- Email: [email protected]
Three-dimensional Doppler Indices for Breast Masses: Systematic Review of Diagnostic Test Accuracy
Journal of Medical Ultrasound (2013) 21, 126e131
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.jmu-online.com
REVIEW ARTICLE
Three-dimensional Doppler Indices for Breast Masses: Systematic Review of Diagnostic Test Accuracy Guilherme L. Martinez 1, Carolina O. Nastri 1,2, ´ H. Miyague 1,2, Wellington P. Martins 1,2* Andre 1
Department of Obstetrics and Gynecology, Medical School of Ribeira˜o Preto, University of Sa˜o Paulo (DGO-FMRP-USP), and 2 Ultrasonography and Retraining Medical School of Ribeira˜o Preto (EURP), Ribeira˜o Preto, Brazil Received 24 January 2013; accepted 8 April 2013 Available online 14 September 2013
KEY WORDS breast neoplasms, Doppler, imaging, review, three-dimensional, ultrasonography
Our objective was to assess the diagnostic accuracy of three-dimensional power Doppler (3DPD) indices (vascularization index, flow index, vascularization flow index) in distinguishing between benign and malignant breast masses by performing a systematic review of the current literature. The following databases were searched from their inception until October 18, 2012: Cochrane Central Register of Controlled Trials, Database of Abstracts of Reviews of Effects, Medical Literature Analysis and Retrieval System Online, Embase, PsycINFO, Cumulative Index to Nursing and Allied Health Literature, and Literatura Latino-Americana e do Caribe em Cie ˆncias da Sau ´de. Studies were considered eligible if they met the following criteria: (1) they evaluated the diagnostic accuracy of at least one of the 3D Doppler indices to differentiate benign from malignant breast masses; (2) cytology or/and histology was used as the reference standard to determine malignancy; (3) the number of true positives, false positives, true negatives, and false negatives could be retrieved. This search resulted in only one eligible study. This study evaluated a relatively large sample size, and we did not have any concern regarding applicability or risk of bias when evaluating its quality. The results show that adding 3DPD indices only slightly improved the diagnostic test accuracy in distinguishing benign and malign breast masses in women with suspicious breast masses requiring biopsy. We concluded that although 3DPD indices seem to be able to help in distinguishing between benign and malignant breast masses, there is still insufficient evidence to suggest using this method in clinical
* Correspondence to: Dr Wellington P. Martins, Department of Obstetrics and Gynecology, Medical School of Ribeira ˜o Preto, University of Sa ˜o Paulo (DGO-FMRP-USP), Avenida dos Bandeirantes, 3900, 8 andar. Ribeira ˜o Preto, Sa ˜o Paulo CEP 14049-900, Brazil. E-mail address: [email protected] (W.P. Martins). 0929-6441/$36 ª 2013, Elsevier Taiwan LLC and the Chinese Taipei Society of Ultrasound in Medicine. All rights reserved. http://dx.doi.org/10.1016/j.jmu.2013.07.007
3-D Doppler Indices for Breast Masses
127
practice. More good-quality studies evaluating the diagnostic accuracy of this method are still required. ª 2013, Elsevier Taiwan LLC and the Chinese Taipei Society of Ultrasound in Medicine. All rights reserved.
Introduction
the value of the incorporation of 3DPD indices during evaluation of breast masses.
Description of condition Objectives Breast cancer is the most frequently diagnosed cancer in women, and the leading cause of cancer death in females worldwide and the second leading cause of death in our society [1]. It accounts for 23% (1.38 million) of the total new cancer cases [1]. Diagnosis in early stages is associated with better survival rates, and the diagnosis of preclinical cases may be beneficial [2]. Breast mass is an early finding in breast cancer patients [3]; however, approximately 90% of the masses are benign tumors, such as fibroadenomas and cysts [4]. Risk factors for malignancy include age, history of breast cancer, and breast cancer in first-degree relatives [4]. In this context, noninvasive tools that accurately separate benign from malignant masses could potentially prevent unnecessary biopsies.
Description of the intervention and how it might work Three-dimensional power Doppler ultrasonography (3DPD) allows the examination and quantification of blood flow within corporal structures. This information can then be quantified to derive three indices of vascularity: the vascularization index (VI), the flow index (FI), and the vascularization flow index (VFI). VI is the ratio of color voxels to all voxels in a given volume of interest; FI is the mean value of all color voxels in the analyzed volume; and VFI is a combination of the two, calculated by multiplying the values together and dividing the result by 100 [5]. Several studies have been published about these indices in the past decade, especially in Ob-Gyn [6], and some authors have demonstrated that these 3DPD indices clearly showed a correlation with real blood flow in experimental models [7e9]. Because angiogenesis plays a fundamental role in the genesis and tumor growth [10], it is plausible that Doppler indices of 3D ultrasound can help discriminate between benign and malignant breast masses. However, several studies reported some limitations of these indices, particularly the dependency of machine settings and attenuation [11e14].
Why it is important to conduct this review Although some studies have already assessed the 3DPD indices in breast masses, there is still no recommendation for its use in clinical practice. The lack of consensus about its clinical usefulness highlights the need for a systematic review and meta-analysis of the diagnostic accuracy of this method in order to obtain more robust evidence about of
We conducted this review to assess the diagnostic accuracy of 3DPD indices (VI, FI, VFI) in distinguishing between benign and malignant breast masses.
Methods Eligibility criteria Studies were considered eligible if they met the following criteria: (1) they evaluated the diagnostic accuracy of at least one of the 3D Doppler indices to differentiate benign from malignant breast masses; (2) cytology or/and histology was used as the reference standard to determine malignancy; (3) the number of true positives, false positives, true negatives, and false negatives could be retrieved by reading the manuscript or contacting the study’s corresponding author.
Information sources and search We searched the following electronic databases: Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE), Medical Literature Analysis and Retrieval System Online (MEDLINE), Embase, PsycINFO; Cumulative Index to Nursing and Allied Health Literature (CINAHL); and Literatura Latino-Americana e do Caribe em Cie ˆncias da Sau ´de (LILACS). In addition, we handsearched the reference list of the included articles. We did not impose any publication date or language limits for the searches. We used the following terms adjusting for each database when necessary: [(breast) or (mammary)] and (Doppler) and [(three-dimensional) or (three dimensional) or (3D) or (3 D) or (3-D)].
