Neonatal erythropoiesis is limited by a relatively inadequate production of erythropoietin. This is likely the result of dependence on the hepatic production of erythropoietin and an incomplete switchover to renal production. The present model of neonatal erythropoiesis suggests that the use of exogenous erythropoietin should correct the early anemia of prematurity that is observed at 6 weeks of age in premature newborns. Randomized, controlled trials of erythropoietin use in very low birthweight infants are reviewed. The data support the conclusion that erythropoietin at doses of > or = 750 u/kg/wk started at less than 7 days of age results in improved reticulocyte counts and hemoglobin levels, but does not reduce the number of infants who will be exposed to blood products. Erythropoietin at doses of > or = 600 u/kg/wk started at an average of 21 days of life improves reticulocyte counts and hemoglobin levels, and reduces the number of infants will will require late transfusion, but does nothing for the bulk of infants who are transfused before that age.