| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Breast Imaging |
1 From the Departments of Radiology (G.S., M.V.), Pathology (X.F., A.C., P.L.F.), and Obstetrics and Gynecology (J.A.V.), Hospital Clínic and University of Barcelona Medical School, Villarroel 170, 08036 Barcelona, Spain. Received August 6, 2003; revision requested October 22; final revision received March 24, 2004; accepted May 12. Supported by grants from the Spanish Health Ministry (FIS 00/0923 and FIS 01/1519). Address correspondence to G.S. (e-mail: gsanta@clinic.ub.es).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Ethics committee approval and informed consent were obtained. A total of 97 invasive breast carcinomas were prospectively studied with unenhanced power Doppler sonography before surgery. Lumpectomy or mastectomy with full axillary nodal dissection was performed. Sonographic tumor size and number of tumor arteries were correlated with axillary nodal status by means of logistic regression analysis. Tumor microvascularization was immunohistochemically assessed in a subset of 55 carcinomas. Sonographic variables were correlated with tumor arteries with a diameter larger than 300 µm and with the density and area of microvascularization. The 
statistic and Bland-Altman agreement limits were used to measure agreement between techniques.
RESULTS: Good agreement of sonographic and histologic findings regarding number of tumor arteries (= 0.66, P < .001) and tumor size (P = .012) was observed. Multivariate analysis showed an independent relationship between probability of axillary metastasis, number of tumor arteries (P = .016), and sonographic tumor size (P = .035). A predictive model of axillary status was developed. The receiver operating characteristic curve was used to determine 0.2324 as the score to classify axillary nodal status. This score indicated high sensitivity (96.1%), low specificity (53.0%), and high negative predictive value (96.1%).
CONCLUSION: The number of arteries in invasive breast carcinoma detected with unenhanced power Doppler sonography and sonographic tumor size are independent predictors of axillary nodal status; these variables could contribute to reliable prediction of absence of axillary involvement on the basis of a mathematic model.
© RSNA, 2004
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Axillary nodal status is one of the most important prognostic factors for survival in patients with breast cancer (10); therefore, axillary lymphadenectomy must be performed to determine the stage of breast cancer. However, since the prevalence of axillary involvement in breast cancer varies between 30% and 35%, most excisions are unnecessary. This has led to increased interest in the use of alternative methods, such as sentinel node biopsy (11–13) or axillary sonography (14,15), in the prediction of axillary status. Some studies based on the sonographic study of primary tumor vascularization have shown that axillary involvement and cell proliferation are more frequent in vascular tumors (16) and that there seems to be a correlation between tumor vascularization and axillary node involvement (17).
On the basis of prior results and the potential of power Doppler sonography in the study of vascular flow, our purpose was to prospectively compare unenhanced power Doppler sonographic findings with histopathologic and immunohistochemical findings and to determine whether tumor arterial vascularization facilitates the prediction of axillary node status.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The mean age of the patients was 60.9 years (range, 30–88 years), and 72 (76%) of 95 were postmenopausal.
Mammography and Interpretation
Mammography was performed with a Senographe DMR imager (GE Medical Systems, Milwaukee, Wis). Breast cancer was suspected on the basis of both mammographic and sonographic findings. Mammographic images were interpreted independently by two radiologists (M.V., G.S.) with 15 and 10 years of experience with breast imaging and sonography. When mammography was performed, the physicians knew the clinical information and the results of the physical examination of the patients. The lesions were assessed with the Breast Imaging Reporting and Data System; 22 had a score of 4, and 75 had a score of 5. The mammographic appearance of the lesions was described, and the location was evaluated.
Sonography
Sonography was performed by using a PowerVision SSA-380 system (Toshiba, Tokyo, Japan) with a 7.5-MHz linear transducer and a 5-MHz Doppler frequency. All sonographic imaging was performed by one of the radiologists (G.S.) who participated in image interpretation. The sonographic examinations were recorded with digital video.
The radiologist measured the lesion in centimeters, including the echogenic rim around the lesion, when present. All lesions were measured in the transverse and sagittal planes, and anteroposterior, transverse, and craniocaudal measurements were obtained. The largest dimension of the tumor was selected for subsequent analysis.
