The Effect of Anemia on Retinopathy of Prematurity in Extremely Low Birth Weight Infants 1

The Effect of Anemia on Retinopathy of Prematurity in Extremely Low Birth Weight Infants 1 Judith A. Englert, MD Richard A. Saunders, MD Dilip Purohit, MD Thomas C. Hulsey, MSPH, SCD Myla Ebeling OBJECTIVE: Numerous risk factors for development of retinopathy of prematurity (ROP)in very low birth weight infants have been identified in the literature. However, the role of anemia in the development of ROP has not been adequately addressed. STUDY DESIGN: We retrospectively examined the medical records of all infants weighing 800 g who were admitted to a university hospital between July 1, 1992 and December 30, 1997. Highest and lowest hemoglobin and hematocrit values and the number of blood transfusions were recorded at each week of life during hospitalization. Gestational age at birth, birth weight, race, sex, oxygen status, history of bronchopulmonary dysplasia, length of hospital stay, and sepsis were also identified as potential risk factors. Data were analyzed using logistic regression to adjust for these confounding variables. RESULTS: Infants were grouped according to ROP status in the following manner: stage 0 to 1 ROP, stage 2 ROP, and stage 3 to threshold ROP. Sex, gestational age at birth, bronchopulmonary dysplasia, ventilator days, length of hospital stay, and number of blood transfusions were significantly associated with severity of ROP by univariate analysis. Using a logistic regression model, only gestational age ( p =0.007)and number of blood transfusions ( p=0.04) remained statistically significant. CONCLUSIONS: Anemia did not affect severity of ROP as an independent risk factor. However, the number of blood transfusions did affect the highest stage of ROP in this group of premature infants. Infants who remained severely anemic (Hgb8 g/dl or Hct25%)for longer periods of time developed milder ROP than less anemic infants. Journal of Perinatology 2001; 21:21± 26. INTRODUCTION Retinopathy of prematurity (ROP)has become increasingly prevalent worldwide as more very low birth weight infants are surviving the neonatal period and remains a major cause of visual morbidity among these children. 1±3 In 1979, Phelps 4 estimated that 163 infants per million live births were blinded because of ROP. More recently, in 1990, Brown et al. 5 estimated that number to have nearly doubled to 301 per million live births, presumably due to the increased survival rate of infants weighing <1000 g at birth. In the United States, ROP is the second leading cause of childhood blindness, following only cortical visual impairment. 6 Numerous retrospective studies have examined potential risk factors contributing to development of ROP. These have included low gestational age at birth, 7±12 low birth weight, 5,10,12 ± 14 male sex, 9 white race, 12,15 bronchopulmonary dysplasia, 5,10,13,16 sepsis, 8,13 acidosis, 7,11,16 oxygen therapy, 14,17 hyperoxia, 2,11,18 hypocarbia, 19 hypoxia, 1,11,19 and maternal use of beta blockers immediately before birth. 14 Recently, a case of neonatal HIV infection after blood transfusion at another institution led to a decrease in blood transfusions in premature infants at that hospital. Many lower risk infants went on to develop severe ROP, raising the concern that this change in transfusion practice and resultant anemia might be etiologically important ( J. O'Neil, personal communication, April, 1998). A subsequent review of the literature revealed that very few studies addressed the effect of anemia on incidence or severity of ROP. Bossi et al. 7 found no association between hemoglobin levels and development of ROP, but did identify a greater frequency of blood transfusions in infants who developed ROP using univariate analysis. A recent report by Brooks et al. 20 showed no difference in ROP incidence or severity between infants receiving blood transfusions to maintain high hematocrit levels (40%)and infants receiving blood transfusions only when clinically symptomatic. Other reports have not addressed anemia specifically. However, they did suggest that frequency of blood transfusions may be a risk factor in the development of ROP in infants weighing less than 1500 g, although this correlation was not substantiated using logistic regression. 