The growth potential of 174 intracranial gliomas was estimated by calculating the bromodeoxyuridine labeling index (BUdR LI). Each patient received a 30-min infusion of BUdR, 200 mg/m2, before tumor removal. Excised tumor specimens were stained immunohistochemically to determine the BUdR LI, or percentage of S-phase cells. A Cox proportional-hazards stepwise model was used to determine the correlation between the BUdR LI and survival. Among patients with glioblastomas, the BUdR LI did not improve the prediction once age was entered in the model. Among patients with malignant or low-grade astrocytomas, the BUdR LI was the best single predictor of survival. The relative predictive abilities of BUdR LI and histopathology were determined by analyzing malignant astrocytoma and glioblastomas together. Distinguishing between malignant astrocytomas and glioblastomas did not significantly improve the prediction of survival once the BUdR LI and age were entered into the model. Equations derived from the model indicate that the probability of survival is a function of age and BUdR LI in patients with glioblastoma or malignant astrocytoma, but is a function of BUdR LI alone in patients with low-grade astrocytoma. The equations also show a substantial difference in the impact of increased BUdR LI on survival among patients with glioblastoma or malignant astrocytoma and those with low-grade astrocytoma. Without highly effective treatments for specific tumor phenotypes, the survival of a patient with an intracranial glioma appears to depend strongly on the proliferative potential of the tumor. Thus, accurate estimates of the proliferative potential are important in predicting the survival of individual patients with gliomas as well as in evaluating the effectiveness of various types of treatment.