The settings for power Doppler sonography were optimized to detect low-velocity or low-volume blood flow. The color box included the lesion and a margin of healthy breast tissue peripheral to the lesion. The color gain was adjusted to a level that allowed detection of small tumor vessels without background noise. To accurately study sonographic vascularization of the tumor, the region of interest was scanned slowly with minimal probe pressure in the transverse plane; thus, all vessels related to the tumor were identified.
The pulsed Doppler sonography study enabled us to determine the arterial or venous nature of each vessel; however, only the number and distribution of tumor arteries were considered. When a vascular branch was observed, two vascular structures were counted. The resistive index was calculated with the following equation: resistive index equals (systolic peak velocity minus end-diastolic peak velocity) divided by systolic peak velocity. The pulsatility index was calculated with the following equation: pulsatility index equals (systolic peak velocity minus end-diastolic peak velocity) divided by mean velocity. Pulsatility index and resistive index were determined for each tumor artery. These indexes were obtained as an average measurement of two waves in each artery. The maximum resistive index and pulsatility index values for each tumor were considered in the subsequent analysis.
The grade of tumor vascularization was based on the number of tumor arteries: In tumors with no detectable flow, arterial vessels were not detected. Tumors were moderately vascularized when one or two arterial vessels were observed. Tumors were highly vascularized when more than two arteries were detected. On the basis of distribution of tumor vessels and reports in the literature (18–20), four patterns of vascularization were described. A radial or afferent pattern was a pattern with afferent vessels to the tumor from the healthy breast parenchyma (Fig 1). A peripheral pattern was a pattern with vessels around the tumor periphery and no evidence of ramifications inside the tumor (Fig 2). An intratumoral pattern was a pattern that showed vessels only inside the tumor. A ramifying pattern was a pattern with an uneven distribution of peripheral and intratumoral vessels (Fig 3).
|
|
|
|
|
Histopathologic Examination and Immunohistochemistry
All specimens were analyzed by a pathologist (P.L.F.) with 15 years of experience in breast pathology who was unaware of the sonographic findings. Tumor size, histologic type and grade, and axillary node status were determined. In patients with axillary involvement, the nodes involved were counted (three or fewer, more than three).
Immunohistochemical analysis of tumor microvascularization and agreement of the sonographic and histologic parameters of vascularization and tumor size were studied. Since we had a time limit for reporting the results, it was not possible to recruit all patients; thus, only the first 55 consecutive cases in the series were included. Since tumor histologic sections demonstrate many more vessels than can be identified with power Doppler sonography, the pathologist and radiologists (G.S., M.V.) established a consensus regarding the minimum tumor artery diameter to be counted by the pathologist and compared it with the number of arterial vessels visible at sonography. We selected 300 µm as the reference value on the basis of previous reports: Delorme et al (21) reported that tumor vascularization depicted with sonography was predominantly peripheral, and Less et al (22) described the microvascular architecture in a mammary carcinoma. The pathologist reviewed all sections of each tumor to determine the number of arteries observed with a diameter equal to or greater than this value.
For the immunohistochemical process, it was not feasible to include the complete tumor for staining; thus, the original hematoxylin-eosin–stained sections of the primary tumors were reviewed by the pathologist. A block was selected that was considered to be representative of the invasive component and included the leading edge of the tumor. The EnVision+ system (Dako, Glostrup, Denmark) was used to stain 4-µm sections with anti-CD34, which is a monoclonal antibody. This antibody is recognized as being the most sensitive for staining endothelial cells (5). The antibody binds to a superficial antigen with a molecular weight of 110 kDa, expressed in hematopoietic progenitor cells and vascular endothelium.
