Mitchell Dennis Schnall MD, Presenter: Nothing to Disclose

Dave Bluemke MD, Abstract Co-Author: Nothing to Disclose

Constantine Gatsonis DPHIL, Abstract Co-Author: Nothing to Disclose

Jeffery D Blume DPHIL, Abstract Co-Author: Nothing to Disclose

Gia Ann DeAngelis MD, Abstract Co-Author: Nothing to Disclose

Nanette D. DeBruhl MD, Abstract Co-Author: Nothing to Disclose

Steven Edward Harms MD, Abstract Co-Author: Nothing to Disclose

Sylvia Heywang-Köbrunner MS, Abstract Co-Author: Nothing to Disclose

Nola Michiko Hylton PhD, Abstract Co-Author: Nothing to Disclose

Christiane Katharina Kuhl MD, Abstract Co-Author: Nothing to Disclose

Etta D. Pisano MD, Abstract Co-Author: Nothing to Disclose

Stanley F. Smazal MD, Abstract Co-Author: Nothing to Disclose

Carol B. Stelling MD, Abstract Co-Author: Nothing to Disclose

Paul T. Weatherall MD, Abstract Co-Author: Nothing to Disclose

Constance Dobbins Lehman MD, Abstract Co-Author: Nothing to Disclose

Stuart Schnitt MD, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

Magnetic resonance (MR) imaging is gaining more widespread use for the detection, diagnosis and staging of breast cancer. Although highly sensitive, concerns exist regarding specificity. This study represents a multi-center clinical trial performed by the IBMC aimed at characterizing the diagnostic performance of Breast MRI.

Patients were eligible if they were referred for breast biopsy for one of the following reasons:1)American College of Radiology category 4 or 5 mammographic assessment, or 2)suspicious clinical or ultrasound finding. MR imaging was performed at one of 14 participating sites prior to breast biopsy. The imaging protocol included a 3D high spatial resolution acquisition. Patients with enhancing findings were invited back for a 15 sec time resolution dynamic study. Imaging results were interpreted at each site blinded to pathologic results.

821 patients completed the study at 14 enrolling sites. The area under the ROC curve pooled over all institutions was 0.88 (95% confidence interval, 0.86 to 0.91,range 0.78 to 0.91). The sensitivity and specificity of MR imaging for breast cancer were 88.1% (95% confidence interval, 84.6 to 91.1%) and 67.4% (95% confidence interval, 62.7 to 71.9%), respectively. MR imaging performance was not significantly affected by breast density, tumor histology, or menopausal status. The positive predictive values for MR imaging and mammography were 72.4% (95% confidence interval, 68.2 to 76.3%) and 52.8% (95% confidence interval, 49.0 to 56.6%), respectively (p<0.005). Predictive imaging features included Margin, Enhancement Intensity, and Qualitative Enhancement Kinetics. Multivariate models were constructed from individual feature ratings with model ROC areas up to .862 (95% confidence interval,.86 to .867).

For patients with suspicious lesions identified prior to planned breast biopsy, breast MR imaging is predictive of diagnosis. Using a BiRads scale, the observed specificity is relatively high, however the observed sensitivity was less than previously reported. Optimized multivariate models using key interpretation features demonstrated equivalent diagnostic accuracy as reader overall assessment.

Schnall, M, Bluemke, D, Gatsonis, C, Blume, J, DeAngelis, G, DeBruhl, N, Harms, S, Heywang-Köbrunner, S, Hylton, N, Kuhl, C, Pisano, E, Smazal, S, Stelling, C, Weatherall, P, Lehman, C, Schnitt, S, et al, , Diagnostic Performance of Breast MRI Observed in International Breast MRI Consortium Trial 6883.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4408404.html