From 1990 to 1993 we performed a prospective study of busulfan (16 mg/kg) and cyclophosphamide (120 mg/kg) in 30 patients with refractory anemia (RA) undergoing related (n = 17) or unrelated (n = 13) donor marrow transplantation. Nineteen patients survive disease free (63% 3-year actuarial disease-free survival [DFS]) and no patient relapsed. These results were compared to those of 38 historical controls with RA treated with cyclophosphamide and total body irradiation, of whom 22 are disease-free survivors and 1 relapsed. After correcting for significant variables between the two treatment groups, we found no statistically significant difference in outcome based on preparative regimen. Combining data from these 68 patients plus 2 additional patients with RA treated before 1993 with busulfan and cyclophosphamide, we identified four variables independently associated with improved survival: younger age, shorter disease duration, lower neutrophil count pretransplant, and lower hematocrit pretransplant. We also found that 15 patients 40 to 55 years of age had a 46% 3-year actuarial DFS and 26 patients receiving unrelated or mismatched related donor marrow had a 50% 3-year actuarial DFS. We conclude that there does not appear to be any significant difference in outcome based on preparative regimen in this patient population. In addition, allogeneic bone marrow transplantation may be a reasonable approach to therapy of RA early after diagnosis. However, whether early intervention with transplantation prolongs survival over that expected without transplantation cannot be ascertained with certainty from available data.