- Email: [email protected]
Effect of protein intake on erythropoiesis during erythropoietin treatment of anemia of prematurity
Effect of protein intake on erythropoiesis during erythropoietin treatment of anemia of prematurity Mark S. Brown, MD, a n d H o w a r d Shapiro, PhD From the Departments of Pediatrics and Research and Development, Presbyterian/St. Luke's Medical Center, Denver, Colorado
Objective: To examine the effect of protein intake on the erythropoietic response of very Iow birth weight infants to treatment with recombinant human erythropoietin (rHuEPO). Study design: Twenty very Iow birth weight infants were enrolled in the study and 19 completed the 6 weeks of study. Weekly absolute reticulocyte counts, protein intakes, and growth, as well as selected markers of protein metabolismmpreal bumin, albumin, and transferrin--were analyzed, fron stores were estimated for each infant to exclude iron deficiency as a cause of anemia. The relationship between protein intake and absolute reticulocyte count was evaluated with a linear breakpoint analysis to account for any plateau in the relationship at higher protein intakes. Results: Adequate iron stores were present in all infants, and transferrin concentrations correlated with measured total iron-binding capacity (r = 0.95, p = 0.000 I). In the rHuEPO-treated infants, absolute reticulocyte count was significantly associated with protein intake up to 3. I gm/kg per day and extending to 3.5 gm/kg per day (p = 0.041 to 0.032); beyond this point there was no Ionger any effect. Moreover, in comparison with the infants who received placebo, the rHuEPOtreated infants had a better daily percent weight gain for a protein intake up to 3.5 gm/kg per day (p = 0.01ö). Conclusions: In VLBW infants treated with rHuEPO, higher protein intake up to 3. I to 3.5 gm/kg per day improved the erythropoietic response, and protein utilization for growth was improved. During treatment with rHuEPO, infants who receive adequate protein to achieve satisfactory growth also receive sufficient protein for erythropoiesis. (J PEDIATR1996; 128:512-7) Like iron, protein is one of the essential nutrients for erythropoiesis. In animals and adult human beings, protein restriction significantly limits erythropoiesis. 13 In preterrn infants with anemia of prematurity, protein supplementation of 9 gm/L to human milk has produced higher hemoglobin
concentrations than with human milk alone, 4 underscoring the erythropoietic requirement for protein during development. This same supplement was used for all infants in a trial of recombinant human erythropoietin in preterm infants. 5 However, the relationship between protein intake and eryth-
Supported by R. W. Johnson Pharmaceutical Research Institute (Raritan, N.J.), and PresbyteriardSt. Luke' s Community Foundation (Denver, Colo.), grant No. 9014. Submitted for publication April 12, 1995; accepted Dec. 2, 1995. Reprint requests: Mark S. Brown, MD, 1601 East 19th Ave., Suite 5300, Denver, CO 80218. Copyright © 1996 by Mosby-Year Book, Inc. 0022-3476/96/$5.00 + 0 9/20/71133
I
512
rHuEPO TIBC VLBW
Recombinant human erythropoietin Total iron-binding capacity Very low birth weight
ropoiesis during rHuEPO treatment has not been more specifically examined to determine the level of protein intake necessary to optimize the erythropoietic response of VLBW infants to rHuEPO.
The Journal of Pediatrics Volume 128, Number 4
As part Of the recently published United States multicenter trial of rHuEPO to treat anemia of prematurity, 6 each site contributed a supplemental study. We chose to study the relationship between protein intake and erythropoiesis during rHuEPO treatment, and hypotbesized that the erythropoietic response to rHuEPO would depend on protein intake to an upper plateau. Iron sufficiency for each infant was assessed to exclude the possibility of iron deficiency confounding this analysis.
METHODS This study was approved by the Food and Drug Administration (investigational new drug No. 4117) and the Presbyterian/St. Luke's Medical Center institutional review board as part of the United States multicenter, double-blind, placebo-coutrolled trial of rHuEPO treatment of anemia of prematurity. 6 Parental consent was obtained. Subjects. Eligibility and exclusion criteria were part of the multicenter study protocol. We enroUed 20 neonates during a 1-year period from December 1991 to November 1992. All infants were born at the medical center. One infant in the placebo group died of necrotizing enterocolitis a week after the study began; a total of 10 rHuEPO-treated infants and 9 placebo-treated infants completed the 6 weeks of study.
Erythropoietin administration and nutritional monitoring. Infants were randomly assigned into blocks of four for each site; rHuEPO (100 U/kg body weight) or au identical volume of placebo was administered subcutaneously in the lateral aspect of the thigh 5 days a week for a total of 6 weeks. When infants were receiving at least 60 kcal/kg per day of feedings, oral iron supplements were given at an initial dose of 3 mg/kg per day and increased to at least 6 mg/kg per day for the remainder of the study. Additional vitamin supplements included vitamin E (15 U/day), folate (50 pf,/ day), and a preparation of vitamins A, C, and D. Total amounts of blood withdrawn and transfused were recorded. For calculations of iron balance between blood taken by phlebotomy and given by transfusion (net blood iron), the amount of iron per gram of hemoglobin was estimated to be 3.4 mg7; the hemoglobin of transfused blood was approximately 230 gin/L, and the hemoglobin of blood obtain by phlebotomy was taken from the most recent determination. Daily enteral and parenteral protein intakes were totaled for each infant individually and averaged for each week as grams per kilogram per day. Protein source, parenteral or enteral, was not discriminated. Human milk was estimated to contain 20 kcal/30 ml and 2.0 gm of protein/100 kcal 8, 9; it was supplemented with Similac Natural Care (Ross Laboratories, Columbus, Ohio) and Similac With Iron infant formula powder (Ross Laboratories) in graduated amounts of kilocalories per kilogram per day to attempt to optimize growth according to the Colorado intrauterine growth
Brown and Shapiro
5 13
chart.l° For a more representative comparison of growth, the percent weight change for each week was calculated. Percent weight change was weight gained or lost for the week expressed as a percentage of weight at the beginning of the week. Laboratory monitoring. Blood was obtained by venipuncture for complete blood cell count, differential, and manual reticulocyte count at the beginning of the study, weeldy, and at the end of the study. Measurements of prealbumin, albumin, transferrin, TIBC, serum iron, and ferritin were made at the beginning of the study (week 0), midway (week 3), and at the end of the study (week 6). The total amount of additional blood taken for the protocol laboratory studies was l 1 rel. Tests were performed with automated instruments in the hospital clinical laboratory, and results were available to the team caring for the infant. Prealbumin and transferrin concentrations were measured as turbidity with specific antibody (Beckman Instruments, Inc., Brea, Calif.). Serum iron and TIBC were measured on the Kodak Ektachem autoanalyzer (Johnson & Johnson Clinical Diagnostics, Rochester, N.Y.). Fen-irin was measured immunologically on the model ACS:IS0 analyzer (Ciba Coming Diagnosfics Corp., Medfield, Mass.). Statistical methods. A two-sample t test was used for comparisons between groups and a paired t test for comparisons within