Original Study

Diagnostic Value of Contrast-Enhanced Spectral

Mammography for Screening Breast Cancer:
Systematic Review and Meta-analysis
Xiao Zhu,1 Jun-ming Huang,2 Kun Zhang,3 Long-jie Xia,1 Lan Feng,4 Ping Yang,2
Meng-ya Zhang,5 Wei Xiao,5 Hui-xia Lin,5 Ying-hua Yu1
Abstract
A new meta-analysis review is needed to evaluate the diagnostic value of contrast-enhanced spectral
mammography (CESM). Eighteen studies were included in the review. The important parameters of CESM
accuracy for breast cancer diagnosis were analyzed. CESM can be considered as a useful triage test for initial
assessment of breast lesions.
Background: Contrast-enhanced spectral mammography (CESM) is a new image examination technology that has
developed over the past few years. As CESM technology keeps improving, a current meta-analysis review is needed
to systematically evaluate the potential diagnostic value of CESM. Methods: A total of 18 studies were included in the
review. Sensitivity, specificity, and other important parameters of CESM accuracy for breast cancer diagnosis were
pooled and analyzed using random-effects models. Summary receiver operating characteristic curves were calculated
for overall accuracy estimation. Results: The summary estimates for CESM in the diagnosis of breast cancer were as
follows: the pooled sensitivity and specificity were 0.89 (95% confidence interval [CI], 0.88-0.91) and 0.84 (95% CI,
0.82-0.85), respectively. Positive likelihood ratio was 3.73 (95% CI, 2.68-5.20), negative likelihood ratio was 0.10
(95% CI, 0.06-0.15), and diagnostic odds ratio was 71.36 (95% CI, 36.28-140.39). The area under the curve was 0.96
(standard error ¼ 0.011). Conclusion: CESM has a high diagnostic accuracy for evaluating breast cancer and can be
considered as a useful test for initial assessment of breast lesions.

Clinical Breast Cancer, Vol. -, No. -, 1-10 ª 2018 Elsevier Inc. All rights reserved.
Keywords: Accuracy, Breast cancer, Contrast-enhanced spectral mammography, Meta-analysis

Introduction ought to be provide the most diagnostic information possible.

Breast cancer is one of the most common causes of cancer death Developing a diagnostic screening method with high accuracy is
among women.1,2 A successful treatment for breast cancer partly therefore been a major goal.
depends on early detection, which ought to be noninvasive and Considering various breast cancer imaging diagnosis methods,
contrast-enhanced breast magnetic resonance imaging is recog-
X.Z., J.-M.H. and K.Z. contributed equally to this article, and all should be considered
first author. nized as one of the most accurate imaging methods. However,
1
because of the relatively long examination time, high cost, and
Department of Breast Surgery of Affiliated Tumor Hospital of Guangxi Medical
University, Guangxi, PR China variable specificity, its use for early screening of breast cancer is
2
Department of Oncology, Panyu Hospital of Chinese Medicine, Guangdong, PR limited.3 Mammography, a breast imaging method, has been
China
3
Deparment of Breast Surgery, The Affiliated Yantai Yuhuangding Hospital of Qing- proven to be a valuable tool for screening breast cancer and its use
dao University, Shandong, PR China
4
is reported to reduce breast cancer mortality.4 Although
National Center for Protein Sciences, Beijing, PR China
5
Graduate School of Guangxi Medical University, Guangxi, PR China mammography has a relatively high sensitivity (75%-85%) for
breast cancer,5 the sensitivity can fall to < 50% when it involves
Submitted: Jan 5, 2018; Revised: May 9, 2018; Accepted: Jun 8, 2018
dense breast parenchyma.6
In order to improve the diagnostic capability of conventional
Address for correspondence: Ying-hua Yu, MD, PhD, Department of Breast Surgery of
Affiliated Tumor Hospital of Guangxi Medical University, No. 71, He Di Lu, Nan- mammography, a new imaging modality, contrast-enhanced spectral
ning, Guangxi 530021, PR China mammography (CESM), has been developed. In combination with
Fax: 86-0771-5312000; e-mail contact: [email protected]
iodine-based contrast agents, it enables accurate detection of

1526-8209/$ - see frontmatter ª 2018 Elsevier Inc. All rights reserved.

