Fifty-one adult patients with recurrent malignant gliomas were treated in a Phase II trial of multidrug chemotherapy (6-thioguanine, dibromodulcitol, procarbazine, 1-(2-chloroethyl)-3-cyclohexyl-1-nitrosourea, 5-fluorouracil, and hydroxyurea). Thirty-one patients underwent radical tumor debulking, before the administration of chemotherapy. Fifty-seven percent of all patients had either an objective radiographic response or stabilization of disease after the institution of therapy. The overall median survival time (MST) was 40 weeks; it was 79 and 33 weeks for anaplastic astrocytoma and glioblastoma patients, respectively. The overall median time to tumor progression (MTP) was 19 weeks--32 weeks for anaplastic astrocytoma patients and 13 weeks for glioblastoma patients. Serious chemotoxicity occurred in 35% of patients without permanent morbidity or mortality. The factors that affected response (including disease stabilization), MTP, and MST were identified through a multivariate statistical analysis. A longer MTP was associated with higher Karnofsky scores, lower grade initial histology, lack of prior chemotherapy, greater degree of myelotoxicity, smaller postoperative tumor volumes, greater extent of surgical resection, and a local versus diffuse recurrence pattern. A longer MST was associated with higher Karnofsky scores, lower grade histology at the time of recurrence, greater degree of myelotoxicity, and lobar versus deep tumor location. Response (including disease stabilization) correlated with higher Karnofsky scores, lower grade histology (initial and current), prior lower grade histology, smaller preoperative tumor volume, longer intervals from the time of initial diagnosis, and absence of prior chemotherapy. These results suggest that, in addition to established prognostic factors such as Karnofsky scores, other factors including prior chemotherapy administration, patterns of tumor recurrence, and tumor location may be important variables to consider in future Phase II-III clinical trials. Of the treatment variables analyzed, greater surgical debulking and smaller postoperative tumor volumes were associated with prolonged MTP but not MST, and greater myelotoxicity had a positive association with all outcomes. The significance of this latter relationship and its relevance to chemotherapy dosing will require further study. Standardization in the design and reporting of clinical trials and the use of computer-assisted tumor volume calculations to assess the extent of surgical resection and the response to therapy are advocated.