Four hundred thirty-eight patients with metastatic breast cancer treated with nine prospective randomized trials were reviewed to analyze, characterize, and compare those with complete regression (CR) (N=49 [11%]) with those with noncomplete regression (non-CR) (N=389 [89%]). Site of dominant disease was identical in the CR and non-CR patients. However, CR was statistically more likely when the disease-free interval and the postmenopausal status was five years or less. Survival and time to progression occurred at sites of prior dominant disease, except that relapses occurred in the CNS in 14%. Six CR patients received prolonged chemotherapy and were believed to be "cured" when treatment was discontinued; however, all except one have had relapses.