Accuracy of breast MRI in predicting pathologic tumor size.

Abstract

Listen

Translate article

1109 Background: MRI use as a preoperative planning tool is increasing in women with breast cancer, yet the correlation between MRI and pathologic size of cancers is unclear. The purpose of this study was to determine the accuracy of MRI in predicting pathologic tumor size, and factors that affect this correlation. Methods: Clinicopathologic and imaging data from 84 patients diagnosed with invasive or in situ breast cancer from September 2010 to October 2011 who had preoperative MRI were reviewed. 12 patients who had neoadjuvant chemotherapy were excluded. MRI detected 147 lesions in the remaining 72 patients. Concordance between MRI and pathology size was determined using Spearman rho coefficients, and factors affecting the accuracy of MRI in predicting tumor size within +/- 0.5 cm were determined. Results: There was a modest correlation between MRI and pathology size for all MRI detected lesions (benign or malignant) with a Spearman coefficient of 0.53. Of the 147 MRI detected lesions, 45 (30.6%) had pathologic and MRI size correlating within +/- 0.5 cm; 76 (51.7%) were overestimated (>0.5cm) by MRI, and 26 (17.7%) were underestimated (>0.5cm). 101 (68.7%) of the 147 lesions were found to be malignant (either with invasive disease or DCIS). In this subgroup, 35 lesions (34.7%) had an MRI size within +/- 0.5 cm of the pathologic size; 40 (39.6%) were overestimated by MRI and 26 (25.7%) were underestimated. Patient age, tumor histology, LVI and grade did not predict concordance between pathologic and MRI size. However, small MRI lesion size more accurately correlated with pathologic tumor size. While 51.1% of tumors that had concordant MRI and pathologic findings within 0.5 cm were <1 cm on MRI, no tumor found to be > 5 cm on MRI was within +/-0.5 cm on final pathology (p=0.001). Conclusions: MRI accurately predicts pathologic tumor size only when the size of the lesion on MRI is <1 cm.