One hundred thirteen evaluable patients with previously untreated stage III breast carcinoma were treated with three monthly cycles of cyclophosphamide (CYC), doxorubicin (DOX), 5-fluorouracil (5-FU), vincristine (VCR), and prednisone (PRED) (CAFVP). Subsequently, 91 (81%) were deemed operable. Patients were then randomized to receive surgery or radiotherapy (RT) to determine which of these modalities afforded better local tumor control. All patients also received 2 additional years of CAFVP in a further attempt to eradicate local disease and systemic micrometastases. Forty-one of the randomized patients have relapsed. Approximately half of the initial relapses in each arm were local. The overall duration of disease control was similar following either modality, with a median of 29.2 months for surgery patients and 24.4 months for RT patients. Similarly, there was no major difference in survival related to randomized treatment with an overall median of 39 months (median follow-up 37 months). Pre- or perimenopausal status and inflammatory disease were associated with shorter disease control and survival. Treatment was generally well tolerated and toxicity was acceptable. This study demonstrates that prolonged control of stage III breast carcinoma can be achieved with combined modality therapy in which cytotoxic chemotherapy precedes and follows treatment directly primarily at the breast tumor, using either surgery or RT. Nevertheless, new regimens must be designed if significant advances that may lead to the cure of this disease are to be achieved.