8,19,21 We present our findings in evaluating the role of anemia and blood transfusions in the development of ROP in extremely low birth weight infants. N. Edgar Miles Center for Pediatric Ophthalmology ( J.A.E., R.A.S. ), Storm Eye Institute, Medical University of South Carolina, Charleston, SC; Department of Pediatrics ( D.P., T.C.H., M.E. ), Medical University of South Carolina, Charleston, SC. 1 This study was supported in part by an unrestricted grant to the Storm Eye Institute from Research to Prevent Blindness, Inc., New York, NY. Address correspondence and reprint requests to Richard A. Saunders, MD, Storm Eye Institute, 171 Ashley Avenue, Charleston, SC 29425. Journal of Perinatology 2001; 21:21 ± 26 # 2001 Nature Publishing Group All rights reserved. 0743-8346/01 $17 www.nature.com/jp PATIENTS AND METHODS We retrospectively reviewed the inpatient medical records of infants weighing 800 g admitted to the Neonatal Intensive Care Unit of a university hospital between July 1, 1992, and December 30, 1997, to determine whether the frequency of anemia or blood 21

Englert et al. Anemia and ROP Table 1 Demographics of Infant Groups Based on Severity of ROP Stage 1 Stage 2 Stage 3 p Value No. (%)(n=23)no. (%)(n=41)no. (%)(n=23) Sex 0.03 male (n=39)6 (15) 18 (46) 15 (38) female (n=48)17 (35) 23 (48) 8 (17) Race 0.14 black (n=60)13 (22) 32 (53) 15 (25) white (n=27)10 (37) 9 (33) 8 (30) EGA (weeks) 0.14 mean SD 27.0 2.0 25.1 1.2 24.8 1.3 median (range)27 (24±31) 25 (23±28) 24 (23±28) Birth weight (g) 0.12 mean SD 703 74 707 70 674 72 median (range)715 (508±792)723 (510±795)686 (514±800) EGA, estimated gestational age. transfusion influenced the severity of ROP. This study was considered exempt from informed consent by the Institutional Review Board for Human Research, and hence informed consent was not obtained. A total of 292 infants meeting our criteria were identified. One hundred seventy infants died before hospital discharge and were excluded from data analysis due to the inability to determine final ROP outcome. Of the remaining 122 surviving infants, inadequate documentation of anemia status or ROP status resulted in exclusion of an additional 35 infants, leaving the medical records of 87 infants suitable for evaluation. Patient demographics are shown in Table 1. Approximately two-thirds of the infants were identified as black and the remainder white. There were more females than males in the group (48 vs. 39, p=0.14). Birth weights of the infants ranged from 508 to 800 g, and estimated gestational age (EGA)at birth ranged from 23 to 31 weeks. Highest and lowest hemoglobin and hematocrit levels as well as total number of blood transfusions received were documented for each week of life during hospitalization. The frequency of anemia was evaluated, with anemia defined as hemoglobin 10 g/dl or hematocrit 30%. The frequency of severe anemia, defined as hemoglobin 8 g/dl and hematocrit 25%, was also evaluated. Due to the likelihood that factors other than the patient's hemoglobin or hematocrit levels, such as pulmonary disease and oxygen therapy, were utilized in the decision to perform blood transfusions, other potential confounding variables were also analyzed. These factors were bronchopulmonary dysplasia, sepsis, ventilator days, number of days of oxygen, and length of hospital stay. When required, a standard volume of replacement blood transfusion consisting of 15 ml/kg of packed red blood cells was given in all cases. Exchange transfusions, which are used to treat hyperbilirubinemia, were excluded from data analysis because elevated serum bilirubin levels, rather than anemia status, was the basis for the decision to initiate transfusion. In addition, the outcome of each ROP examination was recorded. The most severe stage of ROP that developed in the right eye of each infant was used for data analysis. However, in all but one infant, the same level of ROP severity was present in each eye. Data were evaluated using analysis of variance and logistic regression. Infants were grouped according to the highest stage of ROP reached. Due to the small number of patients reaching each stage of ROP, subgroups were combined into groups based on clinical severity and risk for adverse visual outcome: stage 1, stage 2, and stage 3 (including threshold). 22 Variables were cross-tabulated with each ROP group, and differences between the three groups were tested using chi-square for categorical variables and analysis of variance of means for continuous variables. Logistic regression was Figure 1. Eighty- seven premature infants with birth weight 508 to 800 g were grouped according to severity of ROP: Stage 1 (n=23), stage 2 (n=41), and stage 3 (including threshold, n=23). Infants with more severe ROP were more likely to have received multiple blood transfusions ( p= 0.04). Error bars represent standard deviation. 22 Journal of Perinatology 2001; 21:21 ± 26

