OBJECTIVE. The purpose of our study was to evaluate the accuracy of sonography and sonographically guided biopsy in the preoperative diagnosis of metastatic invasion of the axilla in patients with breast carcinoma.
MATERIALS AND METHODS. We performed a MEDLINE search (keywords, “sonography” OR “ultrasound” AND “axillary”) and a manual search of the references of relevant studies and reviews of preoperative diagnosis on sonography of possible axillary metastases. The gold standard required was axillary lymph node dissection; we accepted sentinel node biopsy as an alternative gold standard. Considering the sonographic findings and the results of the sonographically guided biopsy, the sensitivity and specificity were calculated using metaanalysis. We also checked the existence of heterogeneity of the summary results.
RESULTS. Sixteen articles were selected. In sonography of axillae without palpable nodes, and using lymph node size as the criterion for positivity, sensitivity varied between 48.8% (95% confidence interval, 39.6–58%) and 87.1% (76.1–94.3%) and specificity, between 55.6% (44.7–66.3%) and 97.3% (86.1–99.9%). When lymph node morphology was used as the criterion for positivity, sensitivity ranged from 26.4% (15.3–40.3%) to 75.9% (56.4–89.7%) and specificity, from 88.4% (82.1–93.1%) to 98.1% (90.1–99.9%). The results are different if axillae with palpable nodes are included. The sonographically guided biopsy shows a sensitivity that varies between 30.6% (22.5–39.6%) and 62.9% (49.7–74.8%) and a specificity of 100% (94.8–100%). Many of the summary results obtained after meta-analysis show a heterogeneity that disappears, on occasion, on excluding the studies that use a double gold standard.
CONCLUSION. Axillary sonography is moderately sensitive and fairly specific in the diagnosis of axillary metastatic involvement. Sonographically guided biopsy of the sonographically suspicious nodes somewhat increases the specificity, which reaches 100%. Negative sonographic results do not exclude axillary lymph node metastases.
The presence of axillary lymph node metastases in breast cancer is an important factor in assessing prognosis and determines management after surgery . Axillary staging is traditionally performed by means of axillary lymph node dissection.
The traditional use of axillary lymph node dissection has been justified by the need for staging and the possibility of reducing the tumoral load by resecting the possibly affected lymph nodes. Doubt has been cast on the latter aspect with the publication of contradictory results about the effect of lymph node dissection on patient survival. A meta-analysis performed by Orr  and published in 1999 shows a slight improvement in the survival of women who have undergone axillary lymph node dissection. On the other hand, a randomized trial with a long follow-up period performed by Fisher et al.  shows no survival advantage in patients undergoing lymph node dissection. To this has been added the recent introduction of the sentinel node biopsy in the decision to perform an axillary lymph node dissection. One of the problems that arises from the sentinel node biopsy is the existence of false-negative results, which leads to axillary lymph node dissection not being performed in patients with a negative sentinel node biopsy but who in fact have axillary metastases. It has been found that 1–15% of patients with a negative sentinel node biopsy for metastases have other affected lymph nodes in the same region . Most of these false-negatives are due to massive lymph node metastases in the first drainage node [5–7]. Axillary sonography is not used systematically in patients with breast cancer because of its low sensitivity. However, the possibility of detecting some of these metastases and reducing the number of false-negatives at sentinel node biopsy has renewed interest in the use of sonography before sentinel node biopsy .
Our aim was to learn, through a systematic review, the accuracy of sonography in the diagnosis of axillary lymph node metastases in patients with breast cancer, taking into account the different criteria for diagnosis and the contribution of sonographically guided biopsy of suspicious nodes.
Materials and Methods
We performed a literature review starting with an electronic bibliographic search (Ovid-MEDLINE, Ovid Technologies) January 1980 to March 12, 2004. The search was made using the text words “ultrasound” OR “sonography” AND “axillary,” by which we found 367 articles. The selection was limited to articles written in English, Spanish, French, and Italian. After reading the titles, we selected those abstracts that gave data about the diagnostic accuracy of axillary sonography in breast cancer; 31 articles were selected. Then we performed a manual search in the reference lists of the relevant articles previously selected and in the reference lists of the articles reviewed. For the final selection we chose those articles that fulfilled the following criteria for inclusion: patients were diagnosed as having breast cancer and axillary sonography was performed before axillary lymph node dissection or sentinel node biopsy; sonography was performed using transducers of 7 MHz or higher frequencies; sonographic criteria as to size and morphology, or sonographically guided biopsy, were used to classify a lymph node as positive or negative for metastases; the histopathologic result of total axillary lymph node dissection or results of dissection of first (nodes outside the smaller pectoral muscle) and second (behind the smaller pectoral muscle) levels was required as the gold standard, and a sentinel node biopsy was accepted as an alternative gold standard in those studies that used this technique; and results were expressed in terms of sensitivity and specificity. For the inclusion of the results in the meta-analysis, we required that in each study it was possible to establish the number of true-positives, true-negatives, false-positives, and false-negatives.
