A multicenter, randomized, open-label, parallel-group study was conducted to compare the safety and efficacy of perioperative recombinant human erythropoietin (Epoetin alfa) with the safety and efficacy of preoperative autologous donation (PAD) in total joint arthroplasty. A total of 490 patients scheduled for total joint (i.e., hip or knee) surgery and having hemoglobin (Hb) levels > or = 11 to < or = 13 g/dL were randomized to receive weekly doses of subcutaneous Epoetin alfa on preoperative Days -21, -14, and -7, and on the day of surgery, or to participate in a PAD program. The mean baseline Hb level in both groups was 12.3+/-0.6 g/dL, increasing to 13.8 g/dL in the Epoetin alfa-treated group and decreasing to 11.1 g/dL in the PAD group before or on the day of surgery. In the PAD group, 156/219 (71.2%) patients were transfused with autologous blood, and 42/219 (19.2%) patients were transfused with allogeneic blood. A smaller proportion, 27/209 (12.9%) patients, in the Epoetin alfa-treated group were transfused with allogeneic blood (P = .078 compared with the PAD group). Moreover, patients in the PAD group received a total of 325 units of blood (79 allogeneic units and 246 autologous units) compared with patients in the Epoetin alfa group who received a total of 54 units of blood. The mean postoperative Hb level was 11.0 g/dL in the Epoetin alfa-treated group and 9.2 g/dL in the PAD group. Compared with the PAD arm, mean Hb levels measured preoperatively, postoperatively on Day 1, and at discharge visits were significantly greater in the Epoetin alfa-treated arm (P < .0001 ).