screening population is warranted, especially in women
Confidence range 1-5. We also show mean increase in confidence with
the introduction of CEM, and P value for the significance of this with dense breast tissue.
increase. CEM ¼ contrast-enhanced mammography; LE ¼ low
energy.
TAKE-HOME POINTS
CEM in comparison with FFDM. Moreover, there was a - Using CEM in screening setting improves sensi-
significant (mean AUC increase of 5%) improvement in tivity, specificity, and confidence level, especially in
performance of CEM relative to FFDM with improved less experienced breast imagers.
interobserver agreement for CEM interpretation. This is - CEM may have its greatest impact on cancer
similar to a recent screening study reporting improved detection in the screening setting in women with
overall performance of CEM relative to FFDM [22]. dense breast tissue.
Importantly, we found that BPE, which can pose - BPE, which can pose substantial challenges on MR,
substantial challenges on MRI [32,33], did not did not significantly affect the sensitivity or reader
significantly affect the sensitivity or reader confidence, confidence in interpreting CEM and showed
with minimal decrease in specificity. This result, borderline decrease in specificity
combined with the overall strong performance for
specificity, supports the use of CEM for breast cancer
REFERENCES
screening. 1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2018. CA Cancer J
There are several limitations to our study. First, our Clin 2018;68:7-30.
case mix included a higher proportion of malignancies 2. Berry DA, Cronin KA, Plevritis SK, et al. Effect of screening and
adjuvant therapy on mortality from breast cancer. N Engl J Med
(17%) than found in a pure screening population. The 2005;353:1784-92.
case mixture included cases going to biopsy to increase 3. Coccia M. The effect of country wealth on incidence of breast cancer.
statistical power in sensitivity and specificity analysis with Breast Cancer Res Treat 2013;141:225-9.
4. Phi X-A, Tagliafico A, Houssami N, Greuter MJW, de Bock GH.
overall small number of cases. Digital breast tomosynthesis for breast cancer screening and diagnosis
As with all reader studies, readers were aware of study in women with dense breasts—a systematic review and meta-analysis.
BMC Cancer 2018;18:380.
participation, which may have impacted decision making.
5. McCarthy AM, Kontos D, Synnestvedt M, et al. Screening outcomes
This may explain the relatively high sensitivity and low following implementation of digital breast tomosynthesis in a general-
specificity on interpreting LE images, especially for less population screening program. J Natl Cancer Inst 2014 Oct 13;106(11).
6. Expert Panel on Breast Imaging: Mainiero MB, Moy L, Baron P, et al.
experienced readers. However, reduction in false- ACR Appropriateness Criteria Breast Cancer Screening. J Am Coll
positivity with CEM in these readers suggests the po- Radiol 2017;14(11 Suppl):S383-90.
tential of CEM in clinical practice, which can improve 7. Mainiero MB, Moy L, Baron P, et al. American College of Radiology
ACR Appropriateness Criteria Breast Cancer Screening. American
the specificity and result in fewer false-positives, as seen in College of Radiology. Available at: https://acsearch.acr.org/docs/7091
previous studies [23]. 0/Narrative/. Accessed December 16, 2018.
Finally, the case number was small, and many more 8. Wernli KJ, DeMartini WB, Ichikawa L, et al. Patterns of breast
magnetic resonance imaging use in community practice. JAMA Intern
cases are necessary to perform a robust comparison of Med 2014;174:125-32.
CEM versus FFDM or MRI in the screening setting. 9. Stout NK, Nekhlyudov L, Li L, et al. Rapid increase in breast magnetic
This is especially true for evaluating the diagnostic ac- resonance imaging use: trends from 2000 to 2011. JAMA Intern Med
2014;174:114-21.
curacy of CEM as well as how breast tissue density, BPE, 10. Onega T, Hubbard R, Hill D, et al. Geographic access to breast im-
and risk status impact CEM performance. In our aging for US women. J Am Coll Radiol 2014;11:874-82.

1462 Journal of the American College of Radiology

Volume 16 n Number 10 n October 2019
 11. Onega T, Lee CI, Benkeser D, et al. Travel burden to breast MRI and
utilization: are risk and sociodemographics related? J Am Coll Radiol
2016;13:611-9.
12. Ahern CH, Shih Y-CT, Dong W, Parmigiani G, Shen Y. Cost-
effectiveness of alternative strategies for integrating MRI into breast
cancer screening for women at high risk. Br J Cancer 2014;111:
1542-51.
13. Moore SG, Shenoy PJ, Fanucchi L, Tumeh JW, Flowers CR. Cost-
effectiveness of MRI compared to mammography for breast cancer
screening in a high risk population. BMC Health Serv Res 2009;9:9.
