The cell proliferation rate has been correlated to the response of breast carcinomas to preoperative chemotherapy (CT) and to disease outcome. However, this parameter is not yet used to select which tumours should be treated with preoperative CT. Furthermore, there is no consensus in the method used to evaluate cell proliferation. In poor prognosis breast carcinomas (PPBCs) treated by intensive preoperative CT, we compared the predictive value of S phase fraction (SPF), mitotic index (MI) and Ki67. We also evaluated the prognostic significance of the variation of the MI after CT. A series of 55 T2-T4N0N1M0 breast carcinomas were treated with 4 cycles of cyclophosphamide, 5-fluorouracil (5-FU) and doxorubicin. SPF was determined by flow cytometry on pre-therapeutic needle aspiration products. MI and Ki67 were evaluated on pre-therapeutic biopsy samples and on the tumours after CT. Fifteen patients (27%) had a pathological complete response (pCR), whereas 40 (73%) had residual disease. All three proliferative markers were found to have predictive value, but this value was higher for MI than for SPF (P = 0.04) and Ki67 (P = 0.03): the rate of pCR was 50% in cases with MI > 17/3.3 mm2, but was only 7% in cases with MI under this threshold (P = 0.0003). A significant decrease of MI (mean 10.97) was observed after CT (P = 0.001). Furthermore, we observed that even for patients with residual tumour, the variation of MI after CT was a prognostic parameter and overall survival. The sequential analysis of MI in breast cancers treated by preoperative CT thus provides a surrogate for predicting long-term outcome.