We collected data from the selected studies about the methodology used: type of design (prospective or retrospective), selection and number of patients, blind interpretation of the diagnostic tests, and any other data that might prove useful for a later analysis.
Accuracy Measures for Primary Studies and Meta-Analysis
Statistical testing was performed using Meta-DiSc . For each study, the sensitivity, specificity, and exact 95% confidence intervals (CIs) were calculated.
Summary sensitivity and specificity were calculated as a weighted average in which the weight of each study is its sample size. The 95% CIs of individual and pooled sensitivity and specificity were calculated using the F distribution method to compute the exact confidence interval for the binomial proportion. The degree of variability (homogeneity) among study results of sensitivity and specificity was tested using chi-square distribution with K – 1 degrees of freedom (K being the number of studies).
Sixteen studies fulfilled the criteria for inclusion and were considered valid in the final selection [10–25].
Methodology of the Studies
The methodologic characteristics of the studies included, and the sonographic criteria for positivity, are shown in Table 1. Fourteen studies were prospective, although only eight specified that the patients were selected consecutively. The gold standard used was total or first- and second-level axillary lymph node dissection in 11 studies, whereas the other five used as a gold standard the results of the sentinel node biopsy or the axillary lymph node dissection in patients who rejected sentinel node biopsy.
TABLE 1: Methodology and Characteristics of the Studies Selected
Round, oval, or hypoechoic node, or with eccentric cortical lobulation
7.5–12 MHz; stage, Tis–T4
Note–NS = not specified, T = tumor stage.
Lymphadenectomy = axillary lymphadenectomy
For assessment of results
One study always used axillary lymph node dissection after the sentinel node detection technique, regardless of its result . In those studies that used axillary lymph node dissection as the gold standard, only one study specified the methodology used for the histologic processing of the axillary lymph nodes and their criteria for positivity , and another  stated that serial sections were not performed. The remaining studies did not explain the method of processing material for histologic studies. In all the studies in which the sentinel node technique was applied, multiserial sections were made except in one that omitted this information . Only three studies claimed to have performed a blind assessment of the gold standard (axillary lymph node dissection) or the sonography results.
In three studies, the sonographic criterion to establish whether an axillary lymph node was metastatic was a node larger than 5 mm [10, 11, 13], whereas other studies considered any node visible on sonography to be metastatic, regardless of its size . In the remaining studies, various morphologic or structural criteria for the node were applied (a rounded shape, hypoechogenicity, cortical thickening, obliteration of the hilum, lobulation) or a combination of size and morphologic criteria. In the last column of Table 1 we indicate the frequency of the transducer used, the stage (TNM) or tumor size (T) of the patients, and other data of interest. The detailed analysis of some studies allowed us to deduce sensitivity and specificity data, making use of sonographic criteria of size or morphology that had not been analyzed directly by the authors.
In all the studies in which a sonographically guided biopsy was performed, small-caliber needles were used (20–22 gauge), except in one study  in which core-biopsy (14 gauge) was performed, limiting the fine-needle biopsies to technically difficult cases. In two studies [24, 25], biopsy aspiration with a fine needle was performed although the authors did not give the needle caliber. The sonographically guided biopsy criteria were different, depending on the studies: in three studies, any sonographically visible axillary lymph node was biopsied [17, 19, 21], but in the other five studies [20, 22–25] needle biopsy was limited to indeterminate nodes or those suspected of being malignant.
Accuracy of Axillary Sonography
To establish the parameters for accuracy of sonography in the diagnosis of axillary lymph node metastases, we divided the results into three groups: studies that included axillae with both palpable and nonpalpable nodes (Table 2), studies that included only axillae with nonpalpable nodes (Table 3), and studies that included sonographically guided biopsy of the axillary lymph nodes (Table 4).
