The aim of this retrospective study was to assess predictive factors for clinical response to preoperative chemotherapy and prognostic factors for survival. From 1981 to 1992, 936 patients with T2-T3, N0-N1 breast cancer who received 2–6 months (median 4) of preoperative chemotherapy were selected from the Institute Curie database. Preoperative treatment was followed by surgery and/or radiotherapy. Median follow-up was 8.5 years (range 7–211 months). The objective response rate before surgery and/or radiotherapy was 58.3%. In stepwise multivariate analysis (Cox model), favourable prognostic factors for survival were the absence of pathological axillary lymph node involvement (Relative Risk (RR) 1.54; P=0.0004), low histological tumour grade (RR=1.54; P=0.0017), clinical response to preoperative chemotherapy (RR=1.45, P=0.0013), positive progesterone receptor (PR) status (RR=1.56; P=0.0001), smaller tumour size (RR=1.37; P=0.005) and lack of clinical lymph node involvement (RR=1.42; P=0.007). The association of clinical tumour response with survival is independent of the baseline characteristics of the tumour. Clinical response could be used as a surrogate marker for evaluation of the efficacy of neoadjuvant chemotherapy before assessment of the pathological response.