BI-RADS lexicon for US and mammography: interobserver variability and positive predictive value.

PURPOSE To retrospectively evaluate interobserver variability between breast radiologists by using terminology of the fourth edition of the Breast Imaging Reporting and Data System (BI-RADS) to categorize lesions on mammograms and sonograms and to retrospectively determine the positive predictive value (PPV) of BI-RADS categories 4a, 4b, and 4c. MATERIALS AND METHODS Institutional review board approval was obtained; informed consent was not required. This study was HIPAA compliant. Ninety-four consecutive lesions in 91 women who underwent image-guided biopsy comprised 59 masses, 32 calcifications, and three masses with calcification. Five radiologists retrospectively reviewed these lesions. Each observer described each lesion with BI-RADS terminology and assigned a final BI-RADS category. Interobserver variability was assessed with the Cohen kappa statistic. A pathologic diagnosis was available for all 94 lesions; 30 (32%) were malignant and 64 (68%) were benign. Pathologic analysis of benign lesions was performed on tissue obtained with image-guided core-needle biopsy. In cases referred for excisional biopsy after needle biopsy because of atypia or discordance, final surgical pathologic analysis was used for correlation with imaging findings. PPV for category 4 or 5 lesions was determined for all readers combined. RESULTS For ultrasonographic (US) descriptors, substantial agreement was obtained for lesion orientation, shape, and boundary (kappa = 0.61, 0.66, and 0.69, respectively). Moderate agreement was obtained for lesion margin and posterior acoustic features (kappa = 0.40 for both). Fair agreement was obtained for lesion echo pattern (kappa = 0.29). For mammographic descriptors, moderate agreement was obtained for mass shape, mass margin, and calcification distribution (kappa = 0.48, 0.48, and 0.50, respectively). Fair agreement was obtained for calcification description (kappa = 0.32). Slight agreement was obtained for mass density (kappa = 0.18). Fair agreement was obtained for final assessment category (kappa = 0.28). PPVs of BI-RADS category 4 and 5 assignments were as follows: category 4a, six (6%) of 102; category 4b, 17 (15%) of 110; category 4c, 48 (53%) of 91; and category 5, 71 (91%) of 78. CONCLUSION Interobserver agreement with the new BI-RADS terminology is good and validates the US lexicon. Subcategories 4a, 4b, and 4c are useful in predicting the likelihood of malignancy.