The use of intensive chemotherapy in the treatment of myelodysplastic syndromes (MDS) has met with some disappointment, although subgroups of patients have been identified in which the response approaches that of de novo acute myeloid leukaemia (AML). We hypothesized that it is not the FAB classification per se, but the biological behaviour of the blasts as shown by their rate of accumulation that influences the response. We have, therefore, included AML with trilineage dysplasia (AML/TLD) as it represents one extreme of the evolution of MDS to AML. We have analysed the results of intensive chemotherapy in 22 patients (median age 60 years; range 26-77 years) with MDS (14) and AML/TLD (8). Response to treatment was analysed by age, interval from diagnosis to treatment, the number of cytopenias, bone marrow blasts and karyotype. Patients were also divided according to the rate of disease progression, shown by the time from diagnosis to treatment (group A = < 3 months; group B = > 3 months). The overall response rate was 87%; 13 (60%) complete responses (CR) and 6 (27%) partial responses. The rate of disease progression was identified as the most significant predictive factor of achieving CR (p = 0.003) (group A 10/12; group B 3/10). Patients presenting with more than 20% blasts also had a better response (p = 0.031). The combined response rates, however, did not differ significantly between the two groups (group A 92%; group B 80%) as 50% of group B achieved a PR. The failure to normalize blood counts was not related to the number of cytopenias before starting treatment. In all cases, PR was associated with persistence of dysplastic morphology and cytogenetic abnormalities. CR was associated with complete morphological and cytogenetic response except in two patients in group B. Dysplastic morphology re-emerged in patients who achieved CR and of these, all but one acquired a new cytogenetic abnormality. Patients in group B who achieved CR all needed two courses compared with a mean of 1.1 for the other group. The median survival from treatment for both groups was 10 months, however, no patient in group B survived more than 20 months. In comparison 33% in group A were alive at 5 years. The rate of accumulation of blasts predicts the response to chemotherapy and the quality of remissions achieved. Patients with rapidly increasing blasts can achieve complete morphological and cytogenetic remissions, although they eventually have a dysplastic relapse. In contrast, intensive chemotherapy for patients with a slow accumulation of blasts may reduce the blast population but with much less benefit on haemopoiesis.(ABSTRACT TRUNCATED AT 400 WORDS)