272 women with operable breast adenocarcinomas larger than 3 cm were included in a randomized trial. The patients in group A (n = 138) with histological nodal involvement (N+) or a lack of estrogen and progesterone receptors (EPR-) were treated by initial mastectomy and axillary node dissection + adjuvant chemotherapy. Those in group B (n = 134) were treated by initial chemotherapy (the same as in group A) followed by loco-regional treatment, adjusted according to their response to chemotherapy. Prognostic factors were identical in the two groups. In group A, 32 patients received no adjuvant treatment (N- and EPR+), while 104 were given adjuvant chemotherapy (N+ and/or EPR-). Two patients were lost to follow-up. In group B, all patients received initial chemotherapy; 44 were in complete clinical remission and were treated with radiotherapy only; 40 with residual tumor (less than 20 mm) were treated with tumorectomy + axillary node dissection + radiotherapy; 49 with residual tumors (greater than 20 mm) had mastectomies. Conservative treatment was administered to 84 patients in group B (62.6%). EPR-tumors responded better to chemotherapy than did EPR+ ones (p = .003). After a median follow-up of 34 months, isolated local recurrences were more frequent in the group with initial chemotherapy, which, however, experienced a better overall survival (p = 0.04).