Red Blood Cell Transfusions are Independently Associated with Intra- Hospital Mortality in Very Low Birth Weight Preterm Infants Amelia Miyashiro Nunes dos Santos, MD, PhD, Ruth Guinsburg, MD, PhD, Maria Fernanda Branco de Almeida, MD, PhD, Renato S. Procianoy, MD, PhD, Clea Rodrigues Leone, MD, PhD, Sergio Tadeu Martins Marba, MD, PhD, Ligia Maria Suppo de Souza Rugolo, MD, PhD, Humberto Holmer Fiori, MD, PhD, Jose Maria de Andrade Lopes, MD, PhD, and Francisco Eulogio Martinez, MD, PhD, for the Brazilian Network on Neonatal Research* Objective To test the hypothesis that red blood cell (RBC) transfusions in preterm infants are associated with in- creased intra-hospital mortality. Study design Variables associated with death were studied with Cox regression analysis in a prospective cohort of preterm infants with birth weight 1500 g in the Brazilian Network on Neonatal Research. Intra-hospital death and death after 28 days of life were analyzed as dependent variables. Independent variables were infant demographic and clinical characteristics and RBC transfusions. Results Of 1077 infants, 574 (53.3%) received at least one RBC transfusion during the hospital stay. The mean number of transfusions per infant was 3.3 3.4, with 2.1 2.1 in the first 28 days of life. Intra-hospital death oc- curred in 299 neonates (27.8%), and 60 infants (5.6%) died after 28 days of life. After adjusting for confounders, the relative risk of death during hospital stay was 1.49 in infants who received at least one RBC transfusion in the first 28 days of life, compared with infants who did not receive a transfusion. The risk of death after 28 days of life was 1.89 times higher in infants who received more than two RBC transfusions during their hospital stay, com- pared with infants who received one or two transfusions. Conclusion Transfusion was associated with increased death, and transfusion guidelines should consider risks and benefits of transfusion. (J Pediatr 2011 159:371-6). See editorial, p 359 DConcerns espite recent efforts to decrease allogeneic red blood cell (RBC) transfusion thresholds, they remain an important sup- portive and life-saving intervention for neonatal intensive care patients.1-6 about the harm of RBC transfusions are traditionally focused on infections, but these risks have been pro- gressively reduced.7 Currently, lung injury, organ dysfunction, or both, hemolytic transfusion reactions, and transfusion related-sepsis are considered the most frequent causes of morbidity and mortality in adults.8 In the neonatal setting, these ef- fects are rarely suspected, and their effect on infant clinical condition and mortality rate is not fully understood.9,10 There is a growing body of evidence that RBC transfusions are independently associated with short- and long-term mor- tality, on the basis of observational studies in adults11,12 and several plausible biological models of how transfusions may be harmful.13 In adult patients who underwent cardiac surgery, RBC transfusions were strongly associated with a wide range of postoperative morbidity and with increased mortality rates as long as 5 years after surgery.14 Kamper-Jorgensen et al re- ported higher short- and long-term mortality rates in 1 118 261 transfusion recipients than in the general population. The mortality ratio was 17.6 times higher during the first 3 months after the first transfusion and remained higher for trans- fusions recipients 1 to 4 years after the first transfusion (2.1 times) and even 17 years after (1.3 times).11 Kneyber et al reported a higher mortality rate in chil- dren who received a transfusion compared with children and adolescents who did not receive a transfusion.15 However, in the neonatal setting, especially in very low birthweight preterm infants who frequently receive multiple RBC transfusions, the association between RBC transfusions and mortality rate is unknown. Therefore, the objective of this study was to test the hypothesis that RBC transfusions in very low birth weight preterm infants are indepen- dently associated with an increase in intra-hospital mortality rates. From the Department of Pediatrics, Federal University of S~ ao Paulo, S~ ao Paulo, Brazil (A.S., R.G., M.A.) Department of Pediatrics, Federal University of Rio Grande do Sul, Porto Alegre, Brazil (R.P.) Department of Pediatrics, University of S~ ao Paulo, S~ ao Paulo, Brazil (C.L.) Department of Pediatrics, State University of Campinas, Campinas, Brazil (S.M.) Department of Pediatrics, S~ ao Paulo State University, Botucatu, Brazil (L.R.) Department of Pediatrics, Pontifical Catholic University of Rio Grande do Sul, Porto Alegre, Brazil (H.F.) Neonatal Department, Fernandes Figueira Institute of Oswaldo Cruz Foundation, Rio de Janeiro, Brazil (J.L.) and Department of Pediatrics, University of S~ ao Paulo, Ribeir~ ao Preto, Brazil (F.M.) *List of members of the Brazilian Network on Neonatal Research is available at www.jpeds.com (Appendix). The authors declare no conflicts of interest. 0022-3476/$ - see front matter. Copyright �� 2011 Mosby Inc. All rights reserved. 10.1016/j.jpeds.2011.02.040 IVH Intraventricular hemorrhage RBC Red blood cell SNAPPE II Score for Neonatal Acute Physiology���Perinatal Extension 371

