Advertisement

Multicenter Prospective Study of Color Doppler Ultrasound for Breast Masses: Utility of Our Color Doppler Method

Open AccessPublished:March 22, 2019DOI:https://doi.org/10.1016/j.ultrasmedbio.2019.01.021

      Abstract

      The use of color Doppler ultrasound (CD) for distinguishing between benign and malignant breast lesions remains controversial. This study (JABTS BC-04 study) was aimed at confirming the usefulness of our CD diagnostic criteria. We evaluated ultrasound images of 1408 solid breast masses from 16 institutions in Japan (malignant: 839, benign: 569). Multivariate analysis indicated that vascularity (amount of blood flow), vascular flow pattern (“surrounding marginal flow” or “penetrating flow”) and the incident angle of penetrating flow were significant findings for distinguishing between benign and malignant lesions. However, the sensitivity and specificity of B-mode alone did not improve significantly with CD addition (97.6% → 97.9%, 38.3% → 41.5%, respectively). We explored the causes of these negative results and found that age should have been considered when evaluating vascularity. Simulation experiments suggested that specificity is significantly improved when age is taken into consideration (38.3% → 46.0%, p < 0.001) and we thereby improved our diagnostic criteria.

      Key Words

      Introduction

      Breast ultrasound, along with mammography, is an essential examination in the diagnosis of breast cancer. B-Mode ultrasound is a basic examination that has improved markedly because of advances in technologies such as tissue harmonic imaging, spatial compounding and computer technology. Color Doppler ultrasound (CD) and elastography, as adjunct examinations, are used to visualize blood flow and tissue stiffness. The Japan Association of Breast and Thyroid Sonology (JABTS) published the Guidelines for Breast Ultrasound Diagnosis (first edition) in 2004 (
      Japan Association of Breast and Thyroid Sonology (JABTS)
      Guidelines for breast ultrasound: Management and diagnosis.
      ). In these Guidelines, we stated that CD and elastography are tools contributing to efforts to distinguish between benign and malignant breast lesions. The usefulness of CD in diagnosing breast lesions remains controversial (
      • Birdwell R.L.
      • Ikeda D.M.
      • Jeffrey S.S.
      • Jeffrey Jr, R.B.
      Preliminary experience with power Doppler imaging of solid breast masses.
      ,
      • Cosgrove D.O.
      • Kedar R.P.
      • Bamber J.C.
      • al-Murrani B.
      • Davey J.B.
      • Fisher C.
      • McKinna J.A.
      • Svensson W.E.
      • Tohno E.
      • Vagios E.
      Breast diseases: Color Doppler US in differential diagnosis.
      ,
      • Horvath D.E.
      • Cuitino M.J.
      • Pinochet M.A.
      • Sanhueza P.
      Color Doppler in the study of the breast: How do we perform it?.
      ,
      • Lee S.W.
      • Choi H.Y.
      • Baek S.Y.
      • Lim S.M.
      Role of color and power Doppler imaging in differentiating between malignant and benign solid breast masses.
      ,
      • Raza S.
      • Baum J.K.
      Solid breast lesions: Evaluation with power Doppler US.
      ,
      • Sehgal C.M.
      • Arger P.H.
      • Rowling S.E.
      Quantitative vascularity of breast masses by Doppler imaging: Regional variations and diagnostic implications.
      ,
      • Wilkens T.H.
      • Burke B.J.
      • Cancelada D.A.
      • Jatoi I.
      Evaluation of palpable breast masses with color Doppler sonography and gray-scale imaging.
      ). Furthermore, standardized CD diagnostic criteria for breast lesions are as yet lacking. The Doppler pattern was recently removed from the Breast Imaging Reporting and Data System categorization of the American College of Radiology (
      American College of Radiology
      ACR BI-RADS atlas: Breast imaging reporting and data system.
      ). However, to clarify whether CD is effective, large-scale research is necessary. Therefore, the Terminology and Diagnostic Criteria Committee of the JABTS decided to conduct a multicenter study. The main purposes of this study were to examine the usefulness of CD for distinguishing between benign and malignant breast masses and to develop diagnostic criteria based on CD findings. We also examined the relationship between CD findings and elastographic results to clarify whether CD should be performed in addition to elastography.

      Methods

      Study design

      This was a multicenter prospective observational study, and was named JABTS BC-04 study. We enrolled women with pathologically confirmed breast masses who underwent B-mode and CD examinations and women with breast masses that had exhibited no significant change for more than 2 y and had undergone B-mode and CD examinations during the period October 2013 to December 2015. The masses followed up for 2 y were registered as benign. Clinical data and ultrasound images obtained from these patients were collected. The exclusion criteria were as follows: (i) simple cysts; (ii) masses ≥5 cm in diameter; (iii) masses for which vacuum-assisted biopsy had previously been performed; (iv) masses without color Doppler video images. If available, elastographic images were also collected. Because large masses cannot be displayed in the image of an entire tumor, we excluded masses that were ≥5 cm.
      The ultrasound equipment used in this study was a high-resolution system with high-frequency (>10 MHz) transducers. The initial velocity range of color Doppler was set to 3–4 cm/s. The velocity range, the color scan area and the color gain were adjusted to the vascularity. Static and video images were evaluated by a Centralized Image Interpretation Committee.

