To evaluate the prognostic significance of bromodeoxyuridine (BUdR) labeling index (LI) and to estimate tumor proliferative potential, BUdR LI was examined in 98 patients having a primary diagnosis of glioblastoma multiforme (GBM); 55 underwent infusion of 200 mg/m2 of BUdR at the time of the primary resection and 49 underwent infusion at the time of the second resection. The tumors of six patients were labeled at both operations. The tumor specimens were stained with hematoxylin and eosin for histopathology and by immunohistochemical methods to determine the ratio of labeled to unlabeled cells, i.e., BUdR LI. The median BUdR LIs for the primary and recurrent GBM were significantly different at 6.8 and 2.6%, respectively (P < 0.0001). However, there was no significant association between BUdR LI at the first or second operation and survival (log rank, P = 0.12; Cox regression analysis, P = 0.91; log rank, P = 0.55; Cox regression analysis, P = 0.17, respectively). Patients who underwent a second operation within 10 months of the first operation had a lower BUdR LI than did patients with a longer interval between procedures (P = 0.078; Spearman rank correlation, 0.26). The aggressive biological behavior of GBM is dependent on complex cellular kinetics, not simply on the number of cells within the S phase of the cell cycle. Caution should be used when determining prognosis with BUdR LI in GBM.