Anemia and ROP Englert et al. Table 2 Infant Characteristics That Potentially Affected the Decision to Transfuse Blood and Their Relationship to Severity of ROP Stage 1 Stage 2 Stage 3 p Value No. (%)(n=23)no. (%)(n=41)no. (%)(n=23) Bronchopulmonary dysplasia 0.001 present (n=73)14 (19)37 (51)22 (30) absent (n=14)9(64)4(29)1(7) Sepsis 0.46 present (n=55)14 (25)24 (44)17 (31) absent (n=32)9 (28)17 (53)6 (19) Ventilator assistance ( days) 0.004 mean SD 19.7 17.2 40.9 47.9 37.5 19.5 median (range)18 (0±65)32.5 (0±286)45 (0±62) Supplemental oxygen ( days) 0.12 mean SD 37.7 30.7 58.4 49.3 58.1 33.0 median (range)39 (0±126)53.5 (0±287)50 (0±119) Length of hospital stay (days) 0.01 mean SD 77.7 20.7 104.6 83.4 96.2 24.9 median (range)80 (26±121)89 (11±533)100 (25±134) Total number of transfusions 0.0007 mean SD 5.2 4.0 8.9 4.5 9.8 4.2 median (range)4 (1±13)8.5 (0±20)10 (1±19) Number weeks Hct 30% or Hgb 10 g/dl 0.31 mean SD 6.2 2.1 6.3 2.6 7.0 2.6 median (range)6 (2±11)6 (1±12)7 (1±11) Hct, hematocrit; Hgb, hemoglobin. used to examine the association between total number of transfusions, anemia, and severity of ROP controlling for potential confounding variables. To increase the power of the logistic regression, infants with stage 2 ROP and infants with stage 3 to threshold ROP were combined into one group as a dichotomous variable. These two groups were combined because analysis of variance statistics revealed no differences in the characteristics of these groups. Infants with stage 0 to 1 ROP remained the second dichotomous variable for analysis. RESULTS The characteristics of each group based on ROP severity are shown in Table 1. Male infants were more likely to progress to more severe ROP ( p=0.03). Although other reports have indicated that white infants tend to develop worse disease, 12,15 we did not identify race as a significant predictor of ROP severity ( p=0.14). As expected, EGA at birth was strongly correlated with severity of ROP ( p=0.0001), with the more premature infants developing more severe disease. However, birth weight was not correlated with ROP severity, perhaps because the range of birth weights in our study group was narrow (508 to 800 g). Next, other possible risk factors for development of ROP were evaluated. Infants who received a greater number of blood transfusions developed more severe ROP ( p=0.0007)(figure 1). Only one patient in this study did not receive transfused blood, a white female infant with birth weight of 766 g and EGA of 27 weeks who developed stage 2 ROP. There were no identifiable trends regarding infant age at the time of blood transfusion and severity of ROP. However, the decision to transfuse was not determined solely by hemoglobin/hematocrit levels but also by respiratory rate and effort, ventilator/oxygen status, sepsis, and overall health of the infant. The association of these potential confounding variables with the development of ROP is shown in Table 2. Bronchopulmonary Table 3 Potential Risk Factors for Severity of ROP Using Logistic Regression p Value Sex 0.10 Race 0.72 Estimated gestational age at birth 0.007 Birth weight 0.89 Bronchopulmonary dysplasia 0.32 Days on ventilator 0.60 Days receiving supplemental oxygen 0.59 Length of hospital stay 0.53 Anemia (number of weeks Hct 30% or Hgb 10 g/dl)0.17 Total number of blood transfusions 0.04 Hct, hematocrit; Hgb, hemoglobin. Journal of Perinatology 2001; 21:21 ± 26 23

