Effect of blood transfusions on cognitive development in very low birth weight infants.

OBJECTIVE: Preterm infants frequently receive red cell transfusions; however, the effect of transfusions on cognition is unclear. We evaluated the relationship between transfusions and cognitive outcomes in preterm infants enrolled in a randomized trial of erythropoiesis stimulating agents (ESAs). STUDY DESIGN: Preterm infants were randomized to ESAs or placebo during initial hospitalization, and transfusions recorded. Children were evaluated using standard developmental tests of cognition at 18-22 months (56 ESA, 24 placebo) and 3.5-4 years (39 ESA, 14 placebo). RESULTS: Cognitive scores at 18-22 months were inversely correlated with transfusion volume (p = 0.02). Among those receiving ≥1 transfusion, cognitive scores were significantly higher in the ESA-treated group (p = 0.003). At 3.5-4 years, transfusions were not correlated with cognitive scores. CONCLUSIONS: In the placebo group, transfused children had lower cognitive scores than did non-transfused children at 18-22 months. In the ESA group, cognitive scores did not differ by transfusion status, suggesting ESAs might provide neuroprotection.

Exchange transfusion for hemolytic hyperbilirubinemia: could some be averted by emergent administration of an inhibitor of bilirubin production?

OBJECTIVES: The objective of this study is to explore the hypothetical number of neonates where an exchange transfusion (ET) could be prevented by emergency administration of an inhibitor of bilirubin production. STUDY DESIGN: We identified all neonates who received an ET in our NICUs during the past 12 years. We reviewed the indications for ET and recorded the time between ordering and beginning the exchange. RESULTS: Forty-six neonates underwent ET, 37 (80.4%) for hemolytic hyperbilirubinemia (36.9 ± 2.9 weeks gestation and 2.5 ± 2.1 days old at ET). The mean delay period was 7.5 ± 3.5 h. Nine (19.6%) had ET not involving bilirubin. CONCLUSIONS: A trial testing compounds that can inhibit bilirubin production would have about three eligible neonates/years in our system. Since our births are 1% of national, up to 300 neonates/years might qualify for such a study.

Neuroimaging in former preterm children who received erythropoiesis stimulating agents.

BackgroundIn premature children, erythropoiesis-stimulating agents (ESAs) may improve developmental outcome. It is not clear which of the several potential mechanisms are responsible for this improvement. High-resolution MRI and diffusion tensor imaging characterize brain structure and white matter organization, offering possible insight into the long-term effect of ESAs on brain development.MethodsMRI scans were performed at 3.5-4 years of age on former preterm infants treated with ESAs or placebo, and on healthy term controls. Mean cortical thickness, surface area, and fractional anisotropy (FA) were compared across study groups, and were correlated with general IQ measures.ResultsUnivariate analysis found no significant effect of ESAs on cortical thickness (P=0.366), surface area (P=0.940), or FA (P=0.150); however, there was a greater increase in FA among ESA-treated girls. Group analysis found significant correlations between FA and Full-Scale IQ (P=0.044) and Verbal IQ (P=0.036), although there was no significant relationship between Full-Scale IQ and FA among just the preterm children.ConclusionESA treatment may have a preferential effect on white matter development in girls, although factors other than just whole-brain FA are involved in mediating cognitive outcome.

Preschool Assessment of Preterm Infants Treated With Darbepoetin and Erythropoietin.

BACKGROUND: We previously reported improved neurodevelopmental outcomes at 2 years among infants treated with the erythropoiesis-stimulating agents (ESAs) darbepoetin alfa (darbepoetin) or erythropoietin. Here we characterize 4-year outcomes. METHODS: Former preterm infants randomly assigned to receive darbepoetin (10 µg/kg, once per week), erythropoietin (400 U/kg, 3 times/week), or placebo through 35 weeks' postconceptual age were evaluated at 3.5 to 4 years of age. For comparison, healthy children formerly delivered full term (term controls [TCs]) were also recruited. All participants were assessed by using measures of full-scale IQ (FSIQ) and general language from the Wechsler Preschool and Primary Scale of Intelligence, Third Edition, and an overall measure of executive function, on the basis of tests evaluating inhibitory control and spatial working memory. Rates of neurodevelopmental impairment were compared across groups. RESULTS: Multivariate analysis of variance compared children randomly assigned to ESAs (n = 39), placebo (n =14), and TCs (n = 24). FSIQ and performance IQ were significantly higher in the ESA group than in the placebo group (FSIQ: 91.1 ± 17.5 vs 79.2 ± 18.5, P = .036; performance IQ: 93.0 ± 17.0 vs 79.5 ± 19.5, P = .018). Follow-up analyses revealed that the children receiving ESAs performed better than those who received placebo on executive function tasks. The ESA group's performance was below that of TCs, but the results did not reach significance on executive function. The incidence of neurodevelopmental impairment was greater in the placebo group than in the ESA group. CONCLUSIONS: ESA-treated infants had better cognitive outcomes and less developmental impairment at 3.5 to 4 years of age compared with placebo-treated infants. ESAs show promise in improving long-term cognitive outcomes of infants born prematurely.

