Bone marrow harvested from cancer patients for autologous bone marrow reinfusion (ABMR) after myeloablative treatment may be injured, in both its proliferating and stromal cell pools, by either previous treatment or manipulation at the time of harvest. We have examined the relative effects of seven covariates on hematologic recovery after ABMR in children with neuroblastoma (NBL) using univariate and step-up analysis. We measured recovery by times to achieve (1) white blood cell counts greater than 1,000/microL; (2) absolute neutrophil counts greater than 500/microL; and (3) platelet counts greater than 20,000/microL without transfusion. In univariate analysis, recovery was significantly associated with the amount of prior chemotherapy and the interval between last chemotherapy and marrow harvest. Patient sex, the number of granulocyte-macrophage colonies infused, harvest-to-freeze interval, and use of purging were marginally associated. After adjusting for potential confounders in a multivariate model, the amounts of chemotherapy and granulocyte-macrophage colonies infused were independently significant predictors of time to total white blood cell count recovery; chemotherapy courses and chemo-to-harvest interval were predictors of neutrophil count recovery; and sex, use of purging, and harvest-to-freeze interval were marginal predictors of platelet recovery. The speed of hematologic recovery after ABMR seems to depend mainly on pre-existing factors and marginally on manipulation of the marrow after harvest. These factors may affect both proliferating and stromal cell pools.