We have used high-dose therapy followed by randomization to receive or not receive autologous bone marrow infusion in 58 stage IV breast cancer patients (all were estrogen receptor-negative [ER-] or primary hormonal refractory). Patients received a median of four courses of induction chemotherapy and if stable or responding received two courses of intensive therapy with cyclophosphamide 4.5 to 5.25 g/m2, etoposide 750 to 1,200 mg/m2, and cisplatin 120 to 180 mg/m2 (CVP). The complete remission (CR) rate after completion of the induction and intensive phases was 55%. Median progression-free interval from induction is 57 weeks with a projected 2-year progression-free survival of approximately 25%. Six of seven patients followed for greater than 2 years are still progression-free. Three favorable features predicted for improved progression-free survival are the following: (1) absent liver involvement, (2) absent soft tissue involvement, and (3) less than or equal to two disease sites (P values of .001, .013, and .048, respectively). Hormonal and menopausal status did not predict for progression-free survival. Major toxicities were infectious secondary to neutropenia, with a 93% incidence of fever and a 38% incidence of sepsis. The treatment-related mortality rate was 9%, with three infectious, one coronary, and one late hemorrhage-related death of unknown cause. Longer follow-up is still needed to truly evaluate the possibility of long-term disease-free survival, but further studies of this approach to high-dose intensification in poor prognostic groups of stage IV breast cancer appear warranted.