We hypothesize that a complete blood count (CBC) with manual differential from umbilical cord blood is equivalent to a CBC with manual differential obtained from the neonate on admission.
A CBC and manual differential was performed on 174 paired umbilical cord blood and admission blood samples from infants <35 weeks gestation. Paired t-test and Pearson's correlation coefficient were the primary statistical tools used for data analysis.
Cord and admission blood white blood cell (WBC) count, hemoglobin and platelet count all significantly (P<0.0001) correlated with paired neonatal samples (R=0.82, 0.72, 0.76). Admission blood WBC count fell within the variation of WBC count values from currently accepted neonatal admission blood sources. Cord blood hemoglobin was not clinically different than admission hemoglobin (1.0 g dl−1). Cord blood platelet counts were not different from admission blood platelet counts (5800 cells per μl, P=0.23). The immature to total granulocyte ratio was not different between samples (P=0.34).
Umbilical cord blood can be used for admission CBC and differential in premature infants.
The limits of neonatal viability have decreased over the past decades.1, 2 Surfactant and antenatal steroids have improved survival rates dramatically.3 Many centers commonly resuscitate infants at 23 weeks gestation and some resuscitate infants at 22 weeks gestation.4, 5, 6, 7, 8 Decreased gestational age is associated with decreased circulating blood volume. Total circulating blood volume has been estimated to be between 63 and 100 ml kg−1 at birth.4, 5, 6
Approximately 30 000 infants <28 weeks gestation are born annually in the United States.7 Typical admission blood tests require 1.5 to 4 ml of blood. This volume may represent up to a 10% iatrogenic blood loss for the most premature infants. Eliminating the complete blood count (CBC) would decrease the initial blood loss by 0.5 ml and be one step toward minimizing initial neonatal blood loss.
A recent study by Hansen et al.8 found cord blood to be an acceptable replacement source for evaluation of sepsis in full-term babies at risk for infection. Additional recommendations for using umbilical cord blood for sepsis evaluation have been proposed.9 To date, no literature exists on paired comparison between umbilical cord blood and neonatal admission CBC with manual differential (CBC/diff) analysis in premature infants.
Minimizing blood draws remains critically important in all premature infants, especially in extremely low birth weight infants, those with a birth weight <1000 g. Umbilical cord blood as an alternative source for admission laboratory studies may provide physicians with necessary information while nearly eliminating direct neonatal blood draws on admission. Validating common laboratory studies from umbilical cord blood among premature infants is crucial to translating this concept to the bedside. Therefore, we hypothesize that a CBC/diff obtained from umbilical cord blood immediately after delivery is equivalent to a CBC/diff obtained from the infant upon admission to the neonatal intensive care unit.
We conducted a cross-sectional study, in which we collected paired samples of umbilical cord blood in the delivery room (cord blood) and direct neonatal blood collected upon admission (admission blood) for a CBC/diff. To calculate sample size, the study was powered to test the hypothesis that hemoglobin, white blood cell (WBC) count and platelet count were significantly different when drawn from the cord blood as compared with admission blood. Our study sought to reject this hypothesis, therefore demonstrating that cord blood is an acceptable replacement source for admission CBC/diff in premature infants. To show a 1.0 K cell per μl difference in WBC count, a 0.8 g dl−1 difference in hemoglobin, and a 40 K cell per μl difference in platelet count with an α<0.05 and 90% power a sample size of 119, 169 and 44, respectively, was required. Clinically equivalent thresholds for each cell line were determined a priori by considering the coefficient of variation for the laboratory instrument and a non-scientific survey of neonatologists at our institution.
All infants born at The Ohio State University Medical Center from June 2009 to February 2010 with a gestational age less than 35 weeks were eligible for enrollment. Enrollment was continuous until sample size was achieved. Parental plan for cord blood banking was the only exclusion criteria. The study was approved by the Institutional Review Board of the participating institution.
Eligible neonates delivered either by vaginal delivery or cesarean section were cared for by the neonatal resuscitation team as per routine following clamping and cutting of the umbilical cord. Early cord clamping was the routine practice during the study period. The placenta was then delivered by the obstetrician. The obstetric nurse obtained 0.5 ml of umbilical venous cord blood via direct sampling with a sterile needle and syringe. When blood could not be obtained from the umbilical vein, attempts were made to draw blood from a proximal placental vein on the fetal surface of the placenta. The blood was immediately transferred into an EDTA microtainer (BD, Franklin Lakes, NJ, USA).
