The nucleated red blood cell count at birth, the volume of red cell transfusions received, and the risk of developing retinopathy of prematurity

Abstract

BACKGROUND:

Previous studies described an association between the NRBC count at birth and risk of developing retinopathy of prematurity (ROP). Other studies correlated red blood cell (RBC) transfusions with ROP. We are aware of no studies that examined both NRBC count and RBC transfusions, in the same cohort, on ROP risk.

STUDY DESIGN:

We retrospectively analyzed all infants in the Intermountain Health NICUs during the past four years who were born <32 weeks had a NRBC count at birth and had ROP examinations.

RESULTS:

Records of 386 infants demonstrated that both factors are associated with ROP. For every 1000/μL increase in NRBC, severe (grade ≥3) ROP increased by 6.8% (95% CI, 3.0–10.0%). RBC transfusions were associated with ROP incidence and severity (p = 0.001). However, neither factor alone was either necessary or sufficient for ROP.

CONCLUSION:

The NRBC count at birth and the volume of RBC transfusions both influence ROP severity.

INTRODUCTION

Several reports suggest a positive association between the nucleated red blood cell (NRBC) count at preterm birth, and the risk that retinopathy of prematurity (ROP) will develop [1–4]. If that association is genuine and robust, perhaps measuring the NRBC count at every preterm birth, preferably using postnatal umbilical cord blood as recommended by the American Academy of Pediatrics [5], should be a routine practice to help initially risk-stratify neonates for ROP. We judged that to settle that issue a new study was needed, because of small sizes and marked methodological variations in those reports (Table 1). The variations include; (1) the age (time after birth) at which the NRBC count was drawn (the NRBC count typically falls significantly during the first days after birth), (2) the laboratory method used to measure the NRBC count, (3) the criteria used to exclude data from the study, (4) the gestational age of the subjects, (5) the NRBC count level used to define an “elevated” count, (6) the small number of infants with ROP in the reports, and (7) the very small number of infants with severe ROP in the reports.

Table 1.

Features of four reports that support the hypothesis that an elevated NRBC count at preterm birth is associated with the development of ROP.

First Author (ref)	Age when NRBC was drawn	Lab Method	Study Exclusions	Gestational age of subjects	Definition of “elevated” NRBC count	n (%) with “any ROP”	n with “severe ROP”
Lubetzky (1)	<1 h	ANRBC	IDM, PIH, IUGR, abruption, previa, maternal heart, kidney, lung, or other chronic condition, drug, tobacco, alcohol abuse, perinatal infections abnormality in intrapartum monitoring, Apgar < 6. Perinatal blood loss, hemolysis, Hct < 45%, chromosomal anomaly	≤28 weeks	Not stated	23 (NR)	3
Christensen (2)	<1 day	Both manual and ANRBC reported	PIH, Abruption, pH <7.00, 5-min Apgar <7, SGA, IUGR, grade 3 or 4 IVH, PVL, pul hem, ROP requiring surgery, CP.	23–42 weeks	>95th % for gestational age	NR	NR
Niranjan (3)	<1 day	ARNBC	IDM, PIH, HDFN, Apgar <6, IUGR, maternal heart, kidney, lung or chronic condition, drug, tobacco, or alcohol abuse, perinatal infection.	<34 weeks	Not stated	67 (27%)	0
Fevereiro-Martins (4)	<1 week	ARNBC	Major congenital malformation, ophthalmological pathologies not related to ROP, death before ROP screening, absence of informed consent.	<32 weeks	Not stated	172 (37%)	0

NRBC nucleated red blood cell, ANRBC absolute, automated NRBC/μL blood, ROP retinopathy of prematurity, n = number of infants reported, NR not reported, Ref reference number, IDM infant of diabetic mother, PIH pregnancy induced hypertension, pul hem pulmonary hemorrhage, IVH intraventricular hemorrhage, HDFN hemolytic disease of the fetus and newborn, IUGR intrauterine growth restriction, SGA small for gestational age, CP cerebral palsy, Hct hematocrit, PVL periventricular leukomalacia.

