Dose-intensive chemotherapy appears to be important in the treatment of patients with recurrent solid tumors. Expanding upon our prior experience, we report the results of our most recent approach to administering dose-intensive therapy using four cycles of moderately high-dose chemotherapy with hematopoietic cell support for patients with metastatic breast cancer. This outpatient therapy includes high-dose melphalan, thiotepa, and paclitaxel for two cycles followed by mitoxantrone, thiotepa, and paclitaxel for two cycles, with each cycle supported with autologous peripheral blood progenitor cells (PBPCs). Between December 1994 and June 1996, 16 patients with recurrent or refractory breast cancer were enrolled in this prospective study. They had received a median of two previous chemotherapy regimens, with a median of nine prior cycles of chemotherapy. For mobilization of autologous PBPCs, patients received cyclophosphamide, 4 g/m2, followed by granulocyte colony-stimulating factor (G-CSF). PBPCs were collected by apheresis. Each day's collection was divided into four equal fractions, and each fraction was infused after each cycle of combination therapy. Cycles 1 and 2 consisted of melphalan, 80 mg/m2, thiotepa, 300 mg/m2, and paclitaxel, 200 mg/m2. Cycles 3 and 4 were comprised of mitoxantrone, 30 mg/m2, and thiotepa and paclitaxel at the same doses as in the first two cycles. The cyclophosphamide infusion was administered in the hospital, whereas all subsequent infusions of chemotherapy and PBPCs were performed on an outpatient basis. The first seven patients were randomized to receive alternate cycle G-CSF or placebo on day +1 of each cycle. Including the initial pulse of cyclophosphamide, 67 (84%) of a planned 80 total courses of chemotherapy were delivered. Of the planned 64 cycles of high-dose combination chemotherapy, 52 cycles (81%) were delivered. Treatment was discontinued for progressive disease (one patient) or morbidity (five patients). Twelve of 16 patients completed at least three cycles of therapy. Nine patients completed all four cycles. One death resulted from fungal sepsis. In 20 cycles delivered to the first seven patients, day +1 G-CSF versus placebo was administered, with a median WBC recovery of 10 versus 13 days, respectively (P = 0.048 in cycle 1). The median duration of response was almost 9 months, and the median survival was 18 months after therapy. With a median follow-up of 1.5 years and longest follow-up of 4.2 years, two patients continue to be without evidence of disease. The 3-year event-free survival, freedom from progression, and overall survival are 19%, 20%, and 31%, respectively. This four-cycle regimen of high-dose combination therapy supported with hematopoietic progenitor cells is feasible, but it is associated with a range of posttransplant complications. The efficacy of such a treatment would have to be substantially superior to that of other currently available therapies, including single autologous transplant procedures, to justify the prolonged period of treatment, multiple episodes of pancytopenia, and associated toxicities, including infectious risks. G-CSF administration after each PBPC infusion appears to accelerate time to neutrophil recovery but does not affect red cell or platelet engraftment.