Cognitive outcomes of preterm infants randomized to darbepoetin, erythropoietin, or placebo.

BACKGROUND: We previously reported decreased transfusions and donor exposures in preterm infants randomized to Darbepoetin (Darbe) or erythropoietin (Epo) compared with placebo. As these erythropoiesis-stimulating agents (ESAs) have shown promise as neuroprotective agents, we hypothesized improved neurodevelopmental outcomes at 18 to 22 months among infants randomized to receive ESAs. METHODS: We performed a randomized, masked, multicenter study comparing Darbe (10 µg/kg, 1×/week subcutaneously), Epo (400 U/kg, 3×/week subcutaneously), and placebo (sham dosing 3×/week) given through 35 weeks' postconceptual age, with transfusions administered according to a standardized protocol. Surviving infants were evaluated at 18 to 22 months' corrected age using the Bayley Scales of Infant Development III. The primary outcome was composite cognitive score. Assessments of object permanence, anthropometrics, cerebral palsy, vision, and hearing were performed. RESULTS: Of the original 102 infants (946 ± 196 g, 27.7 ± 1.8 weeks' gestation), 80 (29 Epo, 27 Darbe, 24 placebo) returned for follow-up. The 3 groups were comparable for age at testing, birth weight, and gestational age. After adjustment for gender, analysis of covariance revealed significantly higher cognitive scores among Darbe (96.2 ± 7.3; mean ± SD) and Epo recipients (97.9 ± 14.3) compared with placebo recipients (88.7 ± 13.5; P = .01 vs ESA recipients) as was object permanence (P = .05). No ESA recipients had cerebral palsy, compared with 5 in the placebo group (P < .001). No differences among groups were found in visual or hearing impairment. CONCLUSIONS: Infants randomized to receive ESAs had better cognitive outcomes, compared with placebo recipients, at 18 to 22 months. Darbe and Epo may prove beneficial in improving long-term cognitive outcomes of preterm infants.

Neonatal death suspected to be from sepsis was found to be kernicterus with G6PD deficiency.

We cared for a term male infant born to Burmese immigrants. At about 24 hours a total serum bilirubin (TSB) was 9.3 mg/dL, and phototherapy was begun. It was stopped 48 hours later, with a TSB of 10.9 mg/dL, and he was discharged from the hospital with an appointment for a repeat TSB check 48 hours later. A few hours before the appointment he became listless and apneic, and his parents took him to the emergency department of the regional children's hospital, where sepsis was suspected. The TSB was 41 mg/dL. He died 4 hours later, despite intensive care efforts, with opisthotonus and refractory hypotension. Blood drawn before the exchange transfusion had low glucose-6-phosphate dehydrogenase (G6PD) enzymatic activity, and sequencing of the G6PD gene revealed the G6PD Mahidol mutation (c.487G>A). Cultures and postmortem examination did not demonstrate an infectious process, but kernicterus was present. Acute kernicterus can mimic septic shock.

A randomized, masked, placebo-controlled study of darbepoetin alfa in preterm infants.

