Objective: The purpose of this study was to assess the accuracy of 5 sonographic features in the prediction of extranodal extension (ENE) in axillary lymph nodes (ALNs) of patients with biopsy-proven breast cancer.
Methods: A review of our institution's surgical and pathologic database was performed for patients with pathologically proven axillary ENE from October 1, 2003, to October 1, 2007. An equivalent number of patients without ENE were included. All patients had sonograms of ALNs available. A radiologist to whom the study was masked reviewed ALN images with specific attention to unclear margins, node matting, perinodal edema, and hilar effacement or replacement. Univariate and multivariate logistic regression analyses were used to obtain the sensitivity, specificity, and odds ratio (OR) for each feature.
Results: Our review included a total of 131 patients (64 with ENE and 67 without ENE). The respective sensitivity and specificity estimates for each feature were as follows: node matting, 52% and 84%; perinodal edema, 34% and 87%; unclear margins, 64% and 75%; hilar replacement, 71% and 42%; and hilar effacement, 74% and 60%. Univariate analysis showed a statistically significant association between features and ENE with ORs as follows: matting, 5.4; perinodal edema, 3.4; unclear margins, 5.2; and hilar replacement, 4.3. Multivariate analysis showed that matting and unclear margins were independently associated with ENE.
Conclusions: The sonographic features of unclear margins, node matting, perinodal edema, and hilar replacement have a statistically significant association with ENE. The sonographic features of unclear margins, node matting, and perinodal edema predict ENE with high specificity.