https://doi.org/10.1016/j.clbc.2018.06.003 Clinical Breast Cancer Month 2018 -1
 Diagnostic Value of CESM
malignant breast lesions, specifically spotting areas of abnormal the China Biological Medicine Database (CBM-disc), CNKI
vascularization in the breast. In detecting dense breast, CESM is (China National Knowledge Infrastructure Whole Article Data-
better than conventional mammography7 as a result of its sensitivity base), and the VIP China Science and Technology Journal Data-
and specificity of CESM.8,9 Moreover, CESM has almost the same base. We also searched the trial registries of the Cochrane Breast
sensitivity as breast magnetic resonance imaging, but its shorter Cancer Group and the World Health Organization International
examination time makes it a potential alternative to this modality.10 Clinical Trials Registry (http://www.who.int/ictrp/en/) for ongoing
Although a previous meta-analysis accessing diagnostic perfor- and recently completed trials. Abstracts from recent conferences and
mance of CESM for breast cancer has been published, it included the reference lists of articles were checked to screen for additional
only 8 studies.11 Two of these included only breast cancer patients, studies. The 2 reviewers (Y.H.Y. and X.Z.) independently per-
leading to inappropriate exclusions that might result in over- formed the searches up to November 2017.
optimistic estimates of diagnostic accuracy. Furthermore, in the last Our search strategy terms included “breast neoplasms,” “contrast
few years, the diagnostic value of CESM has improved, and it is enhanced spectral mammography” “mammography,” and “contrast
rapidly emerging as a technique for breast cancer diagnosis. enhancement” in combination with “sensitivity,” “specificity,”
Increasingly, new studies are being published about CESM “accuracy,” and their synonyms. Subsequent analyses included
demonstrating its diagnostic value for screening breast cancer. studies with the following criteria: articles written in English or
We therefore conducted a meta-analysis using data from multiple Chinese for the full-text review; studies including at least 10 cases or
studies, including the recently published literature, to systematically patients who underwent CESM; and studies from which adequate
evaluate the diagnostic value of CESM in the diagnosis of breast and reliable estimate of the diagnostic parameters like sensitivity,
cancer. specificity, or other key diagnostic factors could be derived. Letters,
case series, case reports, comments, review articles, and conference
Methods abstracts were excluded because of their limited data sets.
Our study followed the PRISMA guidelines,12 which was also
required in the diagnostic systematic review. The PICOS criteria for Data Extraction and Quality Assessment
inclusion and exclusion of studies were as follows: P, patients who Two authors, X.Z. and K.Z., independently extracted the data
underwent CESM; I, intervention was CESM; C, histopathology or from each study using 2  2 tables. Data retrieved from the reports
clinical follow-up results were comparison tests; O, important were as follows: study design, publication year, participant charac-
accuracy parameters of CESM presented the outcomes; and S, teristics, demographic characteristics, Breast Imaging Reporting
studies about diagnostic accuracy. and Data System (BI-RADS), outcome measures, numbers of true-
positive and false-negative results, or true-negative and false-positive
Search Strategy and Study Selection results (Tables 1 and 2).
To comprehensively assess the accuracy of CESM, we searched Clinical trials are usually assessed by the following criteria:
for relevant studies in PubMed (a free resource providing access to randomization, selection bias of the arms in the study, concealment
many medical databases, such as Medline), the Cochrane library, of allocation, and blinding of outcome.13,14 However, studies

Table 1 Summary of 18 Included Studies

Study Year Geographic Location TP FP FN TN No. of Lesions No. of Patients

Houben8 2017 Netherlands 38 0 32 774 844 879
Li10 2017 USA 62 0 2 2 66 48
Patel30 2017 USA 29 8 1 11 49 45
Mori7 2017 Japan 50 5 8 80 143 72
Richter31 2017 Germany 83 4 1 29 117 105
Cheung32 2016 Taiwan 24 3 3 32 62 62
Luczynska33 2016 Poland 143 0 49 33 225 193
Cheung34 2016 Taiwan 20 6 2 31 59 52
Lalji35 2016 Netherlands 57 42 2 98 199 199
Tennant36 2016 UK 69 5 4 22 100 99
Kamal37 2016 Egypt 160 23 35 43 261 239
Wang38 2016 China 46 10 2 19 77 68
Luczynska39 2015 Poland 114 47 0 30 191 174
Luczynska40 2015 Poland 81 25 0 12 118 102
Kamal41 2015 Egypt 103 42 6 60 211 168
Luczynska42 2014 Poland 114 35 0 24 173 152
Cheung43 2014 Taiwan 67 9 5 19 100 89
Lobbes44 2014 Netherlands 32 10 0 71 113 113

Abbreviations: FN ¼ false negative; FP ¼ false positive; TN ¼ true negative; TP ¼ true positive.