BACKGROUND: A novel erythropoiesis stimulating agent (ESA), darbepoetin alfa (Darbe), increases hematocrit in anemic adults when administered every 1 to 3 weeks. Weekly Darbe dosing has not been evaluated in preterm infants. We hypothesized that infants would respond to Darbe by decreasing transfusion needs compared with placebo, with less-frequent dosing than erythropoietin (Epo). METHODS: Preterm infants 500 to 1250 g birth weight and ≤48 hours of age were randomized to Darbe (10 µg/kg, 1 time per week subcutaneously), Epo (400 U/kg, 3 times per week subcutaneously) or placebo (sham dosing) through 35 weeks' gestation. All received supplemental iron, folate, and vitamin E, and were transfused according to protocol. Transfusions (primary outcome), complete blood counts, absolute reticulocyte counts (ARCs), phlebotomy losses, and adverse events were recorded. RESULTS: A total of 102 infants (946 ± 196 g, 27.7 ± 1.8 weeks' gestation, 51 ± 25 hours of age at first dose) were enrolled. Infants in the Darbe and Epo groups received significantly fewer transfusions (P = .015) and were exposed to fewer donors (P = .044) than the placebo group (Darbe: 1.2 ± 2.4 transfusions and 0.7 ± 1.2 donors per infant; Epo: 1.2 ± 1.6 transfusions and 0.8 ± 1.0 donors per infant; placebo: 2.4 ± 2.9 transfusions and 1.2 ± 1.3 donors per infant). Hematocrit and ARC were higher in the Darbe and Epo groups compared with placebo (P = .001, Darbe and Epo versus placebo for both hematocrit and ARCs). Morbidities were similar among groups, including the incidence of retinopathy of prematurity. CONCLUSIONS: Infants receiving Darbe or Epo received fewer transfusions and fewer donor exposures, and fewer injections were given to Darbe recipients. Darbepoetin and Epo successfully serve as adjuncts to transfusions in maintaining red cell mass in preterm infants.

Linking maternal platelet counts with neonatal platelet counts and outcomes using the data repositories of a multihospital health care system.

It is unclear whether neonates born to women with thrombocytopenia during pregnancy are themselves at increased risk for thrombocytopenia at birth. In the current retrospective study, platelet count reference ranges were developed for pregnant women according to trimester, and correlations were sought between the platelet counts of mothers at delivery and their neonates. During the study period, 92,518 platelet counts were recorded on 41,887 pregnant women. A progressive shift toward lower platelet counts in a similarly shaped histogram occurred during pregnancy, with the lower reference range (2.5 percentile) for platelets during the third trimester being 113 × 10(9)/L. Among 11,797 maternal-neonatal pairs following delivery, no correlation was observed between maternal and neonatal counts. However, if the mother's lowest count was <50 × 10(9)/L, the relative risk of any degree of thrombocytopenia in their neonate was 4.6 (95% confidence interval [CI], 1.8 to 33.3) and the relative risk of severe neonatal thrombocytopenia was 7.8 (95% CI, 1.8 to 33.3). The results of the current study demonstrate that platelet counts >75 × 10(9)/L in pregnant women were not associated with an increased risk of neonatal thrombocytopenia, and maternal platelet counts of <50 × 10(9)/L were accompanied by an almost fivefold risk increase of neonatal thrombocytopenia.

Implementing a program to improve compliance with neonatal intensive care unit transfusion guidelines was accompanied by a reduction in transfusion rate: a pre-post analysis within a multihospital health care system.

BACKGROUND: We previously reported that in the year 2006, approximately 35% of the transfusions administered in the Intermountain Healthcare neonatal intensive care units (NICU) were noncompliant with our transfusion guidelines. In January 2009 we instituted an electronic NICU transfusion ordering and monitoring system as part of a new program to improve compliance with transfusion guidelines. STUDY DESIGN AND METHODS: In the four largest NICUs of Intermountain Healthcare, we performed a pre-post analysis of compliance with transfusion guidelines and transfusion usage. RESULTS: After beginning the new transfusion compliance program all four NICUs had an increase in compliance from 65% to 90%. Accompanying the improved compliance, all four NICUs had a reduction in transfusions administered. Specifically, compared with 2007 and 2008, there were 984 fewer NICU transfusions given in 2009. This included 554 fewer red blood cell (RBC) transfusions, 174 fewer platelet transfusions, and 256 fewer frozen plasma infusions. We calculate that in 2009, a total of 200 NICU patients who in previous years would have received one or more transfusions instead received none. Applying specific Intermountain Healthcare billing data to the observed transfusion reductions, this new program resulted in an annual decrease of $780,074 in blood bank charges (blood administration charges were not included). During the 3-year period, January 2007 through December 2009, we detected no change in NICU demographics, major morbidities, length of hospital stay, or mortality rate. CONCLUSION: Implementing a systemwide NICU program to improve compliance with already-established transfusion guidelines increased compliance from 65% to 90%. Improved compliance with transfusion guidelines was accompanied by a significant reduction in transfusions given, with no increase in NICU length of stay or mortality rate.

