PEDIATRICS Vol. 108 No. 4 October 2001, pp. 934-942

Effects of Early Erythropoietin Therapy on the Transfusion Requirements of Preterm Infants Below 1250 Grams Birth Weight: A Multicenter, Randomized, Controlled Trial

Robin K. Ohls, MD^*, Richard A. Ehrenkranz, MD, Linda L. Wright, MD^¶¶, James A. Lemons, MD^§, Sheldon B. Korones, MD, Barbara J. Stoll, MD^¶, Ann R. Stark, MD^#, Seetha Shankaran, MD^**, Edward F. Donovan, MD, Nicole C. Close, MS^§§, Abhik Das, PhD^||, and for the National Institute of Child Health and Human Development Neonatal Research Network

From the Departments of Pediatrics, ^* University of New Mexico, Albuquerque, New Mexico;  Yale University, New Haven, Connecticut; ^§ Indiana University, Indianapolis, Indiana;  University of Tennessee at Memphis, Memphis, Tennessee; ^¶ Emory University, Atlanta, Georgia; ^# Harvard University, Boston, Massachusetts; ^** Wayne State University, Detroit, Michigan;  University of Cincinnati, Cincinnati, Ohio; ^§§ George Washington University Biostatistics Center, Washington, DC; ^|| Research Triangle Institute, Research Triangle Park, North Carolina; and ^¶¶ National Institute of Child Health and Human Development, Bethesda, Maryland.

	ABSTRACT

Top Abstract Methods Results Discussion Conclusion References

Objectives.  Infants of 1250 g birth weight receive multiple erythrocyte transfusions during their hospitalization. We hypothesized that early erythropoietin (Epo) and iron therapy would 1) decrease the number of transfusions received (infants 401-1000 g birth weight; trial 1) and 2) decrease the percentage of infants who received any transfusions (1001-1250 g birth weight; trial 2).

Methods.  A total of 172 infants in trial 1 and 118 infants in trial 2 were randomized to treatment (Epo, 400 U/kg 3 times weekly) or placebo/control. Therapy was initiated by 4 days after birth and continued through the 35th postmenstrual week. All infants received supplemental parenteral and enteral iron. Complete blood and reticulocyte counts were measured weekly, and ferritin concentrations were measured monthly. Transfusions were administered according to protocol. Phlebotomy losses and transfusion data were recorded.

Results.  Treated and placebo/control infants in trial 1 received a similar number of transfusions (4.3 ± 3.6 vs 5.2 ± 4.2, respectively). A similar percentage of treated and control infants in trial 2 received at least 1 transfusion (37% vs 46%). Reticulocyte counts were higher in treated infants during each week of the study in both trials. Hematocrits were higher among treated infants from week 2 on in both trials. Ferritin concentrations were higher in placebo/controls than in treated infants at weeks 4 and 8 in trial 1 and at week 4 in trial 2. No adverse effects of Epo or supplemental iron occurred.

Conclusion.  The combination of early Epo and iron as administered in this study stimulated erythropoiesis in infants who were 1250 g at birth. However, the lack of impact on transfusion requirements fails to support routine use of early Epo.neonate, intravenous iron, donor exposure.

Critically ill preterm infants experience in the first 1 to 2 weeks after birth daily phlebotomy losses that may equal 5% to 10% of their total blood volume.¹ Such losses and associated anemia typically result in multiple erythrocyte transfusions. This iatrogenic anemia commonly is followed by the anemia of prematurity, prompting additional transfusions. A limited capacity to increase the erythropoietin concentration renders preterm infants less capable of compensating for either of these anemias,^2,3 although erythroid progenitors are present and responsive to erythropoietin.^4,5

Human recombinant erythropoietin (Epo) has been studied as an alternative to transfusions in preterm infants. Previously published trials reported varying degrees of success attributed to initiation of Epo after the first weeks of life,^6,7 insufficient dosing of Epo and iron, and enrollment of larger and healthier preterm infants at low risk for transfusion.^6,8 No adverse effects of Epo administration have been documented in preterm infants.^9,10

We sought to design a study that would optimize the dose and timing of Epo and iron in extremely premature, sick neonates. We hypothesized that administration of Epo and supplemental iron shortly after birth, in association with transfusion guidelines instituted by 96 hours of life, would decrease or eliminate erythrocyte transfusions in critically ill preterm infants. The study was performed concurrently as 2 parallel trials based on birth weight because different primary outcomes were evaluated in each trial. We aimed to determine whether the administration of Epo and supplemental iron to preterm infants, starting in the first 96 hours after birth and continuing until 35 weeks' postmenstrual age (PMA), would decrease the number of transfusions (infants 401-1000 g; trial 1) and decrease the number of infants who received any transfusions (infants 1001-1250 g; trial 2).

	METHODS

Top Abstract Methods Results Discussion Conclusion References

Patient Eligibility

Patients were eligible if they weighed 401 g and 1250 g at birth, were 32 weeks' gestation, were between 24 and 96 hours old at the time of study entry, were likely to survive >72 hours (as determined by the attending neonatologist), and had informed consent from a parent or guardian. Patients were ineligible if they had any of the following: a major congenital anomaly, a positive direct antiglobulin test, evidence of coagulopathy, clinical seizures, systolic blood pressure >100 mm Hg (in the absence of pressor support), or an absolute neutrophil count (ANC) of 500/µL. Randomization was stratified by center and for trial 1 by birth weight (401-750, 751-1000 g) using a permuted block method. All caregivers and investigators (except the research nurses) were masked to the treatment assignment.