Study selection Two independent reviewers (G.L.M. and A.H.M.) scanned the retrieved titles and abstracts, selecting those that clearly do not meet the eligibility criteria; disagreements between reviewers were resolved by consulting a third author (W.P.M.), who also obtained a full copy of the articles. The full articles were then examined for eligibility independently by two reviewers (G.L.M. and A.H.M.), and any disagreements between reviewers were resolved by consulting a third author (WPM).
128
G.L. Martinez et al.
Data collection process
Evaluation of heterogeneity, sensitivity analysis, and publication bias
Data were extracted independently by two authors (G.L.M. and W.P.M.) using a data extraction form designed and pilottested by the authors. Disagreements between reviewers were resolved by consulting a third author (C.O.N.). The names of article authors and titles of the included studies were juxtaposed to seek for duplicate publication. Where articles were related to the same participantsdtotally or partiallydthey were considered as a single study. In such cases, data were extracted from the main article and additional details were derived from secondary articles. We extracted the following items from the included studies: author; publication year; country; number of participants; mean age of participants; number of benign and malignant breast tumors; number of true positives, false positives, true negatives, false negatives.
Quality assessment The quality of the included study was evaluated by the revised tool for the quality assessment of diagnostic accuracy studies (QUADAS-2) [15].
Data synthesis Data was combined using the hierarchical summary receiver operator characteristics method and the bivariate normal random-effects analysis of sensitivity and specificity within the Metandi module in the STATA 11 statistical software (Statacorp LP, College Station, TX, USA). The Metandi plot would be used for graphical representation. However, only one study was included, and the data could not be pooled. The results of the included study were then presented.
Fig. 1
Heterogeneity, sensitivity analysis, and risk of publication bias could not be assessed because only one study was included.
Results Study selection The electronic search was last performed on October 18, 2012, and a total of 662 records were retrieved. We excluded 644 records on the basis of title and abstract: 49 were duplicates and 595 clearly did not meet the eligibility criteria. We obtained 18 full-text articles, from which 12 were excluded: 10 did not report VI, FI, or VFI [16e25], and two full-text articles were related to studies evaluating only confirmed breast cancer tumors [26,27]. Six full-text articles referred to studies that would met our inclusion criteria [28e33]; however, these six full-text articles included an overlapping population of women. This was confirmed by e-mail contact, and these six full-text articles were then considered as a single study (Fig. 1). The article with the largest sample size was considered the main report [33].
Characteristics of the included study The study included in this review was conducted in Taiwan (single center). Two hundred patients were consecutively enrolled in this study during a 13-month period (January 2003 to February 2004). All women had a scheduled biopsydfine
Study flow diagram.
3-D Doppler Indices for Breast Masses
129
needle aspiration cytology, core biopsy, and/or excisional biopsydfor suspicious mass. The result of this biopsy was used as the standard test to discriminate benign and malignant masses. Five patients with a breast mass larger than 3.0 cm were excluded because the authors could not include the tumor plus the 3-mm surrounding tissue in only one Power Doppler 3D dataset. A total of 102 participants with benign breast tumors and 93 with malignant breast tumors were analyzed. All participants were scanned using the Voluson730 ultrasound system (GE Healthcare, Zipf, Austria), which was equipped with a 6 MHz to 12 MHz 3D/4D transducer. Power Doppler static 3D datasets were acquired with 20 sweep angles using fixed preinstalled power Doppler settings: midfrequency, pulse repetition frequency Z 0.9 kHz; gain Z e0.6; wall motion filter Z “low 1”. The probe was held still, and the women were asked to hold their breath for about 20 s, if possible, as the scanner generated the 3D volume. Acquired 3D volumes were stored for subsequent offline analysis. The tumor volume was contoured using VOCAL in the manual mode and a rotation step of 30 : VI, FI, and VFI were estimated from the tumor and also from an outer shell of 3 mm of the surrounding tissue; the latter was not considered in this review. The median age of the participants was 40 years (range, 17e72 years) and 49 years (range, 31e80 years) for benign and malignant breast tumors, respectively. The median tumor diameter was 14.0 mm (range, 5e30 mm) and 18.8 mm (range, 5e30 mm), and the median tumor volume was 0.57 cm3 (range, 0.03e9.09 cm3) and 3.13 cm3 (range, 0.29e16.32 cm3), for benign and malignant breast tumors, respectively.
Accuracy of VI, FI, and VFI for distinguishing between malignant and benign masses
0.32 (range, 0e6.12) for benign and malignant tumors, respectively. The areas under the receiver operating characteristic curves (AUROC) [with its 95% confidence interval (CI)], sensitivity, specificity, accuracy, number of true positives, true negatives, false positives, and false negatives for age, mass diameter, and volume, and for the 3DPD indices are reported in Table 1. All evaluated parameters had some value in differentiating between benign and malignant masses as the 95% CI of the AUROC curves did not overlap 0.50. Individually, mass volume was slightly better than the other parameters: AUROC Z 0.87 (95% CI, 0.83e0.90). However, better performance was achieved when some of these parameters were added up. The multivariate logistic regression analysis showed that adding patient age and breast volume increased the AUROC to 0.91 (95% CI, 0.89e0.93). When adding FI to this model (age þ mass volume þ FI from tumor), a relevant improvement was noticed in terms of sensitivity (72.0% vs. 82.8%, without vs. with FI, respectively) and accuracy (79.5% vs. 85.1%); however, there was still an overlap in the 95% CI of the AUROC (Table 1).
Risk of bias (QUADAS 2) The included study was considered at low risk of bias related to patient selection, index test, reference standard, and flow and timing. No concern regarding applicability on patient selection, index test, or reference standard was identified; in addition, the risk of bias of all four domains was judged to be low. The characteristics of the included study and all QUADAS-2 judgments are presented in Table 2 [33].
Discussion
The median VI was 0.11 (range, 0e6.27) and 1.07 (range, 0e15.45) for benign and malignant tumors, respectively. The median FI was 21.21 (range, 0e38.17) and 29.07 (range, 0e44.05), and the median VFI was 0.03 (range, 0e1.84) and
Summary of main results Only one study could be included in this systematic review. This study evaluated a relatively large sample size, and we
Table 1 Univariate and multivariate logistic regression analyses of the studied parameters in the differentiation of benign and malignant breast masses.
