Abstract Archives of the RSNA, 2012
Dania Daye BS, Presenter: Nothing to Disclose
Sara C. Gavenonis MD, Abstract Co-Author: Research support, Hologic, Inc
Brad M. Keller PhD, Abstract Co-Author: Nothing to Disclose
Ahmed Bilal Ashraf PhD, Abstract Co-Author: Nothing to Disclose
Carolyn J. Mies MD, Abstract Co-Author: Nothing to Disclose
Michael Feldman, Abstract Co-Author: Nothing to Disclose
Mark Alan Rosen MD, PhD, Abstract Co-Author: Consultant, General Electric Company
Despina Kontos PhD, Abstract Co-Author: Research Consultant, Hologic, Inc
Support agreement, Hologic, Inc
To investigate the value of breast MRI tumor features as prognostic markers for breast cancer recurrence as determined by a validated tumor gene expression assay.
Breast DCE-MRI images were retrospectively collected per HIPAA and IRB approval from 61 women diagnosed with estrogen receptor positive/node negative invasive breast cancer. The women had previously undergone Oncotype Dx (GenomicHealth Inc.) profiling of their tumor. The Oncotype DX assay provides prediction of 10-year breast cancer recurrence, using a score stratified into 3 risk categories (risk: low ≤17, medium=18-30, high ≥ 31). Two experienced radiologists assessed 7 tumor-specific features: disease multifocality, lesion size, lesion shape, margin morphology, enhancement amount, enhancement morphology and associated non-mass enhancement (NME). Inter-reader agreement was assessed using Cohen's Kappa. Features were averaged between the two readers to account for inter-reader variability. Multi-class linear discriminant analysis (LDA) was performed to determine the association between tumor features and the Oncotype DX recurrence risk category. Performance of the LDA model was assessed using ROC analysis.
Readers’ agreement is substantial for multifocality (kappa>0.6); moderate for lesion size, shape, margins, enhancement morphology and NME (0.4<kappa<0.6); and fair for enhancement amount (0.2<kappa<0.4). LDA yields two features, multifocality and lesion size, as significant in predicting the recurrence risk categories (p<0.05). A three-class LDA model including multifocality and lesion size has significant agreement in predicting the Oncotype DX recurrence risk categories (kappa=0.41, p=0.01). ROC analysis reveals discriminant capacity in distinguishing between recurrence risk categories (AUC=0.64 for low vs intermediate risk, 0.66 for intermediate vs high risk and 0.76 for low vs high risk).
Breast MRI tumor features may be predictive of breast cancer recurrence. Multifocal and larger lesions tend to have an increased recurrence risk. When combined in an LDA model, MRI tumor features can predict the recurrence risk category as determined by the Oncotype DX assay.
Breast MRI tumor-specific interpretation features may have value in assessing breast cancer prognosis and aid in treatments decisions. Larger clinical studies are needed to validate these findings.
Daye, D,
Gavenonis, S,
Keller, B,
Ashraf, A,
Mies, C,
Feldman, M,
Rosen, M,
Kontos, D,
Breast MRI Tumor Features as Predictive Markers of Breast Cancer Recurrence. Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL.
http://archive.rsna.org/2012/12035454.html