Dose of Study Drug

Treated infants received 400 U/kg Epo 3 times weekly.¹¹ Initial doses were based on birth weight and adjusted weekly based on current weight. Epo or placebo was administered by the research nurse as a 1-hour intravenous infusion or subcutaneously when intravenous access was not available. Infants in the placebo/control group received sham subcutaneous injections when intravenous access was not available. An adhesive bandage covered the true and sham injection sites. The study drug was brought to the bedside in a closed container, and injections were shielded from the caregivers by screens. Treatment continued until discharge, transfer, death, or 35 completed weeks' PMA.

Criteria for Withholding/Stopping the Study Drug

Criteria for withholding the study drug included neutropenia (ANC <500/µL), a hematocrit of >45% (not attributable to transfusion) with a reticulocyte count 200 000 cells/µL, or hypertension (defined as systolic blood pressure >100 mm Hg during the first 2 postnatal weeks and >120 mm Hg thereafter). The study drug was restarted when these conditions resolved. Treatment was stopped when clinical seizures occurred (as defined by the attending neonatologist) or when hypertension or neutropenia persisted.

Iron, Vitamin E, and Folate Supplementation

Treated infants received a weekly intravenous infusion of 5 mg/kg iron dextran until they had an enteral intake of 60 mL/kg/d.^111-13 The iron dextran was either added to the parenteral nutrition solution and administered over 24 hours or diluted in several milliliters of 10% dextrose in water or normal saline and administered over 4 to 6 hours. For maintaining iron stores^12,13 and to assist in masking treatment assignment, placebo/control infants received 1 mg/kg iron dextran once a week, administered in a similar manner. Infants in the placebo/control group did not receive the same dose of parenteral iron as the treated infants because their iron requirements were estimated to be lower during the study.^12,13 Once infants in both groups had an enteral intake of 60 mL/kg/d, they were given enteral iron at a dose of 3 mg/kg/d. Then, enteral iron was increased to 6 mg/kg/d when infants achieved an enteral intake of 120 mL/kg/d. The iron was increased to 10 mg/kg/d enterally or 10 mg/kg/wk intravenously when the serum ferritin fell below 50 ng/mL. The iron dose was decreased by half when the serum ferritin rose above 400 ng/mL. Parenteral iron was not administered whenever Epo was held or stopped.

Study infants received enteral vitamin E (15-25 IU/d) when their enteral intake was at least 60 mL/kg/d. Enteral folate supplements (25-50 µg/d) were provided according to center practice.

Criteria for Transfusion

A strict protocol was used to administer transfusions during the study period (Table 1). Infants did not receive a transfusion solely to replace blood lost through phlebotomy. Whenever possible, designated donor units that were capable of providing at least 4 transfusions were assigned to each infant (available in 6 of the 8 participating centers). Infants who met transfusion criteria received a transfusion of 15 mL/kg packed red blood cells (PRBC) for a hematocrit of >25% or 20 mL/kg PRBC for a hematocrit of 25%.

Laboratory Studies, Data Collection, Power Analysis, and Statistical Analysis

Complete blood counts with differentials and reticulocyte counts were performed weekly by Coulter counter on blood obtained through a central line, by venipuncture, or via capillary blood sampling when central access was not available. Serum ferritin concentrations were measured monthly. Daily phlebotomy losses (determined by laboratory-specified blood volumes and bedside recording and confirmed by the research nurse) and transfusion information (the number of and indications for each transfusion, volume, and donor exposure) were recorded from birth to study completion. Morbidity and mortality information was collected for infants until death, discharge to home, or hospitalization at 120 days and included incidence of late-onset sepsis (a positive blood or cerebral spinal fluid culture obtained in the presence of compatible signs of septicemia after 72 hours of age, or culture-negative clinical infection after 72 hours of age for which the infant received antibiotics for 5 or more days), chronic lung disease (CLD; oxygen administration at 36 weeks' PMA), retinopathy of prematurity (ROP; stage 3 or higher¹⁴), severe intraventricular hemorrhage (IVH; grades 3 and 4¹⁵), patent ductus arteriosus (PDA), and necrotizing enterocolitis (NEC; Bell's stage II¹⁶).

Sample size was estimated using a 2-tailed type I error of 0.05, a power of 90%, and a noncompliance rate of 10%. Preliminary estimates were made using several different reductions of transfusion number (trial 1) or increases in percentage of transfusion-free infants (trial 2). Large sample sizes (>1000/group) would have been needed to demonstrate a significant difference with modest reductions in transfusion number or increases in transfusion-free patients. Although previous reports had demonstrated small but statistically significant benefits,^7,8 clinical practice did not change. We therefore chose an effect size that would lead to changes in clinical practice if a significant benefit were found. For trial 1, we assumed that the number of transfusions per infant in the placebo group would be 8 ± 6 (mean ± standard deviation). This estimate was based on pretrial transfusion data collected from participating network centers. Eighty patients per group were required to detect a minimum reduction from 8 to 4 transfusions. For trial 2, we assumed that 75% of the placebo group would receive at least 1 transfusion (also based on pretrial transfusion data); 100 patients per group were needed to detect a decrease in the percentage of infants who received a transfusion from 75% to 50%.

