Primary chemotherapy in localised breast cancer may prevent tumour spread during surgical treatment and reduce proliferation of micrometastases. A randomised clinical trial, in 196 premenopausal and postmenopausal patients with operable (T2-3, N0-1b) breast cancer, was started in November 1983 at the Institut Curie to compare neoadjuvant and adjuvant regimens of chemotherapy with radiotherapy with or without surgery. The patients have been followed up for 35-70 months (median 54). A neoadjuvant group received two monthly cycles of intravenous doxorubicin/cyclophosphamide/5-fluorouracil before locoregional therapy and four cycles subsequently. Six monthly cycles following locoregional therapy were administered to the adjuvant group. Because of inclusion of postmenopausal and/or node-negative patients, compliance was less than optimal in 39 patients who were analysed separately according to actual dose received. Tumour response, evaluated after two cycles of neoadjuvant chemotherapy, was significantly associated with dose (P = 0.003). Survival showed a slight non-significant advantage for the neoadjuvant group. Survival plotted by actual dose was also similar. Neoadjuvant chemotherapy was safe and at least as effective as the adjuvant regimen. Patients have been accrued to a subsequent larger trial of chemotherapy as first-line treatment.