Neda Isabel Sedora-Roman MD, Presenter: Nothing to Disclose

Vandana Mukesh Dialani MD, Abstract Co-Author: Nothing to Disclose

Richard Earnest Sharpe MD, MBA, Abstract Co-Author: Nothing to Disclose

Shambhavi Venkataraman MD, Abstract Co-Author: Nothing to Disclose

Priscilla Jennings Slanetz MD, MPH, Abstract Co-Author: Nothing to Disclose

Valerie J. Fein-Zachary MD, Abstract Co-Author: Research Consultant, Siemens AG

Jordana Phillips MD, Abstract Co-Author: Nothing to Disclose

Tejas S. Mehta MD, MPH, Abstract Co-Author: Nothing to Disclose

Established biopsy performance benchmarks for mammography (MG) are widely utilized, yet there exists no such performance benchmarks for breast magnetic resonance imaging (MRI). In this study we audited our breast MRI biopsy performance using established MG benchmarks, and reviewed whether these benchmarks could be applied to a breast MRI practice.

We retrospectively reviewed all breast MRIs performed at a large academic center from 10/1/12-9/31/13. Exams were interpreted by 7 radiologists, all with fellowship training and/or more than 10 years of experience in breast imaging. All MRIs reported as BI-RADS 4 or 5, and the associated core and/or surgical biopsy pathology results were tabulated. Using national MG benchmarks, we derived mean performance parameters, including abnormal interpretation rate, positive predictive value (PPV), cancer detection rate, percentage of minimal cancers (Stage 0 and stage 1 cancers <= 1cm) and axillary node-negative invasive cancers.

Of the 725 breast MRIs performed over a one year period, 522 (76.1%) were BI-RADS 1 or 2, 75 (10.3%) BI-RADS 3, 61 (8.4%) BI-RADS 4, 4 (0.6%) BI-RADS 5, and 33 (4.6%) BI-RADS 6. All 65 (100%) patients with BI-RADS 4 or 5 underwent biopsy: 22 (33.8%) were positive for cancer, 31 (47.7%) were high risk lesions (23 atypia; 8 LCIS), and 12 (18.5%) were benign. Abnormal interpretation rate was 3.3% (22/725); PPV for abnormal interpretation (PPV1), biopsy recommended (PPV2), and biopsy performed (PPV3), were all 33.84% (22/65); cancer detection rate was 33/1000 (22/725). Of the 22 cancers, 10 (45.5%) were in-situ and 12 (55.5%) were invasive. 9/12 (75.0%) invasive cancers were less than 1cm with negative lymph nodes, thus the percentage of minimal cancers was 86.4% (19/22). Two cancers were positive for lymph nodes and the largest invasive cancer size was 2.2 cm.

The breast MRI benchmarks closely mimic those of MG, with the expected exception of a higher cancer detection rate for breast MRI, likely resulting from its known higher sensitivity. Our study suggests established national MG benchmarks can be used to reliably audit a breast MRI practice, until which point MRI specific benchmarks are created.

Auditing a breast MRI practice is essential, and accepted mammography benchmarks are reasonable performance evaluation tools, in the absence of MRI specific benchmarks.

Sedora-Roman, N, Dialani, V, Sharpe, R, Venkataraman, S, Slanetz, P, Fein-Zachary, V, Phillips, J, Mehta, T, Proposed Breast Biopsy Performance Benchmarks for MRI Based on an Audit of a Large Academic Center.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14011192.html