For most comparisons, statistical significance was determined by Fisher's exact test or ² test for categorical variables and by the Wilcoxon rank sum test for continuous variables. A 2-sided P value of <.05 was considered statistically significant. Groups were compared on the primary endpoint for trial 1 by fitting a linear regression model to the number of transfusions for each infant. This model included an indicator for treatment assignment as well as effects for study center (indicator variables), baseline number of transfusions, and initial center by treatment interaction. No significant differences were detected in treatment effect by center (P = .6), and the model was refit without the interaction. Statistical significance for the comparison between treatment groups was determined by the F test from this model. For trial 2, statistical significance for comparison of the percentage of infants who received a transfusion in each group was determined by the Mantel-Haenszel ² controlling for study center.

Comparisons of cumulative number of transfusions, reticulocyte counts, and hematocrits over time between the treatment groups were made using linear repeated measures models. Models were fit to each outcome across study weeks using generalized estimating equation methodology, which accounts for the correlation of measures obtained on the same patient and allows for missing data. In addition to the treatment group indicator, models included effects for time (weekly measures starting with week 1), baseline value of the endpoint, and study center. Models were fit initially with an interaction term between week and treatment group to compare the treatment effect at each week and were refit with no interaction term to obtain an overall test between treatment groups. Statistical significance for treatment comparisons from these models was determined by the Wald ² test.

Trial 2 was discontinued after enrollment of 118 infants after the Data and Safety Monitoring Committee reviewed the final results of trial 1 and preliminary results of trial 2. On the basis of these data, the Committee concluded that it was statistically unlikely that trial 2 would demonstrate a significant decrease in the percentage of infants who would receive a transfusion during the study.

	RESULTS

Top Abstract Methods Results Discussion Conclusion References

Of the infants who were eligible for enrollment, 172 (54%) of 318 in trial 1 and 118 (41%) of 288 in trial 2 were randomized. Reasons for nonrandomization in trial 1 were parent refusal (56%), consent not requested (32%), parent unavailable (7%), and physician refusal (5%). Reasons for nonrandomization in trial 2 were parent refusal (62%), consent not requested (29%), parent unavailable (8%), and physician refusal (1%). Decreased availability of research staff at a single site accounted for 76% (trial 1) and 64% (trial 2) of the eligible infants for whom consent was not requested. Results are presented on all 290 patients, 89% of whom in trial 1 and 93% of whom in trial 2 received at least 90% of the expected number of doses.

The treatment and placebo/control groups did not differ significantly in birth weight, gestational age, age at study entry, exposure to antenatal steroid therapy, gender, or race (Table 2). Hematocrit, absolute reticulocyte count, and serum ferritin concentration also were similar at randomization, as were phlebotomy losses and number and volume of transfusions from birth to study entry (Table 2).

Trial 1

During the study period, 73 (84%) of 87 treated infants and 74 (87%) of 85 placebo/control infants received a transfusion (P = .56). A total of 371 transfusions were administered to the treated infants, and 438 were administered to the placebo/control infants. The average number of transfusions received by each infant during the study was similar for treated and placebo/control infants (4.3 ± 3.6 vs 5.2 ± 4.2; P = .09) as was the total volume per infant who received a transfusion (80 ± 53 mL/kg vs 95 ± 63 mL/kg; P = .16; Table 3). Timing of transfusions was examined by adding the number of transfusions received during each week of the study for each infant separately and then calculating the average cumulative number of transfusions per infant among those who were alive at each week (Fig 1, upper panel). By study week 7, the cumulative number of transfusions was significantly higher for placebo/control infants (P < .05 at each time point, weeks 7-10). Transfusions that did not comply with the protocol occurred in 14% of the treated group and 12% of the placebo/control group (P = .5). The distribution of indications for transfusions administered per protocol varied significantly between the groups (P = .04; Table 3). For both groups, most of these noncompliant transfusions occurred because the recipient's hematocrit exceeded the value specified by the transfusion guidelines or because the recipients received a transfusion volume that was less than that specified by the transfusion guidelines.

View larger version (20K): [in this window] [in a new window]   Fig. 1.   The average cumulative number of transfusions received by each infant over time is shown for each treatment group for trial 1 (upper panel) and trial 2 (lower panel). Treated infants are represented by the solid circles and lines, and placebo/control infants are represented by the open circles and dashed lines. The circles indicate means at each study week with bars showing standard deviations. For each treatment group, values were determined by adding the number of transfusions during each week of the study for each infant separately, then averaging the cumulative transfusions per infant among those who were alive at each week. Thus, with this correction for mortality, the denominator used at each week changed over time (the weekly average number of transfusions per infant shown in this figure therefore differs from the overall average based on all study patients presented in Table 3). By study week 7, the cumulative number of transfusions was significantly higher for placebo/control infants. *P < .05 at each time point, weeks 7 to 10. Trial 1 patient numbers on which means are based at each week ranged: Epo, 87 (week 1) to 72 (week 10); placebo/control, 85 (week 1) to 72 (week 10). Trial 2 patient numbers ranged: Epo, 59 (week 1) to 58 (week 8); placebo/control, 59 (week 1) to 55 (week 8).

