To evaluate the accuracy of clinical examination and of three imaging modalities (mammography, ultrasound, and magnetic resonance imaging (MRI)) to assess the tumour response to a pre-operative regimen of concurrent radiochemotherapy for large breast cancers, and to anticipate the eventual complete pathologic response.

Sixty breast cancer women not amenable to conserving surgery at initial presentation were accrued in a phase II study. Treatment consisted of four cycles of 5-fluorouracil–vinorelbine with, starting with the second cycle of chemotherapy, radiotherapy to the breast and regional lymph nodes. Breast-conserving surgery or mastectomy was performed 4 to 6 weeks after completion of irradiation. Imaging assessment was performed before chemotherapy and preoperatively.

Referring to pathologic data, the MRI assessment performed best. A 50% or greater decrease in the largest tumour diameter in MRI was linked to complete pathologic response with 81% sensitivity and 75% specificity. MRI–pathologic correlations helped understand the pitfalls in MRI interpretation that led to overestimating/underestimating some tumour responses.

Compared with the other assessment modalities, MRI substantially improved the prediction of pathologic tumour response. Pitfalls or limits in MRI interpretation in this specific setting were better understood.