In addition, many studies show an association between the number of RBC transfusions received by preterm infants during their NICU stay and their incidence of ROP [6–31]. This has been postulated to be a causative association, due to transfusion-induced replacement of physiological levels of fetal hemoglobin (HbF) with adult hemoglobin (HbA), which releases supraphysiological amounts of oxygen to the developing retina [15, 16, 22, 25].

The aims of the present study were twofold. First, we sought to rigorously assess the association between NRBC count at preterm birth and ROP occurrence, after controlling for well-known risk factors for ROP. We did this by utilizing a large multihospital database and minimizing variations. Specifically, the present database involved, (1) only NRBC counts drawn at birth (first six hours), (2) all NRBC counts were performed on the same clinical laboratory equipment, (3) we had no study exclusions on the basis of maternal, fetal, or neonatal pathologies, (4) we included only infants born <32 weeks, (5) we used NRBC reference intervals to judge whether counts were abnormal for gestational age [4], (6) we separately analyzed the data for cases of “any ROP” and cases of severe (grade≥3) ROP.

The second aim of our investigation was to assess any association between the number and volume of RBC and platelet transfusions received during the NICU stay on ROP occurrence. In doing this we sought interactions between the NRBC count at birth and the number and volume of transfusions received with the incidence and the severity of ROP.

METHODS

Study design and population

This was a retrospective analysis of infants born <32 weeks gestational age, between January 2021 and December 2024, cared for in five Intermountain Health neonatal intensive care units in Utah: McKay-Dee Hospital, Ogden, Primary Children’s Hospital, Salt Lake City, Intermountain Medical Center, Murray, Utah Valley Regional Hospital, Provo, and St. George Regional Hospital, St. George. Data were gathered from the electronic medical record by study personnel and were stored in REDCap (Research Electronic Data Capture, Nashville, Tennessee). Neonates were included in the analysis only if they had an NRBC count drawn within six hours of birth. If they had more than one NRBC count obtained during that period, only the first was used. Neonates all had retinal screening conducted by an ophthalmologist at 31–33 weeks postmenstrual age or 4–6 weeks after birth, whichever was later. Postmenstrual age was calculated as gestational age at birth (weeks+days using the best neonatal estimate) plus the chronological age in weeks+days at the time of each retinal examination.

NRBC counts at birth

NRBC counts drawn within the first six hours after birth were measured using Sysmex Hematology XE or XN analyzers (Sysmex America, Inc., Lincolnshire, IL) with Intermountain Health Laboratory Services standard operating procedures. When available, data were displayed both as NRBC/μL blood and as NRBC/100 WBC [2]. The NRBC counts were linked to deidentified patient information in the Intermountain Health Data Warehouse to obtain patient age (in hours) at sample collection, and gestational age at birth. NRBC values in these patients were compared with NRBC reference intervals appropriate for gestational age [2].

Ophthalmological data

ROP was classified according to the International Classification of Retinopathy of Prematurity (ICROP), third edition (ICROP3) [32], using five stages: (1) A thin demarcation line between vascularized and avascular retina; (2) A ridge at this demarcation line, (3) Abnormal neovascularization extending into the vitreous; (4) Partial retinal detachment; (5) Total retinal detachment.

Ophthalmology exams began no later than 31–33 weeks postmenstrual age, or 4–6 weeks after birth, whichever was later, as recommended in the AAP/AAO/AAPOS guidelines [33]. Subsequent inpatient and outpatient exams were conducted based on the findings of the previous examination. The following were recorded for each eye at each exam: (1) highest stage of ROP in the lowest zone, (2) highest stage of ROP in any zone, (3) presence of plus disease, and whether criteria were met for Type 1 ROP. Retinal examination data were recorded until one of the following endpoints was met: (1) death, (2) full vascularization to the ora serrata or vascularization in zone III, without ROP, on two consecutive examinations, (3) Non-severe ROP, listed according to highest stage in either eye, or (4) Severe ROP in either eye, treated with peripheral retinal ablation, vitreoretinal surgery or bevacizumab (or alternative) injection.