Fourteen treated infants (16%) and 11 placebo/control infants (13%) received no transfusions during the study period. Infants who required no transfusions were more likely to have weighed more at birth, to be of older gestational age, to be female, to have received their first study dose earlier, to have had a higher baseline hematocrit, and to have had lower prestudy phlebotomy loss and were less likely to have had a transfusion before study entry. Smaller percentages of infants who did not receive a transfusion were diagnosed with CLD, PDA, late-onset culture-positive sepsis, and late-onset culture-negative sepsis. A smaller proportion died and hospital stay was significantly shorter for infants who did not receive a transfusion.

Reticulocyte counts in the treated infants remained high during the study (Fig 2), whereas those in the placebo/control group decreased from baseline levels and stayed below baseline. At each week, reticulocyte counts were significantly higher in the treated infants than in placebo/controls (P < .001 overall). Hematocrits decreased somewhat from baseline levels in both groups and remained below baseline during the study (Fig 3). Starting at week 2, hematocrits were significantly greater in treated infants at each week except for week 5 (P < .05 at weeks 2-4, 6-10). There were no significant differences between groups in ANC during the study. Platelet counts were significantly lower during weeks 1, 3, and 8 in the treated group (P < .05) but remained >200 000/µL in all infants.

View larger version (23K): [in this window] [in a new window]   Fig. 2.   Changes in absolute reticulocyte counts during the study period. Average absolute reticulocyte counts in Epo recipients (solid circles and line) and placebo recipients (open circles and dashed line) during trial 1 (upper panel) and trial 2 (lower panel) are shown from study entry (week 0) through week 10 (trial 1) or week 8 (trial 2). The circles indicate means at each study week with bars showing standard deviations. Absolute reticulocyte counts were higher in the treatment group than in the placebo/control group throughout the study in both trials. *P < .001. Trial 1 patient numbers on which means are based at each week starting with week 0 (baseline): Epo, 73 (week 0), 64 (week 5), 23 (week 10); placebo/control, 70 (week 0), 58 (week 5), 20 (week 10). Trial 2 patient numbers: Epo, 53 (week 0), 44 (week 4), 2 (week 8); placebo/control, 51 (week 0), 43 (week 4), 10 (week 8).

View larger version (19K): [in this window] [in a new window]   Fig. 3.   Changes in hematocrits during the study period. Average hematocrits in Epo recipients (solid circles and line) and placebo recipients (open circles and dashed line) during trial 1 (upper panel) and trial 2 (lower panel) are shown from study entry (week 0) through week 10 (trial 1) or week 8 (trial 2). The circles indicate means at each study week with bars showing standard deviations. In both trials, hematocrits decreased in all patients from birth to week 1 of the study and remained lower than birth values in the placebo/control group throughout the study period. Hematocrits were greater in the treatment group in both trials from week 2 onward. *P < .05 except for week 5 in trial 1. Trial 1 patient numbers on which means are based at each week starting with week 0 (baseline): Epo, 87 (week 0), 71 (week 5), 24 (week 10); placebo/control, 82 (week 0), 65 (week 5), 25 (week 10). Trial 2 patient numbers: Epo, 59 (week 0), 50 (week 4), 3 (week 8); placebo/control, 58 (week 0), 49 (week 4), 10 (week 8).

Three infants in the placebo/control group and 3 in the treated group had their study medication held for 1 week because of neutropenia. No infant had treatment stopped because of neutropenia. No study infants had their medication held because of hypertension. Nine placebo/control infants and 3 treated infants had study medication stopped for clinical seizures. Three infants in each group had their study drug held for 1 week for a hematocrit of >45%.

Ferritin concentrations at the beginning of the study were similar for both groups (Table 4) but were greater in the placebo/control group at weeks 4 and 8 (P < .01 at each). The average number of parenteral iron doses per infant was similar in the 2 groups (2.8 ± 1.8 in treated vs 2.7 ± 1.7 in control; P = .67). Likewise, the average age (days since birth) at which oral iron was started was similar (22.1 ± 11.9 in treated vs 21.4 ± 10.1 in control; P = .90). No adverse events were attributed to administration of parenteral or enteral iron. There was a significant difference between treated and placebo/control groups in the percentage of infants who required changes in iron administration (Table 4). More placebo/control than treated infants had their iron dose decreased as a result of a serum ferritin concentration of 400 ng/mL (44% vs 22%; P < .01). The proportion of infants whose iron dose was increased because of a serum ferritin concentration of 50 ng/mL was not significantly different in the 2 groups (P = .08).

The average age at which infants completed the study protocol was similar among treated and placebo/control infants (34.3 ± 3.5 vs 33.7 ± 3.7 weeks' PMA). Two treated and 3 control infants were discharged to home before 35 weeks' PMA, whereas 14 treated and 13 control infants died before 35 weeks. Four infants from each group were transferred to other hospitals before 35 weeks' PMA (PMA at the time of transfer: treated infants, 31.1 ± 3.4; placebo/control infants, 32.0 ± 2.8). Infants who were transferred before 35 weeks' PMA had a similar number of transfusions from the time of transfer to the end of the study (treated infants, 0.7 ± 0.6 transfusions; placebo/control infants, 0.5 ± 1.0 transfusions).