Univariate models Age Breast tumor volume VI from breast tumor FI from breast tumor VFI from breast tumor Multivariate models Age þ Volume Age þ Volume þ FI
AUROC
95% CI
Sens. (%)
Spec. (%)
Accur. (%)
TP
TN
FP
FN
0.79 0.87 0.78 0.82 0.79
0.73e0.84 0.83e0.90 0.72e0.83 0.77e0.86 0.73e0.84
66.7 67.7 51.6 73.1 52.7
77.5 89.2 90.2 72.5 90.2
72.3 79.0 71.8 72.8 72.3
62 63 48 68 49
79 91 92 74 92
23 11 10 28 10
31 30 45 25 44
0.91 0.93
0.89e0.93 0.90e0.94
72.0 82.8
86.3 87.3
79.5 85.1
67 77
88 89
14 13
26 16
95% CI Z 95% confidence interval; Accur. Z accuracy; AUROC Z area under the receiver operating characteristic curve; FN Z false negative; FP Z false positive; Sens. Z sensitivity; Spec. Z specificity; TN Z true negative; TP Z true positive; VI, FI, VFI Z threedimensional power Doppler indices vascularization index, flow index, and vascularization-flow index; Volume Z three-dimensional ultrasound measurement of the mass using manual contouring.
Note. The data in row ‘VFI from breast tumor’ are from “Intra-tumor flow index can predict the malignant potential of breast tumor: dependent on age and volume” by Y-H Hsia, S-J Kuo, W-M Liang, Huang YL, Chen DR., 2008, Ultrasound Med Biol, 34, p. 88e95. Copyright 2007, World Federation for Ultrasound in Medicine & Biology. Adapted with permission.
130
G.L. Martinez et al.
Table 2
QUADAS-2 risk of bias and applicability judgments for the included study.
Study: Hsiao et al [33] Patient selection Description
Risk of bias Concerns regarding applicability Index test Description
Risk of bias Concerns regarding applicability Reference standard Description Risk of bias Concerns regarding applicability Flow and timing Description Risk of bias
A total of 200 women were recruited consecutively during a 13-mo period (January 2003eFebruary 2004). All had been scheduled to undergo excisional or percutaneous needle biopsy on the basis of suspicious mammographic or US findings. Excluded if tumor >3 cm, because it did not fit in one 3D volume (5 women). Low Low Ultrasound setting maintained the same for all scans. The patients had a biopsy scheduled; the reference standard text had not been performed at that point. No threshold was applied. Low Low Women underwent fine needle aspiration cytology, core biopsy and/or excisional biopsy. Low Low Five of 200 women were excluded because of the impossibility of performing the index test. Timing between US and biopsy was not described. Low
did not have any concern regarding applicability or risk of bias when evaluating its quality. However, this included study shows that adding 3DPD indices only slightly improved the diagnostic test accuracy in distinguishing benign and malign breast masses in women with suspicious breast masses requiring biopsy.
Limitations Although we performed an extensive search, only one study could be included in this review. The results obtained are therefore limited by a potential lack of external validity to other healthcare centers. It means that these results are limited by many factors such as the characteristics of the population studied, experience of the observers, and ultrasound machine settings. However, the malignancy rate observed by this study is close to the expected level: 47.7% of the tumors were malignant, which is comparable to the proportion of malignancies (33%) observed in women undergoing breast biopsy after screening mammography [34]. One should still consider the limitations of 3DPD indices, particularly the need for special ultrasound machines and training [6], and the high dependency of machine settings and attenuation [35,36]. New 3DPD indices and standardizations that at least theoretically overcome these limitations are currently being investigated [12,13,37,38]; hopefully, in a few years we will have studies evaluating their clinical utility in breast masses.
Conclusions Although 3DPD indices seem to be able to help in distinguishing between benign and malignant breast masses, there is still
insufficient evidence to suggest using this method in clinical practice. More good quality studies evaluating the diagnostic accuracy of this method are still required.
Acknowledgments The authors acknowledge the following financial support: PhD scholarship from Coordenac ¸˜ ao de Aperfeic ¸oamento de Pessoal de Nı´vel Superior (CAPES), Brazil (AHM), (CON); postdoctoral fellowship from the Conselho Nacional de Desenvolvimento cientı´fico e Tecnolo ´gico (CNPq), Brazil; and salary from the Hospital das Clı´nicas de Ribeira ˜o Preto, Brazil (WPM).