RBC and platelet transfusions

All transfused RBCs and platelets were provided by the American National Red Cross and obtained from volunteer adult donors. The RBC product was irradiated, leukoreduced, and stored in CPDA-1 (rarely [<5%] in Adsol, AS-1). Based on Intermountain Health transfusion guidelines, a volume of 15–20 mL/kg was typically administered over a period of 3–4 h [34]. The platelet product was collected using single or double venous access kits according to the manufacturer’s recommendations (Amicus Cell Separation Platform, Fresenius Kabi, Lake Zurich, IL) and Red Cross standard operating procedures. Irradiation was not performed; rather, donor platelets were subjected to pathogen inactivation using the Intercept system (Cerus Corp, Concord, CA) and stored in 70% platelet additive solution (BAS3, InterSol, Fresenius Kabi) and 30% donor plasma. Platelets were stored at 20–24 °C with agitation. Intermountain Health neonatal intensive care unit (NICU) RBC and platelet transfusion guidelines were in place in all participating hospitals [34, 35].

Statistical analysis

Summary statistics (means, standard deviations, medians, quartiles, and proportions) and basic statistical tests (chi-square test for categorical variables; Student’s t-test for continuous variables when means were used, and Wilcoxon rank-sum test when medians were used) were used for variable summaries and comparisons. We used Poisson regression to estimate the relative risk of ROP per change in NRBC count at birth while controlling for potential confounders of gestational age, birth weight, and mechanical ventilation status. We also used Poisson regression to determine whether RBC and platelet transfusions modified the relationship between NRBC count at birth and ROP. We used robust sandwich estimators to estimate standard errors and confidence intervals generated from Poisson regression models. Statistical analyses were done in the R language and environment for statistical computing (R Foundation, Vienna, Austria).

RESULTS

Dataset inclusion

Between January 1, 2021, and December 31, 2024, 896 neonates who were born <32 weeks gestation were cared for in five Intermountain Health NICUs. As shown in Fig. 1, 455 of these had an NRBC count drawn within six hours of birth and 441 did not. Of those who did, 368 had ROP examinations performed and 87 did not. Of those who were not examined for ROP, 44 were not examined because they died before they qualified for an examination, and 43 because the clinical team judged that they did not qualify for an ROP examination [33]. Thus, of the 896 in the original database, 368 infants qualified for inclusion in this study.

Fig. 1.

Flowchart displaying inclusion of preterm infants in this study.

Clinical characteristics

Of the 368 infants included in the database, 281 did not have ROP and 87 did; 53 stage 1, 16 stage 2, and 18 stage ≥3 (Table 2). Of those with stage ≥3, 6 were treated with only anti-VEGF agent, 4 underwent laser ablation ROP surgery, and 5 had both. None developed retinal detachment or blindness. The group of infants who developed ROP had a lower birth weight and were born at an earlier gestational age than the group who did not. They were also more likely to be on mechanical ventilation on day of life 14 than were those who did not develop ROP. The five-minute Apgar score and the occurrence of IVH did not differ significantly between the groups with severe ROP vs. without ROP. Consequently, we included gestational age, birth weight and mechanical ventilation at day of life 14 as variables in the subsequent regression analyses to control for potential confounding.

Table 2.

Clinical characteristics (mean ± SD or median (Q1, Q3)) of the preterm infants studied who did not vs. did develop ROP, according to their highest ROP stage.

