We designed an ex vivo bone marrow treatment for breast cancer patients receiving high-dose chemotherapy and autologous bone marrow support (ABMS), using 4-hydroperoxycyclophosphamide (4-HC), an active derivative of cyclophosphamide with known activity against breast cancer. This phase I bone marrow purging trial used ficoll-separated mononuclear cells (MNC) (devoid of granulocytes and RBCs), as opposed to the buffy coat. Twenty-five patients with metastatic breast cancer were studied. Patients received three cycles of the Adriamycin (doxorubicin; Adria Laboratories, Columbus, OH), fluorouracil, and methotrexate (Duke AFM) regimen, followed by marrow harvest. An MNC fraction of marrow was prepared and treated with 4-HC in concentrations of 20 micrograms/mL (four patients), 40 micrograms/mL (four patients), 60 micrograms/mL (nine patients), or 80 micrograms/mL (eight patients) and cryopreserved. Patients then received high-dose systemic cyclophosphamide, cisplatin, and carmustine, followed by infusion of the purged marrow. The study end point was marrow engraftment, defined as WBC count greater than 1,000 cells per microliter. At the first three dose levels (20, 40, and 60 micrograms/mL 4-HC), there was no significant delay in time to engraftment (19, 20, and 23 days, respectively) compared with the unpurged historical controls (17 days). At 80 micrograms/mL, engraftment was significantly delayed compared with the lower concentrations (P = .027), and further escalation of 4-HC was not attempted. A significant correlation was observed between the time of leukocyte engraftment and the 4-HC concentration (P = .017). With a methylcellulose-based tissue culture assay, we demonstrated a statistically significant correlation between the colony-forming unit-granulocyte-macrophage (CFU-GM) content in the purged marrow and the days to engraftment. Ninety-five percent of patients responded clinically to the entire program, 55% of them completely. Longer follow-up is required to assess the ultimate benefit of intensive therapy on long-term survival.