      Draft of CD diagnostic criteria for solid breast masses

      After extensively reviewing the medical literature on CD, the Flow Imaging Research Group of the JABTS devised a draft of CD diagnostic criteria for breast masses in 2011 (
      Japan Association of Breast and Thyroid Sonology (JABTS)
      Guidelines for breast ultrasound: Management and diagnosis.
      ). Before launch of the present study, the draft was modified by the Terminology and Diagnostic Criteria Committee of the JABTS. After several meetings, we reached consensus and finalized the updated draft (Table 1). Vascularity was evaluated according to a 4-point scale: 0 = avascular, 1+ = hypovascular, 2+ = moderately vascular, 3+ = hypervascular (Fig. 1a). The categorizations according to the 4-point vascular scale were made subjectively by experienced operators. Vascularity was evaluated only within masses. Avascular was defined as no blood flow signals in the mass. Hypovascular was defined as low blood flow signals in the mass. Hypervascular was defined as extremely abundant blood flow signals in the mass. Moderately vascular was defined as a state between hypovascular and hypervascular. Vascular flow patterns were evaluated taking into consideration the surrounding marginal flow and penetrating flow (Fig. 1b). When surrounding marginal flow and penetrating flow were both present, the dominant flow pattern was chosen. When the two flows were essentially equivalent, the mass was classified as “equivalent.” When both flows were very strong and were essentially equivalent, the mass was classified as “π-type.” When neither flow pattern was present, and flow signals within the mass were imperceptible, the classification was “imperceptible internal flow.” When no flow signals were observed, the mass was classified as avascular. In total, there are eight vascular flow pattern types (Table 2). When penetrating flow was present, the incident angle was also evaluated. The incident angle (
      • Kujiraoka Y.
      • Ueno E.
      • Tohno E.
      • Morishima I.
      • Tsunoda H.
      Incident angle of the penetrating artery of breast tumors.
      ) was classified into three types; near 0°, <45°, ≥45° (Fig. 1c). These angles were judged visually. Disruption of penetrating flow was defined as blood flow signals toward a mass that suddenly disappeared within a mass (Fig. 1d). Doppler spectral analysis was not included in the diagnostic criteria.
      Table 1Draft of color Doppler ultrasound diagnostic criteria for solid breast masses
      Benign←—————————————————————————————————→Malignant
      Vascularity
      0 = avascular, 1+ = hypovascular, 2+ = moderately vascular, 3+ = hypervascular.
      01+2+3+
      Vascular flow pattern
      When both penetrating flow and surrounding marginal flow exist, the vascular flow pattern exhibits the dominant flow.
      Surrounding marginal flowPenetrating flow, π-type
      Both flow patterns are very strong and are equivalent.
      Vascularity of surrounding tissueUnchangedIncrease
      Incident angle≥45°  <45°Near 0°
      Disruption of penetrating flowAbsentPresent
      low asterisk 0 = avascular, 1+ = hypovascular, 2+ = moderately vascular, 3+ = hypervascular.
      When both penetrating flow and surrounding marginal flow exist, the vascular flow pattern exhibits the dominant flow.
      Both flow patterns are very strong and are equivalent.
      Fig. 1.
      Fig. 1(a) Vascularity. (b) Vascular flow pattern. (c) Incident angle. The incident angle, estimated by visual inspection, is classified into three types: near 0°, <45°, ≥45°. (d) Disruption of penetrating flow. Blood flow signals toward a mass suddenly disappear within the mass.
      Table 2Vascular flow pattern and frequency of malignancy
      Vascular patternMalignantBenignTotalFrequency of malignancy
      Penetrating flow alone
      We defined “Penetrating flow alone,” “Penetrating flow dominant” and “π-type” collectively as “dominant penetrating flow” because these vascular patterns were associated with a high frequency of malignancy.
      489 (58.3%)91 (16.0%)58084.3%
      Penetrating flow dominant
      We defined “Penetrating flow alone,” “Penetrating flow dominant” and “π-type” collectively as “dominant penetrating flow” because these vascular patterns were associated with a high frequency of malignancy.
      146 (17.4%)57 (10.0%)20371.9%
      Surrounding marginal flow alone30 (3.6%)114 (20.0%)14420.8%
      Surrounding marginal flow dominant35 (4.2%)101 (17.8%)13625.7%
      Both flows are equivalent33 (3.9%)44 (7.7%)7742.9%
      π-Type
      We defined “Penetrating flow alone,” “Penetrating flow dominant” and “π-type” collectively as “dominant penetrating flow” because these vascular patterns were associated with a high frequency of malignancy.
      22 (2.6%)8 (1.4%)3073.3%
      Imperceptible internal flow48 (5.7%)40 (7.0%)8854.5%
      Avascular36 (4.3%)114 (20.0%)15024.0%
      Total839 (100%)569 (100%)140859.6%
      low asterisk We defined “Penetrating flow alone,” “Penetrating flow dominant” and “π-type” collectively as “dominant penetrating flow” because these vascular patterns were associated with a high frequency of malignancy.
      The judgment policies applied to our diagnostic criteria are as follows. These policies were based on the experts' opinions: (i) Greater blood flow correlated with a higher possibility of malignancy; conversely, less blood flow was associated with a higher possibility of masses being benign. (ii) Surrounding marginal flow suggested a benign mass, whereas penetrating flow indicated malignancy; imperceptible internal flow and avascular flow patterns suggested benignity. (iii) Increased vascularity of surrounding tissue was regarded as suggesting malignancy. (iv) The smaller the incident angle, the higher was the possibility of malignancy. (v) Disruption of penetrating flow suggested malignancy.

      Centralized Image Interpretation Committee

      The Centralized Image Interpretation Committee comprised 26 breast ultrasound specialists working in Japan, including 4 radiologists, 15 breast surgeons and 7 ultrasonographers. Most of these specialists are also members of the Terminology and Diagnostic Criteria Committee or the Flow Imaging Research Group of the JABTS. The 26 ultrasound specialists were divided into 13 pairs. Each pair evaluated the ultrasound images of 100–120 masses. The images were interpreted without access to medical and background information, other than age.
      The remote ultrasound image interpretation system, developed for this study by the Clinical Research Data Center of Tohoku University Hospital, was used at the site of centralized image interpretation. Employing this system, the image readers performed image evaluation on their computers at their convenience. First, B-mode images were evaluated. After evaluation of findings such as shape, margin, depth/width ratio, echogenic halo, internal echoes and posterior echoes, the B-mode category was judged. Then, color Doppler images were evaluated. After evaluation of the color Doppler images, a category combining B-mode and color Doppler (B + CD) was judged. Then, if available, elastographic images were also evaluated, and the elasticity score (
      • Itoh A.
      • Ueno E.
      • Tohno E.
      • Kamma H.
      • Takahashi H.
      • Shiina T.
      • Yamakawa M.
      • Matsumura T.
      Breast disease: Clinical application of US elastography for diagnosis.
      ) was determined. Finally, a category combining B-mode, color Doppler and elastographic findings (B + CD + E) was judged. Changes of the B-mode image judgment after seeing the color Doppler image were not allowed. Similarly, the color Doppler image judgment could not be changed after examining the elastographic image. The evaluation results were sent to the data center via the Internet. Images that the pairs of experts found difficult to judge were assessed by the entire membership of the committee.

