Assessment of Diagnostic Accuracy and Efficiency of Categories 4 and 5 of the Second Edition of the BI-RADS Ultrasound Lexicon in Diagnosing Breast Lesions

Abstract

The purpose of this study was to evaluate the diagnostic accuracy and efficiency of categories 4 and 5 of the second edition of the Breast Imaging Reporting and Data System (BI-RADS) ultrasound (US) lexicon in diagnosing breast lesions. In our retrospective study, 579 lesions in 544 patients were assessed by US as the preliminary diagnosis and classified in subcategories 4a-4c and category 5 based on the second edition of the BI-RADS US lexicon with some obvious changes, such as the redefined margin, new calcification type, associated features and some special cases. Inter-observer agreement was determined. Ultrasound results were compared with the pathologic results for confirmation. Positive predictive values (PPVs) of subcategories 4a-4c were compared with theoretical values using the χ(2) test; the binomial test was used for category 5 lesions. Of the 579 lesions, 212 were confirmed as benign (36.61%), and the remaining 367 lesions were confirmed as borderline/malignant (63.39%). Inter-observer agreement was moderate for subcategories 4a-4c (κ = 0.52), moderate for subcategories 4a-4c and category 5 (κ = 0.56) and substantial for categories 4 and 5 (κ = 0.67). The PPVs for subcategories 4a-4c were 23.74%, 70.67% and 81.25%, respectively. In addition, the total PPV for category 4 was 46.92% (183/390), and the total PPV for category 5 was 97.35% (184/189). Statistical results revealed that the PPVs of subcategories 4a and 4b differed significantly from the theoretical values (p < 0.05); the PPVs of subcategory 4c and category 5 were significantly correlated with the theoretical PPVs (p > 0.05). In conclusion, subcategories 4a and 4b have lower diagnostic efficiency than subcategory 4c and category 5. Inter-observer agreement for subcategories 4a-4c remains to be improved. The most common features of subcategories 4a-4c differ, but overlap. It is recommended that inexperienced doctors in primary hospitals not classify lesions into subcategories in clinical practice.

Keywords: BI-RADS categories 4 and 5; Breast Imaging Reporting and Data System; Breast lesions; Ultrasound; pathology.