Patients with severe aplastic anemia (SAA) and a neutrophil (PMN) count of less than 0.5 x 10(9)/L are exposed to a high risk of early mortality when treated with antilymphocyte globulin (ALG) and steroids, with the major problem being infectious complications. The addition of human recombinant granulocyte colony-stimulating factor (rhG-CSF) to ALG may reduce early mortality by improving neutrophil counts in the short term. To test the feasibility of this approach, the SAA Working Party of the European Group for Blood and Marrow Transplantation (EBMT) designed a pilot study that included rhG-CSF (5 micrograms/kg/d, days 1 through 90), horse ALG (HALG; 15 mg/kg/d, days 1 through 5), methylprednisolone (2 mg/kg/d, days 1 through 5, then tapering the dose), and cyclosporin A (CyA; 5 mg/kg/d orally, days 1 through 180). Patients with newly diagnosed acquired SAA (untreated) and with neutrophil counts of < or = 0.5 x 10(9)/L were eligible. Forty consecutive patients entered this study and are evaluable with a minimum follow up of 120 days: the median age was 16 years (range, 2 to 72 years), the interval from diagnosis to treatment was 24 days, and the median PMN count was 0.19 x 10(9)/L. Twenty-one patients had hemorrhages, and 19 were infected at the time of treatment. Overall, treatment was well tolerated: the median maximum PMN count during rhG- CSF administration was 12 x 10(9)/L (range, 0.4 x 10(9)/L to 44 x 10(9)/L). There were three early deaths (8%) due to infection. Four patients (10%) showed no recovery, whereas 33 patients (82%) had trilineage hematologic reconstitution and became transfusion- independent at a median interval of 115 days from treatment. Median follow up for surviving patients is 428 days (range, 122 to 1,005). Actuarial survival is 92%: 86% and 100% for patients with PMN counts less than 0.2 x 10(9)/L or between 0.2 x 10(9)/L and 0.5 x 10(9)/L, respectively. This study suggests that the addition of rhG-CSF to ALG and CyA is well tolerated, is associated with a low risk of mortality, and offers a good chance of hematologic response. This protocol would appear to be an interesting alternative treatment for SAA patients with a low PMN count who lack an HLA-identical sibling.