Seventy-five patients (77 breasts) with invasive breast cancer were prospectively included. Excision specimens were processed using complete embedding. Microscopic findings were reconstructed and correlated with contrast-enhanced MRI. Tumors were stratified by absence or presence of occult disease ?10 mm from the MRI-visible lesion: BCLE and non-BCLE, respectively. Imaging and pathology characteristics were evaluated for their ability to discriminate between BCLE and non-BCLE. Multivariate binary logistic regression was employed to create a prediction model for BCLE.
At univariate analysis, imaging as well as pathology characteristics were indicative for BCLE (39/77 = 51 % ). At multivariate analysis, a mass on mammography, the absence of tumor washout, positive ER and low quantity of DCIS in the index tumor retained significance (area under ROC curve = 0.87).
Pre-treatment assessment of mammography findings, MRI washout kinetics, ER status and quantity of DCIS in the index tumor has the potential to accurately identify BCLE.