The aim of this study was to assess a potential advantage in survival by neoadjuvant as compared to adjuvant chemotherapy. 414 premenopausal patients with T2-T3 N0-N1 M0 breast cancer were randomised to receive either four cycles of neoadjuvant chemotherapy (cyclophosphamide, doxorubicin, 5-fluorouracil), followed by local-regional treatment (group I) or four cycles of adjuvant chemotherapy after primary irradiation +/- surgery (group II). Surgery was limited to those patients with a persisting mass after irradiation, and aimed to be as conservative as possible. 390 patients were evaluable. With a median follow-up of 54 months, we observed a statistically significant difference (P = 0.039) in survival in favour of the neoadjuvant chemotherapy group. A similar trend was seen when the time to metastatic recurrence was evaluated (P = 0.09). At this stage, no difference in disease-free interval or local recurrence between these two groups could be observed. The mean total dose of chemotherapy administered was similar in both groups. On average, group I had more intensive chemotherapy regimes (doxorubicin P = 0.02) but fewer treatment courses (P = 0.008) as compared to the treated patients in group II. Haematological tolerance was reduced when chemotherapy succeeded to exclusive irradiation. Breast conservation was identical for both groups at the end of primary treatment (82 and 77% for groups I and II, respectively). Of the 191 evaluable patients in the neoadjuvant treatment arm, 65% had an objective response (> 50% regression) following four cycles of chemotherapy. The objective response rate to primary irradiation (55 Gy) was 85%. Improved survival figures in the neoadjuvant treatment arm could be the result of the early initiation of chemotherapy, but we cannot exclude that this difference might be attributable to a slightly more aggressive treatment. So far, the trend in favour of decreased metastases was not statistically significant. The local control appeared similar in both subgroups.