Prevalence and Predictive Value of BI-RADS 3, 4, and 5 Lesions Detected on Breast MRI: Correlation with Study Indication

Abstract

This study aims to determine the prevalence and predictive value of Breast Imaging Reporting and Data System (BI-RADS) 3, 4, and 5 findings on breast magnetic resonance imaging (MRI) and to evaluate the impact of study indication on the predictive value of BI-RADS categories. This institutional review board approved, Health Insurance Portability and Accountability Act (HIPAA) compliant retrospective review of our breast MRI database from 2009 to 2011, of 5778 contrast-enhanced studies in 3360 patients was performed. At our institution, each breast receives an individual BI-RADS assessment. Breast MRI reports and electronic medical records were reviewed to obtain BI-RADS assessment, patient demographics, and outcomes. Univariate analysis was performed with Fisher exact and chi-square tests. A total of 9216 BI-RADS assessments were assigned during the study period: 7879 (85.5%) BI-RADS 1 and 2, 567 (6.2%) BI-RADS 3, 715 (7.8%) BI-RADS 4, and 55 (0.6%) BI-RADS 5 assessments. The frequency of BI-RADS 3, 4, and 5 assessments was higher in studies performed for diagnostic (7.8%, 14.6%, 1.6%, respectively) than screening (5.2%, 4.0%, 0.1%) indications (P < 0.01). A total of 663 BI-RADS 4 and 5 lesions were biopsied with 209 (31.5%) malignant and 454 (68.5%) benign outcomes. The overall cancer rate for BI-RADS 3 findings was 1.9% (11 of 567) with no difference observed by study indication (diagnostic, 1.6%; screening, 2.3%; P = 0.76). The positive predictive value (PPV2) of BI-RADS 4 and 5 was higher for diagnostic (29.1%, 154 of 530) than for screening (22.9%, 55 of 240) indications. Abnormal interpretation rates and PPV2 for MRIs performed for diagnostic indications are higher than for screening indications. Similar to mammography, breast MRI audits should be separated by study indication.