A low-sodium solution for airway care: results of a multicenter trial.

BACKGROUND: Normal saline is sometimes instilled into the endotracheal tube preparatory to airway suctioning, to assist in removing thick secretions. However, saline can damage the antimicrobial properties of airway secretions. We previously described a low-sodium physiologically based solution for airway care and reported a small (n = 60) randomized trial in neonates, which showed trends toward less ventilator-associated pneumonia (VAP) and less chronic lung disease with the new solution. We now report a multicenter trial of that solution. METHODS: We conducted a before-and-after study with a parallel control group, in 4 level-3 neonatal intensive care units (NICUs). During year 1, all 4 NICUs used saline for airway care. During year 2, one NICU used the test solution exclusively while the other NICUs used saline exclusively. The 2 study outcomes were VAP (cases/1,000 ventilator days) and chronic lung disease, defined 3 ways: supplemental oxygen at 28 days; supplemental oxygen at 36 weeks gestation; and supplemental oxygen on hospital discharge. RESULTS: During the study period 1,116 neonates had endotracheal intubation for respiratory management. Of those, 1,029 received the standard saline for airway suctioning, and the 87 in NICU 4 received the test solution. NICU 4 had a decrease in VAP rate, from 4.2 VAPs/1,000 ventilator days with saline, to 1.6 VAPs/1,000 ventilator days with the test solution (P = .04), and also had the lowest prevalence of chronic lung disease (P < .001 for each definition). CONCLUSIONS: The test solution significantly reduced the VAP and chronic lung disease rates.

Is "transfusion-associated necrotizing enterocolitis" an authentic pathogenic entity?

BACKGROUND: Necrotizing enterocolitis (NEC) sometimes occurs after a transfusion, but it is unclear whether this is a chance association or cause and effect. STUDY DESIGN AND METHODS: We compared features of neonates that developed surgical NEC within 48 hours after transfusion with others that developed NEC not preceded by transfusion. We assessed the blood used for transfusion and feeding practices among NEC cases and controls. RESULTS: Forty neonates developed surgical NEC after a transfusion and 72 developed NEC unrelated to a transfusion. Those with NEC after transfusion were born at earlier gestation (mean 27 weeks, 90% confidence interval [CI] 26-28 years vs. mean 30, 90% CI 29-31; p < 0.001) and were of lower birth weight (mean 981 g, 90% CI 835-1128 g vs. mean 1371 g, 90% CI 1245-1496; p < 0.001) and developed NEC later during their neonatal intensive care unit course (day of life: mean 23, 90% CI 20-27 vs. mean 16, 90% CI 13-19; p < 0.001). Transfusions were more prevalent among those that developed NEC (p < 0.001). The blood transfused to those that developed NEC was not older, but those who developed NEC had been fed larger volumes of milk in the 24 hours before and during transfusion (p = 0.04) and were more likely to have been fed a bovine product during that period (p = 0.004). CONCLUSION: Approximately one-third of surgical NEC cases in our system occurred after a transfusion. We speculate that a target area for reducing the prevalence of posttransfusion NEC involves feeding practices immediately before and during RBC transfusion.

The CBC: reference ranges for neonates.

"Normal values" for blood parameters of neonates are generally unavailable, because blood is not usually drawn on healthy, normal neonates to establish normal ranges. Instead, "reference ranges" are used, consisting of the 5th to the 95th percentile values compiled from tests performed on neonatal patients with minimal pathology, under the premise that such ranges approximate normal values. In recent years, we have been seeking to establish reference ranges for various elements of the complete blood count (CBC) of neonates, using the large databases of Intermountain Healthcare, a health care system in the western United States. Establishing these reference ranges has been facilitated by using modern hematology analyzers and electronic data repositories of clinical and laboratory information. The present review brings together several of our recent reports, displaying reference ranges for elements of the CBC among neonates at various gestational and postnatal ages.

Reference ranges for hematocrit and blood hemoglobin concentration during the neonatal period: data from a multihospital health care system.

