Bone marrow purging with cyclophosphamide derivatives (Mafosfamide) requires the establishment of a defined experimental procedure for reliable leukemic cell destruction while sparing normal hematopoietic stem cells to ensure engraftment. We previously defined the granulocyte-macrophage colony-forming unit (CFU-GM) LD95 as being the maximum tolerable dose of drug to use. We now report, in 20 patients with acute non-lymphoblastic leukemia (n = 5), acute lymphoblastic leukemia (n = 5), chronic myelogenous leukemia (n = 5), and non-Hodgkin's lymphoma (n = 5), that the nature of the cells treated (i.e., buffy coat cells or mononuclear cells) significantly influences the accuracy of the LD95 determination, whereas other parameters such as hematocrit or nucleated cell concentration do not. We subsequently define the most reliable experimental procedure for in vitro purging with Mafosfamide: incubation of 2 x 10(7) buffy coat cells/ml with a hematocrit of 5%. We show that the wide individual susceptibility to the drug is not related to any incubation procedure. In a series of 163 patients with hematological malignancies, we confirm the large variation of sensitivity to the drug according to patient susceptibility and diagnosis. These data favor the adjustment of the dose of Mafosfamide on an individual basis, prior to bone marrow purging for autologous bone marrow transplantation.