      Japanese categories

      In Japan, we use Japanese categories: C-1 = normal, C-2 = benign, C-3a = probably benign (observation is recommended), C-3b = probably benign (biopsy is recommended), C-4 = suspicion of malignancy, C-5 = malignant (
      Japan Radiological Society/Japan Society of Radiological Technology, Mammography Guideline Committee
      Mammography guidelines.
      ,
      Japan Association of Breast and Thyroid Sonology (JABTS)
      Guidelines for breast ultrasound: Management and diagnosis.
      ,
      Terminology and Diagnostic Criteria Committee of Japan Society of Ultrasonics in Medicine
      Recall criteria for ultrasound breast cancer screening.
      ).

      Simulation experiments

      Simulation experiments were not planned in the research protocol. However, because unexpected results related to color Doppler diagnostic criteria were obtained during data analysis, we decided to conduct simulation experiments. Employing these simulation experiments, we examined changes in categories when the diagnostic criteria were modified. The reasons and methods used for the simulations are explained in detail in the Results section.

      Informed consent

      The institutional review board or the ethics committee of each hospital approved this prospective observational study. Written informed consent was not required in this trial according to the ethical guidelines for epidemiologic research in Japan (

      Ministry of Health, Labour and Welfare, Japan. Ethical guidelines for epidemiological research. Japan 2012. https://www.niph.go.jp/wadai/ekigakurinri/guidelines.pdf.

      ). There are two reasons for this. First, this trial did not use human biological specimens. Second, CD is conducted as a routine examination for breast cancer diagnosis. However, public disclosure of information obtained in this study is required by all participating hospitals. When a patient refused to allow use of their clinical data, their data were not used.

      Data collection and statistical analysis

      Data collection and statistical analyses were conducted by the Clinical Research Data Center of Tohoku University Hospital. Statistical analysis was conducted using SAS Version 9.4 (SAS Institute, Inc., Cary, NC, USA). Univariate analysis was conducted using χ2-tests. Multivariate analysis was conducted using logistic regression. The correlations between tumor vascularity and hormone receptors/proliferative index status were also examined. The differences in characteristics between the pathologically confirmed masses and the 2-y follow-up masses were also examined. For statistical analysis of sensitivity and specificity, Japanese categories 2 and 3a (observation is recommended) were considered negative, while categories 3b, 4 and 5 (biopsy is recommended) were considered positive.

      Study registration

      This study is registered with the University Hospital Medical Information Network, Japan (No. UMIN000007605).

      Results

      In total, 1578 masses from 16 institutions were registered. Of these, 14 masses were withdrawn by the institutions, 5 were excluded for being simple cysts and 72 were excluded because of inadequate pathologic examination or inadequate images. Of the remaining 1487 masses, 79 complex cystic and solid masses were excluded from this analysis because the sample number was too small. In this study, we analyzed 1408 solid masses (Fig. 2). There were 839 (59.6%) malignant and 569 (40.4%) benign masses. Among the benign masses, 393 were pathologically confirmed masses and 176 were 2-y follow-up masses. The ages of the patients with malignant (818 patients) and benign (547 patients) masses were 57.7 ± 13.7 (mean ± standard deviation, range: 25–96) and 44.8 ± 12.8 (12–87), respectively. Mean ages of the patients with pathologically confirmed benign masses and follow-up masses were 44.1 ± 12.5 (12–83) and 46.3 ± 13.4 (18–87), respectively. Mean sizes of malignant and benign masses were 1.7 ± 0.84 cm (0.4–4.8) and 1.5 ± 0.92 cm (0.3–4.8), respectively. Malignant masses included infiltrating ductal carcinoma (n = 753), ductal carcinoma in situ (n = 53), invasive lobular carcinoma (n = 29) and malignant lymphoma (n = 4). Benign masses included fibroadenoma (n = 179), fibrocystic change (n = 104), intraductal papilloma (n = 43), phyllodes tumor (n = 29), atypical hyperplasia (n = 6), inflammatory change (n = 6), lipoma (n = 4), complicated cyst (n = 3), fat necrosis (n = 2), tubular adenoma (n = 1), hamartoma (n = 1) and other (n = 16). Furthermore, 175 were masses that had exhibited no significant change for more than 2 y.
      Elastographic images were collected for 944 masses. Most (761/944) were strain elastographic images obtained using Hitachi-Aloka devices. Because the systems for performing elastography differ among companies, only elastographic images obtained using Hitachi-Aloka devices were evaluated in this study. Finally, elastographic images of 674 solid masses were analyzed (Fig. 2). There were 415 (61.6%) malignant and 259 (38.4%) benign masses. Mean ages of the patients with malignant and benign masses were 58.6 ± 12.5 and 44.3 ± 12.5, respectively.

      Dominant penetrating flow

      Table 2 outlines the relationship between the vascular flow pattern and the frequency of malignant masses. Among the eight vascular pattern types, “penetrating flow alone,” “penetrating flow dominant” and “π-type” were associated with a high frequency of malignancy. Conversely, “surrounding marginal flow alone,” “surrounding marginal flow dominant” and “avascular” were associated with a low frequency of malignancy. We defined “penetrating flow alone,” “penetrating flow dominant” and “π-type” collectively as “dominant penetrating flow.”

      Differences in characteristics between pathologically confirmed masses and 2-y follow-up masses

      Figure 3 illustrates the differences between the pathologically confirmed (biopsy) masses and those followed up for at least 2 y. The follow-up masses had milder features than the biopsied masses and had an appearance similar to that of typical fibroadenomas (oval shape, circumscribed margin and avascular/hypovascular).
      Fig. 3.
      Fig. 3Comparison of characteristics between biopsy-confirmed masses and 2-y follow-up masses.