2 - Clinical Breast Cancer Month 2018

 Table 2 Characteristics of 18 Included Studies

Location Study Design Patients Patients Lesions Unaware of Final Reference Standard
Study Year (Asia) (Prospective) (Consecutive) (Age > 40 Years) (Included BI-RADS 1-3) Diagnosis (Histopathology Only)
Houben8 2017 No No No Yes Yes Yes No
Li10 2017 No No No No Yes No Yes
Patel30 2017 No No No No No No Yes
Mori7 2017 Yes No No No Yes Yes No
Richter31 2017 No No No No Yes No No
Cheung32 2016 Yes No Yes Yes Yes No Yes
Luczynska33 2016 No No No Yes Yes No Yes
Cheung34 2016 Yes No No No No No Yes
Lalji35 2016 No No Yes Yes Yes Yes No
Tennant36 2016 No No Yes No Yes No Yes
Kamal37 2016 No No Yes No Yes No No
Wang38 2016 Yes Yes Yes No Yes Yes Yes
Luczynska39 2015 No No No Yes Yes Yes Yes
Luczynska40 2015 No No No Yes Yes Yes Yes
Kamal41 2015 No No Yes No Yes Yes No
Luczynska42 2014 No Yes Yes Yes Yes No Yes
Cheung43 2014 Yes No No No Yes Yes Yes
Lobbes44 2014 No No No Yes Yes Yes No

Abbreviation: BI-RADS ¼ Breast Imaging Reporting and Data System.

Clinical Breast Cancer Month 2018

Xiao Zhu et al
-3
 Diagnostic Value of CESM
designed without a control arm, such as diagnostic studies, have no likelihood ratio (NLR), and diagnostic odds ratio (DOR), were
way to assess study quality.14 Therefore, in recent years, a tool, computed. A summary receiver operating characteristic (SROC)
Quality Assessment of Diagnostic Accuracy Studies (QUADAS)- curve, which was plotted based on the sensitivity and specificity
2,15 an improved edition of QUADAS,16 has been developed for value, and the area under the curve (AUC) of the SROC curve18,19
quality assessment of studies of diagnostic accuracy. Quality were also calculated. A random-effects model was used to calculate
assessment is a vital part of any systematic review, and hence these diagnostic accuracy parameters.20
QUADAS-2 is recommended to evaluate the risk of bias in diag- The Cochran Q test (P < .05 or I2 > 50%)21 was used to detect
nostic studies. The QUADAS-2 checklist comprises key domains heterogeneity, which referred to the degree of variability in results
covering patient selection, index test, reference standard, and flow across the studies. We analyzed the effects of the covariates on DOR
and timing domains. Each domain was assessed in terms of risk of (ie, study design, geographic location, patients’ age, BI-RADS,
bias, and the first 3 domains were also analyzed in terms of appli- blinded or not for final diagnosis, kind of reference standard
cability. Each domain was assessed as “yes,” “no,” or “unclear,” with used). The relative DOR was calculated to analyze the influence of
“yes” indicating low risk of bias or low concern regarding applica- these covariates.22,23 We analyzed the potential presence of publi-
bility and “no” indicating high risk of bias or high concern regarding cation bias using funnel plots.24
applicability. Quality assessment was independently cross-checked
by 2 reviewers (Z.X. and K.Z.). If reviewers failed to extract Results
sufficient data from the literature to permit a judgment, those items After independent review, 24 publications were included in the
were categorized as “unclear.” Discrepancies were solved by team analysis. Two of the 24 studies identified were excluded because they
discussion and consensus. The final quality assessments were merely evaluated the value of CESM in monitoring neoadjuvant
recorded in a QUADAS-2 form.15 chemotherapy.25,26 Three studies were not included because they
incorporated only breast cancer patients.27-29 One study was a sys-
Statistical Analysis tematic review and meta-analysis.11 Thus, a total of 18 full-text articles
Analyses were performed by Stata 10.0 (StataCorp, College meeting the analysis criteria were included in our study.7,8,10,30-44
Station, TX) and Meta-DiSc for Windows (XI Cochrane Collo- Figure 1 provides the flowchart for identification of eligible studies.
quium, Barcelona, Spain). Recommended standard methods were
used to carry out diagnostic meta-analyses.17 Characteristics and Quality of Studies
Besides sensitivity and specificity, the other important measures The eligible studies included 2859 patients who ranged in age
of test accuracy, like positive likelihood ratio (PLR), negative from 31 to 75 years with CESM. Two studies were designed

Figure 1 Flowchart Showing Study Exclusion and Inclusion

Abbreviation: CESM ¼ contrast-enhanced spectral mammography.

4 - Clinical Breast Cancer Month 2018

 Xiao Zhu et al
Table 3 Tabular Presentation for QUADAS-2 Results of 18 Included Studies

Risk of Bias Applicability Concerns

Patient Index Reference Flow and Patient Index Reference
Study Selection Test Standard Timing Selection Test Standard
Houben8 H L H H H L L
Li10 H ? L L H L L
Patel30 H ? L L H L L
Mori7 H L ? H H L L
Richter31 H ? L H H L L
Cheung32 L ? L L L L L
Luczynska33 H ? L L H L L
Cheung34 H ? L L H L L
Lalji35 L L ? H L L L
Tennant36 L ? ? L L L L
Kamal37 L ? ? H L L L
Wang38 L L L L L L L
Luczynska39 H L L L H L L
Luczynska40 H L ? L H L L
Kamal41 L L ? H L L L
Luczynska42 H ? ? L H L L
Cheung43 H L H L H L L
Lobbes44 L L ? H L L L

