Positive predictive value of breast MRI in cancer detection.

Abstract

Abstract Abstract #4007 Purpose: 
 Positive predictive values (PPVs) are among the performance benchmarks which have been reported when examining clinical outcomes related to mammography. Breast MRI is increasingly being used as a tool for screening and diagnostic purposes and current performance benchmarks for breast MRI have yet to be established. The purpose of this study is to assess breast MRI performance in our academic clinical practice via comparison of breast MRI PPVs with reported PPV values for mammography. Specifically, we sought to determine the abnormal interpretation rate of breast MRI examinations, and then to calculate the PPVs related to recommendations for further diagnostic evaluation and to recommendations for biopsy of breast lesions visualized on MRI.
 Materials and Methods: 
 This study was approved by our institutional review board and its methodology is in compliance with federal HIPAA regulations. A retrospective search of our institutional breast imaging database identified all breast MRI examinations performed 4/1/07 - 3/31/08. Data abstracted included: patient age, clinical indication for breast MRI, Breast Imaging Reporting and Data System (BI-RADS) assessments for all breast imaging examinations, clinical follow-up, and pathology results. Positive exams were defined as those with an initial BI-RADS 0, 4, or 5. Exams with suspicious findings were defined as those with a final BI-RADS 4 or 5 after additional workup. PPVs were calculated based on American College of Radiology recommendations and using the following definitions: PPV1 = exams with cancer/positive exams, PPV 2 = exams with cancer/exams with suspicious findings, and PPV 3 = cancers/biopsied lesions.
 Results:
 2,600 breast MRI examinations were performed during the study period. After exclusion criteria were applied, 285 positive exams (BI-RADS 0, 4 or 5) were identified for further analysis. The abnormal interpretation rate was 10.96% (285/2600, 95% CI: 9.7% - 12.2%). After non-invasive workup, including mammographic and sonographic correlation, 186 lesions on 177 exams had biopsy recommendations (final BI-RADS 4 or 5). Following biopsy, 55 malignancies were diagnosed (35 invasive carcinomas, 18 DCIS, 2 phyllodes/sarcoma). The overall PPV for a positive breast MRI exam (PPV1) was 17.2% (49/285). The PPV for suspicious findings after non-invasive workup (PPV2) was 27.7% (49/177). The PPV for biopsied lesions (PPV3) was 29.6% (55/186). When stratified by MRI indication, the cancer yield was lower in the screening setting (PPV1 = 9.7%, PPV2 = 19.7%, PPV3 = 19.4%) than in the diagnostic setting (PPV1 = 23.0%, PPV2 = 31.9%, PPV3 = 34.7%).
 Conclusions:
 Diagnostic breast MRI PPVs are comparable to published benchmarks for diagnostic mammography. Screening breast MRI PPVs are lower than published values for screening mammography, and lower when compared with reports from high-risk screening breast MRI trials. This may reflect characteristics of the population undergoing screening breast MRI in our academic clinical practice. Factors such as cancer detection rate and the population risk for breast cancer should be considered when examining PPVs for breast MRI. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 4007.