All study infants were followed for morbidity and mortality through their hospital stay up to 120 days. By hospital discharge or 120 days, 15 treated infants (17%) and 15 placebo/control infants (18%) had died (P = 1.0). The causes of death were similar between treated (respiratory distress/CLD, 3; sepsis/NEC, 8; severe IVH, 1; other, 3) and placebo/control groups (respiratory distress/CLD, 4; sepsis/NEC, 6; severe IVH, 1; other, 4). There were no significant differences between treated and placebo/control infants in medical morbidities or length of hospital stay (Table 5).

Trial 2

During the study, 22 (37%) of 59 treated infants and 27 (46%) of 59 placebo/control infants received a transfusion (P = .25). A total of 58 transfusions were administered to the treated infants, and 62 transfusions were administered to the placebo/control infants. The average number of transfusions received by each infant was similar for treated and placebo/control infants (1.0 ± 1.6 vs 1.1 ± 1.5; P = .42) as was the total volume per infant who received a transfusion (39 ± 26 mL/kg vs 35 ± 24 mL/kg; P = .89; Table 3). The cumulative number of transfusions was similar for both groups at each study week (Fig 1, lower panel). Transfusions that did not comply with the protocol occurred in 17% of the treated group and 29% of the placebo/control group (P = .14). A greater percentage of treated infants received transfusions per protocol while on moderate or significant mechanical ventilation (Table 3) than placebo/control infants (P = .02).

Thirty-seven treated (63%) and 32 placebo/control infants (54%) received no transfusions during the study period. Infants who did not receive a transfusion were more likely to have weighed more at birth, to be of older gestational age, and to have had a higher baseline hematocrit and a lower prestudy phlebotomy loss than infants who received a transfusion and were less likely to have had a transfusion before study entry. Smaller percentages of infants who did not receive a transfusion received a diagnosis of CLD, NEC, or culture-positive or culture-negative late-onset sepsis. A smaller proportion died and hospital stay was significantly shorter for infants who did not receive a transfusion.

After an initial decrease below baseline levels, absolute reticulocyte counts in the treated infants increased and remained high during the study (Fig 2), whereas those in the placebo/control infants remained below baseline. At each week, reticulocyte counts were significantly higher in the treated infants (P < .001 overall). Hematocrits decreased similarly from baseline in both groups in the first week (Fig 3) and continued to drop in the placebo/control infants while remaining unchanged in the treated infants. Hematocrits were greater in treated infants from week 2 onward (P < .05 at each, weeks 2-8). There were no significant differences between groups in ANC during the study. Platelet counts were lower during weeks 3, 4, and 5 in the treated group (P < .05).

None of the infants had treatment held or stopped because of neutropenia, clinical seizures, or hypertension. Eighteen treated infants and 7 placebo/control infants had their study drug held because of a hematocrit of >45% (P < .05).

Ferritin concentrations were similar (P = .10) in the treated and placebo/control groups at the beginning of the study (Table 4) but were greater in the placebo/control group at week 4 (P < .01). The average number of intravenous iron doses per infant was not significantly different in the groups (1.8 ± 1.2 in treated vs 1.5 ± 0.8 in control; P = .27). In addition, the average age (days since birth) at which oral iron was started was similar in the groups (14.6 ± 9.2 in treated and 13.8 ± 5.7 in control; P = .96). No adverse events were attributed to parenteral or enteral iron. No difference was detected in the percentage of infants in each group whose iron dose was decreased because of a serum ferritin concentration of 400 ng/mL (P = 1.0) or in the proportion whose dose was increased because of a serum ferritin concentration of 50 ng/mL (P = .32).

The average age at which infants completed the study protocol was similar among treated and placebo/control infants (34.8 ± 1.8 vs 34.8 ± 1.7 weeks' PMA). Eleven treated and 9 control infants were discharged to home before 35 weeks' PMA, whereas 1 treated and 3 control infants died before 35 weeks. Three treated infants and 6 placebo/control infants were transferred to other hospitals before 35 weeks' PMA (PMA at the time of transfer: treated infants, 33.2 ± 2.3; placebo/control infants, 33.0 ± 0.6). Infants who were transferred before 35 weeks' PMA had a similar number of transfusions from the time of transfer to the end of the study (treated infants, 0.3 ± 0.6 transfusions; placebo/control infants, 0.3 ± 0.6 transfusions).

By hospital discharge or 120 days, 1 treated infant (2%) and 4 placebo/control infants (7%) had died (P = .36). The causes of death were similar between treated (CLD, 1) and placebo/control groups (sepsis/NEC, 4). There were no significant differences between treated and placebo/control infants in medical morbidities or length of hospital stay (Table 5).