Stained Section Analysis
A computer digital analyzer (Microimage; Olympus Europe, Hamburg, Germany) was used for the vascular quantifying process. The procedure was performed by an independent observer (X.F.) who was unaware of the sonographic findings. The stained sections were scanned with a low magnification (original magnification, x20 and x40) to identify the five areas with the most abundant microvessels; these areas were referred to as hot spots. The microvessels in these hot spots were marked for counting with the computer (original magnification, x200). Ramified structures were counted as one microvessel; however, if their course was interrupted, they were counted as two. Parameters registered were the density of microvascularization (number of microvessels per field size on the monitor screen, 0.0521 mm2) and the area of microvascularization (total microvessel area in each field). The highest value for each parameter was used for further statistical analysis (8).
Statistical Analysis
Statistical analysis was performed by using SPSS 10.0 software (SPSS, Chicago, Ill).
Age-related differences in the tumor stage, grade of tumor vascularization, histologic type, and axillary involvement were assessed.
Agreement in the number of arteries was analyzed by using statistics, and weights were defined to penalize disagreement. Values of less than 0.40 indicated positive but poor agreement, values of 0.41–0.75 indicated good agreement, values of 0.76–0.99 indicated excellent agreement, and a value of 1.00 indicated complete agreement (23). Agreement between sonographic and histologic results for tumor size was evaluated with Bland-Altman agreement limits (24). The relationships between tumor microvascularization and sonographic variables and between sonographic and histopathologic variables were assessed with linear regression models. When both of the variables were categorical, the 
2 test or Fisher exact test was used.
In this study, we were interested in the axillary status (metastatic or nonmetastatic), regardless of the number of lymph nodes involved. Thus, forward logistic regression analysis was used to define a predictive model for axillary status. The model was constructed by selecting a random sample containing 77 (79%) of the 97 cases from the series (analysis group); the remaining 20 cases (21%) were reserved for validating the model (validation group). Receiver operating characteristic analysis was used to determine the probability that best classified the axillary status. The predictive values of the model were calculated according to prevalence of axillary involvement in invasive breast carcinoma (30%–35%), as indicated in several series in the literature. In all tests, the P value used to indicate a statistically significance difference was .05.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Sonography
The mean size of the lesions was 1.5 cm (range, 0.4–4.7 cm). Of the 97 tumors, 31 (32%) did not have arterial vascularization, while the remaining 66 (68%) were vascularized. Of the vascularized tumors, 41 (42%) were moderately vascularized, and the remaining 25 (26%) were highly vascularized. There was no significant relationship between age and grade of vascularization (P = .152). The grade of vascularity and axillary node status is shown in Table 1.
|
|
The presence of axillary metastasis was confirmed in 33 (34%) of the 97 cases, which had an average of two positive lymph nodes (range, 1–11). Only five (5%) of the 97 tumors with axillary involvement had more than three metastatic axillary nodes. A high grade of tumor vascularization (more than two arterial vessels) was observed in three of these five tumors. The remaining two tumors were moderately vascularized. Two of these five tumors were included in the subset of 55 tumors; one was highly vascularized, and the other was moderately vascularized.
There was no significant relationship between age and tumor stage (P = .512), histologic type (P = .390), or axillary involvement (P = .569).
Table 3 shows the features of the 55 patients included in the subset used to study tumor microvascularization. Note the similarity of percentages regarding the whole group (Table 1). Two samples were excluded as a result of the scarcity of staining material, and two others were excluded because of the limited staining appeal of the material (colloid). The mean value of the microvascularization density was 272.4 microvessels per square millimeter (range, 115.2–652.7 microvessels per square millimeter), and the mean value of the area of microvascularization was 0.0023 mm2 (range, 0.0009–0.0093 mm2).
|
= 0.6639; P < .001) (Fig 4). In 38 (69%) of the 55 cases, the number of arteries depicted with sonography agreed with the number counted by the pathologist. Disagreement was observed in 17 (31%) cases. In five cases, vascularization was not demonstrated with sonography; however, in each case, one arterial vessel larger than or equal to 300 µm was identified with pathologic analysis. Conversely, in 12 cases, we detected more arteries with sonography than with pathologic analysis. Two of the four tumors with axillary metastases and no detectable flow and eight of the 14 moderately vascularized tumors with axillary metastases (Table 1) were included in this group of 55 lesions used for analysis of agreement. We observed agreement of the number of arteries in these cases, except for two moderately vascularized tumors, in which the number of arteries detected with sonography was greater.
|
|
A significant relationship (P = .03) was also observed between absence or presence of arterial vascularization and histologic grade. Tumors with no detectable flow had low histologic grade compared with vascularized tumors, which had intermediate or high histologic grades.