groups. Stepwise multiple regression was used to assess the relationship between variables when any week was analyzed separately. When weekly data were combined, a repeated measures analysis of varianee was performed with the Statisfical Analysis System Proc Mixed program (SAS System for Windows 3.10, release 6.08, 1991, 1992; SAS Institute Inc., Cary, N.C.). To determine whether there was a relationship between absolute reticulocyte count and protein intake during rHuEPO-stimulated erythropoiesis, absolute reticulocyte counts and protein intakes from week 2 to week 6 were examined for the rHuEPO-treated infants. Week 2 was chosen as the beginning for this analysis because the absolute reticulocyte count initiaUy doubled for almost all infants by this time. In previous studies, the relationships between protein intake, growth, and markers of protein nutrition have been nonlinear.11, la Therefore the relationships between absolute reficulocyte count and protein intake, and between percent weight gain and protein intake, were analyzed with a linear breakpoint method to determine whether there was a plateau in the relationship at a point where there might be maximal benefit from protein intake; this analysis took into consideration the impact of both postconceptional age and repeated measures.13 This type of analysis compares the fitted linear relationships up to and beyond sequential points (in this case increments of protein intake of 0.1 gm/kg per day), to test for a significant difference between the slopes. Model comparisons for best fit for the breakpoint analyses were assessed by Akaike's information criterion. 14
5 14
Brown and Shapiro
The Journal of Pediatrics April 1996
Table. Selected protein and nutritional variables for infauts who received placebo (n = 9) and rHuEPO-treated (n = 10) infants at week 0 and week 6
Weight (gin) Protein (gm/kg/day) Kcal/kg/day Albumin (gm/L) Prealbumin (gin/L) Transferrin (gin/L) TIBC (~unol/L) Ferritin (lag/L) Iron (Nnol/L)
Placebo
rHuEPO
Placebo
rHuEPO
Week 0
Week 0
Week 6
Week 6
p value at week 6 (placebo vs rHuEPO)
950 (198) 2.61 (0.33) 104.7 (18.1) 30.1 (3.8) 80.6 (32.1) 1.51 (0.33) 24.2 (9.3) 382.8 (209.4) 12.8 (5.6)
864 (183) 2.69 (0.45) 113.7 (10.3) 30.0 (1.9) 88.5 (45.1) 1.63 (0.45) 28.6 (8.4) 357.2 (228.9) 15.9 (8.1)
1956 (305) 3.10 (0.47) 119.1 (12.5) 30.4 (6.4) 90.6 (22.1) 1.75 (0.48) 31.2 (12.1) 287.1 (172.1) 12.4 (3.8)
1791 (341) 3.30 (0.54) 122.2 (17.5) 28.0 (3.0) 126.0 (48.6) 1.99 (0.53) 34.5 (8.0) 117.1 (89.6) 14.4 (3.3)
0.30 0.40 0.67 0.29 0.06 0.22 0.57 0.01" 0.23
Data in parenthesesequal 1 SD. rHuEPO, Recombinanthuman erythropoietin;TIBC, total iron-bindingcapacity. *Significantdifference.
The relationship between transferrin concentration and TIBC was examined only for the rHuEPO-treated group to minimize the effect of increased iron stores on the accuracy of measurements of TIBC.15, 16 All analyses were conducted using Statistical Analysis Systems software (SAS Institute Inc.). RESULTS The infants in the two study groups were comparable at birth (placebo group versus rHuEPO treatment group, mean _+ SD): birth weight 889 -+ 249 gm versus 845_+ 155 gin, gestational age 26.4 -+ 1.5 weeks versus 25.7 _+ 1.6 weeks, and at entry: age 16.3 _+ 9.2 days versus 15.6 -+ 8.9 days, weight 950 _+ 198 gm versus 864 _+ 183 gm, and ventilation needs (4 of 10 in placebo group versus 6 of 10 in treatment group). In both groups of infants, dexamethasone was used only at the fime of extubation. The study groups were also comparable with regard to the supplemental nutritional variables at the beginning of the study (week 0, Table) and for weeldy averages of protein and kilocalofie intake per kilogram per day for each week of the study, including week 2. Six of 9 infants in the placebo group and 8 of 10 infants in the treatment group received human milk at some time during the smdy, although most mothers continued to pump their breasts only for the first few weeks after birth. Oral iron supplementation was begun in 17 of the 19 infants by the third week of smdy, and the cumulative oral dose for the 6 weeks of study ranged from 20 to 468 mg/kg. Net blood iron, the difference in iron given by transfusion and taken by phlebotomy, ranged from -56 to 529 mg/kg for 6 weeks. Only four infants had a negative balance, but these infants received a total of 252 to 290 mg/kg of oral iron supplementation for the study period. When the total of net blood iron and oral iron (total iron intake) was correlated to ferritin, serum iron, transferrin, or transferrin saturation in
the rHuEPO-treated infants, the best correlation was to ferritin (r = 0.80, p = 0.006). Iron deficiency was assessed further by examination of ferritin levels, which ranged from 22 to 330 pg/L at 3 and 6 weeks, and by evaluation of the relationship between absolute reticulocyte count and either net blood iron or total iron. The latter relationships were analyzed for each week separately and for the 6-week cumulative total. There was no decrease in absolute reticulocyte counts at lower net blood iron or total iron estimates. The breakpoint in the linear relationship between absolute reticulocyte count and protein intake began at a protein intake of 3.1 gin/kg per day and extended to 3.5 gin/kg per day (p = 0.041 to 0.032); beyond this the relationship was not different from a slope of 0 (Fig. 1). The best fit by Akaike's information criteria 15 occurred at a protein intake of 3.5 gin/kg per day (p = 0.032). There were three sets of data points at which the absolute reticulocyte count was quite high and protein intake was -<3.1 gin/kg per day (Fig. 1); these were from healthier infants who were receiving all-enteral nutrition. On the other hand, there were two sets of data points at which the absolute reticulocyte count was quite low and protein intake was -<3.1 gm&g per day; these were from sicker infants who were being supported by mechanical ventilation and receiving only partial enteral intake. Similar breakpoint modeling was used to analyze the relationship between the percent weight gain and protein intake. The breakpoint for percent weight gain was significant between protein intakes of 2.8 and 3.5 gin/kg per day (p -- 0.017 to 0.0035); the best fit was at 3.0 gm/kg per day (p = 0.0035). In the placebo group there was no breakpoint in the relationship of percent weight gain to protein intake for the same period. Both protein intake and percent weight gain were analyzed with linear regression up to a protein intake of 3.5 grn&g per day, the upper amount of protein intake at which the relationship was still significant for the treatment group, to
The Journal of Pediatrics Volume 128, Number 4
Brown and Shapiro
5 15
700 /Il/l/S/t/Il/« /,l//l/S/l/sb /Sl/I/S//¢S/h Il/l/l/l/Il//, /¢///////~//, ~///////s //, v//l///l////s, /~///l/I/llSll /l/I/////I///~ I/Sl/I///I//h ss/t/l/l/l///, ///////I////¢1 ¢//i///i/i/i/, //l//i///////~ /pl///ll////s //////~/l/l/ /lll//~l/~¢h
•
A
600
O
g 5OO •
B o
400
"$ er
300
õ
200
1t 11111 t/ltllllllltll I/lll/llll tlltlllll ~/llllllll/l/ Il II/I/I
• /I t11111 o~,....,...~AII~I%II;%ZZ /z i z~ ¢ I% / l l% / / I; h
/
Illlll/llllll,
• •
•
/i/1/i/////l// /I/I/I///l/ll ............. 9 // /I/// // /////// // ///I/ /, / l l
8
/// //////sb //,Äl~//4.'/s///, ////// ////h //////~//~/// /////////,4v/// /////////////i //i // // // // // /s//////~ l p / / , /////////////« /////////////¢ /¢///////////¢
•
<
100 0
i1« Ill
/Il /ll//lll/~/ll
I
/ / / / / / / / g / / / / ¢
// // // // // // // // ¢/ //////h/ / ////¢////////, /////////////s ////////////// //////t///¢/h g//////////Sb ////////////h I
2.0
'
'
'
'
I
'
'
'
'
'
I
I
'~"
~'b
'
'
'
3.0
'
I
4.0
.
