Refractoriness occurs in many patients receiving multiple platelet transfusions. We used a sensitive ELISA assay to assess the utility of crossmatching HLA-A,B matched single donor platelets in 51 consecutive, typical refractory patients. Of the 222 transfusions evaluated at 1 to 4 hours posttransfusion, only 17 of 54 (31%) with positive crossmatches had corrected platelet count increments of greater than or equal to 7,500/microL. In contrast, 95 of 168 (57%) of those with negative crossmatches had such increments (P less than .001). Regardless of the results of the crossmatch, HLA-A,B, and ABO matching had independent influences on transfusion outcome. The median corrected 1- to 4-hour increment for crossmatch negative transfusions was 13,300/microL for A/BU grade matches, 9,700 for BX, and 7,800 for C. Increments were 10,000/microL for ABO-identical transfusions and 5,900 for transfusions of platelets ABO incompatible with the recipient's plasma antibodies. When the donor platelets were ABO compatible, but the donor plasma contained ABO antibodies to the recipient's platelets, the increment was intermediate (8,200/microL). The most important factor in predicting platelet survival was the crossmatch, followed by HLA-A,B and ABO, each having independent predictive value. These data demonstrate that the predictive value of a negative crossmatch may be considerably less than that reported in previous studies with stable, less ill patients. In typical refractory patients, there appear to be mechanisms of platelet destruction that are related to HLA-A,B and ABO but are not detected with current crossmatch methods. We hypothesize that soluble plasma HLA-A,B and ABO antigens contribute to the destruction of donor and sometimes recipient platelets by an immune complex or other "innocent bystander" mechanism. With our crossmatching technique, HLA-A,B and ABO match grades remain relevant to platelet transfusion therapy in some refractory patients.