OBJECTIVE: "Reference ranges" are developed when it is impossible or inappropriate to establish "normal ranges" by drawing blood on healthy normal volunteers. Reference ranges for the hematocrit and the blood hemoglobin concentration of newborn infants have previously been reported from relatively small sample sizes by using measurement methods that now are considered outmoded. METHODS: We sought to develop reference ranges for hematocrit and hemoglobin during the neonatal period (28 days) by using very large sample sizes and modern hematology analyzers, accounting for gestational and postnatal age and gender. Data were assembled from a multihospital health care system after exclusion of patients with a high likelihood of an abnormal value and those who were receiving blood transfusions. RESULTS: During the interval from 22 to 40 weeks' gestation, the hematocrit and blood hemoglobin concentration increased approximately linearly. For every week advance in gestational age, the hematocrit increased by 0.64% and the hemoglobin concentration increased by 0.21 g/dL. No difference was seen on the basis of gender. During the 4-hour interval after birth, hematocrit/hemoglobin values of late preterm and term neonates (35-42 weeks' gestation) increased by 3.6% +/- 0.5% (mean +/- SD), those of neonates of 29 to 34 weeks' gestation remained unchanged, and those of <29 weeks' gestation decreased by 6.0% +/- 0.3%. During the first 28 days after birth, an approximately linear decrease in hematocrit/hemoglobin occurred. CONCLUSIONS: The figures presented herein describe reference ranges for hematocrit and blood hemoglobin concentration during the neonatal period, accounting for gestational and postnatal age.

Very high users of platelet transfusions in the neonatal intensive care unit.

BACKGROUND: In neonatal intensive care unit (NICU) practice, a small percentage of the patients receive a large proportion of the platelet (PLT) transfusions administered. This study sought to better define this very-high-user group. To accomplish this, records of all NICU patients in a multihospital health care system who, during a recent 5(1/2)-year period, received 20 or more PLT transfusions were examined. STUDY DESIGN AND METHODS: Electronic medical record repositories of Intermountain Healthcare neonates with dates of birth from January 1, 2002, through June 30, 2007, who received 20 or more PLT transfusions were identified. The causes of the thrombocytopenia were sought, whether each transfusion given was a treatment for bleeding versus prophylaxis was determined, whether each transfusion was compliant with our transfusion guidelines was judged, and the outcomes were tabulated. RESULTS: During this period, 45 patients received 20 or more PLT transfusions (median, 29; range, 20-79). Medical conditions could be categorized into six diagnoses: 1) extracorporeal membrane oxygenation (ECMO) for congenital diaphragmatic hernia (CDH; n = 13), 2) fungal sepsis (n = 8), 3) ECMO for reasons other than CDH (n = 8), 4) necrotizing enterocolitis (n = 7), 5) bacterial sepsis (n = 7), and 6) congenital hyporegenerative thrombocytopenia (n = 2). Nineteen percent of the transfusions were ordered for oozing, bruising, or bleeding and 81 percent for prophylaxis. Thirty-six percent of transfusions were given in violation of our transfusion guidelines. Forty-nine percent of the high users died, but no deaths were due to hemorrhage. All survivors developed chronic lung disease, and all survivors weighing less than 1250 g at birth developed retinopathy of prematurity. CONCLUSIONS: Almost all patients that received 20 or more PLT transfusions had an acquired, consumptive thrombocytopenia. All could have received fewer transfusions had the guidelines already in place been observed. Eighty-one percent fewer PLT transfusions would have been administered had the paradigm been transfusing only if oozing, bruising, or bleeding was present.

Center differences in NEC within one health-care system may depend on feeding protocol.

