Magnetic Resonance Imaging Patterns of Tumor Regression After Neoadjuvant Chemotherapy in Breast Cancer Patients: Correlation With Pathological Response Grading System Based on Tumor Cellularity

Purpose: The objectives of the study were to analyze the tumor shrinkage pattern on magnetic resonance imaging (MRI) after neoadjuvant chemotherapy and to evaluate whether there is any difference in shrinkage pattern between pathological responder and nonresponder groups. In addition, we wanted to compare tumor diameter obtained from MRI with histological diameter according to the tumor shrinkage pattern. Methods: Between July 2008 and December 2010, 55 consecutive patients (56 lesions) with pathologically proven breast cancer who underwent neoadjuvant chemotherapy followed by surgery were retrospectively enrolled. The shrinkage pattern was classified into 4 categories: I (concentric shrinkage without surrounding lesion), II (concentric shrinkage with surrounding lesions), III (shrinkage with residual multinodular lesions, and IV (diffuse contrast enhancement in whole quadrants). Histological regression was scored on a 5-point scale regarding tumor cellularity reduction (Miller-Payne grading system). Patients with Miller-Payne grade 1 or 2 were classified into the nonresponder group, and patients with grade 3, 4, or 5 were in the responder group. Results: Of 56 lesions, pattern I was seen in 29 lesions, pattern II in 13 lesions, pattern III in 5 lesions, and pattern IV in 4 lesions. Three lesions were not visualized on MRI after neoadjuvant chemotherapy, and 2 lesions were increased in size. There was a statistically significant difference in the tumor shrinkage pattern between responder and nonresponder groups (P = 0.017). All 5 lesions with type III shrinkage pattern were found in the responder group, and all 4 lesions with pattern IV were in the nonresponder group. Magnetic resonance imaging diameter of lesions with types I, II, and IV patterns showed significant correlation with the histological diameter. Among them, the correlation factor was highest in pattern IV (P = 0.94, P < 0.001) followed by pattern I (rho = 0.67, P < 0.01) and pattern II (rho = 0.502, P = 0.08). However, in type III shrinkage pattern, tumor size measured on MRI was not significantly correlated with histological size (P = 0.87). Conclusions: Types III and I shrinkage patterns were more frequently observed in the pathological responder group, and type IV was more frequently noted in the nonresponder group. Tumor diameter measured on MRI showed strong correlation with histological diameter in lesions with types I and IV shrinkage patterns, whereas lesions with type III did not show significant correlation. Type II pattern showed similar frequencies between the 2 groups and moderate correlation between sizes obtained from MRI and histology.