	No ROP (n = 281)	Stage 1 ROP (n = 53)	Stage 2 ROP (n = 16)	Stage ≥ 3 ROP (n = 18)	P-value No ROP vs. any ROP	P-value No ROP vs. Stage ≥ 3 ROP
GA (weeks)	28.2 ± 1.7	26.5 ± 1.8	25.0 ± 1.8	24.8 ± 1.5	<0.001	<0.001
Birthweight (grams)	1110 ± 235	940 ± 273	704 ± 144	636 ± 160	<0.001	<0.001
AP 1	5 (3, 7)	4 (2, 6)	4 (1.5, 5)	3 (1.3, 5)	<0.001	0.041
AP 5	8 (7, 8)	7 (6, 8)	7 (6, 8)	7.5 (7, 8)	0.017	0.530
Respiratory mode at 14 days of age
Room air	15 (6.9%)	3 (6.2%)	0 (0%)	0 (0%)	<0.001	<0.001
Nasal Cannula	5 (2.3%)	0 (0%)	0 (0%)	0 (0%)
High-flow NC	70 (32.4%)	7 (14.6%)	1 (6.7%)	0 (0%)
CPAP or NIPPV/NIV-NAVA	89 (41.2%)	21 (43.8%)	4 (26.7%)	2 (12.5%)
Conventional ventilator	15 (6.9%)	9 (18.8%)	5 (33.3%)	4 (25%)
High frequency ventilator	22 (10.2%)	8 (16.7%)	5 (33.3%)	10 (62.5%)
IVH
None	218 (77.6%)	33 (62.3%)	8 (50.0%)	12 (66.7%)	0.008	0.383
Grade 1–2	45 (16.0%)	14 (26.4%)	4 (25.0%)	5 (28.7%)
Grade ≥ 3	18 (6.4%)	6 (11.3%)	4 (25.0%)	1 (5.6%)

P-values represent the results of Student’s t-test (for means), Wilcoxon rank-sum test (for medians), or chi-square test (for counts/percents).

ROP retinopathy of prematurity, GA gestational age, BW birth weight, AP Apgar Score at 1 and 5 min, NC nasal cannula, CPAP continuous positive airway pressure, NIPPV/NIV-NAVA noninvasive positive pressure ventilation/non-invasive ventilation, neurally adjusted ventilatory assist, IVH intraventricular hemorrhage.

The NRBC count at birth and the subsequent diagnosis of ROP

The NRBC count at birth was higher in the group of 87 who developed ROP (median (95% CI); 2859/μL (1867–4482)) than in the group of 281 who did not (1654/μL (1394–1982), p < 0.008). Those with stage 1 ROP had a median NRBC count at birth of 1529/μL (1158–2566); for those with stage 2 it was 3631/μL (2094–6692) and for those with stage ≥3 it was 6064/μL (3077–12478). After controlling gestational age, birth weight and mechanical ventilation at 14 days of life, NRBC count at birth was an independent risk factor for ROP (p = 0.001). For every 1000/μL increase in the NRBC count at birth, the risk of subsequently developing severe (grade≥3) ROP increased by 6.8% (adjusted relative risk=1.068, 95% C.I. = 1.03–1.10).

Blood transfusions and ROP

Red blood cell transfusions and platelet transfusions were tabulated separately, according to both the number of individual transfusions received and the total volume of transfusions received, between birth and 40 weeks post menstrual age (PMA) or hospital discharge. As shown in Table 3, infants who did not develop ROP were more likely to have received no RBC transfusions and no platelet transfusions. Those who developed ROP had a stepwise increase in RBC transfusions as ROP increased in severity. The group of 18 who developed severe ROP had the highest number of RBC transfusions (p < 0.001) and the greatest volume of transfused RBC (p < 0.001). We found a similar but weaker association between the number and volume of platelet transfusions received and the development of ROP. After controlling for gestational age, birth weight, mechanical ventilation at 14 days of life, and NRBC count at birth, the number of RBC transfusions was also an independent risk factor for development of severe ROP (p = 0.007). Number of RBC transfusions did not modify the association observed between NRBC count at birth and the development of severe ROP (p = 0.302).

Table 3.