Abbreviations: H ¼ high risk; L ¼ low risk; ? ¼ unclear risk; QUADAS ¼ Quality Assessment of Diagnostic Accuracy Studies.

prospectively and 16 studies were designed retrospectively. Five Multiple Regression Analysis and Publication Bias
studies were carried out in Asia, and the rest were performed in the Metaregression was applied to assess the following aspects of 18
United States and Europe. One or two outcome measures consid- studies (Table 2): study design (design: prospective or not), de-
ering histopathologic findings, clinical follow-up, or both were mographic characteristics (region: Asia or not), patients (consecutive
adopted in these studies (Tables 1 and 2). or not), patient age (age > 40 years or not), lesions (included BI-
The quality assessment of the incorporated papers by the RADS 1-3 or not), CESM (blinded to final diagnosis or not),
QUADAS-2 tool is depicted in Table 3 and Figure 2. In the “patient and reference standard (histopathology only or not). These cate-
selection” domain, 7 studies32,35-38,41,44 were considered to be at gories did not significantly affect diagnostic accuracy (P > .05), and
relatively low risk of bias, and the rest of the studies were at high risk. none of them showed any definite influence on heterogeneity
In “index test,” 9 studies7,8,35,38-41,43,44 were at low risk of bias, and (Table 4). The funnel plots showed large asymmetry, which indi-
the rest were not. In “reference standard,” 8 studies10,30-34,38,39 were cated a potential for publication bias for CESM (Figure 5).
regarded as low risk; the rest were at high risk. In terms of “flow and
timing,” only 7 studies7,8,31,35,37,41,44 were scored with low risk of Discussion
bias. But in both “index test” and “reference standard” domains, all Feasibility studies have shown that CESM is better than con-
the studies had low concerns regarding applicability. ventional mammography,45 and its sensitivity could possibly be
comparable with magnetic resonance imaging.10,31
Overall Diagnostic Accuracy of CESM A previous meta-analyses about diagnosis of CESM11 reported
Forest plots were created to show sensitivity and specificity values that CESM has a high pool sensitivity (0.98; 95% CI, 0.96-1.00)
with their corresponding 95% confidence intervals (CI) (Figure 3). but very low pool specificity (0.58; 95% CI, 0.38-0.77). The
The pooled sensitivity and specificity was 0.89 (95% CI, 0.88-0.91) present study showed that the pooled sensitivity (0.89, 95% CI
and 0.84 (95% CI, 0.82-0.85), respectively. The PLR was 3.73 0.88-0.91) was in line with previous meta-analyses,11 but the
(95% CI, 2.68-5.20), NLR was 0.10 (95% CI, 0.06-0.15), and pooled specificity (0.84, 95% CI 0.82-0.85) was higher than the
DOR was 71.36 (95% CI, 36.28-140.39). I2 values of sensitivity, previously reported specificity of CESM. It is likely that the
specificity, PLR, NLR, and DOR were 91.9%, 97.0%, 90.7%, development of technology and a deeper experience in the use of a
85.6%, and 66.3%, respectively. new imaging modality such as CESM may increase the diagnostic
As shown in Figure 4, in the SROC curve of CESM, the solid performance. The previous study11 included only 8 articles in the
circle presenting the studies is positioned near the desirable literature, and 2 of them included only breast cancer patients, which
upper left corner, indicating a relatively high level of overall would create inappropriate exclusions and result in overoptimistic
accuracy in breast cancer diagnosis. The AUC was 0.96 (standard estimates of CESM diagnostic accuracy. In addition, the previous
error ¼ 0.011). The maximum joint sensitivity and specificity meta-analyses did not discuss PLR and NLR, which has been pre-
(ie, the Q value) was 0.90 (standard error ¼ 0.016). sented as a Supplemental Figure 1 in the present study.

Clinical Breast Cancer Month 2018 -5

 Diagnostic Value of CESM
Figure 2 Graphical Display for QUADAS-2 Results of 18 Included Studies

Abbreviation: QUADAS ¼ Quality Assessment of Diagnostic Accuracy Studies.