.
.
.
I
.
.
.
.
I
5.0
Protein Intake (g/kg/d) Fig. 1, Absolute reticulocyte counts are plotted against weekly average of daily protein intake from week 2 through week 6 for the rHuEPO-treated infants. By week 2 almost all absolute reticulocyte counts had doubled in the treated infants. A breakpoint analysis was used to analyze the data. 13There was a significant relationship between absolute reticulocyte count and protein intake for the rHuEPO-treated infants starting at protein intakes of 3.1 gin/kg per day and extending to 3.5 gin/kg per day, represented by the vertical shaded area. The regression line is for the best fit at a breakpoint of 3.5 gin/kg per day. 14 All infants were thought to have adequate iron stores, and the analysis was adjusted for postconceptional age and 'repeated measures. There was still considerable variability in absolute reticulocyte counts at similar protein intakes, wbich may have been related to other factors such as complications of prematurity, sources of protein, or estimates of protein intake.
compare the relationship of protein intake to percent weight gain between the placebo and treatment groups. This comparison incorporated repeated measures into a linear regression analysis. The slope of the relationship between protein intake and percent weight gain for the treatment group was slightly steeper than in the placebo group, 6.1% versus 5.3% weight gain per gram per kilogram per day, respectively; the regression lines for the two groups were significantly different (p = 0.016), indicating a greater impact of protein intake on growth during rHuEPO treatment. Analysis of the markers of protein nutrition showed a significant correlation between protein intake and prealbumin for week 6 (r = 0.55, p = 0.014). Transferrin concentration was predicted best by a combination of prealbumin, iron, and postconceptional age (stepwise multiple regression, r e = 0.47, p = 0.03). The linear regression of transferrin on TIBC in the rHuEPO-treated infants was transferrin= 7.5 + 1.001 • TIBC, r = 0.95, p = 0.0001 (Fig. 2). Repeated measures analysis did not significantly improve the fit. DISCUSSION We found that treatment with rHuEPO did not adversely affect growth in VLBW infants and that the erythropoietic
response to rHuEPO improved with increased protein intake. The necessity for protein for erythropoiesis in our analysis was focused on erythroid proliferation and differentiation because erythropoietin was provided exogenously. These observations suggest that in preterm infants protein is additionally and separately important for erythroid formation. Weight gain per gram per kilogram per day of protein intake was slighfly improved in the rHuEPO-treated infants, which may explain the trend to higher prealbumin levels in the rHuEPO-treated infants by the end of the smdy. In support of this latter observation, adults l~eated with dialysis for renal failure have improved nitrogen balance during treatment with rHuEPO.17 Conversely, the European multicenter erythropoietin trial found slower weight gain in the erythropoietin-treated infants, although these infants had more infection complications, including septicemia, meningitis, and septic arthritis.la It is difficult to reconcile these findings with those from our study because nutritional intakes and weight gain per body weight were not analyzed separately. The mechanism by which rHuEPO may have improved growth in our study is not clear, but possibilities include more efficient delivery of oxygen to the tissues from a larger or more consisteut red cell mass, or rHuEPO may be a growth fac-
516
Brown and Shapiro
The Journal of Pediatrics April 1996
3.02.5~. 2.0,E
1.5-
0
0
0
¢.
P- 1.0
0.5 0.0
0
5'o
1;0
1;0
2;0
2;0
3òo
3;0
4'2
24
63
(lag/dL)
;
9
18
2'7
3'6
(lamol/k) TIBC
Fig. 2. Simultaneous measurements oftransferrin and TIBC from
the rHuEPO-treated infants are plotted to assess comparability. The open circles represent values from week 0 and closed circles values from week 6. This analysis was restficted to the rHuEPO-treated infants because higher iron storage states typical ofinfants who have received multiple transfusions during the native anemia of prematurity can interfere with TIBC measurements. I5 For convenience the x-axis is expressed in both conventional units (micrograms per deciliter) and SI units (micromoles per liter). The regression line was transferrin = 7.5 + 1.001 • TIBC; r = 0.95, p = 0.0001.
tor. The possibility of improved protein utilization from rHuEPO treatment would not necessarily have resulted in greater discharge weights or earlier discharge, because individual caloric intake was routinely adjusted to try to maintain intrauterine growth rates.S° In older infants and children with nephrosis, Siimes et al. t9 speculated that during rHuEPO treatment lower protein intake restricted the response of erythropoiesis in deference to body growth. Our analysis could not address such a low fange of protein intake, although at higher intakes this did not seem to be the case. It is reassuring that treatment of anemia of prematurity with rHuEPO did not adversely affect growth and may have improved protein utilization. Although we found a significant relätionship between protein intake and absolute reticulocyte counts, there remained a disconcerting amount of variability in absolute reticulocyte counts for a given protein intake. To strengflaen the power of out analysis, we took into consideration the impact of both postconceptional age and repeated measures. It is doubtful that any of the infants were iron deficient to such a degree as to have significantly restficted erythropoiesis. Our estimate of iron status included consideration of iron stores from transfusion gain and phlebotomy loss, and from
oral iron intake. Moreover, ferritin levels correlated with iron status estimates and were above levels usually associated with iron deficiency, although similar data are not available for rHuEPO treatment. The remaining variability was most likely related to either infant-specific factors affecting metabolism and erythropoiesis--respiratory disease, 2° infection, overall degree of illness-----or to our estimation of protein intake. The estimates of protein intake may have either underestimated or overestimated actual protein intake, especially when human milk was fed. Milk from mothers who deliver before term is known to vary in nutrient concentrations including protein. We did not analyze individual milk samples for protein content but used an estimate drawn from published studies. 8, 9 Differences in protein content of milk would have been dampened because a comparable number ofinfants in both groups were fed human milk and it was fed only during the first weeks of feeding, after which it was supplemented. Another source of variability may have been whether protein was supplied parenterally or enterally. This was not tracked separately, and differences in utilization may have added to the disparity in individual erythropoietic response to protein intake, 21 although to a lesser degree because infants were already receiving enteral nutrition at entry to the smdy. It is difficult to assign an ideal protein intake to a given infant because of this variability. To provide adequate protein for rHuEPO-stimulated erythropoiesis for an individual infant, it would seem prudent to provide protein during treatment with rHuEPO and to adjust protein intake to the amount that achieves optimal ga'owth. Analysis of prealbumin, alburnin, and transferrin concentrations provided another method of examining protein nutrition during rHuEPO treatment. Prealbumin has a shorter half-life, is a short-range marker for protein nutrition,12, 22-24 and was a representative marker of protein nutrition in out infants. Albumin has a longer half-life and is a long-range marker of protein nutrition, extending to 4 weeks and beyond. 