14. Deike-Hofmann K, Koenig F, Paech D, et al. Abbreviated MRI
protocols in breast cancer diagnostics. J Magn Reson Imaging 2019;49:
647-58.
15. Greenwood HI. Abbreviated protocol breast MRI: the past, present,
and future. Clin Imaging 2018;53:169-73.
16. Fallenberg EM, Dromain C, Diekmann F, et al. Contrast-enhanced
spectral mammography versus MRI: Initial results in the detection of
breast cancer and assessment of tumour size. Eur Radiol 2014;24:
256-64.
17. Fallenberg EM, Schmitzberger FF, Amer H, et al. Contrast-
enhanced spectral mammography vs. mammography and MRI—
clinical performance in a multi-reader evaluation. Eur Radiol
2017;27:2752-64.
18. Li L, Roth R, Germaine P, Ren S, Lee M, Hunter K, et al. Contrast-
enhanced spectral mammography (CESM) versus breast magnetic
resonance imaging (MRI): a retrospective comparison in 66 breast
lesions. Diagn Interv Imaging 2017;98:113-23.
19. Patel BK, Gray RJ, Pockaj BA. Potential cost savings of contrast-
enhanced digital mammography. AJR Am J Roentgenol 2017;208:
W231-7.
20. Phillips J, Miller MM, Mehta TS, et al. Contrast-enhanced spectral
mammography (CESM) versus MRI in the high-risk screening setting:
patient preferences and attitudes. Clin Imaging 2017;42:193-7.
21. Covington MF, Pizzitola VJ, Lorans R, et al. The future of contrast-
enhanced mammography. AJR Am J Roentgenol 2018;210:292-300.
22. Sorin V, Yagil Y, Yosepovich A, et al. Contrast-enhanced spectral
mammography in women with intermediate breast cancer risk and
dense breasts. AJR Am J Roentgenol 2018;211:W267-74.
Journal of the American College of Radiology
Clinical Practice Management n Kim et al n Contrast-Enhanced Mammography for Screening
23. Jochelson MS, Pinker K, Dershaw DD, et al. Comparison of screening
CEDM and MRI for women at increased risk for breast cancer: a pilot
study. Eur J Radiol 2017;97:37-43.
24. Zur RM, Pesce LL, Jiang Y. Estimating screening-mammography
receiver operating characteristic (ROC) curves from stratified random
samples of screening mammograms: a simulation study. Acad Radiol
2015;22:580-90.
25. Francescone MA, Jochelson MS, Dershaw DD, et al. Low energy
mammogram obtained in contrast-enhanced digital mammography
(CEDM) is comparable to routine full-field digital mammography
(FFDM). Eur J Radiol 2014;83:1350-5.
26. Lalji UC, Jeukens CRLPN, Houben I, et al. Evaluation of low-energy
contrast-enhanced spectral mammography images by comparing them
to full-field digital mammography using EUREF image quality criteria.
Eur Radiol 2015;25:2813-20.
27. Sickles E, D’Orsi CJ, Bassett LW, et al. ACR BI-RADS
Mammography. In: ACR BI-RADS Atlas, Breast Imaging Reporting
and Data System. Reston, VA: American College of Radiology; 2013.
28. FDA.gov. Contrast enhanced digital mammography 510(k) premarket
notification. Available at: https://www.accessdata.fda.gov/cdrh_docs/
pdf12/K123873.pdf. Accessed January 2, 2019.
29. Badr S, Laurent N, Régis C, Boulanger L, Lemaille S, Poncelet E.
Dual-energy contrast-enhanced digital mammography in routine
clinical practice in 2013. Diagn Interv Imaging 2014;95:245-58.
30. Nodine CF, Mello-Thoms C, Kundel HL, Weinstein SP. Time course
of perception and decision making during mammographic interpre-
tation. AJR Am J Roentgenol 2002;179:917-23.
31. Carney PA, Bogart TA, Geller BM, et al. Association between time
spent interpreting, level of confidence, and accuracy of screening
mammography. AJR Am J Roentgenol 2012;198:970-8.
32. Ray KM, Kerlikowske K, Lobach IV, et al. Effect of background
parenchymal enhancement on breast MR imaging interpretive
performance in community-based practices. Radiology 2018;286:
822-9.
33. DeMartini WB, Liu F, Peacock S, Eby PR, Gutierrez RL,
Lehman CD. Background parenchymal enhancement on breast MRI:
impact on diagnostic performance. AJR Am J Roentgenol 2012;198:
W373-80.
1463