Positive predictive value of the breast imaging reporting and data system

Background: The American College of Radiology has established guidelines fo r outcomes monitoring known as the Breast Imaging Reporting and Data System (BIRADS). These recommendations include calculation of positive predictive values (PPV) and tracking of both benign and malignant histology. We colle cted this data for 688 radiographically guided biopsies and organized it ac cording to the BIRADS assessment categories. The objective was to evaluate the contribution of the BIRAD System when used to stratify PPV, histology, and biopsy modality data according to the overall assessment rating. Study Design: This study included data from 688 image-guided biopsies. Mamm ographic studies were either assigned a BIRADS rating at the time of examin ation or, if the image was taken before our use of BIRADS, examined retrosp ectively and rated. In these retrospective cases, the histologic outcomes o f the biopsy remained unknown to the radiologist until ratings were assigne d. Positive predictive value was calculated for each BIRADS category. Results: The overall PPV for the sample was 0.23. The PPVs increased with i ncreasing level of suspicion as follows: category 1 (0.0), category 2 (0.04 ), category 3 (0.03), category 4 (0.23), category 5 (0.92). Category 1 lesi ons represented 0.1% of the biopsies; category 2, 3.6%; category 3, 46.8%; category 4, 34.0%; and category 5, 15.4%. The most common histologic diagno ses of benign lesions biopsied were fibroadenoma and fibrocystic changes-pr oliferative and nonproliferative. The most common histologic diagnoses of m alignant lesions biopsied were infiltrating ductal carcinoma and ductal car cinoma in situ. Utilization rates of the biopsy techniques varied by BIRADS category. Conclusions: Our study revealed that BIRADS does improve the quality of the risk assessment information by making the PPV more specific to a patient's mammogram rather than simply related to an overall PPV. Our histology anal ysis showed category 3 and category 4 benign biopsies were predominantly be cause of fibrocystic changes. Category 5 lesions were predominantly invasiv e ductal carcinoma. Analysis of biopsy modalities indicated the preferred m ethod for management of radiographically detected lesions evolved from ster eotactic core biopsy to directional, vacuum-assisted biopsy over the course of the study. (J Am Cell Surg 1993;183:34-40. (C) 1993 by the American Col lege of Surgeons).