Relationship between Sonographic and Immunohistochemical Findings
The area of microvascularization showed a significant relationship (P = .016) with pulsatility index in the univariate study, which was not confirmed with multivariate analysis. A trend toward relationship was observed between area of microvascularization and resistive index (P = .052) and tumor size (P = .096); however, they were not significant. There was no significant relationship between area of microvascularization and number of arteries (P = .236), nor was there a significant relationship between density of microvascularization and any of the variables.
Predictive Factors of Axillary Node Metastasis and Logistic Regression Analysis
The univariate analysis showed a significant relationship between the probability of axillary involvement and the number of tumor arteries (P < .001), sonographic tumor size (expressed as a natural logarithm) (P < .001), highest resistive index (P < .001), and highest pulsatility index (P = .02). There was no significant relationship between mammographic appearance of the lesion (P = .207) or tumor location in quadrants (P = .309) and the presence of axillary metastasis.
Multivariate analysis demonstrated that the number of tumor arteries (x) (P = .016) and natural logarithm of tumor size (y) (P = .035) were significantly related to axillary status. On the basis of this analysis, a predictive model of axillary node status (P < .001) was developed according to the following mathematic expression: P = 1/ [1 + e–(–2.311+0.4984x+1.5777y)].
Since the relationship between arterial vascularization and axillary metastasis could be explained indirectly by the relationship between tumor size and presence of axillary metastasis, we specifically investigated the possibility that both variables were covariates. We did not find any significant relationship between them (P = .145); therefore, we concluded that both variables were independently related to axillary status in our series (Figs 3, 5). According to the model, the risk of axillary metastatic involvement was multiplied by 1.65 (95% confidence interval: 1.07, 2.52) for each tumor artery detected and 4.84 (95% confidence interval: 1.02, 23.04) if tumor size was doubled.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
On the basis of sonographic distribution of the tumor arterial vessels, we characterized four sonographic patterns of vascularization as follows: afferent or radial, peripheral, intratumoral, and ramifying. The predominance of afferent and ramifying patterns confirms the observation that tumor vessels depicted with Doppler sonography are found mainly in the tumor periphery (21).
Analysis of the sonographic findings of tumor vascularization and histopathologic parameters of the tumor leads to different observations with possible clinical implications: Histologic tumor size is significantly related to the number of tumor arteries detected with sonography (P < .001). Vascularized tumors generally present intermediate or high histologic grades, which indicates greater clinical aggressiveness than tumors with no detectable flow. The relationship between histologic tumor size and axillary node involvement is a well-known fact (28). In our study we observed that there is also a significant relationship between sonographic tumor size and axillary metastasis (P = .035). This suggests that sonographic tumor size could be a prehistologic parameter related to axillary node status; therefore, tumor size may be of prognostic value. Our results show a relationship between quantity of tumor arterial vascularization and presence of axillary metastasis (P = .016); thus, 88% of the tumors with axillary node involvement are vascularized, and 58% of the tumors with no axillary metastasis show arterial vascularization.
A theory proposed by Byers et al (29) could provide a possible explanation for the relationship between the number of tumor arteries and risk of metastasis. According to these authors, one of the factors involved in neoplastic dissemination is the movement of tumor cells or emboli to local lymph nodes via the lymphatic or venous circulation. In some cases, tumor cells may not be able to enter the fluid in a lymph vessel or vein unless they are detached from the primary tumor mass, either individually or as aggregates, by the laminar flow of the circulatory system. This theory suggests that the more abundant the tumor vascularization, the greater the shear forces induced on tumor cells by laminar flow. Thus, this process partially favors detachment of these cells and the risk of metastases.