Methods This multicenter study enrolled a cohort of preterm infants with birth weight 1500 g born in 8 centers of the Brazilian Network on Neonatal Research. It was a retrospective study, buttheclinicaldata were prospectivelyenteredinthedatabase. The network 8 units were referral centers for high-risk preg- nancies, treating patients of the Brazilian National Health Sys- tem almost exclusively. The 8 centers had 282 intensive and intermediate care beds, varying from 20 to 48 beds per unit, with similar equipment and human resources in the units. Neonates included in the study underwent transfusion according to 8 independent guidelines for allogeneic RBC transfusions defined by each neonatal intensive care unit (Table I available at www.jpeds.com).16 All RBC units were collected from volunteers in citrate-phosphate- dextrose-adenine-1 anticoagulant which were gamma irradiated and depleted of leukocytes before storage. To avoid hyperkalemia, irradiated RBC units stored for as long as 28 days were transfused in small aliquots of 15 mL/kg, slowly in 4 hours.17 Preterm infants born with gestational ages of 23.0 to 36.9 weeks of and birth weights of 400 to 1495 g in the 8 centers from January 2006 to December 2007 were included. Infants with major congenital anomalies were excluded. These data were recorded: number of births in each center, neonatal demographic and clinical characteristics, intra- hospital deaths (death at any time during hospital stay and death after 28 days of life), percentage of neonates receiving at least one RBC transfusion during the first 28 days and dur- ing hospital stay, and number of RBC transfusions performed from 1 to 14 days, 15 to 28 days, and after 28 days of life until discharge from the hospital. For clinical data, these neonatal characteristics were ana- lyzed: gestational age (weeks) birthweight (g) Apgar score in the first and fifth minutes of life (categorized as 7 and 7-10) SNAPPE II18 (Score for Neonatal Acute Physiology��� Perinatal Extension, scored during the first 12 hours of life and stratified in 45 and $45) small for gestational age (birthweight 10th percentile for gestational age according to Alexander et al19) presence of respiratory distress syn- drome any grade of intraventricular hemorrhage (IVH) identified with intracranial ultrasound in the first week of life and early- and late-onset sepsis (clinical sepsis diag- nosed in the first 48 hours and after 48 hours of life, re- spectively). Comparisons in groups were done with the two-tailed c2 test for categorical variables and t tests for continuous vari- ables. To analyze the hazard of death, univariate and multi- variate Cox regression analyses were applied. Differences were considered significant when P values were .05. SPSS software for Windows version 17.0 (SPSS Inc, Chicago, Illi- nois) was used for all statistical procedures. This study was approved by the ethical committee of each center and by the ethical committee of the Universidade Fed- eral de S~ ao Paulo, Brazil, the lead center for this study. Results During the studied period, 1226 infants were born at 23.0 to 36.9 weeks of gestation with birth weight of 400 to 1495 g. Of these infants, 149 (12.2%) with congenital anomalies were excluded. The main demographic characteristics of the 1077 remain- ing patients were: gestational age of 29.0 2.8 weeks (30.2% 28 weeks), birth weight of 1046 288 g (42.3% 1000 g), 40.7% (438/1076) were small for gestational age, 17.5% (188/1075) had a 5-minute Apgar score 7, and 16.4% (172/1046) had a SNAPPE II score $45. For morbidity, re- spiratory distress syndrome was present in 56.0% of patients (593/1058), any grade of IVH was present in 35.2% of pa- tients (324/920), early-onset clinical sepsis was present in 41.9% of patients (448/1068), clinical late-onset sepsis was present in 41.9% of patients (447/1068), and necrotizing en- terocolitis was present in 7.5% of patients (79/1058). The fre- quency of infants with gestational age 28 weeks (P = .003), 1-minute Apgar score 7 (P .001), 5-minute Apgar score 7 (P = .004), SNAPPE II score $45 (P .001), respiratory distress syndrome (P .001), IVH (P .001), early-onset sepsis (P .001), late-onset sepsis (P .001), and necrotizing enterocolitis (P = .007) were different in the centers. Intra-hospital death occurred in 299 infants (27.8%), 29.1%, 25.4%, 36.5%, 28.4%, 22.5%, 25.0%, 27.9% and 25.5% in centers 1, 2, 3, 4, 5, 6, 7 and 8, respectively (c2 test for differences between centers, P = .363). Of the patients studied, 201 (18.7%) died before 14 days of life, 38 (3.5%) died in 15 to 28 days, and 60 (5.6%) died after 28 days. Of the 1077 infants, 574 (53.3%) received at least one RBC transfusion during the hospital stay, 501 (46.5%) received at least one transfusion in the first 28 postnatal days, and 365 (33.9%) received more than one transfusion during their hospital stay. The frequency of infants who received at least one RBC transfusion during hospital stay varied in the centers (c2 test, P .001): center 1, 46.6% (48/103) center 2, 53.1% (69/130) center 3, 48.9% (67/137) center 4, 59.3% (96/162) center 5, 46.5% (33/71) center 6, 47.4% (74/156) center 7, 72.7% (120/165) and center 8, 43.8% (67/153). Of the 1908 transfusions performed, 738 (38.7%) were in the first 14 days of life, 495 (25.9%) were in 15 to 28 days of life, and 675 (35.4%) were after 28 days of life. The mean number of transfusions per infant during hospital stay was 3.3 3.4, with a median of 2 (quartile range, 1 to 4). The mean number of transfusions per infant was 1.3 1.5 in the first 14 days of life 0.9 1.1 in 15 to 28 days and 1.2 2.1 after 28 days until discharge from the hospital. The number of RBC transfusions was higher in infants with these characteristics: 1-minute Apgar score 7 (2.1 3.3 versus 1.3 2.5, P .001), 5-minute Apgar score 7 (2.4 3.8 versus 1.7 2.8, P = .003), SNAPPE II $45 (2.8 4.2 versus 1.6 2.7, P .001), respiratory distress syn- drome (2.1 2.9 versus 1.5 3.1, P = .001), IVH (2.9 3.9 versus 1.4 2.5, P .001), early-onset sepsis (2.5 3.6 THE JOURNAL OF PEDIATRICS www.jpeds.com Vol. 159, No. 3 372 dos Santos et al