References [1] Jemal A, Bray F, Center MM, et al. Global cancer statistics. CA Cancer J Clin 2011;61:69e90. [2] Jemal A, Center MM, DeSantis C, et al. Global patterns of cancer incidence and mortality rates and trends. Cancer Epidemiol Biomarkers Prev 2010;19:1893e907. [3] Elmore JG, Barton MB, Moceri VM, et al. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med 1998;338:1089e96. [4] Morrow M. The evaluation of common breast problems. Am Fam Physician 2000;61:2371e8. 2385. [5] Pairleitner H, Steiner H, Hasenoehrl G, et al. Three-dimensional power Doppler sonography: imaging and quantifying blood flow and vascularization. Ultrasound Obstet Gynecol 1999;14:139e43. [6] Martins WP. Three-dimensional power Doppler: validity and reliability. Ultrasound Obstet Gynecol 2010;36:530e3. [7] Morel O, Pachy F, Chavatte-Palmer P, et al. Correlation between uteroplacental three-dimensional power Doppler
3-D Doppler Indices for Breast Masses
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
indices and true uterine blood flow: evaluation in a pregnant sheep model. Ultrasound Obstet Gynecol 2010; 36:635e40. Raine-Fenning NJ, Nordin NM, Ramnarine KV, et al. Determining the relationship between three-dimensional power Doppler data and true blood flow characteristics: an in-vitro flow phantom experiment. Ultrasound Obstet Gynecol 2008; 32:540e50. Jones NW, Hutchinson ES, Brownbill P, et al. In vitro dual perfusion of human placental lobules as a flow phantom to investigate the relationship between fetoplacental flow and quantitative 3D power Doppler angiography. Placenta 2009; 30:130e5. Folkman J, Watson K, Ingber D, et al. Induction of angiogenesis during the transition from hyperplasia to neoplasia. Nature 1989;339:58e61. Martins WP, Raine-Fenning NJ, Ferriani RA, et al. Quantitative three-dimensional power Doppler angiography: a flow-free phantom experiment to evaluate the relationship between color gain, depth and signal artifact. Ultrasound Obstet Gynecol 2010;35:361e8. Martins WP, Welsh AW, Lima JC, et al. The "volumetric" pulsatility index as evaluated by spatiotemporal imaging correlation (STIC): a preliminary description of a novel technique, its application to the endometrium and an evaluation of its reproducibility. Ultrasound Med Biol 2011;37:2160e8. Welsh AW, Hou M, Meriki N, Martins WP. Spatiotemporal image correlation-derived volumetric Doppler impedance indices from spherical samples of the placenta: intraobserver reliability and correlation with conventional umbilical artery Doppler indices. Ultrasound Obstet Gynecol 2012;40:431e6. Raine-Fenning NJ, Nordin NM, Ramnarine KV, et al. Evaluation of the effect of machine settings on quantitative threedimensional power Doppler angiography: an in-vitro flow phantom experiment. Ultrasound Obstet Gynecol 2008;32: 551e9. 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;155:529e36. LeCarpentier GL, Roubidoux MA, Fowlkes JB, et al. Suspicious breast lesions: assessment of 3D Doppler US indexes for classification in a test population and fourfold cross-validation scheme. Radiology 2008;249:463e70. Huang S-F, Chang R-F, Moon WK, et al. Analysis of tumor vascularity using three-dimensional power Doppler ultrasound images. IEEE Trans Med Imaging 2008;27:320e30. Forsberg F, Goldberg BB, Merritt CRB, et al. Diagnosing breast lesions with contrast-enhanced 3-dimensional power Doppler imaging. J Ultrasound Med 2004;23:173e82. Chang R-F, Huang S-F, Moon WK, et al. Solid breast masses: neural network analysis of vascular features at threedimensional power Doppler US for benign or malignant classification. Radiology 2007;243:56e62. Chang R-F, Huang S-F, Moon WK, et al. Computer algorithm for analysing breast tumor angiogenesis using 3-D power Doppler ultrasound. Ultrasound Med Biol 2006;32:1499e508. Carson PL, Fowlkes JB, Roubidoux MA, et al. 3-D color Doppler image quantification of breast masses. Ultrasound Med Biol 1998;24:945e52.
131 [22] Carson PL, Moskalik AP, Govil A, et al. The 3D and 2D color flow display of breast masses. Ultrasound Med Biol 1997;23: 837e49. [23] Abbattista T, Serri L, Busilacchi P. Three-dimensional sonographic study of breast nodules. J Ultrasound 2007;10:93e8. [24] Yen PL, Wu HK, Tseng HS, et al. Vascular morphologic information of three-dimensional power Doppler ultrasound is valuable in the classification of breast lesions. Clin Imaging 2012;36:267e71. [25] Chang YC, Huang YH, Huang CS, et al. Vascular morphology and tortuosity analysis of breast tumor inside and outside contour by 3-d power Doppler ultrasound. Ultrasound Med Biol 2012;38:1859e69. [26] Chen ST, Kuo SJ, Wu HK, et al. Power Doppler breast ultrasound: association of vascularization and ER/c-erbB-2 co-expression in invasive breast carcinoma. Breast Cancer 2011:1e7. [27] Chen DR, Lien IP, Huang YL. Ultrasonic vascularization evaluation in breast cancer with neoadjuvant chemotherapy. Int J Comp Assist. Radiol Surg 2011;6:S320e1. [28] Hsiao Y-H, Huang Y-L, Liang W-M, et al. Characterization of benign and malignant solid breast masses: harmonic versus nonharmonic 3D power Doppler imaging. Ultrasound Med Biol 2009;35:353e9. [29] Hsiao Y-H, Huang Y-L, Kuo S-J, et al. Characterization of benign and malignant solid breast masses in harmonic 3D power Doppler imaging. Eur J Radiol 2009;71:89e95. [30] Huang Y-L, Kuo S-J, Hsu C-C, et al. Computer-aided diagnosis for breast tumors by using vascularization of 3-D power Doppler ultrasound. Ultrasound Med Biol 2009;35:1607e14. [31] Chen S-T, Hsiao Y-H, Huang Y-L, et al. Comparative analysis of logistic regression, support vector machine and artificial neural network for the differential diagnosis of benign and malignant solid breast tumors by the use of three-dimensional power Doppler imaging. Korean J Radiol 2009;10:464e71. [32] Kuo SJ, Hsiao YH, Huang YL, et al. Classification of benign and malignant breast tumors using neural networks and threedimensional power Doppler ultrasound. Ultrasound Obstet Gynecol 2008;32:97e102. [33] Hsiao Y-H, Kuo S-J, Liang W-M, et al. Intra-tumor flow index can predict the malignant potential of breast tumor: dependent on age and volume. Ultrasound Med Biol 2008;34:88e95. [34] Weaver DL, Rosenberg RD, Barlow WE, et al. Pathologic findings from the Breast Cancer Surveillance Consortium: population-based outcomes in women undergoing biopsy after screening mammography. Cancer 2006;106:732e42. [35] Martins WP, Lima JC, Welsh AW, et al. Three-dimensional Doppler evaluation of single spherical samples from the placenta: intra- and interobserver reliability. Ultrasound Obstet Gynecol 2012;40:200e6. [36] Martins WP, Nastri CO. Reproducibility of 3D power Doppler placental vascular indices. Arch Gynecol Obstet 2011;283: 403e4. [37] Soares CA, Miyague AH, Mauad Filho F, et al. The fractional moving blood volume (FMBV) concept can be applied to 3D power Doppler quantification. Ultrasound Obstet Gynecol 2013;41:98e9. [38] Martins WP, Welsh AW, Falkensammer P, et al. Re: spatiotemporal imaging correlation (STIC): technical notes about STIC triggering and choosing between power Doppler or highdefinition color flow. Ultrasound Med Biol 2013;39:549e50.