	DISCUSSION

Top Abstract Methods Results Discussion Conclusion References

This was the first large, multicenter, placebo-controlled study designed to optimize early dosing of Epo and iron in critically ill premature infants. In this study, although erythropoiesis was induced in treated infants, as evidenced by significantly elevated reticulocyte counts and higher hematocrits, the numbers of transfusions and donor exposures were not significantly reduced. In trial 1, the average number of transfusions per infant during the study period was reduced in the treated infants but not to the level of statistical significance (P = .09). Analysis of the cumulative number of transfusions per infant at each study week indicated that the transfusion profiles were not parallel over time and that in fact the cumulative number was lower in the treated infants during the last 4 weeks of the study (P < .05 at each of these weeks). This, however, did not translate into an overall difference in average number of transfusions for the entire study period. In trial 2, the proportion of treated and placebo/control infants who received a transfusion was not significantly different (P = .25).

The minimal impact of Epo on reducing transfusion requirements might be attributable to a variety of reasons, including phlebotomy losses that exceeded infants' ability to maintain a specified hematocrit,^7,17 starting hematocrits that were lower than expected, shortened red cell survival with Epo treatment,¹⁸ decreased adherence to transfusion guidelines, and less effective administration of Epo. Despite achieving elevated rates of erythropoiesis, we were unable to demonstrate a significant impact on transfusions because of the combination of high phlebotomy losses and compliance with strict transfusion guidelines. For example, if infants had been transfused at higher hematocrits, then a significant difference between Epo-treated and placebo/control infants likely would have been noted. The results of trial 1 are similar to the US multicenter trial in that infants who received Epo in that study began the study with an average of 3.5 transfusions per infant, then received an additional 1.1 transfusions during the study,⁷ for a total of approximately 4.6 transfusions. Both trial 1 and trial 2 differ from the multicenter study of Maier et al⁸ in that infants in this study were much smaller and sicker and had greater phlebotomy losses. In a previous study, serum Epo concentrations in preterm infants who received similar intravenous dosing averaged 1600 to 1800 mU/mL.¹⁹ It is possible that a different dosing strategy that achieves lower peak serum concentrations over a more prolonged period of time would be more efficacious.^7,11,20

No differences were noted between groups in side effects or morbidities. Although previously published studies have observed a decreased incidence of PDA,^6,21 a statistically significant decrease was not noted among treated infants in either trial of the present study. Other previously reported side effects of Epo include hypertension and seizures in adults and neutropenia in neonates.¹⁰ A greater incidence of hypertension or seizures was not observed in treated infants, and neutropenia did not occur more commonly.

Studies that evaluated Epo in adult and pediatric patients reported limited erythropoiesis when iron supplementation was not adequate.^22,23 Previous neonatal studies evaluated the use of both enteral and parenteral iron in a wide dosing range and noted evidence of iron deficiency when the amount of iron supplementation was low.^6,13,24 We administered early supplemental iron in an effort to maximize the amount of iron available for erythropoiesis. In both treated and placebo/control infants, serum ferritin concentrations remained at greater than or equal to baseline levels, presumably reflecting adequate administration of iron throughout the study. We speculate that ferritins were lower in Epo recipients because Epo stimulated the incorporation of iron into red cells.

The side effects of parenteral iron reported previously include infection, iron overload, hypersensitivity reactions, and, rarely, anaphylaxis. None of these effects were noted in our study. Past studies that reported increased incidence of infection in infants who received parenteral iron were based on intramuscular routes and doses of >100 mg/d.^25,26 In this study, there were no differences in late-onset sepsis episodes between treated and placebo/control groups in either trial. A similar incidence of culture-positive late-onset sepsis was reported by Stoll et al²⁷ in the National Institute of Child Health and Human Development network population. Other investigators have suggested that iron overload, transfusions, and oxidant injury play a role in the pathogenesis of neonatal morbidities.^28,29 No difference in neonatal morbidities was noted between treated and placebo/control infants. In addition, the incidence of morbidities such as ROP and CLD is similar to those reported previously by the National Institute of Child Health and Human Development Neonatal Research Network.³⁰ It should be noted, however, that this study was not powered to determine differences in any of these secondary outcomes.

The transfusion guidelines chosen for this study are conservative compared with some of the previous trials.^6,8 Hematocrits were maintained at a lower level without an increase in morbidities or hospital stay. This underscores the impact on transfusion number that follows the institution of conservative transfusion guidelines.^31-33 The donor exposures in trial 1 were lower than previous studies in extremely low birth weight infants as a result of the availability of designated donor units.³⁴ Exposures in trial 2 were higher than anticipated because infants who received transfusions generally received >1 transfusion, which often occurred after the initial donor unit had expired. We speculate that future development of noninvasive laboratory monitoring and improved neonatal blood banking practices should help to reduce phlebotomy losses and result in decreased transfusions.^35,36

Previous analyses have provided conflicting estimates of cost-effectiveness.^18,37 A specific cost-benefit analysis was not performed. We found that the early initiation of Epo therapy provided no additional benefit over the reduction in blood transfusions that was achieved by using conservative transfusion guidelines. This suggests that Epo therapy as administered in our study is not cost-effective.

	CONCLUSION

Top Abstract Methods Results Discussion Conclusion References

Despite evidence of significantly increased erythropoiesis, the administration of Epo and supplemental iron did not decrease the average number of transfusions or donor exposures per infant in infants of 1000 g birth weight or decrease the percentage of infants who had a birth weight of 1001 to 1250 g and received a transfusion. There were no statistically significant differences in neonatal morbidities between groups in either trial, and no adverse effects of Epo or iron supplementation were noted.

