Cortical morphologic features of axillary lymph nodes as a predictor of metastasis in breast cancer: in vitro sonographic study.

OBJECTIVE: The purpose of this study was in vitro sonographic-pathologic correlation of findings in dissected axillary lymph nodes from breast cancer patients undergoing axillary lymph node dissection and classification of the sonographic appearance of the nodes on the basis of cortical morphologic features to facilitate early recognition of metastatic disease.
MATERIALS AND METHODS: High-resolution sonography was used for in vitro examination of 171 lymph nodes from 19 axillae in 18 patients with unknown nodal status who underwent axillary lymph node dissection for early infiltrating breast cancer. The images were evaluated by two blinded observers, and discordant readings were referred to a third blinded observer. Each lymph node was classified as one of types 1-6 according to cortical morphologic features. Types 1-4 were considered benign, ranging from hyperechoic with no visible cortex to thickened generalized hypoechoic cortical lobulation. Type 5 (focal hypoechoic cortical lobulation) and type 6 (hypoechoic node with absent hilum) nodes were considered metastatic. The reference standard for metastatic disease was histopathologic evaluation of sectioned nodes by a single pathologist blinded to sonographic findings. Largest nodal diameter also was measured.
RESULTS: Interobserver agreement was 77% for classification of nodal morphology (types 1-6) and 88% for characterization of a node as benign or malignant. Sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy of cortical shape in prediction of metastatic involvement of axillary nodes were 77%, 80%, 36%, 96%, and 80%. Type 4 nodes had the most false-negative findings (four of 36). Node size ranged from 0.2 to 3.8 cm, and subcentimeter nodes of all types were detected.
CONCLUSION: In breast cancer, axillary lymph nodes can be classified according to cortical morphologic features. Predominantly hyperechoic nodes (types 1-3) can be considered benign. Generalized cortical lobulation (type 4) is uncommonly a false-negative finding, but metastasis, if present, is invariably detected at sentinel node mapping. The presence of asymmetric focal hypoechoic cortical lobulation (type 5) or a completely hypoechoic node (type 6) should serve as a guideline for universal performance of fine-needle aspiration for preoperative staging of breast cancer. This classification, when verified with larger samples, may serve as a useful clinical guideline if proven with results of in vivo studies.