      Color Doppler features contributing to the differential diagnosis between benign and malignant

      Univariate analysis indicated vascularity, vascular flow pattern (dominant penetrating flow vs. not), vascularity of surrounding tissue, the incident angle and disruption of penetrating flow to be statistically significant findings distinguishing benign from malignant masses (Table 3). Furthermore, findings that significantly affected malignancy by multivariate analysis were vascularity (3+), vascular flow pattern (dominant penetrating flow) and the incident angle (near 0°) (Table 4). If none of these three findings was present (n = 561), the frequency of malignancy was 28.5%. If one finding was present (n = 518), the frequency of malignancy was 73.0%; two findings (n = 280), 90.7%; and all three findings (n = 49), 95.9%. In the 561 masses with none of these three findings, shape was important for distinguishing benign from malignant masses. Of the oval masses (possibly including two or three undulations), 91% had benign histology, while 51% of non-oval masses were malignant (p < 0.001). Furthermore, the benign mass frequencies by decade for the oval mass group were 96.6% (28/29), 96.8% (61/63), 90.8% (99/109), 93.2% (41/44), 84.4% (27/32) and 72.0% (18/25) for those <30, 30–39, 40–49, 50–59, 60–69 and ≥70 y or older, respectively.
      Table 3Univariate analysis of color Doppler features and benign/malignant differential diagnosis
      Color Doppler featuresMalignant (839)Benign (569)Pearson's χ2-test
      Vascularityp < 0.001
       037 (4.4%)114 (20.0%)
       1+220 (26.2%)241 (42.4%)
       2+398 (47.4%)187 (32.9%)
       3+184 (21.9%)27 (4.7%)
      Vascular flow patternp < 0.001
       Dominant penetrating flow
      Includes “penetrating flow alone,” “penetrating flow dominant” and “π-type.”
      657 (78.3%)156 (27.4%)
       Other182 (21.7%)413 (72.6%)
      Vascularity of surrounding tissuep < 0.001
       Increased580 (69.1%)216 (38.0%)
       Unchanged259 (30.9%)353 (62.0%)
      Incident anglep < 0.001
       Near 0°186 (22.2%)15 (2.6%)
       ≤45°409 (48.7%)143 (25.1%)
       >45°55 (6.6%)113 (19.9%)
       Not determinable189 (22.5%)298 (52.4%)
      Disruption of penetrating flowp < 0.001
       Present278 (33.1%)30 (5.3%)
       Absent561 (66.9%)539 (94.7%)
      low asterisk Includes “penetrating flow alone,” “penetrating flow dominant” and “π-type.”
      Table 4Multivariate analysis of color Doppler features and benign/malignant differential diagnosis (logistic regression)
      FindingAdjusted odds ratio95% Confidence intervalp
      Vascularity
       2+ vs. 1+2.561.1065.9270.028
       3+ vs. 1+10.0223.43129.270<0.001
      Vascular flow pattern
       Dominant penetrating flow
      Includes “penetrating flow alone,” “penetrating flow dominant” and “π-type.”
      vs other
      2.8321.0777.4430.045
      Vascularity of surrounding tissue
       Increased vs. unchanged2.1480.7707.7160.144
      Incident angle
       ≤45° vs. >45°1.3540.6822.6900.387
       Near 0° vs. >45°3.8891.18712.7360.025
      Disruption of penetrating flow
       Present vs. absent1.0360.4042.6560.942
      low asterisk Includes “penetrating flow alone,” “penetrating flow dominant” and “π-type.”
      Vascularity was higher in estrogen receptor-negative (p < 0.001), progesterone receptor-negative (p < 0.001) and HER2-positive (p < 0.001) masses (Table 5). The vascularity of malignant masses increased as the Ki-67 labeling index increased (p < 0.001). Histologic type was not associated with vascularity. Tumor diameter averages increased as vascularity increased in both malignant and benign masses (p < 0.001). However, features of vascularity differed between benign and malignant masses depending on patient age (Fig. 4). Although mean patient ages did not differ with the vascularity of malignant masses, age was significantly lower as vascularity increased in those with benign masses (p < 0.001). As for patients with benign masses, their mean ages with respect to vascularity classes of 0, 1+, 2+ and 3+ were 52.0, 44.6, 41.5 and 38.5 y, respectively. Figure 5 illustrates the relationship between vascularity and patient age in benign and malignant masses. The frequency of moderately vascular or hypervascular masses increased as age decreased, but only in those with benign masses. Although the moderately vascular or hypervascular masses accounted for 21.6% of benign masses in patients ≥50 y of age, they comprised 44.0% of benign masses in patients <50 y of age (p < 0.001).
      Table 5Vascularity and factors related to breast cancer
      Vascularity
      Factor01+2+3+Totalp
      Pearson's χ2 test.
      Estrogen receptor
       Positive31 (4.5%)192 (27.9%)329 (47.9%)135 (19.7%)687 (100%)<0.001
       Negative6 (4.5%)21 (15.7%)60 (44.8%)47 (35.1%)134 (100%)
      Progesterone receptor
       Positive26 (4.3%)167 (27.4%)296 (48.6%)120 (19.7%)609 (100%)<0.001
       Negative11 (5.2%)45 (21.4%)93 (44.3%)61 (29.0%)210 (100%)
      HER2
       Positive3 (3.3%)17 (18.9%)40 (44.4%)30 (33.3%)90 (100%)<0.001
       Negative31 (4.6%)181 (27.1%)316 (47.4%)139 (20.8%)667 (100%)
      low asterisk Pearson's χ2 test.
      Fig. 4.
      Fig. 4Relationships between vascularity and tumor diameter (a) and between vascularity and age (b). CI = confidence interval.
      Fig. 5.
      Fig. 5Vascularity and age. Only benign masses exhibited increases in vascularity as age decreased, in the moderately vascular or hypervascular category.