The DOR is the ratio of the odds of a positive test relative to the conclusive shifts from pretest to posttest probability, indicating high
odds of a negative test, and the value ranges from 0 to infinity. accuracy.49,50 In the present study, the PLR value of 3.73 suggests
DOR is considered as a single indicator of test accuracy.46 Higher that patients with breast cancer, compared to those without the
values of DOR indicate higher accuracy, except when DOR is equal disease, have an approximately 4-fold higher chance of being disease
to 1. DOR equal to 1 refers to the index test that does not positive on CESM. This PLR value is not satisfactory. The reason
distinguish patients with the disease and those without it. In the might be that there were several kinds of benign lesions with
present meta-analysis, the DOR of CESM is relatively high (71.36). enhancement,8,30 which results in a false-positive result and reduce
It indicated that the likelihood of a correct diagnosis of true breast the PLR value. The high pool sensitivity of 89% and a PLR value of
cancers is 71.36:1 if CESM indicates a positive result. 3.73 suggests that the enhancing area of breast lesion should be
The SROC curve and the corresponding AUC value are used to sampled by biopsy to exclude malignant mass. On the other hand,
estimate the overall diagnostic performance. When the AUC value is NLR was found to be 0.10, which is not low enough to completely
greater than 0.90, the performance of the index test is considered to rule out the presence of breast malignant lesions. If the CESM result
have a good diagnostic capability.47 Therefore, in the present study, was negative, the probability that the patient has breast cancer is
the location of solid circles in the SROC curves and the AUC value approximately 10%. In terms of NLR, the CESM indicate the
of 0.96 indicated that CESM is an excellent breast imaging unsatisfactory robustness. In other words, a negative CESM result
modality. should be interpreted with caution when this method is used
Likelihood ratios (LRs) are considered to be more clinically independently for the detection of breast cancer.
meaningful than SROC curves and DOR.48,49 LRs evaluate According to QUADAS-2 quality assessment criteria, the do-
whether a negative or positive result changes the probability of an mains “patient selection,” “index test,” and “flow and timing” were
existing disease state. LRs of > 10 or < 0.1 often produce contributing potentially high risks of bias to our review. In the

6 - Clinical Breast Cancer Month 2018

 Xiao Zhu et al
Figure 3 Forest Plot of Estimates of Sensitivity and Specificity for CESM in Diagnosis of Breast Cancer. Studies are indicated by
authors’ names. Point estimates of sensitivity and specificity from each study are shown as solid circles. Error bars are 95%
confidence intervals

Abbreviation: CESM ¼ contrast-enhanced spectral mammography.

Clinical Breast Cancer Month 2018 -7

 Diagnostic Value of CESM
studies. In order to explore the potential source of heterogeneity, we
Figure 4 SROC Curves for CESM. Each study is represented by
each solid circle; size of circle indicates size of included obvious variables in metaregression analysis: study design,
study. SROC curves summarize overall diagnostic demographic characteristics, consecutive patients or not, patient’s
accuracy age, lesions (included BI-RADS 1-3 or not), CESM analysis
(blinded to final diagnosis or not), reference standard (histopa-
thology only or not). However, these obvious different aspects did
not contribute to the heterogeneity. Other various aspects might
influence the heterogeneity, such as tumor size, tumor type,
metastasis, tumor, node, metastasis classification system staging, and
different research protocols. However, these aspects could not be
analyzed in the present study because of partial loss of data.
Funnel plots were constructed to estimate publication bias.52 The
funnel plot asymmetry in the present analysis indicated the presence
of significant publication bias among the included studies. It should
be of concern that studies with statistically significant results are
more likely to be published. In addition, the larger trials could show
fewer diagnostic effects than smaller studies due to more case-mix
differences.
We adopted a comprehensive search and analysis strategy to assess
the accuracy of CESM for breast cancer diagnosis. However, there are
still limitations in our study. First, we have only included those articles
Abbreviations: CESM ¼ contrast-enhanced spectral mammography; SROC, summary receiver
operating characteristic. published in either Chinese or English for full-text analysis. Such se-
lection might lead to publication bias. Second, the major source of
domain “patient selection,” studies are recommended to enroll participants included trials that were based on highly selected case
consecutive patients with suspected disease. However, the majority series. This aspect may make the study prone to selection bias. Third,
of the studies included were designed retrospectively, and the number of the breast lesions included in our studies was 3108,
CESM was performed on selected populations (such as patients in which is relatively low to adequately assess the real performance of
BI-RADS 3-4 classification), which would lead to the maximum CESM. Hence, high-quality and large randomized controlled trials are
proportion of cases with breast cancer (Tables 1 and 2). Because required to analyze the diagnostic accuracy of CESM in breast cancers.
breast cancer prevalence is generally low, this method of patient
selection could not really reflect how CESM performs on popula- Conclusion
tion screening. However, these selection biases may be acceptable in Our meta-analysis review showed that CESM has good sensitivity
early clinical studies. In the domain “index test,” we could not and specificity in the diagnosis of breast cancer. CESM may be
ignore the fact that 9 studies did not report whether CESM results considered a useful test for initial breast lesion assessment.
were blinded regarding the final diagnosis (Table 3, Figure 2). These
unknown subjects may relate the potential bias to the subjectivity of Clinical Practice Points
CESM interpretation. However, the included studies all had low  CESM is a new image examination technology that has devel-
concerns regarding applicability in both the “index test” and oped over the past few years.
“reference standard” domains (Table 3, Figure 2).  In detecting lesions in dense breast tissue, CESM is better than
In addition to the summary estimate of test performance, conventional mammography because of its sensitivity and spec-
exploring the sources of heterogeneity is still an important issue of ificity. Moreover, CESM seems to have almost the same sensi-
meta-analysis.51 In the present meta-analysis, the Q test and I2 tivity as BMRI, but shorter examination time makes it a
statistic indicated significant heterogeneity among the included potential alternative to BMRI.

