The prognostic power of three proliferation estimation methods, Ki-67 (MIB-1) and PCNA immunohistochemistry, and flow cytometry (S-phase and S + G2/M fractions, respectively), were evaluated in 50 cases of astrocytoma. Each proliferation index showed a strong association with the grade of malignancy (grades I-IV). The MIB-1 labelling index (LI) provided additional information, as it could be used for the discrimination of grade II and grade III astrocytomas (P = 0.0357). All three proliferation estimation methods also had strong prognostic potential (MIB-1 LI: P < 0.0001; PCNA Li: P < 0.0001; S-phase: P = 0.0004; S + G2/M: P = 0.0124). According to the receiver operating characteristics (ROC) curve, the MIB-1 LI showed generally the best sensitivity and specificity in placing the patients correctly into groups of survivors and non-survivors, which was further confirmed in the multivariate analysis. Only 4 per cent of the patients having high MIB-1 scores (> 15.3 per cent) were alive after 2-years' follow-up. In contrast, 72 per cent of patients with tumours of low proliferation activity survived. It appears that Ki-67 (MIB-1) immunolabelling using archival paraffin-embedded samples is of value in predicting prognosis in astrocytic tumours.