To determine the clinical relevance of in vitro drug chemoresistance in childhood acute myeloid leukemia, we used an MTT assay to test leukemic cells from 132 newly diagnosed children. Patients were diagnosed according to the French-American-British (FAB) classification as follows: M0 (n = 12), M1 (n = 16), M2 (n = 53), M4 (n = 17), M5 (n = 19) and M7 (n = 15). The results revealed that, compared to leukemic cells from complete-responders (n = 107), those from non-responders who failed induction therapy (n = 17) were 1.4 to 5.0 times more resistant in vitro to cytarabine (P = 0.005), melphalan (P = 0.003), etoposide (P = 0.011), L-asparaginase (P = 0.017), aclarubicin (P = 0.026) and dexamethasone (P = 0.039). For seven other drugs tested, the median lethal dose of 70% and leukemic cell survival of non-responders were higher than those of complete-responders, but the difference was not statistically significant. We sought correlations between FAB subtypes and in vitro drug resistance. Leukemias of the FAB M4 and M5 subtype were more sensitive to L-asparaginase (P = 0.01, P = 0.0036) than those of the FAB M2 subtype. FAB M5 leukemia was more sensitive to etoposide than were the FAB M2, M4 and M7 subtypes (P = 0.001, P = 0.034, P = 0.023, respectively). By contrast, FAB M5 leukemia was significantly more resistant to prednisolone and dexamethasone than were the FAB M0, M1, M2, M4 and M7 subtypes. We sought correlations between in vitro drug resistance and long-term clinical outcome, but found no associations in this case. These results suggest that in vitro resistance to cytarabine, melphalan, etoposide, L-asparaginase, aclarubicin and dexamethasone might represent factors that can predict response to the early course of therapy. Selecting an appropriate anti-cancer drug according to the FAB classification together with drug sensitivity testing may contribute to improved prognoses in childhood acute myeloid leukemia.