The usefulness of histopathological grading in predicting the prognosis of patients with ependymomas is controversial. To clarify the discrepancy between the histological malignancy and the prognosis of these tumors, we estimated the proliferative potential of 32 intracranial and intraspinal ependymomas and correlated the findings with the clinical behavior. Each patient received an intraoperative infusion of bromodeoxyuridine (BUdR, 200 mg/m2 i.v.) before tumor removal; the BUdR labeling index (LI), or percentage of BUdR-labeled cells, was determined immunohistochemically in excised specimens. The mean BUdR LI (+/- SD) of intracranial malignant ependymomas was 4.1 +/- 2.8%. Nonmalignant intracranial and intraspinal ependymomas and subependymomas had mean LIs of 1.5 +/- 0.9%, 1.1 +/- 0.3%, and less than 1%, respectively. Overall, 44% of the tumors recurred. There were no statistically significant differences in the recurrence rates of intracranial and intraspinal ependymomas, including subependymomas (43% and 44%, respectively), or of intracranial ependymomas with LIs greater than 1.0% and less than 1.0% (67% and 44%, respectively). However, the early recurrence rate (within 24 months after treatment) of tumors with LIs greater than 1.0% was higher than that of tumors with LIs of less than 1.0% (100% vs. 25%, P less than 0.05). The BUdR LI also showed a statistically significant inverse correlation with the time to recurrence. These findings indicate that BUdR LI reflects the proliferative potential of individual ependymomas and can be used to help predict the recurrence and estimate the prognosis of these tumors.