Table 4 Metaregression of Effects of Different Studies’ Aspects on Diagnosis Value of Contrast-Enhanced Spectral Mammography

Covariate No. of Studies Coefficient P RDOR 95% Confidence Interval

Prospective design 2 1.44 .26 4.23 0.29, 62.13
Consecutive patients 7 1.57 .06 0.21 0.04, 1.04
Including BI-RADS 1-3 16 0.57 .63 1.77 0.13, 24.22
Geographic (Asia) 5 0.54 .52 0.58 0.09, 3.61
Age > 40 years 8 1.13 .13 3.08 0.66, 14.32
Unaware of diagnosis 9 0.19 .81 0.83 0.15, 4.61
Histopathology only 11 0.11 .90 0.90 0.15, 5.50

Abbreviations: BI-RADS ¼ Breast Imaging Reporting and Data System; RDOR, relative diagnostic odds ratio.

8 - Clinical Breast Cancer Month 2018

 Xiao Zhu et al
Figure 5 Funnel Graph for Assessment of Potential Publication References
1. Freitas SA, Mackenzie R, Wylde DN, et al. All-cause mortality for life insurance
Bias in CESM. Funnel graph plots log of DOR against applicants with a history of breast cancer. J Insur Med 2017; 47:6-22.
SE of log of DOR (indicator of sample size). Eighteen 2. Rocha-Brischiliari SC, Oliveira RRD, Andrade L, et al. The rise in mortality from
circles represent 18 studies in meta-analysis. Line in breast cancer in young women: trend analysis in Brazil. PLoS One 2017; 12:
center indicates summary diagnostic odds ratio e0168950.
3. Kim HR, Jung HK, Ko KH, Kim SJ, Lee KS. Mammography, US, and MRI for
preoperative prediction of extensive intraductal component of invasive breast
cancer: interobserver variability and performances. Clin Breast Cancer 2016; 16:
305-11.
4. Shetty MK. Screening for breast cancer with mammography: current status and an
overview. Indian J Surg Oncol 2010; 1:218-23.
5. Loberg M, Lousdal ML, Bretthauer M, Kalager M. Benefits and harms of
mammography screening. Breast Cancer Res 2015; 17:63.
6. Boyd NF, Guo H, Martin LJ, et al. Mammographic density and the risk and
detection of breast cancer. N Engl J Med 2007; 356:227-36.
7. Mori M, Akashi-Tanaka S, Suzuki S, et al. Diagnostic accuracy of contrast-
enhanced spectral mammography in comparison to conventional full-field digital
mammography in a population of women with dense breasts. Breast Cancer 2017;
24:104-10.
8. Houben IPL, Van de Voorde P, Jeukens C, et al. Contrast-enhanced spectral
mammography as work-up tool in patients recalled from breast cancer
screening has low risks and might hold clinical benefits. Eur J Radiol 2017;
94:31-7.
9. Bhimani C, Matta D, Roth RG, et al. Contrast-enhanced spectral
mammography: technique, indications, and clinical applications. Acad Radiol
2017; 24:84-8.
10. Li L, Roth R, Germaine P, et al. Contrast-enhanced spectral mammography
(CESM) versus breast magnetic resonance imaging (MRI): a retrospective com-
Abbreviations: CESM ¼ contrast-enhanced spectral mammography; DOR, diagnostic odds parison in 66 breast lesions. Diagn Interv Imaging 2017; 98:113-23.
ratio; SE ¼ standard error. 11. Tagliafico AS, Bignotti B, Rossi F, et al. Diagnostic performance of contrast-
enhanced spectral mammography: systematic review and meta-analysis. Breast
2016; 28:13-9.
 As CESM technology keeps improving, a new meta-analysis re-
12. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting
view is needed to systematically evaluate the potential diagnostic systematic reviews and meta-analyses of studies that evaluate health care
interventions: explanation and elaboration. PLoS Med 2009; 6:e1000100.
value of CESM. 13. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of ran-
 The important parameters of CESM accuracy for breast cancer domized clinical trials: is blinding necessary? Control Clin Trials 1996; 17:1-12.
14. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in
diagnosis were analyzed. epidemiology: a proposal for reporting. Meta-analysis of Observational Studies in
 The pooled sensitivity and specificity were 0.89 (95% CI, Epidemiology (MOOSE) group. JAMA 2000; 283:2008-12.
0.88e0.91) and 0.84 (95% CI, 0.82e0.85), respectively. PLR 15. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for
