Glioblastoma is the commonest neuroectodermal tumor and the most malignant in the range of cerebral astrocytic gliomas. The prognostic utility of various biological markers for glioblastomas has been broadly tested but the results obtained are regarded as controversial. In the present study, 302 glioblastoma specimens were studied to evaluate a possible association between clinical outcome and expression of some immunohistochemical variables. Furthermore, tumors examined were subdivided on the three cytological subsets--small-cell (SGB), pleomorphic-cell (PGB) and gemistocytic (GGB). Immunohistochemical variables differed between various subsets: the number of p53-positive tumors was found to be prevailed among the PGB, whereas the number of tumors with EGFR and mdm2 positivity was significantly greater in SGB. GGB contained significantly lowest mean proliferating cell nuclear antigen (PCNA) labeling index (LI), greater number of p21ras positive cases, and higher mean apoptotic index (AI). Survival time in patients with SGB, EGFR and mdm2-positivity and PCNA LI >40% was found to be significantly shorter, whereas presence of p21ras and AI >0.5% were associated with prolonged survival. Multivariate analysis revealed that survival time is associated with SGB, EGFR-positivity, and AI (p = 0.0023, p = 0.0035 and p = 0.0029 respectively). We conclude that although some immunohistochemical variables were found to be significant for glioblastoma outcome, they appear to be closely related to biology of single cytological subsets. Furthermore, these variables exhibited no prognostic value when they were analyzed within each cytological subset separately. Therefore, the glioblastoma subdivision on three cytological subsets proposed by us is carrying some element of rationality but, undoubtedly, requires further prospective studies.