RÉSUMÉ
We reviewed erythrocyte transfusions, given after 2 weeks of age to premature infants, to derive patient selection criteria for trials of administration of recombinant human erythropoietin to maintain hematocrit during the anemia of prematurity. We first studied 122 infants, less than 36 weeks of gestational age, who remained in a level 3 nursery longer than 4 weeks. Multiple logistic regression analysis was used to select those variables associated with more than two transfusions after the second week; these included gestational age, 5-minute Apgar score, transfusion during the first week, and patent ductus ligation. The best predictor, alone or combined, was gestational age less than 30 weeks. This predictor was applied to a second population of 44 premature infants from a level 2 nursery, and we found a 94% probability of excluding those infants who did not receive a transfusion. This information will allow selection of premature infants who are candidates for administration of recombinant human erythropoietin.