the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011; 155:
was 3.73 (95% CI, 2.68e5.20), NLR was 0.10 (95% CI, 529-36.
0.06e0.15), and DOR was 71.36 (95% CI, 36.28e140.39). 16. Whiting PF, Weswood ME, Rutjes AW, Reitsma JB, Bossuyt PN, Kleijnen J.
Evaluation of QUADAS, a tool for the quality assessment of diagnostic accuracy
The area under the curve (AUC) was 0.96 (standard error ¼ studies. BMC Med Res Methodol 2006; 6:9.
0.011). 17. Deville WL, Buntinx F, Bouter LM, et al. Conducting systematic reviews of
diagnostic studies: didactic guidelines. BMC Med Res Methodol 2002; 2:9.
 Our meta-analysis review showed that CESM has good sensi- 18. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a
tivity and specificity in the diagnosis of breast cancer. diagnostic test into a summary ROC curve: data-analytic approaches and some
additional considerations. Stat Med 1993; 12:1293-316.
 CESM could be considered as a useful triage test for initial breast 19. Lau J, Ioannidis JP, Balk EM, et al. Diagnosing acute cardiac ischemia in the
lesions assessment. emergency department: a systematic review of the accuracy and clinical effect of
current technologies. Ann Emerg Med 2001; 37:453-60.
20. Irwig L, Tosteson AN, Gatsonis C, et al. Guidelines for meta-analyses evaluating
diagnostic tests. Ann Intern Med 1994; 120:667-76.
21. Deeks JJ, Higgins JP, Altman DG. Analysing data and undertaking meta-analyses.
Acknowledgments In: Julian Higgins PT, Green S, eds. Cochrane Handbook for Systematic Reviews of
The authors thank Hong-Yu Chen (Guangxi Medical University) Interventions. New York: Wiley-Blackwell; 2008:243-96.
22. Suzuki S, Moro-oka T, Choudhry NK. The conditional relative odds ratio pro-
and Chiliang Chen (University of California) for helpful discus- vided less biased results for comparing diagnostic test accuracy in meta-analyses.
sions. This work was supported by Guangxi Colleges and Univer- J Clin Epidemiol 2004; 57:461-9.
23. Westwood ME, Whiting PF, Kleijnen J. How does study quality affect the results
sities Science and Technology Research Projects (project of a diagnostic meta-analysis? BMC Med Res Methodol 2005; 5:1.
KY2015LX054), Postgraduate Course Construction of Guangxi 24. Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected by a
simple, graphical test. BMJ 1997; 315:629-34.
Medical University Project (project YJSB2017013), Innovation 25. Iotti V, Ravaioli S, Vacondio R, et al. Contrast-enhanced spectral mammography
Project of Guangxi Graduate Education (project YCBZ2017041), in neoadjuvant chemotherapy monitoring: a comparison with breast magnetic
resonance imaging. Breast Cancer Res 2017; 19:106.
Humanities Base Project of Guangxi Medical University (project 26. Barra FR, de Souza FF, Camelo R, Ribeiro ACO, Farage L. Accuracy of contrast-
2016RWB04), and International Communication of Guangxi enhanced spectral mammography for estimating residual tumor size after neo-
adjuvant chemotherapy in patients with breast cancer: a feasibility study. Radiol
Medical University Graduate Education Project. Bras 2017; 50:224-30.
27. Jochelson MS, Dershaw DD, Sung JS, et al. Bilateral contrast-enhanced dual-
energy digital mammography: feasibility and comparison with conventional digital
Disclosure mammography and MR imaging in women with known breast carcinoma. Radi-
The authors have stated that they have no conflict of interest. ology 2013; 266:743-51.
28. Fallenberg EM, Dromain C, Diekmann F, et al. Contrast-enhanced spectral
mammography: does mammography provide additional clinical benefits or can
Supplemental Data some radiation exposure be avoided? Breast Cancer Res Treat 2014; 146:371-81.
29. Fallenberg E, Dromain C, Diekmann F, et al. Contrast-enhanced spectral
Supplemental figure accompanying this article can be found in mammography versus MRI: initial results in the detection of breast cancer and
the online version at https://doi.org/10.1016/j.clbc.2018.06.003. assessment of tumour size. Eur Radiol 2014; 24:256-64.