As a result of the relationship found between sonographic findings (tumor size and arterial vascularization) and some histologic parameters with prognostic value (tumor size, histologic grade, and presence of axillary metastases), we thought the information obtained from the sonographic study of invasive breast tumors might be useful as a predictor of axillary node status in these patients. Thus, we quantified the risk of axillary metastasis on the basis of scores obtained with multivariate analysis. According to this model, the significant independent predictors of lymph node malignancy were sonographic tumor size and the number of tumor arteries. In our opinion, this description offers a new contribution to this study.
The receiver operating characteristic curve was used to choose the best cutoff value for deciding which patients could reliably forego axillary dissection. Bearing in mind the fact that when predicting axillary involvement the consequences of a false-negative diagnosis are more serious than those of a false-positive diagnosis, we selected a score that ensured high sensitivity, even though it decreased specificity. Thus, the score of 0.2324 indicated high sensitivity (96.1%), high negative predictive value (96.1%), and low specificity (53.0%). Low specificity limits the utility of the model as a predictor of axillary status. Thus, we believe that the real effectiveness of this model is based on its high negative predictive value, which allows extremely reliable prediction of the absence of axillary metastases.
It should be noted that the limited size of the sample partially explains the lower sensitivity and negative predictive value observed in the validation group. Moreover, it is quite likely that this fact would also explain that the risk of axillary metastases was independently related to sonographic tumor size and the number of tumor arteries. Larger series would probably demonstrate the dependency of both variables.
Another limitation is related to the sonographic technique. The study was performed by using unenhanced power Doppler sonography. Bearing in mind our results in the subset of 55 tumors, there were 12 tumors in which power Doppler sonography depicted some vessels not counted at histologic evaluation; this probably means that unenhanced power Doppler sonography is able to depict vessels smaller than 300 µm. Conversely, there were five tumors with no detectable flow at sonography that had 300-µm arterial vessels at pathologic analysis. It is known that sonographic detection of vessels depends not only on vessel size but also on flow velocity. It is very likely that extremely low-velocity flow in those vessels could have made detection impossible, which, according to our conclusions, could represent an underevaluation of the risk of axillary involvement. Thus, although none of these five tumors showed axillary involvement and the role of sonographic contrast agents in the evaluation of breast cancer is not clearly defined (30,31), we believe that detection of vessels in lesions that show no vascular flow would improve with use of sonographic contrast agents. This idea could be best addressed in another study.
Finally, since vascular staining for the microvascularization evaluation was performed in a subset of 55 lesions, the absence of relationship between density and area of microvascularization with regard to sonographic parameters of vascularization could be related to this issue. However, other authors (6) did not find any relationship between the density of microvascularization and sonographic parameters of tumor flow either.
In summary, power Doppler sonography is a feasible noninvasive diagnostic technique that is capable of depicting arterial vessels with a minimum diameter of 300 µm that are significantly related to breast tumor size and usually in a peripheral location. The information provided by power Doppler sonography of invasive breast carcinoma is of clinical value, since the number of tumor arteries and sonographic tumor size were independent predictors of axillary node involvement. We have defined a mathematic model that allows us to predict axillary status on the basis of both variables. The real effectiveness of this model is its high negative predictive value, which facilitates highly reliable prediction of the absence of axillary metastatic involvement. Moreover, the possibility of achieving greater specificity by using power Doppler sonography as an adjunct to sentinel node biopsy should not be ruled out. Further studies are required to analyze the benefits that may be obtained by using both techniques.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, all authors; study concepts and design, all authors; literature research, G.S., X.F., A.C.; clinical studies, J.A.V.; data acquisition and analysis/interpretation, all authors; statistical analysis, G.S., P.L.F.; manuscript preparation and definition of intellectual content, all authors; manuscript editing, G.S., M.V., P.L.F.; manuscript revision/review and final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  B. Escudier, N. Lassau, E. Angevin, J. C. Soria, L. Chami, M. Lamuraglia, E. Zafarana, V. Landreau, B. Schwartz, E. Brendel, et al. Phase I Trial of Sorafenib in Combination with IFN {alpha}-2a in Patients with Unresectable and/or Metastatic Renal Cell Carcinoma or Malignant Melanoma Clin. Cancer Res., March 15, 2007; 13(6): 1801 - 1809. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