TABLE 2: Diagnostic Accuracy of Axillary Sonography in Patients with Breast Carcinoma (Studies That Include Axillae with Both Palpable and Nonpalpable Nodes)
Note–Numbers in parentheses are 95% confidence intervals. TP = true-positive, TN = true-negative. FP = false-positive, FN = false-negative.
Includes only patients in whom processable material was obtained at fine-needle aspiration biopsy
Retrospective, so excluded from the meta-analyses
Includes studies in which gold standard was axillary lymph node dissection
Using chi-square test
Includes studies in which gold standard was axillary lymph node dissection or sentinel node biopsy
Includes all patients who were selected initially regardless of whether fine-needle aspiration biopsy was performed
Studies that included palpable and nonpalpable nodes—In these studies (Table 2), if the size (> 5 mm) or visibility on sonography of the node was used as the criterion for positivity, sensitivity oscillated between 66.1% (95% CI, 52.6–77.9%) and 72.7% (49.8–89.3%), with no heterogeneity found between the results, both when we include only those studies that used axillary lymph node dissection as the gold standard and when we also include the studies that used sentinel node biopsy. Specificity varies between 44.1% (34.3–54.3%) and 97.9% (88.7–99.9%), with heterogeneity found between the results, even when we exclude studies with sentinel node biopsy.
If the morphology of the lymph node was used as a criterion for sonographic positivity, sensitivity oscillated between 54.7% (95% CI, 41.7–67.2%) and 92.3% (74.9–99.1%) and specificity, between 80.4% 73.9–86.2%) and 97.1% (90–99.6%). Heterogeneity does not exist for either parameter when studies with sentinel node biopsy are excluded.
Studies that included only nonpalpable nodes—In these studies (Table 3), if the size on sonography of the node (> 5 mm) or its visibility was used as a criterion for positivity, sensitivity varies from 48.8% (95% CI, 39.6–58%) to 87.1% (76.1–94.3%) and specificity, from 55.6% (44.7–66.3%) to 97.3% (86.1–99.9%). The chi-square test shows heterogeneity in both, even when studies with sentinel node biopsy are excluded.
If the morphology of the node was used as the criterion for positivity, sensitivity varies from 26.4% (95% CI, 15.3–40.3%) to 75.9% (56.4–89.7%) and specificity, from 88.4% (82.1–93.1%) to 98.1% (90.1–99.9%). The results are heterogeneous in all groups except specificity when studies with sentinel node biopsy are excluded.
Sonographically guided biopsy—Although studies using sonographically guided biopsy are included in Table 4, studies with retrospective selection of patients were excluded from the meta-analysis.
Counting only those patients in whom it was possible to perform a sonographically guided biopsy and obtain material for histologic analysis of axillary lymph nodes, the sensitivity of sonography varies between 43.5% (95% CI, 33–54.7%) and 94.9% (88.5–98.3%) and specificity, between 96.9% (91.3–99.4%) and 100% (96.2–100%). Logically, heterogeneity is shown only in the sensitivity results.
When all patients are included just as they were initially selected (considering as false-negatives those women in whom sonography did not find nodes, and those in whom nodes were not suspicious or processable material was not obtained on needle biopsy and metastatic invasion was later shown), sensitivity descends with respect to the previous group and varies from 30.6% (95% CI, 22.5–39.6%) to 62.9% (49.7–74.8%) and specificity is always 100% (94.8–100%). The average sensitivity result excluding studies with sentinel node biopsy is from a sole study and therefore does not show heterogeneity.
The studies selected had different degrees of methodologic quality. We have included only the prospective studies, and many of them were consecutive. There is no explicit reference to the manner of selecting patients, and in most studies there does not seem to have been a blind interpretation of the results (gold standard and sonography). The gold standards used were two: in most studies the standard was axillary lymph node dissection, whereas in others it was the combination of biopsy of the sentinel node and, when this was not accepted, axillary lymph node dissection. In the case of axillary lymph node dissection, neither the positivity criterion nor the method of processing the nodes was specified. These aspects have a decisive influence because the use of more exhaustive histology sections permits a greater percentage of metastases to be discovered [26, 27].
Our review shows that axillary sonography is moderately sensitive and quite specific in the diagnosis of axillary metastatic involvement in patients with breast cancer. The different studies show variable results (Tables 2, 3, 4). With the aim of reducing this variability, we performed analyses according to whether the axillary nodes were palpable. The existence of palpable axillary nodes is often used to decide whether to perform a sentinel node biopsy or an axillary lymph node dissection, although axillary palpation has low sensitivity and specificity . Within these groups we subdivided the studies according to the two sonographic criteria used to diagnose a node as positive: size (> 5 mm) and morphology (round, hypoechoic, with loss of central hilum, eccentric cortical hypertrophy). Even so, the results continued to be variable in each group by the heterogeneity tests performed in the meta-analysis.