Admission blood CBC/diff was obtained by the neonatal intensive care unit treatment team. The source and time of collection were documented in the physician and nursing notes. The sample (0.5 ml) was placed into an EDTA microtainer (BD).
Both cord blood and admission blood CBC/diff results were obtained by the inpatient laboratory on the same instrument (Beckman Coulter AcT diff2, Brea, CA, USA). Manual differential was performed by laboratory personnel. Maternal and neonatal demographics, timing of blood draws and sepsis diagnosis were abstracted from maternal and neonatal charts and laboratory records. Presence of premature rupture of membranes, preterm labor and oligohydramnios, were collected from antepartum and intrapartum labor and delivery records. Chorioamnionitis was determined by placental pathology confirming the histological diagnosis or, in the absence of placental pathology, when the obstetric chart indicated the clinical diagnosis. Funisitis was not abstracted. The ratio of immature to total granulocytes (I:T ratio) was calculated by dividing the percentage of immature (band+myelocytes+metamyelocytes) by total (band+myelocytes+metamyelocytes+neutrophils) granulocytes.
Continuous paired parametric data were analyzed using paired t-test, Pearson's correlation, and simple and stepwise multivariable linear regression. Continuous unpaired parametric data were analyzed using student t-test, whereas unpaired proportions were analyzed using a two-sample test of proportion. Unpaired non-parametric data were analyzed using the Wilcoxon rank-sum test. Paired dichotomous data were analyzed using McNemar's test. Paired non-parametric data were analyzed using Wilcoxon signed-rank test. ANOVA was used for parametric categorical data. Statistical analysis was completed by using the Stata-IC 11.0 statistical package (StataCorp, College Station, TX, USA).
We enrolled 174 infants during the study period among 368 eligible infants. The majority of those not enrolled were because of lack of an attempt to collect cord blood (Figure 1). Each infant enrolled had paired samples of cord blood and admission blood sent to the laboratory for CBC/diff analysis. Demographics were only different with respect to antenatal steroid administration between enrolled and un-enrolled patients. There was no difference among remaining neonatal (birth weight, gestational age, gender, apgar scores, umbilical catheterization) or maternal (delivery type, preterm labor, premature rupture of membranes, oligohydramnios, maternal group B strep colonization, pre-eclampsia, maternal diabetes) characteristics. The median time to obtain the umbilical cord blood sample was 4 min (IQR: 0 to 10 min), while the median time to obtain neonatal admission blood samples was 68 min (IQR: 53 to 92 min).
A WBC count and hemoglobin was obtained on all paired samples. A platelet count was obtained on 166 paired samples (eight unable to be determined due to platelet clumping). In all, 12 admission samples and 21 cord blood samples contained fibrin strands. These samples were excluded from WBC and platelet analysis,10, 11 leaving 142 paired WBC samples and 141 paired platelet samples for analysis. Admission blood was drawn from umbilical arterial catheter (UAC, n=45), umbilical venous catheter (UVC, n=68), radial artery (n=29) or capillary sampling (n=28). There were four samples with an undocumented admission blood source.
The mean difference in WBC count between cord blood and admission blood was 2.4 K cells per μl (Table 1). There was a high correlation (Figure 2) between cord blood and admission blood WBC counts (r=0.82, P<0.0001). One-way ANOVA demonstrated significant differences in mean WBC count stratified by admission blood source (P<0.001). The difference between admission blood and cord blood varied by gestational age, only after stratifying by admission blood source. There was no confounding by birth weight, time to admission labs, 1- or 5-min apgar, chorioamnionitis or antenatal steroid administration.
Admission band, neutrophil and lymphocyte counts were significantly higher than paired cord blood samples (P<0.001). However, the basophil count decreased slightly in the admission blood differential, whereas there was no significant difference in monocyte, eosinophil, myelocyte or metamyelocyte counts.
The I:T ratio was not different between umbilical cord blood samples and admission blood samples (P=0.34). There were 19 infants with an elevated I:T ratio from admission blood. Six of these 19 infants also had an elevated I:T ratio from cord blood. There were 12 infants with an elevated I:T ratio from cord blood. Six of these 12 infants also had an elevated I:T ratio from admission blood. Only one patient with paired samples had culture-positive early-onset sepsis. Both admission blood and umbilical cord blood I:T ratios were <0.20 in this patient.