We tabulated the incidence of necrotizing enterocolitis (NEC) during a recent 4-year period among three neonatal intensive care units (NICUs) within a single health-care system. We then sought associations to explain differences in NEC incidence between the centers. Between January 1, 2002, and December 31, 2005, 6787 neonates were admitted to the three NICUs. The incidence of NEC (Bell's stage II or higher) among these patients was correlated with birthweight, gestational age, maternal and neonatal demographics, and various events and practices. These events and practices included feeding practices, the management of patent ductus arteriosus, rates of systemic bacterial and fungal infection, transfers to the regional children's hospital for surgical treatment, and mortality rate. Bell's stage II or higher NEC was documented in 131 of 6787 NICU patients. The incidence was 7.4% among those with birthweights <750 g (16 of 217), 6.9% among those of birthweights 750 to 1250 g (36 of 519), and 1.3% (79 of 6051) among those with birthweights >1250 g. Center A had an incidence of NEC significantly higher than the other two, accounting for 72% of the total cases (94 of 131). Among patients <1250 g, Center A had a rate of NEC of 14.5%; Centers B (2.3%) and C (2.3%) had lower rates ( P<0.0001). After controlling for gestational age, birthweight, small for gestational age status, and Apgar scores, the overall odds ratio of developing NEC in Center A, compared with the other two, was 21.6 (95% confidence interval, 14.7 to 31.6). This difference could not be accounted for by differences in maternal or neonatal demographic characteristics, bed occupancy rates, or a higher incidence of culture-proven nosocomial bacterial or fungal infections. Although the incidence of NEC was significantly higher at Center A, the percentage of patients with NEC transferred to the children's hospital for surgical evaluation and treatment was similar. The mortality rate of patients who developed NEC was similar among the three hospitals. Centers B and C utilize standardized feeding guidelines. During each of the 4-year study periods, one of three NICUs within the same health-care system had a higher incidence of NEC than the other two. Once NEC developed, the outcome was similar in all three NICUs. The higher incidence in Center A could not be explained by differences in demographics, socioeconomics, or systemic nosocomial infections. Similarities in feeding practices between Centers B and C suggest to us that these may be responsible, at least in part, for the differences in the incidence of NEC. Changing the feeding practices at Center A to those at Centers B and C is planned to test this theory.

The effect of instituting a prehospital-discharge newborn bilirubin screening program in an 18-hospital health system.

OBJECTIVE: Kernicterus is a rare but devastating condition. The prevention of bilirubin-induced brain injury is based on the detection of infants at risk for developing severe hyperbilirubinemia. In an 18-hospital health system, Intermountain Health Care (IHC), we initiated a program of predischarge bilirubin screening of all neonates and coupled this with a results assessment using a percentile-based nomogram. Data during 2 periods of time, before versus after initiating the program, were compared to assess the effect of the program on significant hyperbilirubinemia and rehospitalization. METHODS: We conducted a historic cohort study involving all neonates delivered at > or =35 weeks' gestation, within IHC's 18-hospital system, during 2 periods of time: March 1, 2001, to December 31, 2002, versus January 1, 2003, to December 31, 2004. A bilirubin screening program, instituted in December 2002, called for a total serum bilirubin (TSB) or transcutaneous bilirubin measurement to be performed on every neonate either at the recognition of clinical jaundice or before discharge regardless of whether jaundice was observed. For nonjaundiced neonates, the nursery staff was encouraged to obtain the screening TSB at the same time they obtained the state-mandated newborn screen for inborn errors of metabolism. Bilirubin values were plotted on an hour-specific nomogram and the corresponding percentile was used to guide evaluation, therapy, and follow-up. This study compared TSB data and readmission data for a 2-year period before versus a 2-year period after implementing the program. RESULTS: The study involved 101272 neonates: 48789 in period 1 and 52483 in period 2. Before the program, 1 in every 77 neonates born at an IHC hospital had 1 or more serum bilirubin levels >20 mg/dL. After initiating the program, the incidence fell to 1 in 142 and the number of neonates with a level >25 mg/dL fell from 1 in 1522 before to 1 in 4037 after. The rate of hospital readmission with a primary diagnosis of jaundice fell from 0.55% in period 1 to 0.43% in period 2. CONCLUSIONS: Initiating a program of bilirubin screening in a multihospital health system, coupled with evaluating the results using a percentile-based nomogram, reduced the proportion of neonates with significant hyperbilirubinemia and reduced the rate of hospital readmissions with jaundice.

Single-dose darbepoetin administration to anemic preterm neonates.

OBJECTIVE: Darbepoetin is longer acting and more potent than recombinant erythropoietin (rEpo). In certain situations, preterm neonates might benefit from rEpo, and for such patients darbepoetin would require fewer doses at a lower cost. However, the proper dose and dosing interval have not been established. STUDY DESIGN: We performed a prospective trial in two level III Neonatal Intensive Care Units. Patients <32 weeks gestation at birth, with a birth weight (BW) <1500 g, were eligible for participation if they were >21-days-old and had a hemoglobin (Hgb) concentration