      Sensitivity and specificity of B-mode and color Doppler

      The sensitivity and specificity for B-mode alone (B) were 97.6% and 38.3% (Table 6). The sensitivity and specificity of B + CD were 97.9% and 41.5%. The sensitivity and specificity of B-mode alone were not significantly improved by adding color Doppler. However, in patients ≥50 y of age, the specificity of B-mode alone was significantly improved by adding color Doppler (p = 0.049). The specificity of B + CD was lower in patients <50 y of age (39.8%) than in those ≥50 (45.7%). This meant that we tended to place patients <50 y in a higher category than those ≥50 y.
      Table 6Sensitivity and specificity
      Simulation experiment
      B aloneB + CDp
      Compared with B-mode alone.
      B + CD + Ep
      Compared with B-mode alone.
      Simulated B + CDp
      Compared with B-mode alone.
      Simulated B + CD + Ep
      Compared with simulated B + CD.
      Color Doppler analysis (n = 1408)
      All patients
       Sensitivity97.6% (819/839)97.9% (821/839)0.68397.7% (820/839)0.808
       Specificity38.3% (218/569)41.5% (236/569)0.06646.0% (262/569)<0.001
      ≥50 y of age
       Sensitivity98.2% (558/568)98.6% (560/568)0.564
       Specificity40.1% (65/162)45.7% (74/162)0.049
      <50 y of age
       Sensitivity96.3% (261/271)96.3% (261/271)1.00095.9% (260/271)0.655
       Specificity37.6% (153/407)39.8% (162/407)0.29946.2% (188/407)<0.001
      Elastographic analysis (n = 674)
      All patients
       Sensitivity98.8% (410/415)99.0% (411/415)0.65598.6% (409/415)0.70698.8% (410/415)1.00097.8% (406/415)0.157
       Specificity42.1% (109/259)39.0% (101/259)0.25852.9% (137/259)<0.00148.3% (125/259)0.00469.1% (179/259)<0.001
      ≥50 years of age
       Sensitivity98.6% (290/294)98.6% (290/294)1.00098.3% (289/294)0.655
       Specificity43.7% (31/71)43.7 (31/71)1.00059.2% (42/71)0.005
      <50 years of age
       Sensitivity99.2% (120/121)100% (121/121)99.2% (120/121)1.00099.2% (120/121)1.00096.7% (117/121)0.083
       Specificity41.5% (78/188)37.2% (70/188)0.21750.5% (95/188)0.01050.0% (94/188)<0.00172.9% (137/188)<0.001
      B = B-mode; B + CD = combination of B-mode and color Doppler; B + CD + E = combination of B-mode, color Doppler and elastography; ns = not significant.
      low asterisk Compared with B-mode alone.
      Compared with simulated B + CD.

      Sensitivity and specificity of a combination of B-mode, color Doppler and elastography

      For B-mode alone, the respective sensitivity and specificity were 98.8% and 42.1% in the 674 masses (Table 6). The sensitivity and specificity of B + CD were 99.0% and 39.0%, and those of B + CD + E were 98.6% and 52.9%. The specificity of B + CD + E was significantly improved relative to the sensitivity of B-mode alone (p < 0.001).

      Simulation experiments for specificity using the modified color Doppler evaluation concept

      Simulation experiments were not planned in the study protocol. However, as explained above, we found conducting such experiments to be highly informative.
      This study found the vascularity of benign masses to clearly be far greater in younger patients, whereas the vascularity of malignant masses was unrelated to age (Fig. 4b). This was an unexpected result. Our original draft for the CD diagnostic criteria did not include the concept of age. Therefore, we did not consider patient age when we evaluated color Doppler images. Because we believed that abundant vascularity was a finding suggestive of malignancy, many benign masses in younger patients were judged as raising the suspicion of malignancy in this study. This tendency for overestimation may have reduced the specificity of B + CD in patients <50 y of age. We realized that if this observation were found to be accurate, our diagnostic criteria would have to be modified. To assess whether the specificity of B + CD improves when the vascularity evaluation takes age into consideration, we designed and conducted simulation experiments.

      Color Doppler analysis (n = 1408)

      We considered the following study results when we set the conditions for the simulation experiments. In patients <50 y of age, vascularity 2+ did not, whereas vascularity 3+ did, suggest malignancy. More than 90% of masses with two or all three of the findings (vascularity 3+, dominant penetrating flow and incident angle near 0°) were malignant. Furthermore, in patients <40 y of age, 97% of oval masses with none of these three findings were benign. Finally, we employed the following two conditions for color Doppler analysis.
      • Condition 1. In patients <50 y of age, if vascularity is 3+, or if both “dominant penetrating flow” and “incident angle near 0°” are confirmed, the B-mode category is upgraded by one level (e.g., from C-3a to C-3b, maximum: C-5).
      • Condition 2. In patients <40 y of age and with oval masses, if “vascularity 3+,” “dominant penetrating flow” and “incident angle near 0°” are all confirmed to be absent, the B-mode category is downgraded by one level (minimum: C-2).
      By employment of these conditions and CD findings, the B-mode category was mechanically converted into a “simulated B + CD category.” Then, sensitivity and specificity were calculated. In patients ≥50 y of age, B + CD categories as judged by the Centralized Image Interpretation Committee were used without modification. Specificity in patients <50 y of age improved with color Doppler, from 37.6% to 46.2% (p < 0.001, Table 6). In all patients, the specificity of B-mode was also significantly improved by color Doppler (from 38.3% to 46.0%, p < 0.001).

      Elastographic analysis (n = 674)

      Similarly, we conducted simulation experiments focusing on elastographic findings. The B-mode category was mechanically converted to a “simulated B + CD category” using conditions 1 and 2. We employed conditions 3 and 4 to convert the “simulated B + CD category” into the “simulated B + CD + E category” using the elasticity score. In patients 50 y of age and older, B + CD + E categories were used without modification. The sensitivity and specificity for each were then calculated.
      • Condition 3. If the elasticity score is 1 or 2, the “modified B + C category” is downgraded by one level (minimum: C-2).
      • Condition 4. If the elasticity score is 4 or 5, the “modified B + C category” is upgraded by one level (maximum: C-5).
      The specificities of B-mode alone, simulated B + CD and simulated B + CD + E in patients <50 y of age were 41.5%, 50.5% and 72.9%, respectively. The specificity of B-mode alone was significantly improved by color Doppler (p < 0.001). Furthermore, the specificity of B + CD was significantly improved by elastography (p < 0.001). The specificities of B-mode, simulated B + CD and simulated B + CD + E in all 674 patients were 42.1%, 48.3% and 69.1%, respectively. The specificity of B-mode alone was significantly improved by color Doppler (p = 0.004). Furthermore, the specificity of B + CD was also significantly improved by elastography (p < 0.001).