To lessen as much as possible the sources of variability, we performed a second meta-analysis, excluding the studies that used sentinel node biopsy as a gold standard. With this procedure the heterogeneity disappeared in most but not in all cases. Sentinel node biopsy may give false-negative results, which may lead to verification bias. Other factors may determine residual heterogeneity. One of these could be the type of patients selected (e.g., different staging) or the way in which they were selected (sometimes consecutively and sometimes not), and independently of this, the place in which the selection was performed: in almost all cases, selection of patients took place in the surgery department, but in a few cases selection took place in the radiology department  and in other studies the place of selection was not recorded. The different equipment used may also have an influence (with greater resolution resulting with the use of more up-to-date apparatus); different transducer frequencies were also used. Also, small variations existed in the criteria applied by different authors to determine on sonography whether the node was benign or malignant.
On occasion, our results are contradictory. Sensitivity should be greater when palpable and nonpalpable nodes are included than when considering only nonpalpable nodes, because the palpable ones, being larger, should be detected better on sonography. However, if we compare the sensitivities in Tables 2 and 3, we observe the opposite tendency. Equally, sensitivity should be greater when size criteria are used as compared with morphologic criteria, because the latter require the node to be seen and to have certain sonographic characteristics. This hypothesis is borne out in those articles that used data based on both size and morphologic criteria: thus Bonnema et al. , Damera et al. , and Deurloo et al.  found greater sensitivity for sonography when they used size as a positivity criterion than when they used morphologic criteria. This same finding, although with heterogeneity, is observed in the average results of the series that include only nonpalpable nodes (Table 3) but not in the average results considering all the nodes, both palpable and nonpalpable (Table 2).
On the contrary, although it shows heterogeneity in some groups, specificity is more uniform in tendency, being greater when morphologic criteria are used than when only size is considered.
In an attempt to improve the results of sonography, sonographically guided biopsy of the axillary nodes has been introduced. The criteria applied as indications for needle biopsy are different in the various studies because in some cases only those nodes suspected of being malignant on sonography are biopsied [20, 22–25], and in other cases all the nodes are biopsied independently of their appearance on sonography or their size [17, 19, 21]. The use of sonographically guided biopsy notably increases specificity, which reaches as high as 100%, although sensitivity is reduced because it is necessary to have visualized the node previously (in cases in which all nodes were biopsied) or to fulfill the sonographic criteria for malignancy (in those cases in which only nodes suspected of being malignant on sonography are biopsied), and a needle biopsy must be technically possible. Therefore, sensitivity decreases if we consider all the axillae studied on sonography and not only those cases in which it was possible to perform the biopsy. Those axillae should be considered negative if lymph nodes were not visualized, they were not deemed suspicious, or needle biopsy was not technically possible. True-negatives were those that were found to be free of metastasis according to the gold standard, and false-negatives were those found to be metastases. The specificity of the sonographically guided biopsy is greater (100%) than that of sonography alone (96.5% using morphologic criteria).
However, the number of women who benefit from the results obtained with biopsy is small because there are ultimately few patients in whom a biopsy is performed. In the series of Bonnema et al. , sonographically guided biopsy was performed in only 81 women of the initial 150, and in the series of Sapino et al. , in only 95 of 267.
Furthermore, we cannot consider all the sonographically guided biopsy results to be equally useful. Only those with positive findings have a high predictive value (100%). Negative results are not reliable because of the high proportion of false-negatives. That is to say, a negative sonographically guided biopsy does not exclude lymph node metastasis. This is reflected in Table 4, which shows an average sensitivity of sonographically guided biopsy of 89.8%. This figure indicates that 10.2% of patients with a negative biopsy result (false-negative) had axillary metastases (none of the articles mentions whether the metastases were micro- or macrometastases). If, instead of considering only those cases in which it was possible to perform a biopsy, we take into account the total number of axillae initially studied with sonography, we obtain an average sensitivity of 62.9% and, therefore, 37.1% false-negatives (among these are included patients with invisible or not sonographically suspected lymph nodes and in whom no biopsy was performed but who were found to have metastases).