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.jmu-online.com
REVIEW ARTICLE
Three-dimensional Doppler Indices for Breast Masses: Systematic Review of Diagnostic Test Accuracy Guilherme L. Martinez 1, Carolina O. Nastri 1,2, ´ H. Miyague 1,2, Wellington P. Martins 1,2* Andre 1
Department of Obstetrics and Gynecology, Medical School of Ribeira˜o Preto, University of Sa˜o Paulo (DGO-FMRP-USP), and 2 Ultrasonography and Retraining Medical School of Ribeira˜o Preto (EURP), Ribeira˜o Preto, Brazil Received 24 January 2013; accepted 8 April 2013 Available online 14 September 2013
KEY WORDS breast neoplasms, Doppler, imaging, review, three-dimensional, ultrasonography
Our objective was to assess the diagnostic accuracy of three-dimensional power Doppler (3DPD) indices (vascularization index, flow index, vascularization flow index) in distinguishing between benign and malignant breast masses by performing a systematic review of the current literature. The following databases were searched from their inception until October 18, 2012: Cochrane Central Register of Controlled Trials, Database of Abstracts of Reviews of Effects, Medical Literature Analysis and Retrieval System Online, Embase, PsycINFO, Cumulative Index to Nursing and Allied Health Literature, and Literatura Latino-Americana e do Caribe em Cie ˆncias da Sau ´de. Studies were considered eligible if they met the following criteria: (1) they evaluated the diagnostic accuracy of at least one of the 3D Doppler indices to differentiate benign from malignant breast masses; (2) cytology or/and histology was used as the reference standard to determine malignancy; (3) the number of true positives, false positives, true negatives, and false negatives could be retrieved. This search resulted in only one eligible study. This study evaluated a relatively large sample size, and we did not have any concern regarding applicability or risk of bias when evaluating its quality. The results show that adding 3DPD indices only slightly improved the diagnostic test accuracy in distinguishing benign and malign breast masses in women with suspicious breast masses requiring biopsy. We concluded that although 3DPD indices seem to be able to help in distinguishing between benign and malignant breast masses, there is still insufficient evidence to suggest using this method in clinical
* Correspondence to: Dr Wellington P. Martins, Department of Obstetrics and Gynecology, Medical School of Ribeira ˜o Preto, University of Sa ˜o Paulo (DGO-FMRP-USP), Avenida dos Bandeirantes, 3900, 8 andar. Ribeira ˜o Preto, Sa ˜o Paulo CEP 14049-900, Brazil. E-mail address: [email protected] (W.P. Martins). 0929-6441/$36 ª 2013, Elsevier Taiwan LLC and the Chinese Taipei Society of Ultrasound in Medicine. All rights reserved. http://dx.doi.org/10.1016/j.jmu.2013.07.007
3-D Doppler Indices for Breast Masses
127
practice. More good-quality studies evaluating the diagnostic accuracy of this method are still required. ª 2013, Elsevier Taiwan LLC and the Chinese Taipei Society of Ultrasound in Medicine. All rights reserved.
Introduction
the value of the incorporation of 3DPD indices during evaluation of breast masses.
Description of condition Objectives Breast cancer is the most frequently diagnosed cancer in women, and the leading cause of cancer death in females worldwide and the second leading cause of death in our society [1]. It accounts for 23% (1.38 million) of the total new cancer cases [1]. Diagnosis in early stages is associated with better survival rates, and the diagnosis of preclinical cases may be beneficial [2]. Breast mass is an early finding in breast cancer patients [3]; however, approximately 90% of the masses are benign tumors, such as fibroadenomas and cysts [4]. Risk factors for malignancy include age, history of breast cancer, and breast cancer in first-degree relatives [4]. In this context, noninvasive tools that accurately separate benign from malignant masses could potentially prevent unnecessary biopsies.
Description of the intervention and how it might work Three-dimensional power Doppler ultrasonography (3DPD) allows the examination and quantification of blood flow within corporal structures. This information can then be quantified to derive three indices of vascularity: the vascularization index (VI), the flow index (FI), and the vascularization flow index (VFI). VI is the ratio of color voxels to all voxels in a given volume of interest; FI is the mean value of all color voxels in the analyzed volume; and VFI is a combination of the two, calculated by multiplying the values together and dividing the result by 100 [5]. Several studies have been published about these indices in the past decade, especially in Ob-Gyn [6], and some authors have demonstrated that these 3DPD indices clearly showed a correlation with real blood flow in experimental models [7e9]. Because angiogenesis plays a fundamental role in the genesis and tumor growth [10], it is plausible that Doppler indices of 3D ultrasound can help discriminate between benign and malignant breast masses. However, several studies reported some limitations of these indices, particularly the dependency of machine settings and attenuation [11e14].
Why it is important to conduct this review Although some studies have already assessed the 3DPD indices in breast masses, there is still no recommendation for its use in clinical practice. The lack of consensus about its clinical usefulness highlights the need for a systematic review and meta-analysis of the diagnostic accuracy of this method in order to obtain more robust evidence about of
We conducted this review to assess the diagnostic accuracy of 3DPD indices (VI, FI, VFI) in distinguishing between benign and malignant breast masses.
Methods Eligibility criteria Studies were considered eligible if they met the following criteria: (1) they evaluated the diagnostic accuracy of at least one of the 3D Doppler indices to differentiate benign from malignant breast masses; (2) cytology or/and histology was used as the reference standard to determine malignancy; (3) the number of true positives, false positives, true negatives, and false negatives could be retrieved by reading the manuscript or contacting the study’s corresponding author.
Information sources and search We searched the following electronic databases: Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE), Medical Literature Analysis and Retrieval System Online (MEDLINE), Embase, PsycINFO; Cumulative Index to Nursing and Allied Health Literature (CINAHL); and Literatura Latino-Americana e do Caribe em Cie ˆncias da Sau ´de (LILACS). In addition, we handsearched the reference list of the included articles. We did not impose any publication date or language limits for the searches. We used the following terms adjusting for each database when necessary: [(breast) or (mammary)] and (Doppler) and [(three-dimensional) or (three dimensional) or (3D) or (3 D) or (3-D)].