      Improved diagnostic criteria

      Based on the results of the simulation experiments, we updated and revised our CD diagnostic criteria (Table 7).
      Table 7Revised draft of color Doppler ultrasound diagnostic criteria for solid breast masses
      In patients <40 y of age and with oval masses, if “vascularity 3+,” “dominant penetrating flow” and “incident angle near 0°” are all absent, the probability of a mass being benign is very high.
      Benign←——————————————————————————————→Malignant
      Vascularity
      0 = avascular, 1+= hypovascular, 2+ = moderately vascular, 3+ = hypervascular.
       ≥50 y of age01+2+3+
       <50 y of age01+2+3+
      Vascular flow patternSurrounding marginal flow (dominant or alone)Penetrating flow (dominant or alone)


      Incident angle


      ≥45°
      <45°π-type
      Penetrating flow and surrounding marginal flow were both very strong and were equivalent.


      near 0°
      Disruption of penetrating flowAbsentPresent
      low asterisk In patients <40 y of age and with oval masses, if “vascularity 3+,” “dominant penetrating flow” and “incident angle near 0°” are all absent, the probability of a mass being benign is very high.
      0 = avascular, 1+= hypovascular, 2+ = moderately vascular, 3+ = hypervascular.
      Penetrating flow and surrounding marginal flow were both very strong and were equivalent.

      Discussion

      Numerous studies have analyzed the usefulness of CD in breast masses. The early devices used had poor performance. The first Doppler feature used to differentiate between benign and malignant masses was the detection of vascularization in the lesion.
      • Cosgrove D.O.
      • Kedar R.P.
      • Bamber J.C.
      • al-Murrani B.
      • Davey J.B.
      • Fisher C.
      • McKinna J.A.
      • Svensson W.E.
      • Tohno E.
      • Vagios E.
      Breast diseases: Color Doppler US in differential diagnosis.
      reported that 99% of malignant masses contained blood vessels, whereas only 3% of benign masses had Doppler signals. However, technological advances have allowed blood flow to be recognized even in benign lesions (
      • Birdwell R.L.
      • Ikeda D.M.
      • Jeffrey S.S.
      • Jeffrey Jr, R.B.
      Preliminary experience with power Doppler imaging of solid breast masses.
      ). The second Doppler feature used to differentiate between benign and malignant masses was the distribution of vascularization in the lesion.
      • Raza S.
      • Baum J.K.
      Solid breast lesions: Evaluation with power Doppler US.
      reported the presence of penetrating vessels to be a diagnostic criterion for malignancy. On the other hand, the presence of peripheral vessels was found to be a criterion allowing a mass to be diagnosed as benign. In this way, the diagnostic criteria have changed with advancements in the CD devices.
      Because of the recent technological innovations in ultrasound devices, both penetrating vessels and peripheral vessels are now visible in many breast masses. For this reason, we employed the “dominant flow pattern” concept in this study. When both penetrating flow and peripheral flow were visible, the vascular patterns were divided into three types; “penetrating flow dominant,” “the two flow types equivalent” and “surrounding marginal flow dominant.” In this study, the frequencies of malignancy for these types were 71.9%, 42.9% and 20.8%, respectively (Table 2). As for malignant masses, 75.7% exhibited the “penetrating flow alone” or “penetrating flow dominant” pattern. On the other hand, 57.8% of benign masses had the “surrounding marginal flow alone,” “surrounding marginal flow dominant” or “avascular” pattern.
      In this study, we employed a 4-point vascular scale. However, we did not know whether this 4-point scale or a 3-point scale would be better. The difference between the 4-point scale and the 3-point scale is whether or not “moderately vascular” is included. Because the vascularity scale requires a subjective judgment, the 4-point scale might be difficult to use. We decided that for this study, if there were no difference between moderately vascular and hypervascular, we would employ the 3-point scale. However, the results revealed differences between moderately and hypervascular lesions (Fig. 4). Therefore, we decided to continue using the 4-point scale.
      One of the problems with this study is the patient composition. Many patients with benign masses visit the outpatient breast clinic. However, the number of benign masses was small, and the number of typical fibroadenomas registered was also small in this study. In many institutions, biopsy is not conducted on typical benign masses, and a 2-y follow-up period for such lesions would be unlikely in Japan. However, many typical benign masses can be diagnosed only using B-mode, with CD and elastography not playing particularly large diagnostic roles. In this sense, benign masses requiring registration in this study were of the atypical type. Therefore, the quality of the study results might be almost the same as if typical benign masses had been included.
      Multivariate analysis revealed vascular flow pattern, in addition to vascularity and the incident angle, to be a significant finding allowing benign masses to be distinguished from malignant masses. The incidence angle, as reported by
      • Kujiraoka Y.
      • Ueno E.
      • Tohno E.
      • Morishima I.
      • Tsunoda H.
      Incident angle of the penetrating artery of breast tumors.
      in 2005, is a relatively new concept in the field of CD. Kujiraoka et al. reported the incident angle of breast cancer to be significantly lower than that of benign masses. They accurately measured the angle using a protractor. However, we visually assessed the incident angle because the purpose of our criteria is ease of application in daily clinical examinations. We selected 0° and 45° as standards that can easily be judged visually. Although only 2.6% of benign masses had an "incident angle near 0°," 22.2% of malignant masses met the "incident angle near 0°" criterion.
      According to the results of the present study, patient age is an important factor that should be taken into consideration when evaluating the vascularity of masses. The vascularity of benign masses in younger women was higher than that in older women, while this feature was not observed in malignant masses regardless of age (Fig. 4, Fig. 5). Because our prior knowledge and experience suggested hypervascularity to be associated with malignancy, regardless of age, many benign masses in young women were miscategorized as ≥3a. However, the simulation experiments revealed that taking age into consideration would reduce such overestimation. It is very important when using CD, with the aim of making an accurate differential diagnosis of benign versus malignant, to recognize this difference in vascularity between the breast masses of younger and older women. Therefore, the relationship between vascularity and age should be incorporated into the color Doppler diagnostic criteria.
