Two hundred patients who were scheduled for a major elective orthopaedic operation were enrolled in a prospective study and were randomly assigned to one of three treatment groups. Group 1 consisted of sixty patients who received recombinant human erythropoietin, 300 international units per kilogram of body weight per day; Group 2, seventy-one patients who received recombinant human erythropoietin, 100 international units per kilogram of body weight per day; and Group 3, sixty-nine patients who received a placebo. A total of fifteen doses was given subcutaneously, beginning ten days before the operation and extending through the fourth postoperative day. Patients who declined or were unable to donate autologous blood preoperatively were included in the study and were maintained on iron supplementation orally. The decision to transfuse red blood cells depended on the physician; however, physicians were encouraged not to do so if the hematocrit was more than 0.27 (27 per cent), unless the clinical symptoms warranted it. Of the 185 patients who were evaluable with regard to efficacy, significantly fewer patients received homologous red-blood-cell transfusions in Groups 1 and 2 (17 per cent [nine] and 25 per cent [sixteen], respectively) than in Group 3 (54 per cent [thirty-six]) (p < 0.001). When the patients were stratified into two groups on the basis of the pre-treatment hemoglobin level (more than 100 to 130 grams per liter or more than 130 grams per liter), we found that patients who had received a placebo and had a baseline hemoglobin level of more than 100 to 130 grams per liter were at significantly higher risk for transfusion (78 per cent [twenty-one]) than those who had received a placebo and had a baseline level of more than 130 grams per liter (36 per cent [fourteen]). For patients who had a baseline hemoglobin level of more than 100 to 130 grams per liter, the higher dose of recombinant human erythropoietin appeared somewhat more effective than the lower dose, with 14 per cent (three) of the patients in Group 1 and 39 per cent (nine) in Group 2 needing a transfusion; however, the difference was not significant (p = 0.09). For patients who had a baseline hemoglobin level of more than 130 grams per liter, the two doses of recombinant human erythropoietin produced similar results, with 14 per cent (four) of the patients in Group 1 and 11 per cent (four) in Group 2 needing a transfusion; this was in contrast to a rate of transfusion of 36 per cent (fourteen) in Group 3 (the patients who received the placebo) (p = 0.03). The recombinant human erythropoietin was generally well tolerated, although one patient, who did not have a history of hypertension, had an increase in blood pressure, from a baseline level of 142/78 millimeters of mercury (18.93/10.40 kilopascals) to a level of 220/100 millimeters of mercury (29.33/13.33 kilopascals), after ten days of treatment with the higher dose. These data suggest that recombinant human erythropoietin, administered before and after major orthopaedic operations, can minimize the need for homologous red-blood-cell transfusion.