24 There was no difference in albumin between placebo and treated infants by the end of the study, indicating that treatment with rHuEPO probably did not significantly drain protein metabolism. Transferrin is a mid-range indicator of protein nutrition 12and in preterm infants has correlated with either protein nutritional state, 12 iron status, 25 or gestational age at birth, 26' 27 but not postuatal age. 28 Thus not only protein nutrition and iron balance but also maturity add complexity to the interplay of the factors regulating transferrin levels in preterm infants. W e used a multiple stepwise analysis to examine a combination of factors that may affect transferrin levels in preterm infants to explain some of the variability in the transferrin values. Prealbumin, iron, and postconceptional age accounted for 47% of the variability in transferrin values in all infants by week 6. In addition, trans-
The Journal of Pediatrics VoIume 128, Number 4
ferrin was comparable to TIBC as an estimate of iron-carrying capacity and offers significant advantages over measurement of TIBC in V L B W infants. Measurements of TIBC have greater phlebotomy requirements, have larger variation, are more laborious, and can produce erroneous results when obtained shortly after a transfusion or when there are higher amounts of body iron. 15 Additional factors other than protein intake may also either limit or enhance the erythropoietic response and, in mm, need study during rHuEPO treatment in preterm infants. The current information on the relationship between protein intake, growth, and erythropoiesis during treatment with rHuEPO is specific to V L B W infants at the dose schedule used. This analysis was secondary to the primary intent of this study; therefore these observations should be viewed as prefiminary and need to be tested in a prospective manner. We thank Kim Dohreu for expertise with graphics and T. Heaberlin, P. Dillon, S. Aschfort, and R. Zatopek for technical assistance with data collection and ding administration. We thank Barbara Quissell, MD, Peter Honeyfield, MD, Delphine Eichorst, MD, Bruce Reddix, MD, Jeffrey B. Hanson, MD, and the nurses in the neonatal intensive care unit for their support in conducting this study and for their care of the study infants. REFERENCES
1. Okano M, Ohnota H, Sasaki R. Protein deficiency impairs erythropoiesis in rats by reducing serum erythropoietin concentration and the population size of erythroid precursor cells. J Nutr 1992;122:1376-83. 2. Anagnostou A, Schade S, Ashkinaz M, Barone J, Fried W. Effect of protein deprivation on erythropoiesis. Blood 1977; 50:1093-7. 3. Catchatourian R, Eckerling G, Fried W. Effect of short-term protein deprivation on hemopoietic functions of healthy volunteers. Blood 1980;55:625-8. 4. Rönnholm KAR, Siimes MA. Haemoglobin concentration depends on protein intake in small preterm infants fed human milk. Arch Dis Child 1985;60:99-104. 5. Bechensteen AG, Hägä P, Halvorsen S, et al. Erythropoietin, protein, and iron supplementation and the prevention of anaemia of prematurity. Arch Dis Child 1993:69:19-23. 6. Shannon KM, Keith JF llI, Mentzer WC, et al. Recombinant human erythropoietin Stimulates erythropoiesis and reduces erythrocyte transfusions in very low birth weight preterm infants. Pediatrics 1995;95:1-8. 7. Oski FA. Iron requirements of the premature infant. In: Tsang RC, ed. Vitamin and mineral requirements in preterm infants. New' York: Marcel Dekker, 1985:9-21. 8. Gross SJ, David RJ, Banman L, Tomarelli RM. Nutritional composition of milk produced by mothers delivering preterm. J PED~T~ 1980;96:641-4. 9. Chessex P, Reichman B, Verellen G, et al. Quality of growth in premamre infants fed their own mothers' milk. J PEDIATR 1983;102:107-12. 10. Lubchenco LO, Hansman C, Dressler M, Boyd E. Intrauterine growth as estimated from liveborn birth-weight data at 24 to 42 weeks of gestation. Pediatrics 1963;32:793-800.
Brown and Shapiro
5 17
11. Moskowitz SR, Pereira G, Spitzer A, Heaf L, Amsel J, Watkins JB. Prealbumiu as a biochemical marker of nutritional adequacy in premature infants. J P~OIATR1983;102:749-53. 12. Polberger SKT, Fex GA, Axelsson IE, Räihä NCR. Eleven plasma proteins as indicators of protein nutrition status in VLBW infants. Pediatrics 1990;86:916-21. 13. Jones RH, Molitoris BA. A statistical method for determining the breakpoint of two lines. Anal Biochem 1984; 141:28790. 14. Akaike H. A new look at the statistical model identification. IEEE Transactions on Automatic Control 1974;AC-19: 716-23. 15. Pootrakul P, Josephson B, Huebers HA, Finch CA. Quantitation of ferritin iron in plasma, an explanation for non-transferrin iron. Blood 1988;71:1120-3. 16. Brown MS. Effect of transfusion and phlebotomy on serum ferritin levels in low birth weight infants. J Perinatol (in press). 17. Canand B, Bouloux C, Rivory JP, et al. Erythropoietin-induced changes in protein nutrition: quantitative assessment by urea kinetic modeling analysis. Blood Purif 1990;8:301-8. 18. Maler RF, Obladen M, Scigalla P, et al. The effect of epoetin beta (recombinant human erythropoietin) on the need for transfusion in very-low-birth-weight infants. N Engl J Med 1994;330:1173-8. 19. Siimes MA, Rönnholm KAR, Antikainen M, Holmberg C. Factors limiting the erythropoietin response in rapidly growing infants with congenital nephrosis on a peritoneal dialysis regimen after nephrectomy. J PEDIATR1992;120:44-8. 20. Soubasi V, Kremenopoulos G, Diamandi E, Tsantali C, Tsakiris D. In which neonates does early recombinant erythropoietin treatment prevent anemia of prematurity? Results of a randomized controlled study. Pediatr Res 1993;34:675-9. 21. Socha J, Eggermont E, Carchon H, Devlieger H, Eeckels R. Plasma prealbumin in low birth weight infants. Acta Paediatr Belg 1977;30:171-4. 22. Giacoia GP, Watson S, West K. Rapid tumover transport proteins, plasma albumin, and growth in low birth weight infants. J Patenter Enteral Nutr 1984;8:367-70. 23. Georgieff MK, Sasanow SR, Pereira GR. Serum transthyretin levels and protein intake as predictors of weight gain velocity in premature infants. J Pediatr Gastroenterol Nutr 1987 ;6:775-9. 24. Yoder MC, Anderson DC, Gopalakrishna GS, Douglas SD, Polin RA. Comparison of serum fibronectiu, prealbumin, and albumin concentrations during nutritional replefion in proteincalorie malnourished infants. J Pediatr Gastrenterol Nutr 1987; 6:84-8. 25. Chockalingam UM, Murphy E, Ophoven JC, Weisdorf SA, Georgieff MK. Cord transferrin and ferritin values in newbom infants at risk for prenatal uteroplacental insufficiency and chronic hypoxia. J t~OIATR 1987;111:283-6. 26. Chockalingam U, Murphy E, Ophoven JC, Georgieff MK. The influence of gestational age, size for dates, and prenatal steroids on cord transferrin levels in newboru infants. J Pediatr Gastroenterol Nutr 1987;6:276-80. 27. Scott PH, Berger HM, Kenward C, Scott P, Wharton BA. Effect of gestational age and intranterine nutrition on plasma transferrin and iron in the newboru. Arch Dis Child 1975; 50:796-8. 28. Georgieff MK, Amamath UM, Murphy EL, Ophoven JJ. Serum transferrin levels in the longitudinal assessment of proteinenergy status in preterm infants. J Pediatr Gastroenterol Nutr 1989;8:234-9.