      As noted above, the relationship between vascularity and age surprised us, because this relationship had not been described in recent reports. However, Cosgrove et al. made the same observation in 1993. After their study, to our knowledge, the relationship between vascularity and age was not examined again in subsequent small-scale investigations (
      • Birdwell R.L.
      • Ikeda D.M.
      • Jeffrey S.S.
      • Jeffrey Jr, R.B.
      Preliminary experience with power Doppler imaging of solid breast masses.
      ,
      • Chao T.C.
      • Lo Y.F.
      • Chen S.C.
      • Chen M.F.
      Color Doppler ultrasound in benign and malignant breast tumors.
      ,
      • Chatterjee S.
      • Ray A.K.
      • Karim R.
      • Biswas A.
      Color Doppler to characterize malignant breast lesion.
      ,
      • Cho N.
      • Jang M.
      • Lyou C.Y.
      • Park J.S.
      • Choi H.Y.
      • Moon W.K.
      Distinguishing benign from malignant masses at breast US: Combined US elastography and color Doppler US—Influence on radiologist accuracy.
      ,
      • Gupta K.
      • Chandra T.
      • Kumaresan M.
      • Venkatesan B.
      • Patil B.A.
      Role of color Doppler for assessment of malignancy in solid breast masses: A prospective study.
      ,
      • Kook S.H.
      • Park H.W.
      • Lee Y.R.
      • Lee Y.U.
      • Pac W.K.
      • Park Y.L.
      Evaluation of solid breast lesions with power Doppler sonography.
      ,
      • Lee S.W.
      • Choi H.Y.
      • Baek S.Y.
      • Lim S.M.
      Role of color and power Doppler imaging in differentiating between malignant and benign solid breast masses.
      ,
      • Ozdemir A.
      • Ozdemir H.
      • Maral I.
      • Konuş O.
      • Yücel S.
      • Işik S.
      Differential diagnosis of solid breast lesions: Contribution of Doppler studies to mammography and gray scale imaging.
      ,
      • Raza S.
      • Baum J.K.
      Solid breast lesions: Evaluation with power Doppler US.
      ,
      • Svensson W.E.
      • Pandian A.J.
      • Hashimoto H.
      The use of breast ultrasound color Doppler vascular pattern morphology improves diagnostic sensitivity with minimal change in specificity.
      ). Therefore, the relationship between blood flow and age was overlooked for more than 20 y. We regard this relationship as likely being one of the factors explaining the inconsistencies in results among studies conducted to date. The relationship between vascularity and age might be affected by the hormonal environment. We can reasonably hypothesize that the vascularity of benign breast masses is affected by estrogen, whereas that of malignant breast masses is not.
      A combination of B-mode and elastography has been documented to have higher diagnostic accuracy than either modality alone (
      • Cho N.
      • Jang M.
      • Lyou C.Y.
      • Park J.S.
      • Choi H.Y.
      • Moon W.K.
      Distinguishing benign from malignant masses at breast US: Combined US elastography and color Doppler US—Influence on radiologist accuracy.
      ,
      • Parajuly S.S.
      • Lan P.Y.
      • Yan L.
      • Gang Y.Z.
      • Lin L.
      Breast elastography: A hospital-based preliminary study in China.
      ,
      • Sadigh G.
      • Carlos R.C.
      • Neal C.H.
      • Dwamena B.A.
      Ultrasonographic differentiation of malignant from benign breast lesions: A meta-analytic comparison of elasticity and BIRADS scoring.
      ,
      • Sadigh G.
      • Carlos R.C.
      • Neal C.H.
      • Dwamena B.A.
      Accuracy of quantitative ultrasound elastography for differentiation of malignant and benign breast abnormalities: A meta-analysis.
      ,
      • Stoian D.
      • Timar B.
      • Craina M.
      • Bernad E.
      • Petre I.
      • Craciunescu M.
      Qualitative strain elastography–strain ratio evaluation—An important tool in breast cancer diagnostic.
      ). In our research as well, diagnostic performance was improved by elastography. However, our goal was to ascertain whether we need to add color Doppler to elastography. If diagnostic performance does not improve even when color Doppler is added to elastography, color Doppler is unnecessary. The results of the simulation experiments were particularly interesting. When the specificity of B + CD increased (all patients: 39.0%→48.3%), the specificity of B + CD + E had an even greater increase (52.9%→69.1%). This indicates that if the diagnostic performance of color Doppler increases, the diagnostic performance of the combination of color Doppler and elastography also increases and to a greater degree. This suggests that color Doppler and elastography are independent parameters. Therefore, adding color Doppler to elastography is clinically meaningful. In this study, we failed to show the usefulness of the color Doppler diagnostic criteria we had initially prepared. However, we made important observations; the relationship between vascularity and age, and the relationship between color Doppler results and elastography. On the basis of these observations, we revised our color Doppler diagnostic criteria. Now, we are preparing a new study to confirm the improved color Doppler diagnostic criteria.
      Regarding the improved diagnostic criteria (Table 7), although “disruption of penetrating flow” was not a statistically significant finding by multivariate analysis, we decided to retain this criterion. This is because masses with this finding were evaluated as avascular. Avascular is a pattern that suggests a mass is benign. However, most of these masses are malignant. Because it is important to distinguish masses with “disruption of penetrating flow” from other avascular masses, we also retained this finding as one of the criteria.
      A limitation of this study is that we relied, in part, on results obtained in simulation experiments. It is necessary to conduct further research to confirm the usefulness of the improved diagnostic criteria. We launched a new study in April 2018, entitled Multicenter Study to Confirm the Usefulness of the Color Doppler Ultrasound Diagnostic Criteria for Breast Masses. This study was named CD-CONFIRM study, and registered with the University Hospital Medical Information Network, Japan (No. UMIN000032298).

