The treatment and prognosis of patients with cerebral astrocytic tumours are currently guided by histopathological classification. This study evaluates immunohistochemistry using Ki-67, an antibody to a nuclear protein expressed in proliferating cells, and DO-7, an antibody to the product of the tumour suppressor gene p53, as prognostic indicators for these tumours. Immunohistochemistry with Ki-67 has been correlated with the behaviour of many different tumours, but its value as a prognostic indicator in astrocytic tumours is diminished by the conflicting results of previous studies. Immunohistochemistry with antibodies to the p53 protein has been used as a prognostic indicator in melanomas and some carcinomas, but the relation between prognosis and accumulation of this protein in astrocytic tumours has not been clarified. We have tested the hypothesis that survival is correlated with Ki-67 immunolabelling indices (LIs) and patterns of p53 immunolabelling in the cerebral astrocytic tumours of a large cohort of patients (n = 123) for whom clinical indices were well documented. Astrocytic tumours were divided into three histological types: fibrillary astrocytoma (n = 24), anaplastic astrocytoma (n = 31), and glioblastoma (n = 68). Histological type and patient age were independent predictors of survival. Median Ki-67 LIs differed significantly (P < 0.0001) between the types of astrocytic tumour, and tumours with a Ki-67 LI < 2% had a significantly (P < 0.0001) better prognosis. Ki-67 LI as a continuous variable carried a significant (P = 0.0043) unadjusted hazard to survival which was lost when adjusted for other variables, notably histological type. By contrast, no relation was found between survival and three categories of p53 labeling (p53-negative, p53 LI < 40%, and p53 LI > 60%). The results indicate that, whereas Ki-67 immunohistochemistry predicts survival in patients with astrocytic tumours, conventional histological appraisal remains the best guide to prognosis, and immunohistochemistry for p53 has no value in the assessment of these tumours.