Objective: To examine the effect of protein intake on the erythropoietic response of very Iow birth weight infants to treatment with recombinant human erythropoietin (rHuEPO). Study design: Twenty very Iow birth weight infants were enrolled in the study and 19 completed the 6 weeks of study. Weekly absolute reticulocyte counts, protein intakes, and growth, as well as selected markers of protein metabolismmpreal bumin, albumin, and transferrin--were analyzed, fron stores were estimated for each infant to exclude iron deficiency as a cause of anemia. The relationship between protein intake and absolute reticulocyte count was evaluated with a linear breakpoint analysis to account for any plateau in the relationship at higher protein intakes. Results: Adequate iron stores were present in all infants, and transferrin concentrations correlated with measured total iron-binding capacity (r = 0.95, p = 0.000 I). In the rHuEPO-treated infants, absolute reticulocyte count was significantly associated with protein intake up to 3. I gm/kg per day and extending to 3.5 gm/kg per day (p = 0.041 to 0.032); beyond this point there was no Ionger any effect. Moreover, in comparison with the infants who received placebo, the rHuEPOtreated infants had a better daily percent weight gain for a protein intake up to 3.5 gm/kg per day (p = 0.01ö). Conclusions: In VLBW infants treated with rHuEPO, higher protein intake up to 3. I to 3.5 gm/kg per day improved the erythropoietic response, and protein utilization for growth was improved. During treatment with rHuEPO, infants who receive adequate protein to achieve satisfactory growth also receive sufficient protein for erythropoiesis. (J PEDIATR1996; 128:512-7) Like iron, protein is one of the essential nutrients for erythropoiesis. In animals and adult human beings, protein restriction significantly limits erythropoiesis. 13 In preterrn infants with anemia of prematurity, protein supplementation of 9 gm/L to human milk has produced higher hemoglobin
concentrations than with human milk alone, 4 underscoring the erythropoietic requirement for protein during development. This same supplement was used for all infants in a trial of recombinant human erythropoietin in preterm infants. 5 However, the relationship between protein intake and eryth-
Supported by R. W. Johnson Pharmaceutical Research Institute (Raritan, N.J.), and PresbyteriardSt. Luke' s Community Foundation (Denver, Colo.), grant No. 9014. Submitted for publication April 12, 1995; accepted Dec. 2, 1995. Reprint requests: Mark S. Brown, MD, 1601 East 19th Ave., Suite 5300, Denver, CO 80218. Copyright © 1996 by Mosby-Year Book, Inc. 0022-3476/96/$5.00 + 0 9/20/71133
I
512
rHuEPO TIBC VLBW
Recombinant human erythropoietin Total iron-binding capacity Very low birth weight
ropoiesis during rHuEPO treatment has not been more specifically examined to determine the level of protein intake necessary to optimize the erythropoietic response of VLBW infants to rHuEPO.
The Journal of Pediatrics Volume 128, Number 4
As part Of the recently published United States multicenter trial of rHuEPO to treat anemia of prematurity, 6 each site contributed a supplemental study. We chose to study the relationship between protein intake and erythropoiesis during rHuEPO treatment, and hypotbesized that the erythropoietic response to rHuEPO would depend on protein intake to an upper plateau. Iron sufficiency for each infant was assessed to exclude the possibility of iron deficiency confounding this analysis.
METHODS This study was approved by the Food and Drug Administration (investigational new drug No. 4117) and the Presbyterian/St. Luke's Medical Center institutional review board as part of the United States multicenter, double-blind, placebo-coutrolled trial of rHuEPO treatment of anemia of prematurity. 6 Parental consent was obtained. Subjects. Eligibility and exclusion criteria were part of the multicenter study protocol. We enroUed 20 neonates during a 1-year period from December 1991 to November 1992. All infants were born at the medical center. One infant in the placebo group died of necrotizing enterocolitis a week after the study began; a total of 10 rHuEPO-treated infants and 9 placebo-treated infants completed the 6 weeks of study.
Erythropoietin administration and nutritional monitoring. Infants were randomly assigned into blocks of four for each site; rHuEPO (100 U/kg body weight) or au identical volume of placebo was administered subcutaneously in the lateral aspect of the thigh 5 days a week for a total of 6 weeks. When infants were receiving at least 60 kcal/kg per day of feedings, oral iron supplements were given at an initial dose of 3 mg/kg per day and increased to at least 6 mg/kg per day for the remainder of the study. Additional vitamin supplements included vitamin E (15 U/day), folate (50 pf,/ day), and a preparation of vitamins A, C, and D. Total amounts of blood withdrawn and transfused were recorded. For calculations of iron balance between blood taken by phlebotomy and given by transfusion (net blood iron), the amount of iron per gram of hemoglobin was estimated to be 3.4 mg7; the hemoglobin of transfused blood was approximately 230 gin/L, and the hemoglobin of blood obtain by phlebotomy was taken from the most recent determination. Daily enteral and parenteral protein intakes were totaled for each infant individually and averaged for each week as grams per kilogram per day. Protein source, parenteral or enteral, was not discriminated. Human milk was estimated to contain 20 kcal/30 ml and 2.0 gm of protein/100 kcal 8, 9; it was supplemented with Similac Natural Care (Ross Laboratories, Columbus, Ohio) and Similac With Iron infant formula powder (Ross Laboratories) in graduated amounts of kilocalories per kilogram per day to attempt to optimize growth according to the Colorado intrauterine growth
Brown and Shapiro
5 13
chart.l° For a more representative comparison of growth, the percent weight change for each week was calculated. Percent weight change was weight gained or lost for the week expressed as a percentage of weight at the beginning of the week. Laboratory monitoring. Blood was obtained by venipuncture for complete blood cell count, differential, and manual reticulocyte count at the beginning of the study, weeldy, and at the end of the study. Measurements of prealbumin, albumin, transferrin, TIBC, serum iron, and ferritin were made at the beginning of the study (week 0), midway (week 3), and at the end of the study (week 6). The total amount of additional blood taken for the protocol laboratory studies was l 1 rel. Tests were performed with automated instruments in the hospital clinical laboratory, and results were available to the team caring for the infant. Prealbumin and transferrin concentrations were measured as turbidity with specific antibody (Beckman Instruments, Inc., Brea, Calif.). Serum iron and TIBC were measured on the Kodak Ektachem autoanalyzer (Johnson & Johnson Clinical Diagnostics, Rochester, N.Y.). Fen-irin was measured immunologically on the model ACS:IS0 analyzer (Ciba Coming Diagnosfics Corp., Medfield, Mass.). Statistical methods. A two-sample t test was used for comparisons between groups and a paired t test for comparisons within groups. Stepwise multiple regression was used to assess the relationship between variables when any week was analyzed separately. When weekly data were combined, a repeated measures analysis of varianee was