We conclude that the combination of Epo and iron stimulates erythropoiesis, maintains hematocrits at a higher level, and is safe in very low birth weight infants. However, when conservative transfusion guidelines are followed, the combination of Epo and iron as used in this trial does not have a significant impact on the number of transfusions administered to very low birth weight infants. On the basis of the results of this study, the early use of Epo and iron to reduce transfusion number and exposure of infants to the risks of multiple blood transfusions is not warranted.

	ACKNOWLEDGMENTS

This study was supported by a grant from the National Institute of Child Health and Human Development, National Institutes of Health, through cooperative agreements with the authors' institutions: U10 HD27881 (R. K. O.); U10 HD27871 (R. A. E.); U10 HD27856 (J. A. L.); U10 HD21415 (S. B. K.); U10 HD27851 (B. J. S.); U10 HD34167 (A. R. S.); U10 HD21385 (S. S.); U10 HD27853 (E. F. D.); U01 HD19897 (N. C. C.); U01 HD36790 (A. D.); and General Clinical Research Centers M01 RR 00997 (R. K. O.); M01 RR 06022 (R. A. E.); M01 RR 00750 (J. A. L.); M01 RR02635, M01 RR 02172, M01 RR 01032 (A. R. S.); and M01 RR 08084 (S. S.); and generous grants from Ortho-Biotech and Schein Pharmaceuticals.

We thank Nellie I. Hansen for assistance in statistical analysis.

	FOOTNOTES

Received for publication Sep 14, 2000; accepted Jan 29, 2001.

Reprint requests to (R.K.O.) University of New Mexico School of Medicine, 2211 Lomas St, Albuquerque, NM 87131. E-mail: rohls{at}unm.edu

	ABBREVIATIONS

Epo, human recombinant erythropoietin; PMA, postmenstrual age; ANC, absolute neutrophil count; PRBC, packed red blood cells; CLD, chronic lung disease; ROP, retinopathy of prematurity; IVH, intraventricular hemorrhage; PDA, patent ductus arteriosus; NEC, necrotizing enterocolitis.

