It has been well established that erythropoietin (EPO) in the dosage range of 500 units/kilo/week and perhaps slightly lower doses will produce a brisk reticulocyte response in infants with anemia of prematurity. Controlled clinical trials to demonstrate that this therapy can result in significant reductions in transfusion in these babies face several complex issues of experimental design. 1. Should the study population be relatively bigger, healthier babies (< 1500 grams birth weight, not on ventilatory support) who have lower transfusion requirements, or smaller sicker infants (< 1250 grams birth weight and on ventilators) who have higher transfusion requirements? These infants will need adequate nutrition and liberal supplementation with iron if they are to respond adequately, but the sicker smaller infants will take longer to meet these nutritional goals. 2. Timing is important because spontaneous recovery occurs at about 35 to 36 weeks of corrected gestational age, so to be effective, therapy must start before 33 weeks of gestational age and preferably earlier than that. 3. Since the end point is transfusion, the criteria used for transfusions become a critical issue. If liberal transfusion criteria are used, the study will be doubly biased in favor of EPO efficacy. There will be an increased number of transfusion events in the control population and spontaneous recovery from the anemia of prematurity will be overly suppressed in the control population. It's likely that the current transfusion criteria are too liberal thus introducing these biases to experimental design.(ABSTRACT TRUNCATED AT 250 WORDS)