performed with the Statisfical Analysis System Proc Mixed program (SAS System for Windows 3.10, release 6.08, 1991, 1992; SAS Institute Inc., Cary, N.C.). To determine whether there was a relationship between absolute reticulocyte count and protein intake during rHuEPO-stimulated erythropoiesis, absolute reticulocyte counts and protein intakes from week 2 to week 6 were examined for the rHuEPO-treated infants. Week 2 was chosen as the beginning for this analysis because the absolute reticulocyte count initiaUy doubled for almost all infants by this time. In previous studies, the relationships between protein intake, growth, and markers of protein nutrition have been nonlinear.11, la Therefore the relationships between absolute reficulocyte count and protein intake, and between percent weight gain and protein intake, were analyzed with a linear breakpoint method to determine whether there was a plateau in the relationship at a point where there might be maximal benefit from protein intake; this analysis took into consideration the impact of both postconceptional age and repeated measures.13 This type of analysis compares the fitted linear relationships up to and beyond sequential points (in this case increments of protein intake of 0.1 gm/kg per day), to test for a significant difference between the slopes. Model comparisons for best fit for the breakpoint analyses were assessed by Akaike's information criterion. 14
5 14
Brown and Shapiro
The Journal of Pediatrics April 1996
Table. Selected protein and nutritional variables for infauts who received placebo (n = 9) and rHuEPO-treated (n = 10) infants at week 0 and week 6
Weight (gin) Protein (gm/kg/day) Kcal/kg/day Albumin (gm/L) Prealbumin (gin/L) Transferrin (gin/L) TIBC (~unol/L) Ferritin (lag/L) Iron (Nnol/L)
Placebo
rHuEPO
Placebo
rHuEPO
Week 0
Week 0
Week 6
Week 6
p value at week 6 (placebo vs rHuEPO)
950 (198) 2.61 (0.33) 104.7 (18.1) 30.1 (3.8) 80.6 (32.1) 1.51 (0.33) 24.2 (9.3) 382.8 (209.4) 12.8 (5.6)
864 (183) 2.69 (0.45) 113.7 (10.3) 30.0 (1.9) 88.5 (45.1) 1.63 (0.45) 28.6 (8.4) 357.2 (228.9) 15.9 (8.1)
1956 (305) 3.10 (0.47) 119.1 (12.5) 30.4 (6.4) 90.6 (22.1) 1.75 (0.48) 31.2 (12.1) 287.1 (172.1) 12.4 (3.8)
1791 (341) 3.30 (0.54) 122.2 (17.5) 28.0 (3.0) 126.0 (48.6) 1.99 (0.53) 34.5 (8.0) 117.1 (89.6) 14.4 (3.3)
0.30 0.40 0.67 0.29 0.06 0.22 0.57 0.01" 0.23
Data in parenthesesequal 1 SD. rHuEPO, Recombinanthuman erythropoietin;TIBC, total iron-bindingcapacity. *Significantdifference.
The relationship between transferrin concentration and TIBC was examined only for the rHuEPO-treated group to minimize the effect of increased iron stores on the accuracy of measurements of TIBC.15, 16 All analyses were conducted using Statistical Analysis Systems software (SAS Institute Inc.). RESULTS The infants in the two study groups were comparable at birth (placebo group versus rHuEPO treatment group, mean _+ SD): birth weight 889 -+ 249 gm versus 845_+ 155 gin, gestational age 26.4 -+ 1.5 weeks versus 25.7 _+ 1.6 weeks, and at entry: age 16.3 _+ 9.2 days versus 15.6 -+ 8.9 days, weight 950 _+ 198 gm versus 864 _+ 183 gm, and ventilation needs (4 of 10 in placebo group versus 6 of 10 in treatment group). In both groups of infants, dexamethasone was used only at the fime of extubation. The study groups were also comparable with regard to the supplemental nutritional variables at the beginning of the study (week 0, Table) and for weeldy averages of protein and kilocalofie intake per kilogram per day for each week of the study, including week 2. Six of 9 infants in the placebo group and 8 of 10 infants in the treatment group received human milk at some time during the smdy, although most mothers continued to pump their breasts only for the first few weeks after birth. Oral iron supplementation was begun in 17 of the 19 infants by the third week of smdy, and the cumulative oral dose for the 6 weeks of study ranged from 20 to 468 mg/kg. Net blood iron, the difference in iron given by transfusion and taken by phlebotomy, ranged from -56 to 529 mg/kg for 6 weeks. Only four infants had a negative balance, but these infants received a total of 252 to 290 mg/kg of oral iron supplementation for the study period. When the total of net blood iron and oral iron (total iron intake) was correlated to ferritin, serum iron, transferrin, or transferrin saturation in
the rHuEPO-treated infants, the best correlation was to ferritin (r = 0.80, p = 0.006). Iron deficiency was assessed further by examination of ferritin levels, which ranged from 22 to 330 pg/L at 3 and 6 weeks, and by evaluation of the relationship between absolute reticulocyte count and either net blood iron or total iron. The latter relationships were analyzed for each week separately and for the 6-week cumulative total. There was no decrease in absolute reticulocyte counts at lower net blood iron or total iron estimates. The breakpoint in the linear relationship between absolute reticulocyte count and protein intake began at a protein intake of 3.1 gin/kg per day and extended to 3.5 gin/kg per day (p = 0.041 to 0.032); beyond this the relationship was not different from a slope of 0 (Fig. 1). The best fit by Akaike's information criteria 15 occurred at a protein intake of 3.5 gin/kg per day (p = 0.032). There were three sets of data points at which the absolute reticulocyte count was quite high and protein intake was -<3.1 gin/kg per day (Fig. 1); these were from healthier infants who were receiving all-enteral nutrition. On the other hand, there were two sets of data points at which the absolute reticulocyte count was quite low and protein intake was -<3.1 gm&g per day; these were from sicker infants who were being supported by mechanical ventilation and receiving only partial enteral intake. Similar breakpoint modeling was used to analyze the relationship between the percent weight gain and protein intake. The breakpoint for percent weight gain was significant between protein intakes of 2.8 and 3.5 gin/kg per day (p -- 0.017 to 0.0035); the best fit was at 3.0 gm/kg per day (p = 0.0035). In the placebo group there was no breakpoint in the relationship of percent weight gain to protein intake for the same period. Both protein intake and percent weight gain were analyzed with linear regression up to a protein intake of 3.5 grn&g per day, the upper amount of protein intake at which the relationship was still significant for the treatment group, to
The Journal of Pediatrics Volume 128, Number 4
Brown and Shapiro
5 15
700 /Il/l/S/t/Il/« /,l//l/S/l/sb /Sl/I/S//¢S/h Il/l/l/l/Il//, /¢///////~//, ~///////s //, v//l///l////s, /~///l/I/llSll /l/I/////I///~ I/Sl/I///I//h ss/t/l/l/l///, ///////I////¢1 ¢//i///i/i/i/, //l//i///////~ /pl///ll////s //////~/l/l/ /lll//~l/~¢h
•
A
600
O
g 5OO •
B o
400
"$ er
300
õ
200
1t 11111 t/ltllllllltll I/lll/llll tlltlllll ~/llllllll/l/ Il II/I/I
• /I t11111 o~,....,...~AII~I%II;%ZZ /z i z~ ¢ I% / l l% / / I; h
/
Illlll/llllll,
• •
•
/i/1/i/////l// /I/I/I///l/ll ............. 9 // /I/// // /////// // ///I/ /, / l l
8
/// //////sb //,Äl~//4.'/s///, ////// ////h //////~//~/// /////////,4v/// /////////////i //i // // // // // /s//////~ l p / / , /////////////« /////////////¢ /¢///////////¢
•
<
100 0
i1« Ill
/Il /ll//lll/~/ll
I
/ / / / / / / / g / / / / ¢
// // // // // // // // ¢/ //////h/ / ////¢////////, /////////////s ////////////// //////t///¢/h g//////////Sb ////////////h I
2.0
'
'
'
'
I
'
'
'
'
'
I
I
'~"
~'b
'
'
'
3.0
'
I
4.0
.
