We have determined the relationships between busulfan average concentration at steady-state and (1) rejection of graft, and (2) regimen-related toxicity, and have evaluated the dependence of busulfan clearance/F on body size and age. Patients received 16-30 mg/kg of busulfan followed by cyclophosphamide in doses of 120, 150, 174, or 200 mg/kg or 8 g/m2 in preparation for autologous, syngeneic or allogeneic grafts varying in compatibility from HLA-matched siblings to HLA-partially-matched unrelated donors. In a multivariate Cox time-to-rejection analysis, only busulfan concentration remained a significant determinant of rejection, P = 0.0154. An average concentration of busulfan at steady-state of at least 200 ng/ml was needed to avoid rejection of a matched-sibling graft, while 600 ng/ml was needed to avoid rejection of HLA-partially-matched related or HLA-matched unrelated donor grafts. The toxicity of the cytoreductive regimen correlated with busulfan average concentration at steady-state (rs = 0.717). Busulfan clearance/F expressed relative to body weight, ideal body weight or surface area declined with age during the first decade of life. Over the entire span of age, the coefficient of variation in clearance/F for all ages was similar when clearance/F was expressed in absolute terms (ml/min) and when adjusted for body surface area; the coefficient of variation was greater for clearance/F when expressed relative to total or ideal body weight. We conclude that busulfan concentration in plasma is an important determinant of graft survival and regimen-related toxicity, and that the variability of busulfan pharmacokinetics with age precludes the use of a fixed dose for all ages and indications.