BACKGROUND The reported immunomodulatory effects of transfusion raise concern about the potential for virus activation and tumor growth in human immunodeficiency virus (HIV)-infected patients. In the absence of "standards" of transfusion practice for such patients, a survey of transfusion policies among institutions specializing in the care of HIV-infected patients was performed to delineate current practices. METHODS A survey developed by the Transfusion Practices Committee of the American Association of Blood Banks was sent to 47 AIDS clinical trial units and 14 regional hemophilia centers in North America. RESULTS Forty-three percent of centers completed the survey. Most centers observed more than 200 HIV-infected patients each. The key findings were that 1) 81 percent of centers used identical red cell transfusion criteria for HIV-infected and noninfected patients; 2) 52 percent used recombinant human erythropoietin as initial treatment for zidovudine-induced anemia, while 46 percent used recombinant human erythropoietin for anemia not associated with zidovudine; 3) 35 percent of centers used white cell-reduced blood components in lieu of cytomegalovirus (CMV)-seronegative components when administering transfusion(s) to CMV-seronegative patients; 4) 27 percent gamma-radiated cellular components, but no case of graft-versus-host disease had been observed; 5) > 85 percent of centers used monoclonal factor VIII for pediatric and adult hemophiliacs infected with HIV; 6) approximately one-third of centers routinely white cell-reduced cellular components; and 7) the most common reasons for white cell reduction included reduction of febrile reactions and CMV risk, reduction of platelet alloimmunization, and delay of immunomodulatory consequences of transfusion. CONCLUSIONS There is marked heterogeneity in transfusion practice for HIV-infected patients. Modification of cellular components to achieve different objectives is routine in many centers.