Englert et al. Anemia and ROP Table 4 Relationship of Severe Anemia to Highest Stage of ROP Observed Stage 1 Stage 2 Stage 3 p Value ( univariate analysis) p Value ( logistic regression) Number of weeks of severe anemia 0.09 0.03 mean SD 2.7 1.7 1.8 1.5 2.3 1.6 median (range)3 (0±6)1 (0±6)2 (0±6) Mean number of transfusions 5.2 4.0 8.9 4.5 9.8 4.2 0.0007 0.12 Severe anemia is defined as hematocrit 25% or hemoglobin 8 g/dl. dysplasia was significantly associated with ROP severity ( p=0.001). More severe ROP was also associated with a greater number of ventilator days ( p=0.004)and longer hospital stays ( p=0.01). A possible correlation between anemia and severity of ROP was also evaluated. All babies developed anemia at some time during their hospitalization, with anemia being defined as hemoglobin 10 g/dl or hematocrit 30% in this study. The mean lowest hemoglobin and hematocrit level for each week of life was similar for all groups, with lowest levels occurring between 5 and 6.5 weeks of age. There was no correlation between the frequency of anemia and ROP severity (Table 2). Maximum hemoglobin and hematocrit levels for each week of life were also similar for all groups and did not appear to affect the development of ROP. Logistic regression showed only two characteristics that remained statistically significant when controlling for potential confounding variables: EGA at birth ( p=0.007)and total number of transfusions ( p=0.04)in these extremely low birth weight infants (Table 3). The relationship between the duration of severe anemia (hemoglobin 8 g/dl or hematocrit 25%)and the development of ROP was then analyzed. Infants who were severely anemic for longer periods of time and received fewer transfusions developed milder ROP (stage 0 to 1)than infants who were severely anemic for shorter periods of time and received more transfusions (ROP stage 2 or greater)(table 4). Logistic regression analysis showed that severe anemia was significantly associated with ROP severity ( p=0.03). However, the number of transfusions no longer remained a significant risk factor ( p=0.12), indicating that these two variables were highly codependent. Our sample size was too small to determine whether delaying blood transfusion improved ROP outcome. Only 11 infants with a minimum hematocrit between 25% and 30% did not undergo transfusion, (one infant with stage 0 to 1 ROP, four infants with stage 2 ROP, and six infants with stage 3 to threshold ROP). All other infants received transfusion when they reached this level of anemia. DISCUSSION The relationship of anemia to ROP is difficult to study due to many factors affecting blood hemoglobin and hematocrit values, especially blood transfusions. Once an infant has received a blood transfusion, the anemia is temporarily resolved and hemoglobin/hematocrit values increase. This codependency was overcome in our study by separately analyzing hemoglobin/hematocrit values and number of blood transfusions each week of life during hospitalization. In addition, the decision to transfuse is affected by other variables, such as the presence of lung disease, oxygen status, and overall health of the infant. Data were analyzed using logistic regression to control for these confounding variables. We found only three other studies addressing anemia as a possible risk factor in ROP development. Bossi et al. 7 evaluated the association between anemia and ROP in infants weighing less than 1500 g. The medical records of 53 infants with ROP and 53 infants without ROP were analyzed, matching for EGA, birth weight, birth date, and hospitalization at the same neonatal unit. The mean blood hemoglobin values for the first 35 days of hospitalization were identical between the two groups. However, the frequency of blood transfusions was significantly different. Infants with ROP received 3.8 4.4 transfusions whereas babies without ROP received 2.5 3.7 transfusions ( p<0.01). A decade later, Alter et al. 23 found no difference between infants with stage 2ROP andinfants with stage 3 ROP when evaluating hemoglobin levels during the first week of life ( p=0.059)or number of blood transfusions ( p=0.072). A recent prospective study by Brooks et al. 20 found no association between anemia or blood transfusions and ROP incidence or severity in infants evaluated during a 6-week period (day of life 29 to 71). Of the 50 infants prospectively enrolled, only 34 infants completed the study. Sixteen infants in group 1 received blood transfusions only if certain medical criteria were met, maintaining average hematocrit levels at 33.9%. Eighteen infants in group 2 received blood transfusions to maintain hematocrit levels 40% (average 41.8% over 6 weeks). Despite large differences in anemia ( p=0.0001)and blood transfusions between the two groups (average 2.8 vs. 5.7 transfusions, p=0.0007), there was no difference in ROP incidence (83%vs. 73%,p=0.38)or ROP severity ( p=0.32). However, small sample size makes it difficult to draw clinically useful conclusions regarding ROP risk factors. 