RBC transfusions and platelet transfusions (number and volume/kg birth wt.) administered between birth and either hospital discharge or 40 weeks PMA (which ever came first) among preterm infants who did not develop ROP, vs. those who did develop ROP, according to their highest ROP stage.

	No ROP (n = 281)	Stage 1 ROP (n = 53)	Stage 2 ROP (n = 16)	Stage ≥ 3 ROP (n = 18)	P -value, No ROP vs. any ROP	P- value, No ROP vs. Stage ≥ 3 ROP
RBC Tx (n) *	0 (0, 0)	0 (0, 2)	3 (2, 4.3)	4.5 (3.3, 8.3)	<0.001	<0.001
RBC Tx volume(/kg) **	0 (0, 0)	0 (0, 24)	60 (41.5, 103.5)	121 (98.5, 196.5)	<0.001	<0.001
Platelet Tx (n) *	0 (0, 0)	0 (0, 0)	0 (0, 1)	0 (0, 3)	<0.001	<0.001
Platelet Tx volume(/kg) **	0 (0, 0)	0 (0, 0)	0 (0, 21.8)	0.5 (0, 58.5)	<0.001	<0.001

ROP retinopathy of prematurity, PMA post-menstrual age, RBC red blood cell, Tx transfusion, n = number of transfusions received.

^*Median and 25–75 percentiles.

^**Volume of RBC and platelet Tx/kg were calculated based on birth weight.

Outliers, where the NRBC count at birth did not predict ROP

Forty-six of the infants who were born with a high NRBC count (>5000/μL) [2] did not develop ROP. These 46 received no RBC transfusions and no platelet transfusions (Table 4). Six of the infants who were born with a normal NRBC count (<2000/μL) [2] developed severe ROP. This group had received multiple RBC transfusions (median 3, 25^th to 75^th percentile 2 and 6) and they had received a large volume of adult donor RBC (median 58 mL/kg, 44–126). Thus, neonates born with a high NRBC, and were therefore at an elevated risk for developing ROP, might have been “protected” from ROP by receiving no transfusions. Moreover, neonates born with a normal NRBC count, and therefore at lower risk for developing ROP, might have been “disposed” to develop ROP by receiving multiple RBC transfusions.

Table 4.

The number and volume of RBC transfusions and platelet transfusions (median and 25th – 75th percentile values) received between birth and hospital discharge or 40 weeks PMA (which ever came first), as a potential explanation for the “outliers” in the dataset.

“Outlier” Group	RBC Tx (n)	RBC Tx (volume/kg)	Platelet Tx (n)	Platelet Tx (volume/kg)
(1) The NRBC was elevated (>5000/μL), but ROP did not develop (Stage 0 or 1) (n = 46)	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)
(2) The NRBC was normal (<2000/μL), but severe ROP developed (n = 6)	3 (2, 6)	58 (44, 126)	0 (0, 3)	0 (0, 71)
P - Value	0.002	0.003	0.032	0.030

NRBC nucleated red blood cell, ROP retinopathy of prematurity, RBC red blood cells, Tx transfusion, n number, kg kilogram birth weight.

DISCUSSION

Our present analysis supports the conclusion reached in four previous reports [1–4]; namely, an elevated NRBC count at birth is associated with an increased risk that ROP will occur. In making this connection, we issue the caveat that a statistical association is not always equivalent to a causative association [36]. In fact, we suspect that a high number of NRBC in the blood at birth does not, itself, cause ROP. Rather, we predict that the high NRBC count is a biomarker of antecedent fetal hypoxia, and that severe hypoxia in utero is the culprit, perhaps through the mechanism of disrupting retinal vascular growth [37–39].