The results of our review have some limitations as to their applicability, such as the inconsistency of the sensitivity results and the methodologic flaws in some studies that reduce their internal validity. The methodologic flaws and the presence of bias is a common feature affecting publications about diagnostic tests, as was shown in a study that reviewed the quality of research on diagnostic testing .
In conclusion, our review shows that axillary sonography cannot be used in isolation as a method for deciding whether to perform axillary lymph node dissection. If it is decided to use axillary sonography before a sentinel node biopsy, once axillae with palpable nodes have been excluded, we recommend using morphologic sonographic criteria for malignancy. When suspected metastatic axillary nodes are visualized, a sonographically guided biopsy can be performed. This procedure adds some specificity compared with the use of sonography alone (100% vs 96.5%), but it entails a certain degree of aggression and extra resources. In fact, with the use of sonography alone, applying morphologic criteria to establish malignancy, approximately half of the axillae with metastases would be detected (48.4% sensitivity) with a high specificity (96.5%). The results of sonography would allow patients with sonographically positive axillae to be directed to axillary lymph node dissection. The remaining patients with sonographically negative axillae would be candidates for sentinel node biopsy, improving the negative predictive value of the sentinel node biopsy because of the lower prevalence of metastases and thereby increasing the certainty of this technique.
Fisher B, Jeong JH, Anderson S, Bryant J, Fisher ER, Wolmark N. Twenty-five-year follow-up of a randomized trial comparing radical mastectomy, total mastectomy, and total mastectomy followed by irradiation. N Engl J Med 2002; 347:567-575
Bobin JY, Spirito C, Isaac S, et al. Le marquage lymphatique et la biopsie du ganglion sentinelle axillaire dans 243 cancers invasifs du sein sans ganglion palpable: expérience du centre hospitalier Lyon-Sud. Ann Chir 2000; 125:861-870
de Kanter AY, van Geel AN, Paul MA, et al. Controlled introduction of the sentinel node biopsy in breast cancer in a multi-centre setting: the role of a coordinator for quality control. Eur J Surg Oncol 2000; 26:652-656
Zamora J, Muriel A, Abraira V. Meta-DiSc for Windows: a software package for the meta-analysis of diagnostic tests. XI Cochrane Colloquium. Barcelona 2003. Available at: www.hrc.es/investigacion/metadisc.htlm. Accessed May 1, 2005
Lam WW, Yang WT, Chan YL, Stewart IE, Metreweli C, King W. Detection of axillary lymph node metastases in breast carcinoma by technetium-99m sestamibi breast scintigraphy, ultrasound and conventional mammography. Eur J Nucl Med 1996; 23:498-503
Yang WT, Ahuja A, Tang A, Suen M, King W, Metreweli C. High resolution sonographic detection of axillary lymph node metastases in breast cancer. J Ultrasound Med 1996; 16:241-246; erratum in: J Ultrasound Med 1996; 15:644
Verbanck J, Vandewiele I, De Winter H, Tytgat J, van Aelst F, Tanghe W. Value of axillary ultrasonography and sonographically guided needle biopsy of axillary nodes: a prospective study in 144 consecutive patients. J Clin Ultrasound 1997; 25:53-56
Bonnema J, van Geel AN, van Ooijen B, et al. Ultrasound-guided aspiration biopsy for detection of nonpalpable axillary node metastases in breast cancer patients: new diagnostic method. World J Surg 1997; 21:270-274
Krishnamurthy S, Sneige N, Bedi DG, et al. Role of ultrasound-guided fine-needle aspiration of indeterminate and suspicious axillary lymph nodes in the initial staging of breast carcinoma. Cancer 2002; 95:982-988
Kuenen-Boumeester V, Menke-Pluymers M, de Kanter AY, Obdeijn IM, Urich D, Van Der Kwast TH. Ultrasound-guided fine needle aspiration cytology of axillary lymph nodes in breast cancer patients: a preoperative staging procedure. Eur J Cancer 2003; 39:170-174
Bedrosian I, Bedi D, Kuerer HM, et al. Impact of clinicopathological factors on sensitivity of axillary ultrasonography in the detection of axillary nodal metastases in patients with breast cancer. Ann Surg Oncol 2003; 10:1025-1030
Fisher B, Wolmark N, Bauer M, Redmon C, Gebhardt M. The accuracy of clinical nodal staging and of limited axillary dissection as a determinant of histologic and nodal status in carcinoma of the breast. Surg Gynecol Obstet 1981; 152:765-772