Since 1998, there has been a marked increase in incidence of pure red cell aplasia secondary to development of anti-erythropoietin antibodies (Ab+ PRCA) in patients who have chronic kidney disease (CKD) and receive recombinant erythropoietin. The relationship between incidence of Ab+ PRCA and specific erythropoietin products has not been examined rigorously. Manufacturers provided data regarding exposure to erythropoietin products and incidence of Ab+ PRCA between January 1998 and March 2003 in patients with CKD. Assuming a Poisson distribution, a maximum likelihood estimate for the Poisson rate parameter was calculated for each product. A test for homogeneity of Poisson rates was conducted to compare likelihood estimates between products. Global incidence of Ab+ PRCA was relatively low. Likelihood estimates were not significantly different for Epogen, Procrit, and Aranesp, independent of their formulation or route of administration. Eprex lacking human serum albumin (HSA) and administered subcutaneously was associated with the greatest risk of Ab+ PRCA. HSA-containing Eprex administered subcutaneously was associated with a lower risk than HSA-free Eprex administered subcutaneously, but this risk exceeded that of intravenous Epogen and intravenous HSA-free Eprex. NeoRecormon administered subcutaneously was associated with less risk than subcutaneous HSA-free Eprex but more risk than intravenous Epogen. HSA-free Eprex should not be administered subcutaneously to patients with CKD due to increased risk of Ab+ PRCA. Although the subcutaneous administration of HSA-containing Eprex is riskier than intravenous Epogen and intravenous HSA-free Eprex, and the use of subcutaneous NeoRecormon is riskier than intravenous Epogen, there is currently no evidence that other products are safer.