Meta-analysis of controlled clinical trials studying the efficacy of rHuEPO in reducing blood transfusions in the anemia of prematurity.

BACKGROUND: Recombinant human EPO (rHuEPO) has not gained broad acceptance in the treatment of the anemia of prematurity, because its efficacy in diminishing RBC transfusions is questionable. Meta-analysis was used to investigate the extent and reasons for variation in the results of published clinical trials. STUDY DESIGN AND METHODS: Prospective controlled trials published from 1990 through 1999 were retrieved; 21 met the criteria for meta-analysis. Calculated across these studies were the summary OR of RBC transfusion in treated neonates as compared with controls and the summary mean difference between controls and treated neonates in the volume of RBCs transfused and the number of RBC transfusions per infant. Twelve study descriptors were examined as possible reasons for the variation in results.
RESULTS: Results of 21 eligible studies varied widely (p<0.001 for the Q test statistic), and this variation persisted in most analyses when studies were stratified by individual study descriptors. When the difference in volume of RBCs transfused was the outcome measure, variation was modest across the four studies with highly desired characteristics (i.e., high blindness and design quality scores, "conservative" transfusion criteria, and the majority of neonates weighing <1 kg at birth), and treatment with rHuEPO reduced RBC transfusions by an average of 11.0 mL per kg (p<0.001).
CONCLUSION: Benefit from rHuEPO is detected across high-quality studies using conservative RBC transfusion criteria. However, there is extreme variation overall in the findings of available trials, and-until this variation is accounted for-it is premature to recommend rHuEPO as standard treatment for the anemia of prematurity.