24 Other reports addressing an association between blood transfusions and ROP development without analyzing anemia had variable findings. Mittelman and Cronin 25 found that infants weighing less than 1360 g at birth who developed ROP received more blood transfusions but also more oxygen therapy. The ratio of transfusions to days of supplemental oxygen administration was actually greater in those infants without ROP than in those infants who developed 24 Journal of Perinatology 2001; 21:21 ± 26

Anemia and ROP Englert et al. ROP, which is opposite our results. Sacks et al. 21 found that infants weighing less than 1251 g with ROP received more replacement blood transfusions than infants without ROP by univariate analysis, but this association was not significant when controlling for oxygen therapy. Similar to our findings, Cooke et al. 26 found that only gestational age and frequency of blood transfusions were independently associated with development of ROP. They reported a significant difference in median frequency of blood transfusions between 92 infants weighing less than 1500 g without ROP and 92 infants with ROP (1 vs. 7 transfusions). They also found differences in median frequency of blood transfusions between infants with stage 1 to 3 ROP and infants with threshold disease (6 transfusions vs. 16 transfusions). Although the association of blood transfusions and ROP development has been shown in some studies, this does not prove a cause-and-effect relationship. However, there is speculation that transfusions containing adult hemoglobin with its lower affinity for oxygen than fetal hemoglobin F cause greater oxygen delivery to the retina, increasing the risk for developing ROP. 2 In addition, packed red blood cell transfusions contain large amounts of iron, which can rapidly increase serum iron levels in premature infants. Free iron can produce free radicals, which have been enzymatically linked to ROP development. 27,28 To test the hypothesis that increased iron load promotes the development of ROP, Inder et al. 29 performed a prospective study evaluating serum iron concentrations in infants with ROP and those without ROP who weighed less than 1250 g at birth. By 7 days of age, infants with ROP had significantly higher mean serum iron levels than infants without ROP. By 28 days of age, these values were no longer significantly different. However, total blood transfusion volume remained significantly different between the two groups at 7 and 28 days of age, suggesting no relationship between iron concentration and transfusion frequency. Hesse et al. 30 found that, although blood transfusions were an independent risk factor for ROP, the iron load did not appear to be associated with ROP development during the first 6 months of life in infants weighing less than 1500 g. CONCLUSION Our study is the first to assess duration of anemia as a risk factor for ROP by evaluating serial hemoglobin and hematocrit levels in extremely low birth weight infants. Anemia (Hgb10 g/dl or Hct30%)was not significantly associated with severity of ROP in this group of infants. Using univariate analysis, our data confirm known risk factors of gestational age, blood transfusions, ventilator days, male sex, and presence of bronchopulmonary dysplasia. Only EGA at birth and frequency of blood transfusions remained risk factors for ROP severity using logistic regression. Those children who remained severely anemic (Hgb8 g/dlorhct25%)for longer periods of time and received fewer transfusions were less likely to develop severe ROP. Our findings suggest that delaying blood transfusions in otherwise asymptomatic premature infants does not increase their risk of developing clinically significant ROP. References 1. Ben- Sira I, Nissenkorm I, Kremer I. Retinopathy of prematurity. Surv Ophthalmol 1988;33:1 ± 16. 2. Bossi E, Koerner F. Retinopathy of prematurity. Intensive Care Med 1995;21:241 ± 246. 3. Gilbert C, Rahi J, Eckstein M, et al. Retinopathy of prematurity in middleincome countries. Lancet 1997;350:12 ± 14. 4. Phelps DL. Retinopathy of prematurity: an estimate of vision loss in the United States Ð 1979. Pediatrics 1981;67:924 ±926. 