The clinical application of recombinant human G-CSF in patients with acute myeloid leukemia (AML) has been controversial because it stimulates the in vitro proliferation of leukemic cells. In order to explore the possibility of predicting in vivo leukemic proliferation after G-CSF administration to AML patients by using in vitro assays, we investigated the leukemic blasts of 30 AML patients, including 14 patients who received G-CSF for severe infection associated with neutropenia following chemotherapy (G-CSF group) and 16 patients who did not (control group). Of the 14 patients in the G-CSF group, 9 showed an increase of leukemic blasts in the peripheral blood during G-CSF administration, while 11 of the 16 control patients developed leukemic resurgence. In the G-CSF group, the frequency of leukemic resurgence among patients whose blasts showed dose-dependent proliferation after addition of G-CSF to cultures was similar to that among patients whose blasts did not. In addition, there were no significant differences between the G-CSF and control groups in [3H]thymidine incorporation by leukemic cells and leukemic colony formation after the addition of G-CSF to cultures. The G-CSF receptor affinity of leukemic blasts was significantly higher in the patients with leukemic resurgence (mean dissociation constant [Kd]: 55 pM in the G-CSF group and 63 pM in the control group) than in those without it (101 pM and 96 pM, respectively), and the number of G-CSF receptors per cell was significantly lower when leukemic resurgence occurred (200 in the G-CSF group and 260 in the control group) than when it did not (3400 and 3030, respectively). Immunophenotyping (for CD2, CD7, CD10, CD13, CD19, CD33, CD34, CD71, HLA-DR, glycophorin A and the G-CSF receptor) revealed no significant differences between blasts from the patients with and without leukemic resurgence in the G-CSF group. Thus, we conclude that the in vivo leukemic resurgence during G-CSF administration after chemotherapy for AML was not correlated with the in vitro responsiveness of leukemic blasts to this cytokine or with blast phenotyping data. Leukemic resurgence is likely to occur in patients whose leukemic blasts have fewer numbers of G-CSF receptors with a high affinity irrespective of whether patients receive G-CSF.