In 80 patients with Ph-positive chronic myelogenous leukaemia the main clinical, haematological and cytogenetical data were recorded at diagnosis of blast crisis and evaluated for prognostic significance. At the time of the analysis 73 patients had died, with a median survival of 4-8 months from diagnosis of blast crisis for the whole series. When analysed as a time-dependent variable, the achievement of a favourable response to chemotherapy resulted in a longer patient's survival. On the other hand, the univariate analysis identified six pretreatment characteristics associated with a poorer prognosis: a longer chronic phase, presence of extramedullary blastic involvement, a platelet count below 200 x 10(9)/l, a less marked leucocytosis, a blood blast cell percentage higher than 10%, and presence of trisomy 8. The latter parameters were included in a multiple regression model together with the blast cell phenotype (lymphoid versus non-lymphoid), and only four of them (trisomy 8, duration of chronic phase, platelet count, and leucocyte count) retained their prognostic influence. When the therapeutical response was also included in the regression model, it proved to be the most important prognostic variable, followed by trisomy 8, length of chronic phase, extramedullary disease, and platelet count, whereas the leukocyte count lost its predictive value. Thus, in spite of the short overall survival of blast crisis patients, the identification of prognostic factors in such a haematological condition may be of interest, especially in the interpretation of new therapeutical approaches.