.
.
.
I
.
.
.
.
I
5.0
Protein Intake (g/kg/d) Fig. 1, Absolute reticulocyte counts are plotted against weekly average of daily protein intake from week 2 through week 6 for the rHuEPO-treated infants. By week 2 almost all absolute reticulocyte counts had doubled in the treated infants. A breakpoint analysis was used to analyze the data. 13There was a significant relationship between absolute reticulocyte count and protein intake for the rHuEPO-treated infants starting at protein intakes of 3.1 gin/kg per day and extending to 3.5 gin/kg per day, represented by the vertical shaded area. The regression line is for the best fit at a breakpoint of 3.5 gin/kg per day. 14 All infants were thought to have adequate iron stores, and the analysis was adjusted for postconceptional age and 'repeated measures. There was still considerable variability in absolute reticulocyte counts at similar protein intakes, wbich may have been related to other factors such as complications of prematurity, sources of protein, or estimates of protein intake.
compare the relationship of protein intake to percent weight gain between the placebo and treatment groups. This comparison incorporated repeated measures into a linear regression analysis. The slope of the relationship between protein intake and percent weight gain for the treatment group was slightly steeper than in the placebo group, 6.1% versus 5.3% weight gain per gram per kilogram per day, respectively; the regression lines for the two groups were significantly different (p = 0.016), indicating a greater impact of protein intake on growth during rHuEPO treatment. Analysis of the markers of protein nutrition showed a significant correlation between protein intake and prealbumin for week 6 (r = 0.55, p = 0.014). Transferrin concentration was predicted best by a combination of prealbumin, iron, and postconceptional age (stepwise multiple regression, r e = 0.47, p = 0.03). The linear regression of transferrin on TIBC in the rHuEPO-treated infants was transferrin= 7.5 + 1.001 • TIBC, r = 0.95, p = 0.0001 (Fig. 2). Repeated measures analysis did not significantly improve the fit. DISCUSSION We found that treatment with rHuEPO did not adversely affect growth in VLBW infants and that the erythropoietic
response to rHuEPO improved with increased protein intake. The necessity for protein for erythropoiesis in our analysis was focused on erythroid proliferation and differentiation because erythropoietin was provided exogenously. These observations suggest that in preterm infants protein is additionally and separately important for erythroid formation. Weight gain per gram per kilogram per day of protein intake was slighfly improved in the rHuEPO-treated infants, which may explain the trend to higher prealbumin levels in the rHuEPO-treated infants by the end of the smdy. In support of this latter observation, adults l~eated with dialysis for renal failure have improved nitrogen balance during treatment with rHuEPO.17 Conversely, the European multicenter erythropoietin trial found slower weight gain in the erythropoietin-treated infants, although these infants had more infection complications, including septicemia, meningitis, and septic arthritis.la It is difficult to reconcile these findings with those from our study because nutritional intakes and weight gain per body weight were not analyzed separately. The mechanism by which rHuEPO may have improved growth in our study is not clear, but possibilities include more efficient delivery of oxygen to the tissues from a larger or more consisteut red cell mass, or rHuEPO may be a growth fac-
516
Brown and Shapiro
The Journal of Pediatrics April 1996
3.02.5~. 2.0,E
1.5-
0
0
0
¢.
P- 1.0
0.5 0.0
0
5'o
1;0
1;0
2;0
2;0
3òo
3;0
4'2
24
63
(lag/dL)
;
9
18
2'7
3'6
(lamol/k) TIBC
Fig. 2. Simultaneous measurements oftransferrin and TIBC from
the rHuEPO-treated infants are plotted to assess comparability. The open circles represent values from week 0 and closed circles values from week 6. This analysis was restficted to the rHuEPO-treated infants because higher iron storage states typical ofinfants who have received multiple transfusions during the native anemia of prematurity can interfere with TIBC measurements. I5 For convenience the x-axis is expressed in both conventional units (micrograms per deciliter) and SI units (micromoles per liter). The regression line was transferrin = 7.5 + 1.001 • TIBC; r = 0.95, p = 0.0001.
tor. The possibility of improved protein utilization from rHuEPO treatment would not necessarily have resulted in greater discharge weights or earlier discharge, because individual caloric intake was routinely adjusted to try to maintain intrauterine growth rates.S° In older infants and children with nephrosis, Siimes et al. t9 speculated that during rHuEPO treatment lower protein intake restricted the response of erythropoiesis in deference to body growth. Our analysis could not address such a low fange of protein intake, although at higher intakes this did not seem to be the case. It is reassuring that treatment of anemia of prematurity with rHuEPO did not adversely affect growth and may have improved protein utilization. Although we found a significant relätionship between protein intake and absolute reticulocyte counts, there remained a disconcerting amount of variability in absolute reticulocyte counts for a given protein intake. To strengflaen the power of out analysis, we took into consideration the impact of both postconceptional age and repeated measures. It is doubtful that any of the infants were iron deficient to such a degree as to have significantly restficted erythropoiesis. Our estimate of iron status included consideration of iron stores from transfusion gain and phlebotomy loss, and from
oral iron intake. Moreover, ferritin levels correlated with iron status estimates and were above levels usually associated with iron deficiency, although similar data are not available for rHuEPO treatment. The remaining variability was most likely related to either infant-specific factors affecting metabolism and erythropoiesis--respiratory disease, 2° infection, overall degree of illness-----or to our estimation of protein intake. The estimates of protein intake may have either underestimated or overestimated actual protein intake, especially when human milk was fed. Milk from mothers who deliver before term is known to vary in nutrient concentrations including protein. We did not analyze individual milk samples for protein content but used an estimate drawn from published studies. 8, 9 Differences in protein content of milk would have been dampened because a comparable number ofinfants in both groups were fed human milk and it was fed only during the first weeks of feeding, after which it was supplemented. Another source of variability may have been whether protein was supplied parenterally or enterally. This was not tracked separately, and differences in utilization may have added to the disparity in individual erythropoietic response to protein intake, 21 although to a lesser degree because infants were already receiving enteral nutrition at entry to the smdy. It is difficult to assign an ideal protein intake to a given infant because of this variability. To provide adequate protein for rHuEPO-stimulated erythropoiesis for an individual infant, it would seem prudent to provide protein during treatment with rHuEPO and to adjust protein intake to the amount that achieves optimal ga'owth. Analysis of prealbumin, alburnin, and transferrin concentrations provided another method of examining protein nutrition during rHuEPO treatment. Prealbumin has a shorter half-life, is a short-range marker for protein nutrition,12, 22-24 and was a representative marker of protein nutrition in out infants. Albumin has a longer half-life and is a long-range marker of protein nutrition, extending to 4 weeks and beyond. 