Letter to the Editor: “Diagnostic accuracy of cone-beam breast computed tomography: a systematic review and diagnostic meta-analysis”

by Yueqiang Zhu, Yue Ma, Aidi Liu, Zhaoxiang Ye (zhuyueqiang1985@126.com, yezhaoxiang@163.com)

Diagnostic accuracy of cone-beam breast computed tomography: a systematic review and diagnostic meta-analysis

Dear Editor,

We read with considerable interest Dr. Uhlig and colleagues’ article “Diagnostic accuracy of cone-beam breast computed tomography: a systematic review and diagnostic meta-analysis” published in the March 2019 issue of European Radiology [1]. The authors aimed to review the published evidence on cone-beam breast computed tomography (CBBCT) and summarize its diagnostic accuracy for breast lesion assessment. We appreciate their comprehensive review on the clinical utilization of CBBCT and diagnostic performance of current studies. The authors stated that they have made a comprehensive electronic literature search with unrestricted concerning date, language and region, manually searched reference lists of reviews and conference proceedings to identify gray literature, and contacted international CBBCT experts to identify unpublished data. But in fact, maybe due to language or literature retrieval methods, two of our related studies published on Chinese journals in 2018 were not included in the meta-analysis [2, 3]. Our two studies provided sufficient information to allow for computation of sensitivity and specificity, and fulfilled the inclusion criteria of Uhlig’s study. In our opinion, when our studies are stratified and analyzed, the pooled sensitivity and specificity maybe changed. Thus, we present the main results of these studies to help the literature review and meta-analysis more perfect. In addition, one more research on diagnostic accuracy of CBBCT published in 2017 from Korea is eligible and should also be included [4].

StudyⅠ [2]
The objective of studyⅠ was to compare the diagnostic efficiency of CBBCT and mammography (MG). CBBCT and MG images were retrospectively evaluated in 160 patients (median age of 47 years, range 35-74 years) with 165 breasts from May 2012 to August 2014. Independent double-blind reading was performed by two radiologists with 3 and 5 years of experience with CBBCT. Sixty-seven breasts underwent non-contrast (NC) CBBCT only, and 98 breasts underwent both NC- and contrast-enhanced (CE) CBBCT. Pre- and post-contrast CBBCT images were read in pair. Final conclusions were reached in consensus by discussion in case of discrepancy. Sensitivity, specificity and AUC were compared between the two modalities. In the 165 breast (78 benign, 87 malignant), 24 were ACR type a/b and 141 were type c/d. The interobserver agreement (Kappa) between two readers was 0.866 for MG and 0.768 for CBBCT. With BI-RADS 4b as cutoff value, the sensitivity for CBBCT (83.3%) was higher than MG (70%), and also for the dense group (83.1% vs. 69%). The specificity was higher for MG in comparison to CBBCT (98.7% vs. 97.3%), while be equivalent in dense group (98.6% vs. 98.6%). The AUC was lower with MG than with CBBCT in both all (0.923 vs. 0.959, p < 0.05) and dense group (0.919 vs. 0.973, p < 0.05). All in all, CBBCT showed higher sensitivity and accuracy in the diagnosis of breast malignant tumors compared to MG.

Study Ⅱ [3]
Similar to the study of Wienbeck et al [5] in 2018, our Study Ⅱ focused on ACR type c/d breasts. The purpose of Study Ⅱ was to compare the diagnostic value of MG, NC-CBBCT and CE-CBBCT in dense breast tissue. The study included 80 patients with 80 breasts (median age of 47 years, range 35-66 years) who underwent MG, NC- and CE-CBBCT examinations from May 2012 to August 2014. Two independent blinded readers evaluated all images. If there was disagreement between the two readers, discussion was performed until reaching agreement. ΔCT values were measured. In the 80 breasts (40 benign, 40 malignant), 72 were ACR type c and 8 were type d. AUC of MG, NC-CBBCT and CE-CBBCT were 0.934 (95% CI: 0.855-0.977), 0.971 (95% CI: 0.907-0.996) and 0.975 (95% CI: 0.912-0.997), respectively. With BI-RADS 4b as cutoff value, sensitivity of three modalities were 72.5%, 75% and 92.5%, respectively; specificity of three modalities were all 100%. ΔCT of malignant tumor (71.682 ± 27.765 HU) was obviously higher than that of benign tumor (42.825 ± 33.701HU) (t = -4.180, p < 0.01). In conclusion, the diagnosis of breast malignant tumors by CE-CBBCT showed the highest sensitivity and accuracy in assessment of lesions with dense breast tissue.