There was significant correlation (Figure 3) between cord blood and admission blood hemoglobin (r=0.72, P<0.0001). The mean difference in cord blood as compared with admission blood hemoglobin was 1.0 g dl−1. Capillary admission samples were 2.9 g dl−1 higher than cord blood samples. Radial arterial, UVC and UAC samples were 0.7, 0.4 and 0.6 g dl−1 higher than cord blood hemoglobin samples, respectively. There was no confounding by gestational age, birth weight, time to admission labs, 1- or 5-min apgar score, chorioamnionitis or antenatal steroid administration. Only one cord blood sample had a hematocrit 65. The paired admission sample was also 65. There were two capillary admission samples that were 65. The paired cord blood samples were both <65.
The platelet count was not significantly different (Table 1) between the cord blood samples and admission blood samples (6 K cells per μl, P=0.23). The interpair difference did not vary by admission blood source (P=0.18). There was a high correlation (Figure 4) between cord blood samples and admission blood samples (r=0.76, P<0.0001). There was no confounding by gestational age, birth weight, time-to-admission labs, 1- or 5-min apgar, chorioamnionitis, or antenatal steroid administration. There were seven infants with admission platelet count 100. Five of these also had a cord blood platelet count 100. Both paired cord samples that were >100 were paired with capillary admission samples. However, no clotting or fibrin was noted in either of these admission samples. There were 14 cord blood samples with a platelet count 100. Five of these were also 100 on paired admission blood sample, whereas nine paired admission blood samples were >100.
In this study, CBC/diff results from umbilical cord blood were compared with paired direct neonatal admission blood. Admission blood WBC count obtained via capillary sampling demonstrated the greatest interpair difference compared with cord blood (4.6 K cells per μl), whereas that from the UVC was the smallest interpair difference (1.2 K cells per μl). The admission blood hemoglobin was 1.0 g dl−1 higher than the paired cord blood hemoglobin. This difference was further minimized to 0.6 g dl−1 after eliminating hemoglobin obtained from a capillary source. There was no statistical or clinical difference between cord blood and admission blood platelet count (6 K cells per μl).
Hansen et al.8 demonstrated high correlation between cord blood and admission blood CBC and I:T ratio in term neonates. Other studies have reported CBC values from cord blood of term neonates without paired admission blood samples, and found the values fell within the normal neonatal admission reference range.12, 13 Our results in preterm infants are consistent with these findings.
Obtaining admission laboratory studies is necessary to provide optimal patient care. Blood loss from laboratory studies is associated with increased transfusion rates in extremely low birth weight infants.14 Recent trials on delayed cord clamping and umbilical cord milking in preterm infants have shown decreased intraventricular hemorrhage, increased admission hematocrit and decreased transfusions.15, 16, 17, 18
Our results demonstrate an increase in admission blood WBC count and band, segmented neutrophil, and lymphocyte counts as compared with paired cord blood samples. Neutrophils increase rapidly reaching their peak at about 12 h of life in term infants and 18 h in preterm infants.19, 20 We speculate the increase in admission blood WBC count compared with cord blood WBC count is due to this natural increase in WBC count in preterm infants. Because of the low variability in the time between cord blood and admission blood draws (median 68 min, interquartile range 53 to 92 min), we were unable to show an increase in interpair WBC count difference with respect to an increase in time-to-admission blood draw from birth. Previous studies evaluated the effect of sample source on WBC count and demonstrated higher leukocyte and neutrophil concentrations in capillary specimens when compared with arterial samples.21, 22 Our data are consistent with these findings.
The difference we report between cord blood and admission blood WBC count (2.4 K cells per μl) is less than the difference between currently accepted neonatal blood sources for admission CBC. Standardizing admission WBC count to a single source may provide more reliable values for interpretation by the treatment team. We suggest that umbilical cord blood is the most widely available, accessible and reliable source to be used as a single source for admission WBC count. Adopting a strategy of using cord blood for admission labs would have resulted in 13 infants not receiving antibiotics who would have received them compared with admission labs and six infants receiving antibiotics that otherwise would not have, if the only criteria was an elevated I:T ratio. In practice, however, the I:T ratio was rarely the only factor in antibiotic initiation.
