For the treatment of childhood solid tumors, we performed a pilot feasibility study of consecutive high-dose therapies, in which each course was followed by transplantation with granulocyte colony-stimulating factor-mobilized peripheral blood cells which had been separated into CD34-positive and -negative fractions by an Isolex system (Baxter). Positive selection of CD34+ cells has been associated with inevitable cell loss. To overcome this loss, CD34+ cells that had migrated into the negative fraction were saved and used for the first transplant, which was followed by a second transplant after a 3- to 5-month interval. In this phase I feasibility study, the results in six children were evaluated for safety and engraftment. Multi-drug cytoreductive regimens using ranimustine (MCNU), melphalan, thiotepa, carboplatin, cyclophosphamide or VP-16 were comparable between the two transplant procedures in terms of their intensity. The number of CD34+ cells in the 'CD34(+) fraction' was 3.31 x 10(6)/kg (0.63-4.3 x 10(6)/kg), while this number in the 'CD34(-) fraction' could not be evaluated correctly due their scarcity (<0.1%). The median numbers of infused MNC and CFU-GM were, respectively, 4.2 x 10(6)/kg and 1.75 x 10(5)/kg in the CD34(+) fraction, and 4.8 x 10(8)/kg and 3.35 x 10(5)/kg in the CD34(-) fraction. The number of days required to achieve an ANC >0.5 x 10(9)/l and a platelet count >20 x 10(9)/l and >50 x 10(9)/l were, respectively, 14.5, 15.0 and 19.5 in the first transplant with CD34- cells, and 13.5, 18.0 and 25.0 in the second transplant with CD34+ cells, with no essential difference between the two treatments. Although the small number of patients, the variation in clinical status and treatment, and the short follow-up invalidate any evaluation of the therapeutic benefit of this strategy, the encouraging results support the feasibility of this strategy, which enables an escalation of dose intensity with an improved cost/benefit ratio.