Our data, and our statistical modeling, indicates that among preterm births, the higher the NRBC count is at birth, the more likely it is that ROP will develop. After controlling for gestational age, birth weight, and mechanical ventilation at 14 days of life, this association remained robust. However, our interpretation must be moderated by two observations. First, we sometimes saw that the NRBC count at birth was normal, yet the infant developed severe ROP. Thus, we conclude that an elevated NRBC count at birth is not necessary for the development of ROP. Second, we sometimes saw that the NRBC count at birth was elevated, yet the neonate did not go on to develop ROP. Thus, we conclude that an elevated NRBC count at birth is not sufficient for the development of ROP. The patterns we found suggest to us that the pathogenesis of ROP usually involves both a prenatal insult (such as severe fetal hypoxia) and a postnatal insult (such as excessive delivery of oxygen to the developing retina). However, sometimes only one insult (only prenatal or only postnatal) is identified.

We propose that in most cases of ROP, severe fetal hypoxia is the relevant prenatal insult, and repeated transfusion of adult red blood cells, raising the percentage of Hgb A and diminishing the percentage of Hgb F during the period of retinal vascular development, is one of the principal postnatal insults. Surely, other insults can be involved in ROP pathogenesis [40–43]. However, we hypothesize that identifying an elevated NRBC count at birth and avoiding adult RBCs transfusions in those patients might be one way to reduce ROP incidence and/or severity.

Routinely obtaining an NRBC count at birth could be accomplished at every delivery below a certain gestational age (such as below 29 weeks) using the method recently advocated by the AAP Committee of the Fetus and Newborn [5]. Specifically, blood could be drawn postnatally from the umbilical vein on a double-clamped cord segment. A key action statement in the AAP publication is that postnatal umbilical cord blood sampling can be used to reliably obtain a CBC and differential count. Doing this would permit the NRBC count at preterm birth to be considered, along with other factors, as part of an initial ROP risk assessment.

Avoiding RBC transfusions in preterm infants who are at risk for developing ROP could be accomplished, in part, by utilizing the combination of delayed umbilical cord clamping [44] and restrictive transfusion guidelines [45]. RBC transfusions could be reduced further by employing erythropoietic stimulating factors such as Darbepoetin along with supplemental iron [46, 47]. When, despite performing delayed cord clamping, restrictive transfusion practices, and ESA administration, an RBC transfusion is deemed necessary, it could be administered as fetal (umbilical cord blood) RBC as in the BORN and U-BET studies, thereby preserving the high concentrations of HbF during the period of retinal vascular development [21, 26, 48–50].

We recognize limitations in our report. For instance, in our database not every neonate born <32 weeks gestation had a NRBC count obtained in the first six hours after birth. Differences in the infants who did vs. who did not have an order for an immediate CBC could have introduced bias into the analysis. Also, some variation in NRBC counts might exist based on the laboratory equipment used for the measurement. Ours were all performed on Sysmex Hematology analyzers, but perhaps different analyzers at other centers could result in some differences in NRBC count. We also recognize the complexity of ROP pathogenesis [40–43], and we know that our studying only two elements, the NRBC count at birth and blood transfusions given, is overly simplistic. However, our analysis, in the context of the other studies that examined these two aspects of ROP pathogenesis, strengthens the case for routinely performing a NRBC measurement at the birth of all preterm neonates, and for considering ways to maintain the high physiological HgF levels of infants judged to be at risk for ROP, during their period of retinal vascular development.

We encourage NICUs to consider routinely measuring the NRBC count of infants born at very early gestational ages, where ROP can be problematic, and doing this utilizing otherwise discarded umbilical cord blood, as recommended by the AAP [5, 51]. We also encourage NICUs to establish transfusion stewardship groups to establish, support, and monitor restrictive RBC transfusion practices [34, 47]. Whether these two efforts will reduce ROP incidence or severity in a meaningful way can only be determined by large studies. However, the global burden of ROP, for society, families, and individuals warrants novel efforts aimed at decreasing the incidence and severity of this unfortunate complication of premature birth.

DATA AVAILABILITY

Inquiries regarding data access can be addressed to the corresponding author.

CODE AVAILABILITY

Computer code used for statistical analysis is available upon written request to the corresponding author.