When stratified by sample source, capillary hemoglobin was significantly higher than paired cord blood hemoglobin. It is well known that hematocrit and hemoglobin values are 10–25% higher when the sample is obtained from a capillary source as compared with a central source.22, 23, 24, 25 Moreover, the difference in capillary and venous hematocrit is more pronounced for premature infants.24, 26 We report no difference between hemoglobin from cord blood and admission blood obtained from a UVC, UAC or radial artery. These results are consistent with previously published studies, which reported unpaired cord blood hemoglobin values within normal neonatal reference values.27, 28
We demonstrated admission blood hemoglobin to be 0.6 g dl−1 higher than paired cord blood hemoglobin after eliminating capillary specimens. Christensen et al.29 evaluated over 30 000 hematocrit values and reported a 3.6% increase in hematocrit in the first four hours of life among infants 35 weeks gestation, no change in the hematocrit among infants from 29 to 34 weeks gestation, and a 6% fall in hematocrit among infants 28 weeks gestation. Our data were limited to infants <35 weeks gestation. Therefore, the slight increase in hemoglobin in admission blood as compared with cord blood does not seem to be explained by a natural increase in hemoglobin in our patient population. However, we do not consider this difference to be clinically significant. Adopting the strategy of using cord blood for admission labs would have resulted in two ‘missed cases’ of polycythemia. However, both these cases were determined to be false positives when central hematocrit was measured. In essence cord blood measure of hemoglobin/hematocrit seems to be more accurate than neonatal capillary testing of these measures.
In our study, there was no difference in platelet count between cord blood and admission blood. Admission blood source did not have any effect on the interpair difference in platelet count. However, among admission blood samples a higher percentage of capillary specimens had fibrin strands (29%) than radial artery (6.9%), UVC (1.5%) or UAC (2.2%) specimens. Several studies report conflicting results regarding variation in platelet count by source.22, 30, 31 We observed a statistically insignificant (P=0.11) number of discordant pairs (8/141), in which the cord blood platelet count was low (<100 K cells per μl), whereas the paired admission platelets were normal. Adopting a strategy of using cord blood for admission labs would result in a clinically important number of false-positive cases of thrombocytopenia. Nine of the 14 cases of thrombocytopenia (100) from cord blood had normal platelet counts from neonatal blood. Although a minority of infants demonstrated thrombocytopenia from cord blood, any thrombocytopenia from cord blood labs should likely be repeated from direct neonatal sampling.
The infants who would benefit most from admission laboratories being obtained from cord blood, infants <1000 g, had the lowest enrollment rate. Among infants less than 27 weeks gestation, only 33% were successfully enrolled (18/54). This was lower than the enrollment rate of infants between 27 and 31 weeks gestation (47/85, 55%) and those >31 weeks gestation (110/227, 48%). Clotted cord blood samples (5/54, 9.3%) and inability to obtain cord blood despite an attempt (7/54, 13%) remained a challenge among infants less than 27 weeks. However, five of 10 infants less than 24 weeks gestation including three of the four most premature infants were successfully enrolled. Experience and improved techniques may improve the ability to obtain cord blood samples in the most premature infants.
Umbilical cord blood is an acceptable replacement source for admission CBC/diff in preterm infants. The difference between cord blood and admission blood WBC count is less than the difference between admission blood capillary and arterial sample group means. Umbilical cord blood hemoglobin and platelet counts are not significantly different than admission blood hemoglobin and platelet values compared with our a priori threshold. However, owing to a clinically important rate of false-positive thrombocytopenia, cases of thrombocytopenia from cord blood should be repeated from direct neonatal sampling.
Utilization of umbilical cord blood for initial laboratory evaluation of premature infants may prevent iatrogenic blood loss from these infants and yield results more quickly. Studies validating additional admission laboratory studies from cord blood are necessary to further limit admission laboratory blood loss. Once validated, randomized controlled trials should be conducted to assess whether umbilical cord blood, when used for admission laboratory tests in premature infants, can improve neonatal outcomes.
We would like to acknowledge the obstetric nursing staff of The Ohio State University Medical Center, for umbilical cord blood collection, Philip Binkley, MD, MPH, Courtney Lynch, PhD, MPH and Juli Richter, MD, PharmD for reviewing the manuscript before submission. This study was funded through an intramural grant provided by Nationwide Children's Hospital.