In this prospective randomized multicenter trial 93 patients, median age 72 years, with RAEB-t (n=25) and myelodysplastic syndrome (MDS)-AML (n=68) were allocated to a standard induction chemotherapy regimen (TAD 2+7) with or without addition of granulocyte-macrophage-CSF (GM-CSF). The overall complete remission (CR) rate was 43% with no difference between the arms. Median survival times for all patients, CR patients, and non-CR patients were 280, 550, and 100 days, respectively, with no difference between the arms. Response rates were significantly better in patients with serum lactate dehydrogenase (S-LDH) levels </=9.5 microkat/l, bone marrow cellularity </=70%, and WBC counts <4.0 x 10(9)/l, but S-LDH was the only variable independently associated with response by logistic regression analysis. Cox's regression analysis identified four significant prognostic factors for survival: bone marrow cellularity, S-LDH, cytogenetic risk group (International Prognostic Scoring System), and age. Only bone marrow cellularity (P=0.01) and S-LDH (P=0.0003) retained statistical significance in the log-rank test. Severe adverse events were significantly more common in the GM-TAD arm (P=0.01). Thus, addition of GM-CSF to chemotherapy showed no clinical benefit in terms of response but carried an increased risk for side effects. We present a clinically useful tool to predict response to chemotherapy and survival in elderly patients with transforming MDS, favoring patients with features of less proliferative disease.