24 There was no difference in albumin between placebo and treated infants by the end of the study, indicating that treatment with rHuEPO probably did not significantly drain protein metabolism. Transferrin is a mid-range indicator of protein nutrition 12and in preterm infants has correlated with either protein nutritional state, 12 iron status, 25 or gestational age at birth, 26' 27 but not postuatal age. 28 Thus not only protein nutrition and iron balance but also maturity add complexity to the interplay of the factors regulating transferrin levels in preterm infants. W e used a multiple stepwise analysis to examine a combination of factors that may affect transferrin levels in preterm infants to explain some of the variability in the transferrin values. Prealbumin, iron, and postconceptional age accounted for 47% of the variability in transferrin values in all infants by week 6. In addition, trans-
The Journal of Pediatrics VoIume 128, Number 4
ferrin was comparable to TIBC as an estimate of iron-carrying capacity and offers significant advantages over measurement of TIBC in V L B W infants. Measurements of TIBC have greater phlebotomy requirements, have larger variation, are more laborious, and can produce erroneous results when obtained shortly after a transfusion or when there are higher amounts of body iron. 15 Additional factors other than protein intake may also either limit or enhance the erythropoietic response and, in mm, need study during rHuEPO treatment in preterm infants. The current information on the relationship between protein intake, growth, and erythropoiesis during treatment with rHuEPO is specific to V L B W infants at the dose schedule used. This analysis was secondary to the primary intent of this study; therefore these observations should be viewed as prefiminary and need to be tested in a prospective manner. We thank Kim Dohreu for expertise with graphics and T. Heaberlin, P. Dillon, S. Aschfort, and R. Zatopek for technical assistance with data collection and ding administration. We thank Barbara Quissell, MD, Peter Honeyfield, MD, Delphine Eichorst, MD, Bruce Reddix, MD, Jeffrey B. Hanson, MD, and the nurses in the neonatal intensive care unit for their support in conducting this study and for their care of the study infants. REFERENCES
1. Okano M, Ohnota H, Sasaki R. Protein deficiency impairs erythropoiesis in rats by reducing serum erythropoietin concentration and the population size of erythroid precursor cells. J Nutr 1992;122:1376-83. 2. Anagnostou A, Schade S, Ashkinaz M, Barone J, Fried W. Effect of protein deprivation on erythropoiesis. Blood 1977; 50:1093-7. 3. Catchatourian R, Eckerling G, Fried W. Effect of short-term protein deprivation on hemopoietic functions of healthy volunteers. Blood 1980;55:625-8. 4. Rönnholm KAR, Siimes MA. Haemoglobin concentration depends on protein intake in small preterm infants fed human milk. Arch Dis Child 1985;60:99-104. 5. Bechensteen AG, Hägä P, Halvorsen S, et al. Erythropoietin, protein, and iron supplementation and the prevention of anaemia of prematurity. Arch Dis Child 1993:69:19-23. 6. Shannon KM, Keith JF llI, Mentzer WC, et al. Recombinant human erythropoietin Stimulates erythropoiesis and reduces erythrocyte transfusions in very low birth weight preterm infants. Pediatrics 1995;95:1-8. 7. Oski FA. Iron requirements of the premature infant. In: Tsang RC, ed. Vitamin and mineral requirements in preterm infants. New' York: Marcel Dekker, 1985:9-21. 8. Gross SJ, David RJ, Banman L, Tomarelli RM. Nutritional composition of milk produced by mothers delivering preterm. J PED~T~ 1980;96:641-4. 9. Chessex P, Reichman B, Verellen G, et al. Quality of growth in premamre infants fed their own mothers' milk. J PEDIATR 1983;102:107-12. 10. Lubchenco LO, Hansman C, Dressler M, Boyd E. Intrauterine growth as estimated from liveborn birth-weight data at 24 to 42 weeks of gestation. Pediatrics 1963;32:793-800.
Brown and Shapiro
5 17
11. Moskowitz SR, Pereira G, Spitzer A, Heaf L, Amsel J, Watkins JB. Prealbumiu as a biochemical marker of nutritional adequacy in premature infants. J P~OIATR1983;102:749-53. 12. Polberger SKT, Fex GA, Axelsson IE, Räihä NCR. Eleven plasma proteins as indicators of protein nutrition status in VLBW infants. Pediatrics 1990;86:916-21. 13. Jones RH, Molitoris BA. A statistical method for determining the breakpoint of two lines. Anal Biochem 1984; 141:28790. 14. Akaike H. A new look at the statistical model identification. IEEE Transactions on Automatic Control 1974;AC-19: 716-23. 15. Pootrakul P, Josephson B, Huebers HA, Finch CA. Quantitation of ferritin iron in plasma, an explanation for non-transferrin iron. Blood 1988;71:1120-3. 16. Brown MS. Effect of transfusion and phlebotomy on serum ferritin levels in low birth weight infants. J Perinatol (in press). 17. Canand B, Bouloux C, Rivory JP, et al. Erythropoietin-induced changes in protein nutrition: quantitative assessment by urea kinetic modeling analysis. Blood Purif 1990;8:301-8. 18. Maler RF, Obladen M, Scigalla P, et al. The effect of epoetin beta (recombinant human erythropoietin) on the need for transfusion in very-low-birth-weight infants. N Engl J Med 1994;330:1173-8. 19. Siimes MA, Rönnholm KAR, Antikainen M, Holmberg C. Factors limiting the erythropoietin response in rapidly growing infants with congenital nephrosis on a peritoneal dialysis regimen after nephrectomy. J PEDIATR1992;120:44-8. 20. Soubasi V, Kremenopoulos G, Diamandi E, Tsantali C, Tsakiris D. In which neonates does early recombinant erythropoietin treatment prevent anemia of prematurity? Results of a randomized controlled study. Pediatr Res 1993;34:675-9. 21. Socha J, Eggermont E, Carchon H, Devlieger H, Eeckels R. Plasma prealbumin in low birth weight infants. Acta Paediatr Belg 1977;30:171-4. 22. Giacoia GP, Watson S, West K. Rapid tumover transport proteins, plasma albumin, and growth in low birth weight infants. J Patenter Enteral Nutr 1984;8:367-70. 23. Georgieff MK, Sasanow SR, Pereira GR. Serum transthyretin levels and protein intake as predictors of weight gain velocity in premature infants. J Pediatr Gastroenterol Nutr 1987 ;6:775-9. 24. Yoder MC, Anderson DC, Gopalakrishna GS, Douglas SD, Polin RA. Comparison of serum fibronectiu, prealbumin, and albumin concentrations during nutritional replefion in proteincalorie malnourished infants. J Pediatr Gastrenterol Nutr 1987; 6:84-8. 25. Chockalingam UM, Murphy E, Ophoven JC, Weisdorf SA, Georgieff MK. Cord transferrin and ferritin values in newbom infants at risk for prenatal uteroplacental insufficiency and chronic hypoxia. J t~OIATR 1987;111:283-6. 26. Chockalingam U, Murphy E, Ophoven JC, Georgieff MK. The influence of gestational age, size for dates, and prenatal steroids on cord transferrin levels in newboru infants. J Pediatr Gastroenterol Nutr 1987;6:276-80. 27. Scott PH, Berger HM, Kenward C, Scott P, Wharton BA. Effect of gestational age and intranterine nutrition on plasma transferrin and iron in the newboru. Arch Dis Child 1975; 50:796-8. 28. Georgieff MK, Amamath UM, Murphy EL, Ophoven JJ. Serum transferrin levels in the longitudinal assessment of proteinenergy status in preterm infants. J Pediatr Gastroenterol Nutr 1989;8:234-9.