Study selection Two independent reviewers (G.L.M. and A.H.M.) scanned the retrieved titles and abstracts, selecting those that clearly do not meet the eligibility criteria; disagreements between reviewers were resolved by consulting a third author (W.P.M.), who also obtained a full copy of the articles. The full articles were then examined for eligibility independently by two reviewers (G.L.M. and A.H.M.), and any disagreements between reviewers were resolved by consulting a third author (WPM).
128
G.L. Martinez et al.
Data collection process
Evaluation of heterogeneity, sensitivity analysis, and publication bias
Data were extracted independently by two authors (G.L.M. and W.P.M.) using a data extraction form designed and pilottested by the authors. Disagreements between reviewers were resolved by consulting a third author (C.O.N.). The names of article authors and titles of the included studies were juxtaposed to seek for duplicate publication. Where articles were related to the same participantsdtotally or partiallydthey were considered as a single study. In such cases, data were extracted from the main article and additional details were derived from secondary articles. We extracted the following items from the included studies: author; publication year; country; number of participants; mean age of participants; number of benign and malignant breast tumors; number of true positives, false positives, true negatives, false negatives.
Quality assessment The quality of the included study was evaluated by the revised tool for the quality assessment of diagnostic accuracy studies (QUADAS-2) [15].
Data synthesis Data was combined using the hierarchical summary receiver operator characteristics method and the bivariate normal random-effects analysis of sensitivity and specificity within the Metandi module in the STATA 11 statistical software (Statacorp LP, College Station, TX, USA). The Metandi plot would be used for graphical representation. However, only one study was included, and the data could not be pooled. The results of the included study were then presented.
Fig. 1
Heterogeneity, sensitivity analysis, and risk of publication bias could not be assessed because only one study was included.
Results Study selection The electronic search was last performed on October 18, 2012, and a total of 662 records were retrieved. We excluded 644 records on the basis of title and abstract: 49 were duplicates and 595 clearly did not meet the eligibility criteria. We obtained 18 full-text articles, from which 12 were excluded: 10 did not report VI, FI, or VFI [16e25], and two full-text articles were related to studies evaluating only confirmed breast cancer tumors [26,27]. Six full-text articles referred to studies that would met our inclusion criteria [28e33]; however, these six full-text articles included an overlapping population of women. This was confirmed by e-mail contact, and these six full-text articles were then considered as a single study (Fig. 1). The article with the largest sample size was considered the main report [33].
Characteristics of the included study The study included in this review was conducted in Taiwan (single center). Two hundred patients were consecutively enrolled in this study during a 13-month period (January 2003 to February 2004). All women had a scheduled biopsydfine
Study flow diagram.
3-D Doppler Indices for Breast Masses
129
needle aspiration cytology, core biopsy, and/or excisional biopsydfor suspicious mass. The result of this biopsy was used as the standard test to discriminate benign and malignant masses. Five patients with a breast mass larger than 3.0 cm were excluded because the authors could not include the tumor plus the 3-mm surrounding tissue in only one Power Doppler 3D dataset. A total of 102 participants with benign breast tumors and 93 with malignant breast tumors were analyzed. All participants were scanned using the Voluson730 ultrasound system (GE Healthcare, Zipf, Austria), which was equipped with a 6 MHz to 12 MHz 3D/4D transducer. Power Doppler static 3D datasets were acquired with 20 sweep angles using fixed preinstalled power Doppler settings: midfrequency, pulse repetition frequency Z 0.9 kHz; gain Z e0.6; wall motion filter Z “low 1”. The probe was held still, and the women were asked to hold their breath for about 20 s, if possible, as the scanner generated the 3D volume. Acquired 3D volumes were stored for subsequent offline analysis. The tumor volume was contoured using VOCAL in the manual mode and a rotation step of 30 : VI, FI, and VFI were estimated from the tumor and also from an outer shell of 3 mm of the surrounding tissue; the latter was not considered in this review. The median age of the participants was 40 years (range, 17e72 years) and 49 years (range, 31e80 years) for benign and malignant breast tumors, respectively. The median tumor diameter was 14.0 mm (range, 5e30 mm) and 18.8 mm (range, 5e30 mm), and the median tumor volume was 0.57 cm3 (range, 0.03e9.09 cm3) and 3.13 cm3 (range, 0.29e16.32 cm3), for benign and malignant breast tumors, respectively.
Accuracy of VI, FI, and VFI for distinguishing between malignant and benign masses
0.32 (range, 0e6.12) for benign and malignant tumors, respectively. The areas under the receiver operating characteristic curves (AUROC) [with its 95% confidence interval (CI)], sensitivity, specificity, accuracy, number of true positives, true negatives, false positives, and false negatives for age, mass diameter, and volume, and for the 3DPD indices are reported in Table 1. All evaluated parameters had some value in differentiating between benign and malignant masses as the 95% CI of the AUROC curves did not overlap 0.50. Individually, mass volume was slightly better than the other parameters: AUROC Z 0.87 (95% CI, 0.83e0.90). However, better performance was achieved when some of these parameters were added up. The multivariate logistic regression analysis showed that adding patient age and breast volume increased the AUROC to 0.91 (95% CI, 0.89e0.93). When adding FI to this model (age þ mass volume þ FI from tumor), a relevant improvement was noticed in terms of sensitivity (72.0% vs. 82.8%, without vs. with FI, respectively) and accuracy (79.5% vs. 85.1%); however, there was still an overlap in the 95% CI of the AUROC (Table 1).
Risk of bias (QUADAS 2) The included study was considered at low risk of bias related to patient selection, index test, reference standard, and flow and timing. No concern regarding applicability on patient selection, index test, or reference standard was identified; in addition, the risk of bias of all four domains was judged to be low. The characteristics of the included study and all QUADAS-2 judgments are presented in Table 2 [33].
Discussion
The median VI was 0.11 (range, 0e6.27) and 1.07 (range, 0e15.45) for benign and malignant tumors, respectively. The median FI was 21.21 (range, 0e38.17) and 29.07 (range, 0e44.05), and the median VFI was 0.03 (range, 0e1.84) and
Summary of main results Only one study could be included in this systematic review. This study evaluated a relatively large sample size, and we
Table 1 Univariate and multivariate logistic regression analyses of the studied parameters in the differentiation of benign and malignant breast masses.