Clinical Breast Cancer Month 2018 -9

 Diagnostic Value of CESM
30. Patel BK, Naylor ME, Kosiorek HE, et al. Clinical utility of contrast-enhanced 41. Mohamed Kamal R, Hussien Helal M, Wessam R, Mahmoud Mansour S,
spectral mammography as an adjunct for tomosynthesis-detected architectural Godda I, Alieldin N. Contrast-enhanced spectral mammography: impact of the
distortion. Clin Imaging 2017; 46:44-52. qualitative morphology descriptors on the diagnosis of breast lesions. Eur J Radiol
31. Richter V, Hatterman V, Preibsch H, et al. Contrast-enhanced spectral mammog- 2015; 84:1049-55.
raphy in patients with MRI contraindications. Acta Radiol 2018; 59:798-805. 42. Luczynska E, Heinze-Paluchowska S, Dyczek S, Blecharz P, Rys J, Reinfuss M.
32. Cheung YC, Juan YH, Lin YC, et al. Dual-energy contrast-enhanced spectral Contrast-enhanced spectral mammography: comparison with conventional
mammography: enhancement analysis on BI-RADS 4 non-mass micro- mammography and histopathology in 152 women. Korean J Radiol 2014; 15:
calcifications in screened women. PLoS One 2016; 11:e0162740. 689-96.
33. Luczynska E, Niemiec J, Hendrick E, et al. Degree of enhancement on contrast 43. Cheung YC, Lin YC, Wan YL, et al. Diagnostic performance of dual-energy
enhanced spectral mammography (CESM) and lesion type on mammography contrast-enhanced subtracted mammography in dense breasts compared to
(MG): comparison based on histological results. Med Sci Monit 2016; 22:3886-93. mammography alone: interobserver blind-reading analysis. Eur Radiol 2014; 24:
34. Cheung YC, Tsai HP, Lo YF, Ueng SH, Huang PC, Chen SC. Clinical utility of 2394-403.
dual-energy contrast-enhanced spectral mammography for breast microcalcifications 44. Lobbes MB, Lalji U, Houwers J, et al. Contrast-enhanced spectral mammography
without associated mass: a preliminary analysis. Eur Radiol 2016; 26:1082-9. in patients referred from the breast cancer screening programme. Eur Radiol 2014;
35. Lalji UC, Houben IP, Prevos R, et al. Contrast-enhanced spectral mammography 24:1668-76.
in recalls from the Dutch breast cancer screening program: validation of results in a 45. Luczynska E, Heinze S, Adamczyk A, Rys J, Mitus JW, Hendrick E. Comparison
large multireader, multicase study. Eur Radiol 2016; 26:4371-9. of the mammography, contrast-enhanced spectral mammography and ultraso-
36. Tennant SL, James JJ, Cornford EJ, et al. Contrast-enhanced spectral mammog- nography in a group of 116 patients. Anticancer Res 2016; 36:4359-66.
raphy improves diagnostic accuracy in the symptomatic setting. Clin Radiol 2016; 46. Glas AS, Lijmer JG, Prins MH, Bonsel GJ, Bossuyt PM. The diagnostic
71:1148-55. odds ratio: a single indicator of test performance. J Clin Epidemiol 2003; 56:
37. Kamal RM, Helal MH, Mansour SM, et al. Can we apply the MRI BI-RADS 1129-35.
lexicon morphology descriptors on contrast-enhanced spectral mammography? 47. Swets JA. Measuring the accuracy of diagnostic systems. Science 1988; 240:
Br J Radiol 2016; 89:20160157. 1285-93.
38. Wang Q, Li K, Wang L, Zhang J, Zhou Z, Feng Y. Preclinical study of diagnostic 48. Deeks JJ. Systematic reviews of evaluations of diagnostic and screening tests. BMJ
performances of contrast-enhanced spectral mammography versus MRI for breast 2001; 323:1188.
diseases in China. SpringerPlus 2016; 5:763. 49. Guyatt G, Rennie D, Meade MO, Cook DJ. Users’ Guides to the Medical
39. Luczynska E, Niemiec J, Ambicka A, et al. Correlation between blood and Literature: A Manual for Evidence-Based Clinical Practice. Chicago, IL: AMA
lymphatic vessel density and results of contrast-enhanced spectral mammography. Press; 2002.
Pol J Pathol 2015; 66:310-22. 50. Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. BMJ 2004; 329:168-9.
40. Luczynska E, Heinze-Paluchowska S, Hendrick E, et al. Comparison between 51. Petitti DB. Approaches to heterogeneity in meta-analysis. Stat Med 2001; 20:
breast MRI and contrast-enhanced spectral mammography. Med Sci Monit 2015; 3625-33.
21:1358-67. 52. Sterne JA, Harbord RM. Funnel plots in meta-analysis. Stata J 2004; 4:127-41.

10 - Clinical Breast Cancer Month 2018

 Xiao Zhu et al
Supplemental Figure 1 Forest Plot of Estimates of PLR and NLR for CESM in Diagnosis of Breast Cancer. Studies are indicated by
author names. Point estimates of PLR and NLR from each study are shown as solid circles. Error bars are 95%
confidence intervals

Abbreviations: CESM ¼ contrast-enhanced spectral mammography; NLR ¼ negative likelihood ratio; PLR ¼ positive likelihood ratio.

Clinical Breast Cancer Month 2018 - 10.e1