Our CE-CBBCT protocol is as follows: After initial NC-CBBCT scan, 90 ml non-ionic contrast media (Iohexol, Omnipaque® 300, GE Healthcare) was single-shot intravenously injected at a rate of 2 ml/s using a power injector, followed by a 30 ml saline solution chaser. CE-CBBCT scan was obtained 2 min after contrast media administration. The radiation dose is 5.8 mGy in NC-CBBCT. If the patient accepted both NC- and CE-CBBCT, the dose means one fold increase.

Erratum

We carefully read the relevant literatures cited in the meta-analysis and found some minor mistakes in the manuscript, listed as follows:

In Results
1. The total number of patients should be 651, not 559.
2. The study of Aminololama-Shakeri et al [6] used proprietary CBBCT at University of California, Davis (UCD), not San Diego (UCSD).
3. In the study of Aminololama-Shakeri et al [6], the majority of patients had ACR type c/d breasts (72%, 28/39), not type a/b.
4. He et al [7] reported diagnostic accuracy of NC-CBBCT among 212 patients, and NC-&CE-CBBCT among 120 patients. Specifically, 120 patients with ACR type c/d (98 type c and 32 type d) was chosen as subgroup accepted CE-CBBCT for further evaluation.
5. The percent of breast lesions presented as microcalcifications in study by He et al [7] was 75% (331/442) in MG and 69% (306/442) in CBBCT, not 14%. In fact, 14% (62/442) was the lesion only manifested as calcification in MG.
6. The research of Wienbeck et al [5] in 2018 included 41 patients with dense or very dense breast tissue (63% ACR type c, 37% ACR type d), not 73% ACR type c and 27% ACR type d.
7. Microcalcifications were evident in 65% of cases in the study of Zhao et al [8], not 63%. In their study, 85 breast masses were confirmed by histopathology, including 45 malignant and 40 benign, 67.7% malignant and 63.2% benign tumors coexisted with microcalcifications.

In Discussion
8. The proportion of premenopausal women ranged from 32-68%, not 32-48%.
9. The lowest proportion of microcalcifications was 16%, not 14%.

Supplement

After reading these papers, we found that more valuable information could be extracted from the six diagnostic accuracy researches, listed as follows.

ΔCT value
Aminololama-Shakeri et al [6] found that all DCIS enhanced on CE-CBBCT. The mean enhancement value was 90 ± 53 HU for malignant calcifications and 33 ± 30 HU for benign calcifications (p < 0.0001). The enhancement value of CE-BCBCT in the study by He et al [7] was 48 HU (20-169 HU) for benign lesions and 124 HU (35-275 HU）for malignant lesions. The enhancement difference between benign and malignant masses on CE-CBBCT has been reported previously [9, 10]. Diagnostic accuracy could increase if ΔCT value served as a quantitative tool for assessment.

Cutoff value
In the study by Aminololama-Shakeri et al [6], a conspicuity score (CS) for each lesion was assigned for each modality by two independent observers. Cole et al [11], He at al [7] and Zhao et al [8] analyzed lesions with cutoff value of BI-RADS 4, 4a and 4b, respectively. Wienbeck et al [5, 12] dichotomized the BI-RADS score, labelling BI-RADS 1 and 2 as negative readings, and BI-RADS 3, 4 and 5 as positive readings. Different diagnostic criteria were used in various studies. If they are adjusted to one standard, the results may be different.

Amendments and more supplementary information extracted from our [2, 3] and research by Jung et al [4] are shown in Fig. 1 and Table 1.