Conventional treatment of high-grade glioma includes maximal surgical resection followed by external radiation therapy. Despite this treatment, the prognosis for patients is poor. High doses of chemotherapy might be another way to increase the response rate and median survival. Increasing doses of BCNU might be more effective, but also provokes unacceptable myelotoxicity. This dose-limiting toxicity can be circumvented by using autologous blood stem cell rescue. We report our experience of high-dose BCNU followed by transplantation of autologous hematopoietic stem cells in 114 patients with high-grade gliomas. Of the 114 gliomas, 78 were glioblastoma multiforme (GM) (68%), 24 anaplastic astrocytomas (AA) (21%), and 12 anaplastic oligodendrogliomas (OD) (11%). Complete surgical resection was performed for 22 patients (18 GM and 4 AA). The median age was 44 years (range 17-65). A total of 84 patients received autologous hematopoietic stem cells from bone marrow harvest, while 30 patients received granulocyte colony-stimulating factor followed by apheresis and received peripheral blood progenitor cells (PBPC). High dose of BCNU (800 mg/m(2)) was given at least 1 month after neurosurgery. Bone marrow or PBPC was transplanted 48-72 h after chemotherapy. Radiotherapy was started approximately 40 days after transplantation to a total of 60 Gy. Median follow-up was 89 months (19-163). The overall survival (OS) was, respectively, 12 months for GM, 37 months for OD and 81 months for AA. Histological type appeared to be the main discriminating factor, with a worse prognosis for GM. Within the GM population, age, completeness of surgery, and response appeared to be one important prognostic factors. The AA and OD populations were small to reliably assess prognostic factors. On multivariate analysis, the main prognostic factors were histologic type, quality of surgery, and age (P<0.005). Five of 114 patients had lethal complications from the procedure. Four of these patients had a Karnovsky performance score (KPS) of 60%. The protocol thus appears to be feasible but patients should be selected for KPS more than 70%. We observed long-term survivors, although the OS and the time to treatment failure seem to be comparable to that described for other treatment. Additional pilot studies are unlikely to reveal more than a modest benefit from this procedure and therefore a randomized study should be performed.