Accuracy of MRI for treatment response assessment after taxane- and anthracycline-based neoadjuvant chemotherapy in HER2-negative breast cancer.

Background: Studies suggest that MRI is an accurate means for assessing tumor size after neoadjuvant chemotherapy (NAC). However, accuracy might be dependent on the receptor status of tumors. MRI accuracy for response assessment after homogenous NAC in a relative large group of patients with stage II/III HER2-negative breast cancer has not been reported before.

Methods: 250 patients from 26 hospitals received NAC (docetaxel, adriamycin and cyclophosphamide) in the context of the NEOZOTAC trial. MRI was done after 3 cycles and post-NAC. Imaging (RECIST 1.1) and pathological (Miller and Payne) responses were recorded. Accuracy measures were calculated and MRI and pathologically assessed tumor sizes were correlated. Tumor size over- and underestimation were quantified.

Results: Accuracy of MRI for determining pathological complete response (pCR) was 76%. The ROC-curve of MRI response and pCR had an area under the curve value of 0.63 (95% C.I. 0.52-0.74). The correlation coefficient of MRI and histopathological tumor measurements was 0.46 (p < 0.001). Correlations were different for ER-positive (r = 0.40, p < 0.001) and ER-negative (r = 0.76, p < 0.001) breast tumors. MRI under- and overestimated the tumor size in 47% and 40% of all patients. In cases of substantial tumor size underestimation (>2 cm), surgical margins were more often tumor positive compared to the rest of the patients (33% vs.12%, p = 0.005).

Conclusion: MRI measurements correlated moderately with tumor size on the surgical specimen. Only in ER-negative breast tumors, MRI tumor sizes correlated sufficiently with residual tumor size on the pathological specimen. Therefore, post-NAC MRI should be interpreted with caution.