We have previously reported that repeated cycles of high-dose therapy (HDT), can be supported by unmanipulated autologous PBPC. Here we investigate whether purified CD34+ cells, obtained by immunomagnetic separation using the Isolex 300i device, can support such therapy.Twenty-nine consecutive patients with metastatic breast cancer had PBPC mobilized and harvested following chemotherapy and G-CSF (10 microg/kg per day). Patients with > 4.0 x 10(6)/kg CD34+ cells in the apheresis product underwent CD34-selection using the Isolex 300i (v2.0) device. All cells collected were equally divided into three aliquots and cryopreserved. Patients who did not achieve this threshold had unmanipulated cells collected and stored. Patients subsequently received three cycles of HDT with paclitaxel (175 mg/m2), thiotepa (300 mg/m2) and either ifosfamide (10 g/m2) or cyclophosphamide (4 g/m2). It was intended for patients to receive CD34-selected cells to support each of the three cycles of HDT (i.e 1/3 for each cycle) and to compare hemopoietic recovery between patients receiving CD34-selected cells or unmanipulated cells.Thirteen of the 29 patients (45%) did not mobilize sufficient CD34+ cells to undergo CD34-selection. The remaining 16 patients underwent CD34-selection with a median purity of 84.3% (range: 16.3-96.1%) and yield of 34% (range: 1-60%). Fifteen of these patients proceeded to HDT and 42 of the planned 45 cycles were administered. Nine patients had all three HDT cycles supported by CD34-selected cells. The median number of CD34-selected cells (x 10(6)/kg) infused per cycle was 1.5 (range: 0.04-3.01). Three of the 15 patients required infusion of 'back-up' unmanipulated cells because of delayed neutrophil recovery. Of the 13 patients whose PBPCs did not undergo CD34+ cell selection, 11 proceeded to HDT with a median of 3.2 x 10(6)/kg (range: 2.0-4.4) unselected cells infused per cycle and 31 of 33 planned cycles were delivered. When hemopoietic recovery was compared between cycles of HDT supported by CD34-selected (n = 34) and unmanipulated cells (n = 31), there was a modest slowing in the patients receiving CD34-selected cells; time to ANC > 1.0 x 10(9)/L = 11 days versus 10 days (P = 0.0122) and platelets > 20 x 10(9)/L = 14 days versus 13 days (P = 0.0009). No difference in recovery to 50 x 10(9)/L was observed (P = 0.54).We have demonstrated that Isolex 300i CD34-selected cells are capable of supporting multiple cycles of HDT. However, we were unable to acquire sufficient CD34+ cells to perform this processing in 45% (13/29) of patients and further improvements in yield are required to overcome the modest delay in neutrophil and platelet recovery.