Compared with conventional transfusion regimes a strong reduction in HLA alloimmunization and refractoriness to platelet transfusions is obtained when both red blood cell concentrates (RBCs) and platelet concentrates (PCs) are depleted of leukocytes by filtration. Because most of the leukocyte contamination is introduced by transfusion of RBCs, filtration of RBCs appears rational, but uncertainty exists regarding the degree of leukocyte-depletion of PCs needed for the prevention of HLA alloimmunization and refractoriness. We conducted a prospective trial and randomized patients with acute leukemia to receive leukocyte-depleted PCs prepared either by centrifugation (mean leukocyte count 35 x 10(6)/PC of 6 U) or by filtration (mean leukocyte count less than 5 x 10(6)/PC of 6 U). Both groups received RBCs that were filtered after prior removal of the buffy coat. Clinical refractoriness occurred in 46% (12 of 26) of the evaluable patients that were transfused with centrifuged PCs and only in 11% (3 of 27) in the filtered group (P less than .005). De novo anti-HLA antibodies were detected in 42% (11 of 26) patients in the centrifuged group and only in 7% (2 of 27) of the patients receiving filtered PCs (P less than .004). In 8 of 11 alloimmunized patients in the centrifuged group antibodies were detected in the first 4 weeks of transfusion therapy while none of the patients in the filtered group became immunized against HLA antigens during that period. We conclude that for the prevention of HLA alloimmunization and refractoriness to platelet transfusions from random donors, both RBCs and PCs have to be leukocyte- depleted by filtration.