5. Brown DR, Biglan AW, Stretavsky MMA. Retinopathy of prematurity: the relationship with intraventricular hemorrhage and bronchopulmonary dysplasia. J Pediatr Ophthalmol Strabismus 1990;27:268 ± 271. 6. Steinkuller PG, Du L, Gilbert C, Foster A, Collins ML, Coats DK. Childhood blindness. J AAPOS 1999;3:26 ±32. 7. Bossi E, Koerner F, Zulauf M. Retinopathy of prematurity: a risk factor analysis with univariate and multivariate statistics. Helv Paediatr Acta 1984;39:307 ± 317. 8. Gunn TR, Easdown J, Outerbridge EW, Aranda JV. Risk factors in retrolental firbroplasia. Pediatrics 1980;65:1096 ± 1100. 9. Nodgaard H, Andreasen H, Hansen H, Sorensen HT. Risk factors associated with retinopathy of prematurity ( ROP)in Northern Jutland, Denmark 1990± 1993. Acta Ophthalmol Scand 1996;74:306 ± 310. 10. HolmstroÈm G, Broberger U, Thomassen P. Neonatal risk factors for retinopathy of prematurity Ð a population- based study. Acta Ophthalmol Scand 1998;76:204 ±207. 11. Prendiville A, Schulenburg WE. Clinical factors associated with retinopathy of prematurity. Arch Dis Child 1988;63:522 ±527. 12. Schaffer DB, Palmer EA, Plotsky DF, et al. Prognostic factors in the natural course of retinopathy of prematurity. Ophthalmology 1993;100:230 ± 237. 13. Purohit DM, Ellison RC, Zierler S, Miettinen OS, Nadas AS. Risk factors for retrolental fibroplasia with 3,025 premature infants. Pediatrics 1985;76:339± 344. 14. Gallo JE, Jacobson L, Broberger U. Perinatal factors associated with retinopathy of prematurity. Acta Paediatr 1993;82:829 ± 834. 15. Saunders RA, Donahue ML, Christmann LM, et al. Racial variation in retinopathy of prematurity. Arch Ophthalmol 1997;115:604 ± 608. 16. Biglan AW, Brown DR, Reynolds JD, Milley JR. Risk factors associated with retrolental fibroplasia. Ophthalmology 1984;91:1504 ± 1511. 17. Chirico G, the Italian ROP Study Group. Italian multicentre study on retinopathy of prematurity. Eur J Pediatr 1997;156:939 ± 943. 18. Shahinian L, Malachowski N. Retrolental fibroplasia Ð a new analysis of risk factors based on recent cases. Arch Ophthalmol 1978;96:70 ± 74. 19. Shohat M, Reisner SH, Krikler R, Nissenkorn I, Yassur Y, Ben- Sira I. Retinopathy of prematurity: incidence and risk factors. Pediatrics 1983;72: 159± 163. 20. Brooks SE, Marcus DM, Gillis D, Pirie E, Johnson MH, Bhatia J. The effect of blood transfusion protocol on retinopathy of prematurity: a prospective, randomized study. Pediatrics 1999;104:514 ± 518. 21. Sacks LM, Schaffer DB, Anday EK, Peckham GJ, Delivoria- Papadopoulos M. Retrolental fibroplasia and blood transfusion in very low- birth- weight infants. Pediatrics 1981;68:770 ±74. Journal of Perinatology 2001; 21:21 ± 26 25

Englert et al. Anemia and ROP 22. Cryotherapy for retinopathy of prematurity cooperative group: Multicenter trial of cryotherapy for retinopathy of prematurity: preliminary results. Arch Ophthalmol 1988;106:471 ± 79. ( Identical publication in Pediatrics 1988;81: 697±706). 23. Alter D, Garcia-Valenzuela E, Kim Y, Kammer J, Viana M, Shapiro M. Hemoglobin levels, blood transfusions, and other neonatal risk factors associated with progression of retinopathy of prematurity. Invest Opthalmol Visual Sci 1998;39( Suppl 4):S820. 24. Saunders RA, Purohit D, Hulsey TC. The effects of blood transfusion protocol on retinopathy of prematurity ( Letter to the Editor). Pediatrics. [In press]. 25. Mittelman D, Cronin C. The relationship of blood transfusions to retrolental fibroplasia. Ann Ophthalmol 1983;15:376 ±378. 26. Cooke RWI, Clark D, Hickey-Dwyer M, Weindling AM. The apparent role of blood transfusions in the development of retinopathy of prematurity. Eur J Pediatr 1993;152:833 ±836. 27. Sullivan JL. Iron, plasma antioxidants and the ``oxygen radical disease of prematurity.'' Am J Dis Child 1988;142:1341 ± 1344. 28. Hittner HM, Rudolph AJ, Kretzer FL. Suppression of severe retinopathy of prematurity with vitamin E supplementation. Ophthalmology 1984;91: 1512± 1523. 29. Inder TE, Clemett RS, Austin NC, Graham P, Darlow BA. High iron status in very low birth weight infants is associated with an increased risk of retinopathy of prematurity. J Pediatr 1997;131:541 ± 544. 30. Hesse L, Eberl W, Schlaud M, Poets CF. Blood transfusion Ð iron load and retinopathy of prematurity. Eur J Pediatr 1997;156:465 ± 470. 26 Journal of Perinatology 2001; 21:21 ± 26