Univariate models Age Breast tumor volume VI from breast tumor FI from breast tumor VFI from breast tumor Multivariate models Age þ Volume Age þ Volume þ FI
AUROC
95% CI
Sens. (%)
Spec. (%)
Accur. (%)
TP
TN
FP
FN
0.79 0.87 0.78 0.82 0.79
0.73e0.84 0.83e0.90 0.72e0.83 0.77e0.86 0.73e0.84
66.7 67.7 51.6 73.1 52.7
77.5 89.2 90.2 72.5 90.2
72.3 79.0 71.8 72.8 72.3
62 63 48 68 49
79 91 92 74 92
23 11 10 28 10
31 30 45 25 44
0.91 0.93
0.89e0.93 0.90e0.94
72.0 82.8
86.3 87.3
79.5 85.1
67 77
88 89
14 13
26 16
95% CI Z 95% confidence interval; Accur. Z accuracy; AUROC Z area under the receiver operating characteristic curve; FN Z false negative; FP Z false positive; Sens. Z sensitivity; Spec. Z specificity; TN Z true negative; TP Z true positive; VI, FI, VFI Z threedimensional power Doppler indices vascularization index, flow index, and vascularization-flow index; Volume Z three-dimensional ultrasound measurement of the mass using manual contouring.
Note. The data in row ‘VFI from breast tumor’ are from “Intra-tumor flow index can predict the malignant potential of breast tumor: dependent on age and volume” by Y-H Hsia, S-J Kuo, W-M Liang, Huang YL, Chen DR., 2008, Ultrasound Med Biol, 34, p. 88e95. Copyright 2007, World Federation for Ultrasound in Medicine & Biology. Adapted with permission.
130
G.L. Martinez et al.
Table 2
QUADAS-2 risk of bias and applicability judgments for the included study.
Study: Hsiao et al [33] Patient selection Description
Risk of bias Concerns regarding applicability Index test Description
Risk of bias Concerns regarding applicability Reference standard Description Risk of bias Concerns regarding applicability Flow and timing Description Risk of bias
A total of 200 women were recruited consecutively during a 13-mo period (January 2003eFebruary 2004). All had been scheduled to undergo excisional or percutaneous needle biopsy on the basis of suspicious mammographic or US findings. Excluded if tumor >3 cm, because it did not fit in one 3D volume (5 women). Low Low Ultrasound setting maintained the same for all scans. The patients had a biopsy scheduled; the reference standard text had not been performed at that point. No threshold was applied. Low Low Women underwent fine needle aspiration cytology, core biopsy and/or excisional biopsy. Low Low Five of 200 women were excluded because of the impossibility of performing the index test. Timing between US and biopsy was not described. Low
did not have any concern regarding applicability or risk of bias when evaluating its quality. However, this included study shows that adding 3DPD indices only slightly improved the diagnostic test accuracy in distinguishing benign and malign breast masses in women with suspicious breast masses requiring biopsy.
Limitations Although we performed an extensive search, only one study could be included in this review. The results obtained are therefore limited by a potential lack of external validity to other healthcare centers. It means that these results are limited by many factors such as the characteristics of the population studied, experience of the observers, and ultrasound machine settings. However, the malignancy rate observed by this study is close to the expected level: 47.7% of the tumors were malignant, which is comparable to the proportion of malignancies (33%) observed in women undergoing breast biopsy after screening mammography [34]. One should still consider the limitations of 3DPD indices, particularly the need for special ultrasound machines and training [6], and the high dependency of machine settings and attenuation [35,36]. New 3DPD indices and standardizations that at least theoretically overcome these limitations are currently being investigated [12,13,37,38]; hopefully, in a few years we will have studies evaluating their clinical utility in breast masses.
Conclusions Although 3DPD indices seem to be able to help in distinguishing between benign and malignant breast masses, there is still
insufficient evidence to suggest using this method in clinical practice. More good quality studies evaluating the diagnostic accuracy of this method are still required.
Acknowledgments The authors acknowledge the following financial support: PhD scholarship from Coordenac ¸˜ ao de Aperfeic ¸oamento de Pessoal de Nı´vel Superior (CAPES), Brazil (AHM), (CON); postdoctoral fellowship from the Conselho Nacional de Desenvolvimento cientı´fico e Tecnolo ´gico (CNPq), Brazil; and salary from the Hospital das Clı´nicas de Ribeira ˜o Preto, Brazil (WPM).
