This study describes a novel application of HLAMatchmaker to determine platelet compatibility in 16 alloimmunized patients with aplastic anemia refractory to random donor platelet transfusions. HLAMatchmaker is a software algorithm that predicts HLA compatibility by identifying immunogenic epitopes represented by amino acid triplets in antibody-accessible regions of human leukocyte antigen (HLA) molecules and determines the number of triplet mismatches (TMMs) and highly immunogenic triplet mismatches (HIMMs). Corrected count increments (CCIs) and molecular HLA typing were available for 523 transfusions. Conventional compatibility assessment based on cross-reactive group (CREG) determination was not predictive of transfusion outcome. Low HIMMs and TMMs numbers were associated with a higher likelihood of satisfactory (CCIs > or = 8) compared with unsatisfactory (CCIs < 8) outcomes (median HIMMs = 4 vs 6, p2 value < .001; median TMMs = 11 vs 13, p2 value < .001). Although receiver operator characteristic curves revealed that HIMMs or TMMs number are not powerful predictors of individual transfusion outcome, a threshold of at least 3 HIMMs or at least 9 TMMs appeared to be associated with successful transfusions. Triplet-matched transfusions were successful, regardless of CREG matching. Our data validate HLAMatchmaker for platelet transfusions and demonstrate its potential to refine and expand donor selection for HLA-alloimmunized patients.