      Conclusions

      Color Doppler was found to be useful for distinguishing between benign and malignant masses. With the dominant penetrating flow pattern especially, vascularity and the incident angle were significant and useful findings. Adding CD did not improve the diagnostic performance of B-mode alone in this study. However, if we include age among the color Doppler diagnostic criteria, the diagnostic performance of B-mode appears probably to be improved by adding color Doppler. Furthermore, our results suggested that color Doppler and elastography are independent parameters. If this is true, we need to make optimal use of color Doppler as well as elastography.

      Acknowledgments

      We thank all of the patients who consented to the use of their ultrasound images for this study. We also thank all participating investigators and their support staff who contributed to this study. Finally, we extend very special thanks to Hiroko Yaegashi for developing the remote ultrasound image interpretation system.

      Conflict of interest disclosure

      The authors have no conflicts of interest to declare.

      References

        • American College of Radiology
        ACR BI-RADS atlas: Breast imaging reporting and data system.
        5th edition. Author, Reston, VA2013
        • Birdwell R.L.
        • Ikeda D.M.
        • Jeffrey S.S.
        • Jeffrey Jr, R.B.
        Preliminary experience with power Doppler imaging of solid breast masses.
        AJR Am J Roentgenol. 1997; 169: 703-707
        • Chao T.C.
        • Lo Y.F.
        • Chen S.C.
        • Chen M.F.
        Color Doppler ultrasound in benign and malignant breast tumors.
        Breast Cancer Res Treat. 1999; 57: 193-199
        • Chatterjee S.
        • Ray A.K.
        • Karim R.
        • Biswas A.
        Color Doppler to characterize malignant breast lesion.
        International J Life Sci Med Res. 2013; 2: 24-27
        • Cho N.
        • Jang M.
        • Lyou C.Y.
        • Park J.S.
        • Choi H.Y.
        • Moon W.K.
        Distinguishing benign from malignant masses at breast US: Combined US elastography and color Doppler US—Influence on radiologist accuracy.
        Radiology. 2012; 262: 80-90
        • Cosgrove D.O.
        • Kedar R.P.
        • Bamber J.C.
        • al-Murrani B.
        • Davey J.B.
        • Fisher C.
        • McKinna J.A.
        • Svensson W.E.
        • Tohno E.
        • Vagios E.
        Breast diseases: Color Doppler US in differential diagnosis.
        Radiology. 1993; 189: 99-104
        • Gupta K.
        • Chandra T.
        • Kumaresan M.
        • Venkatesan B.
        • Patil B.A.
        Role of color Doppler for assessment of malignancy in solid breast masses: A prospective study.
        Int J Anat Radiol Surg. 2017; 6: RO59-RO65
        • Horvath D.E.
        • Cuitino M.J.
        • Pinochet M.A.
        • Sanhueza P.
        Color Doppler in the study of the breast: How do we perform it?.
        Rev Chil Radiol. 2011; 17: 1-9
        • Itoh A.
        • Ueno E.
        • Tohno E.
        • Kamma H.
        • Takahashi H.
        • Shiina T.
        • Yamakawa M.
        • Matsumura T.
        Breast disease: Clinical application of US elastography for diagnosis.
        Radiology. 2006; 239: 341-350
        • Japan Association of Breast and Thyroid Sonology (JABTS)
        Guidelines for breast ultrasound: Management and diagnosis.
        Nankodo, Tokyo2004 ([In Japanese])
        • Japan Association of Breast and Thyroid Sonology (JABTS)
        Guidelines for breast ultrasound: Management and diagnosis.
        3rd edition. Nankodo, Tokyo2014 ([In Japanese])
        • Japan Radiological Society/Japan Society of Radiological Technology, Mammography Guideline Committee
        Mammography guidelines.
        Igaku Shoin, Tokyo1999 ([In Japanese])
        • Kook S.H.
        • Park H.W.
        • Lee Y.R.
        • Lee Y.U.
        • Pac W.K.
        • Park Y.L.
        Evaluation of solid breast lesions with power Doppler sonography.
        J Clin Ultrasound. 1999; 27: 231-237
        • Kujiraoka Y.
        • Ueno E.
        • Tohno E.
        • Morishima I.
        • Tsunoda H.
        Incident angle of the penetrating artery of breast tumors.
        in: Ueno E. Shiina T. Kubota M. Sawai K. Research and development in breast ultrasound. Springer-Verlag, Tokyo2005: 72-75
        • Lee S.W.
        • Choi H.Y.
        • Baek S.Y.
        • Lim S.M.
        Role of color and power Doppler imaging in differentiating between malignant and benign solid breast masses.
        J Clin Ultrasound. 2002; 30: 459-464
      1. Ministry of Health, Labour and Welfare, Japan. Ethical guidelines for epidemiological research. Japan 2012. https://www.niph.go.jp/wadai/ekigakurinri/guidelines.pdf.

        • Ozdemir A.
        • Ozdemir H.
        • Maral I.
        • Konuş O.
        • Yücel S.
        • Işik S.
        Differential diagnosis of solid breast lesions: Contribution of Doppler studies to mammography and gray scale imaging.
        J Ultrasound Med. 2001; 20: 1091-1101
        • Parajuly S.S.
        • Lan P.Y.
        • Yan L.
        • Gang Y.Z.
        • Lin L.
        Breast elastography: A hospital-based preliminary study in China.
        Asian Pac J Cancer Prev. 2010; 11: 809-814
        • Raza S.
        • Baum J.K.
        Solid breast lesions: Evaluation with power Doppler US.
        Radiology. 1997; 203: 164-168
        • Sadigh G.
        • Carlos R.C.
        • Neal C.H.
        • Dwamena B.A.
        Ultrasonographic differentiation of malignant from benign breast lesions: A meta-analytic comparison of elasticity and BIRADS scoring.
        Breast Cancer Res Treat. 2012; 133: 23-35
        • Sadigh G.
        • Carlos R.C.
        • Neal C.H.
        • Dwamena B.A.
        Accuracy of quantitative ultrasound elastography for differentiation of malignant and benign breast abnormalities: A meta-analysis.
        Breast Cancer Res Treat. 2012; 134: 923-931
        • Sehgal C.M.
        • Arger P.H.
        • Rowling S.E.
        Quantitative vascularity of breast masses by Doppler imaging: Regional variations and diagnostic implications.
        J Ultrasound Med. 2000; 19: 427-440
        • Stoian D.
        • Timar B.
        • Craina M.
        • Bernad E.
        • Petre I.
        • Craciunescu M.
        Qualitative strain elastography–strain ratio evaluation—An important tool in breast cancer diagnostic.
        Med Ultrason. 2016; 18: 195-200
        • Svensson W.E.
        • Pandian A.J.
        • Hashimoto H.
        The use of breast ultrasound color Doppler vascular pattern morphology improves diagnostic sensitivity with minimal change in specificity.
        Ultraschall Med. 2010; 31: 466-474
        • Terminology and Diagnostic Criteria Committee of Japan Society of Ultrasonics in Medicine
        Recall criteria for ultrasound breast cancer screening.
        J Med Ultrason. 2016; 43: 301-313
        • Wilkens T.H.
        • Burke B.J.
        • Cancelada D.A.
        • Jatoi I.
        Evaluation of palpable breast masses with color Doppler sonography and gray-scale imaging.
        J Ultrasound Med. 1998; 17: 109-115