References [1] Jemal A, Bray F, Center MM, et al. Global cancer statistics. CA Cancer J Clin 2011;61:69e90. [2] Jemal A, Center MM, DeSantis C, et al. Global patterns of cancer incidence and mortality rates and trends. Cancer Epidemiol Biomarkers Prev 2010;19:1893e907. [3] Elmore JG, Barton MB, Moceri VM, et al. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med 1998;338:1089e96. [4] Morrow M. The evaluation of common breast problems. Am Fam Physician 2000;61:2371e8. 2385. [5] Pairleitner H, Steiner H, Hasenoehrl G, et al. Three-dimensional power Doppler sonography: imaging and quantifying blood flow and vascularization. Ultrasound Obstet Gynecol 1999;14:139e43. [6] Martins WP. Three-dimensional power Doppler: validity and reliability. Ultrasound Obstet Gynecol 2010;36:530e3. [7] Morel O, Pachy F, Chavatte-Palmer P, et al. Correlation between uteroplacental three-dimensional power Doppler
3-D Doppler Indices for Breast Masses
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
indices and true uterine blood flow: evaluation in a pregnant sheep model. Ultrasound Obstet Gynecol 2010; 36:635e40. Raine-Fenning NJ, Nordin NM, Ramnarine KV, et al. Determining the relationship between three-dimensional power Doppler data and true blood flow characteristics: an in-vitro flow phantom experiment. Ultrasound Obstet Gynecol 2008; 32:540e50. Jones NW, Hutchinson ES, Brownbill P, et al. In vitro dual perfusion of human placental lobules as a flow phantom to investigate the relationship between fetoplacental flow and quantitative 3D power Doppler angiography. Placenta 2009; 30:130e5. Folkman J, Watson K, Ingber D, et al. Induction of angiogenesis during the transition from hyperplasia to neoplasia. Nature 1989;339:58e61. Martins WP, Raine-Fenning NJ, Ferriani RA, et al. Quantitative three-dimensional power Doppler angiography: a flow-free phantom experiment to evaluate the relationship between color gain, depth and signal artifact. Ultrasound Obstet Gynecol 2010;35:361e8. Martins WP, Welsh AW, Lima JC, et al. The "volumetric" pulsatility index as evaluated by spatiotemporal imaging correlation (STIC): a preliminary description of a novel technique, its application to the endometrium and an evaluation of its reproducibility. Ultrasound Med Biol 2011;37:2160e8. Welsh AW, Hou M, Meriki N, Martins WP. Spatiotemporal image correlation-derived volumetric Doppler impedance indices from spherical samples of the placenta: intraobserver reliability and correlation with conventional umbilical artery Doppler indices. Ultrasound Obstet Gynecol 2012;40:431e6. Raine-Fenning NJ, Nordin NM, Ramnarine KV, et al. Evaluation of the effect of machine settings on quantitative threedimensional power Doppler angiography: an in-vitro flow phantom experiment. Ultrasound Obstet Gynecol 2008;32: 551e9. 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;155:529e36. LeCarpentier GL, Roubidoux MA, Fowlkes JB, et al. Suspicious breast lesions: assessment of 3D Doppler US indexes for classification in a test population and fourfold cross-validation scheme. Radiology 2008;249:463e70. Huang S-F, Chang R-F, Moon WK, et al. Analysis of tumor vascularity using three-dimensional power Doppler ultrasound images. IEEE Trans Med Imaging 2008;27:320e30. Forsberg F, Goldberg BB, Merritt CRB, et al. Diagnosing breast lesions with contrast-enhanced 3-dimensional power Doppler imaging. J Ultrasound Med 2004;23:173e82. Chang R-F, Huang S-F, Moon WK, et al. Solid breast masses: neural network analysis of vascular features at threedimensional power Doppler US for benign or malignant classification. Radiology 2007;243:56e62. Chang R-F, Huang S-F, Moon WK, et al. Computer algorithm for analysing breast tumor angiogenesis using 3-D power Doppler ultrasound. Ultrasound Med Biol 2006;32:1499e508. Carson PL, Fowlkes JB, Roubidoux MA, et al. 3-D color Doppler image quantification of breast masses. Ultrasound Med Biol 1998;24:945e52.
131 [22] Carson PL, Moskalik AP, Govil A, et al. The 3D and 2D color flow display of breast masses. Ultrasound Med Biol 1997;23: 837e49. [23] Abbattista T, Serri L, Busilacchi P. Three-dimensional sonographic study of breast nodules. J Ultrasound 2007;10:93e8. [24] Yen PL, Wu HK, Tseng HS, et al. Vascular morphologic information of three-dimensional power Doppler ultrasound is valuable in the classification of breast lesions. Clin Imaging 2012;36:267e71. [25] Chang YC, Huang YH, Huang CS, et al. Vascular morphology and tortuosity analysis of breast tumor inside and outside contour by 3-d power Doppler ultrasound. Ultrasound Med Biol 2012;38:1859e69. [26] Chen ST, Kuo SJ, Wu HK, et al. Power Doppler breast ultrasound: association of vascularization and ER/c-erbB-2 co-expression in invasive breast carcinoma. Breast Cancer 2011:1e7. [27] Chen DR, Lien IP, Huang YL. Ultrasonic vascularization evaluation in breast cancer with neoadjuvant chemotherapy. Int J Comp Assist. Radiol Surg 2011;6:S320e1. [28] Hsiao Y-H, Huang Y-L, Liang W-M, et al. Characterization of benign and malignant solid breast masses: harmonic versus nonharmonic 3D power Doppler imaging. Ultrasound Med Biol 2009;35:353e9. [29] Hsiao Y-H, Huang Y-L, Kuo S-J, et al. Characterization of benign and malignant solid breast masses in harmonic 3D power Doppler imaging. Eur J Radiol 2009;71:89e95. [30] Huang Y-L, Kuo S-J, Hsu C-C, et al. Computer-aided diagnosis for breast tumors by using vascularization of 3-D power Doppler ultrasound. Ultrasound Med Biol 2009;35:1607e14. [31] Chen S-T, Hsiao Y-H, Huang Y-L, et al. Comparative analysis of logistic regression, support vector machine and artificial neural network for the differential diagnosis of benign and malignant solid breast tumors by the use of three-dimensional power Doppler imaging. Korean J Radiol 2009;10:464e71. [32] Kuo SJ, Hsiao YH, Huang YL, et al. Classification of benign and malignant breast tumors using neural networks and threedimensional power Doppler ultrasound. Ultrasound Obstet Gynecol 2008;32:97e102. [33] Hsiao Y-H, Kuo S-J, Liang W-M, et al. Intra-tumor flow index can predict the malignant potential of breast tumor: dependent on age and volume. Ultrasound Med Biol 2008;34:88e95. [34] Weaver DL, Rosenberg RD, Barlow WE, et al. Pathologic findings from the Breast Cancer Surveillance Consortium: population-based outcomes in women undergoing biopsy after screening mammography. Cancer 2006;106:732e42. [35] Martins WP, Lima JC, Welsh AW, et al. Three-dimensional Doppler evaluation of single spherical samples from the placenta: intra- and interobserver reliability. Ultrasound Obstet Gynecol 2012;40:200e6. [36] Martins WP, Nastri CO. Reproducibility of 3D power Doppler placental vascular indices. Arch Gynecol Obstet 2011;283: 403e4. [37] Soares CA, Miyague AH, Mauad Filho F, et al. The fractional moving blood volume (FMBV) concept can be applied to 3D power Doppler quantification. Ultrasound Obstet Gynecol 2013;41:98e9. [38] Martins WP, Welsh AW, Falkensammer P, et al. Re: spatiotemporal imaging correlation (STIC): technical notes about STIC triggering and choosing between power Doppler or highdefinition color flow. Ultrasound Med Biol 2013;39:549e50.