The authors describe the results of multimodality therapy in 27 patients with biopsy-proven primary central nervous system (CNS) lymphoma treated between 1976 and 1986. Treatment included surgical resection (15 patients), radiotherapy (27 patients), and chemotherapy (nine patients). Actuarial survival rates for the 27 patients at 1, 2, and 5 years after diagnosis were 70%, 54%, and 45%, respectively. Nine patients were recurrence-free at 8 to 106 months follow-up. A multivariate risk analysis identified five factors which had a favorable impact on survival: (1) age less than 60 years (P less than 0.02); (2) preoperative Karnofsky performance score greater than or equal to 70 (P less than 0.02); (3) presence of strictly hemispheric tumor (P less than 0.0003); (4) whole-brain radiation dose between 4000 and 5000 cGy (P less than 0.05); and (5) addition of chemotherapy to radiotherapy (P less than 0.002). Patients with complete tumor resolution on computed tomography 6 months after beginning treatment also had longer survival (P less than 0.01). The presence of malignant cells on cerebrospinal fluid cytologic examination correlated with an increased risk of distant metastasis (P less than 0.05). In those patients whose disease eventually recurred, the administration of an additional therapeutic modality significantly increased the length of postrecurrence survival (P less than 0.05). Although surgical resection provided no increase in survival, the addition of chemotherapy to postoperative cranial irradiation significantly enhanced the duration of survival. Our experience suggests that pretreatment clinical and diagnostic factors can help in predicting survival and in planning treatment.