Dose intensity may be an important determinant of the outcome in cancer chemotherapy, but is often limited by cumulative haematological toxicity. The availability of haematopoietic growth factors such as granulocyte colony-stimulating factor (G-CSF) and of peripheral blood progenitor cell (PBPC) transplantation has allowed the development of a new treatment strategy in which several courses of high-dose combination chemotherapy are administered for the treatment of solid tumours. PBPCs were mobilised before chemotherapy using 12 or 30 micrograms kg-1 day-1 G-CSF (Filgrastim) for 10 days, and were collected by 2-5 leucaphereses. The yields of mononuclear cells, colony-forming units and CD34-positive cells were similar at the two dose levels of Filgrastim, and the numbers of PBPCs were sufficient for rescue following multiple cycles of chemotherapy. High-dose chemotherapy (cyclophosphamide 2.5 g m-2 for 2 days, etoposide 300 mg m-2 for 3 days and cisplatin 50 mg m-2 for 3 days) was administered sequentially for a median of three cycles (range 1-4) to ten patients. Following the 30 evaluable cycles, the median duration of leucopenia < or = 0.5 x 10(9) l-1 and < or = 1.0 x 10(9) l-1 was 7 and 8 days respectively. The median time of thrombopenia < or = 20 x 10(9) l-1 was 6 days. There was no cumulative haematological toxicity. The duration of leucopenia, but not of thrombopenia, was inversely related to the number of reinfused CFU-GM (granulocyte-macrophage colony-forming units). In the majority of patients, neurotoxicity and ototoxicity became dose limiting after three cycles of therapy. However, the average dose intensity delivered was about three times higher than in a standard regimen. The complete response rate in patients with small-cell lung cancers was 66% (95% CI 30-92%) and the median progression-free survival and overall survival were 13 months and 17 months respectively. These results are encouraging and should be compared, in a randomised fashion, with standard dose chemotherapy.