	REFERENCES

Top Abstract Methods Results Discussion Conclusion References

1. Ohls RK, Harcum J, Schibler KR, Christensen RD The effect of erythropoietin on the transfusion requirements of preterm infants 750 grams: a randomized, double-blind, placebo-controlled study. J Pediatr 1997; 131:661-665 [CrossRef][Medline]
2. Stockman JA III, Graeber JE, Clark DA, McClellan K, Garcia JF, Kavey REW Anemia of prematurity: determinants of the erythropoietin response. J Pediatr 1984; 105:786-792 [CrossRef][Medline]
3. Brown MS, Garcia JF, Phibbs RH, Dallman PR Decreased response of plasma immunoreactive erythropoietin to "available oxygen" in anemia of prematurity. J Pediatr 1984; 105:793-798 [CrossRef][Medline]
4. Shannon KM, Naylor GS, Torkildson JC, Circulating erythroid progenitors in the anemia of prematurity. N Engl J Med 1987; 317:728-733 [Abstract]
5. Rhondeau SM, Christensen RD, Ross MP, Rothstein G, Simmons MA Responsiveness to recombinant human erythropoietin of marrow erythroid progenitors from infants with the "anemia of prematurity." J Pediatr 1988; 112:935-940 [CrossRef][Medline]
6. Meyer MP, Meyer JH, Commerford A, Recombinant human erythropoietin in the treatment of the anemia of prematurity: results of a double-blind, placebo-controlled study. Pediatrics 1994; 93:918-923 [Abstract/Free Full Text]
7. Shannon KM, Keith JF, Mentzer WC, Recombinant human erythropoietin stimulates erythropoiesis and reduces erythrocyte transfusions in very low birth weight preterm infants. Pediatrics 1995; 95:1-8 [Abstract/Free Full Text]
8. Maier RF, Obladen M, Scigalla P, The effect of epoetin beta (recombinant human erythropoietin) on the need for transfusion in very low birth weight infants. N Engl J Med 1994; 330:1173-1178 [Abstract/Free Full Text]
9. Soubasi V, Kremenopoulos G, Diamanti E, Tsantali C, Sarafidis K, Tsakiris D Follow-up of very low birth weight infants after erythropoietin treatment to prevent anemia of prematurity. J Pediatr 1995; 127:291-297 [CrossRef][Medline]
10. Ohls RK The use of erythropoietin in neonates. Clin Perinatol 2000; 3:681-696
11. Ohls RK, Veerman MW, Christensen RD Pharmacokinetics and effectiveness of recombinant erythropoietin administered to preterm infants by continuous infusion in parenteral nutrition solution. J Pediatr 1996; 128:518-523 [CrossRef][Medline]
12. Friel JK, Andrews WL, Hall MS, Intravenous iron administration to very-low-birth-weight newborns receiving total and partial parenteral nutrition. J Parenter Enteral Nutr 1995; 19:114-118 [Abstract/Free Full Text]
13. Meyer MP, Haworth C, Meyer JH, Commerford A A comparison of oral and intravenous iron supplementation in preterm infants receiving recombinant erythropoietin. J Pediatr 1996; 129:258-263 [CrossRef][Medline]
14. The Committee for the Classification of Retinopathy of Prematurity An international classification for retinopathy of prematurity. Arch Ophthalmol 1984; 102:1130-1134 [Abstract/Free Full Text]
15. Papile LA, Burstein J, Burstein R, Koffler H Incidence and evolution of subependymal and intraventricular hemorrhage: a study of infants with birth weights less than 1500 grams. J Pediatr 1978; 92:529-534 [CrossRef][Medline]
16. Walsh MC, Kliegman RM Necrotizing enterocolitis: treatment based on staging criteria. Pediatr Clin North Am 1986; 33:179-201 [Medline]
17. Goodnough LT, Skikne B, Brugnara C Erythropoietin, iron, and erythropoiesis. Blood 2000; 96:823-833 [Abstract/Free Full Text]
18. Brugnara C Reticulocyte cellular indices: a new approach in the diagnosis of anemias and monitoring of erythropoietic function. Crit Rev Clin Lab Sci 2000; 37:93-130 [CrossRef][Medline]
19. Cogar AA, Hartenberger CH, Ohls RK Endothelin concentrations in preterm infants treated with human recombinant erythropoietin. Biol Neonate 2000; 77:105-108 [CrossRef][Medline]
20. Brown MS, Keith JF III Comparison between two and five doses a week of recombinant human erythropoietin for anemia of prematurity: a randomized trial. Pediatrics 1999; 104:210-215 [Abstract/Free Full Text]
21. Ohls RK, Osborne KA, Christensen RD Efficacy and cost analysis of treating very low birth weight infants with erythropoietin during their first two weeks of life: a randomized, placebo controlled trial. J Pediatr 1995; 126:421-426 [CrossRef][Medline]
22. Eschbach JW, Kelly MR, Haley RN, Abels RI, Adamson JW Treatment of the anemia of progressive renal failure with recombinant human erythropoietin. N Engl J Med 1989; 321:158-163 [Abstract]
23. Skikne BS, Cook JD Effect of enhanced erythropoiesis on iron absorption. J Lab Clin Med 1992; 120:746-751 [Medline]
24. Bechensteen AG, Hågå P, Halvorsen S, Erythropoietin, protein, and iron supplementation and the prevention of anaemia of prematurity. Arch Dis Child 1993; 69:19-23 [Abstract/Free Full Text]
25. Barry DMJ, Reeve AW Increased incidence of gram-negative neonatal sepsis with intramuscular iron administration. Pediatrics 1977; 60:908-912 [Abstract/Free Full Text]
26. Becroft DMO, Dix MR, Farmer K Intramuscular iron-dextran and susceptibility of neonates to bacterial infections. Arch Dis Child 1977; 52:778-781 [Abstract/Free Full Text]
27. Stoll BJ, Gordon T, Korones SB, Late-onset sepsis in very low birth weight neonates: a report from the National Institute of Child Health and Human Development Neonatal Research Network. J Pediatr 1996; 129:63-71 [CrossRef][Medline]
28. Berger HM, Lindeman JH, van Zoeren-Grobben D, Houdkamp E, Schrijver J, Kanhai HH Iron overload, free radical damage, and rhesus haemolytic disease. Lancet 1990; 335:933-936 [CrossRef][Medline]
29. Hesse L, Eberl W, Schlaud M, Poets CF Blood transfusion. Iron load and retinopathy of prematurity. Eur J Pediatr 1997; 156:465-470 [CrossRef][Medline]
30. Stevenson DK, Wright LL, Lemons JA, Very low birth weight outcomes of the National Institute of Child Health and Human Development Neonatal Research Network, January 1993 through December 1994. Am J Obstet Gynecol 1998; 179:1632-1639 [CrossRef][Medline]
31. Bifano EM, Curran TR Minimizing donor blood exposure in the neonatal intensive care unit. Current trends and future prospects. Clin Perinatol 1995; 22:657-669 [Medline]
32. Widness JA, Seward VJ, Kromer IJ, Burmeister LF, Bell EF, Strauss RG Changing patterns of red cell transfusion in very low birth weight infants. J Pediatr 1996; 29:680-687
33. Maier RF, Sonntag J, Walka MM, Guosheng L, Metze BC, Obladen M Changing practices of red blood cell transfusions in infants with birth weights less than 1000 g. J Pediatr 2000; 136:220-224 [CrossRef][Medline]
34. Liu EA, Mannino FL, Lane TA Prospective, randomized trial of the safety and efficacy of a limited donor exposure transfusion program for premature neonates. J Pediatr 1994; 125:92-96 [CrossRef][Medline]
35. Strauss RG, Burmeister LF, Johnson K, Cress G, Cordle D Feasibility and safety of AS-3 red blood cells for neonatal transfusions. J Pediatr 2000; 136:215-219 [CrossRef][Medline]
36. Widness JA, Kulhavy JC, Johnson KJ, Clinical performance of an in-line point-of-care monitor in neonates. Pediatrics 2000; 106:497-504 [Abstract/Free Full Text]
37. Shireman TI, Hilsenrath PE, Strauss RG, Widness JA, Mutnick AH Recombinant human erythropoietin vs transfusions in the treatment of the anemia of prematurity. A cost-benefit analysis. Arch Pediatr Adolesc Med 1994; 148:582-588 [Abstract/Free Full Text]