Objective Preterm infants often need red blood cell (RBC) transfusions. The aim of this study was to determine whether haemoglobin levels before transfusion were associated with regional cerebral tissue oxygen saturation (rcSO2) and fractional tissue oxygen extraction (FTOE) and whether RBC transfusions were associated with rcSO2 and FTOE during the 24-h period thereafter.
Design Prospective observational cohort study.
Setting Third level neonatal intensive care unit.
Patients Thirty-three preterm infants (gestational age 25–34 weeks, birth weight 605–2080 g) were included.
Main Outcome Measures RcSO2 was measured during a 1-h period, before, 1 h after and 24 h after a 15 ml/kg RBC transfusion in 3 h. Using rcSO2 and transcutaneous arterial oxygen saturation (tcSaO2) values, FTOE was calculated: FTOE=(tcSaO2−rcSO2)/tcSaO2.
Results Forty-seven RBC transfusions were given. RcSO2 and FTOE correlated strongly with haemoglobin before transfusion (r=0.414 and r=−0.462, respectively, p<0.005). TcSaO2 did not correlate with haemoglobin before transfusion. 24 h after transfusion, rcSO2 increased from a weighted mean of 61% to 72% and FTOE decreased from a weighted mean of 0.34 to 0.23. The decrease in FTOE was strongest in the group with haemoglobin below 6.0 mmol/l (97 g/l). The decrease in FTOE was already present 1 h after transfusion and remained unchanged at 24 h after transfusion.
Conclusion Following RBC transfusion, cerebral tissue oxygen saturation increases and FTOE decreases. The data suggest that cerebral oxygenation in preterm infants may be at risk when haemoglobin decreases under 6 mmol/l (97 g/l).
Statistics from Altmetric.com
There is considerable controversy in neonatology as when to transfuse preterm infants with anaemia with red blood cells (RBC). As a result of frequent blood sampling and an immature haematopoietic system, preterm infants become progressively anaemic.1 In an effort to limit the risks associated with RBC transfusion, many neonatal units have adopted more restrictive guidelines for transfusing preterm infants.1 Two recently published randomised clinical trials both found that patients received fewer RBC transfusions when using restrictive guidelines.2 3 The first study randomly assigned 100 preterm infants with birth weights of 500–1300 g to either a restricted transfusion group, with haematocrit thresholds between 0.22 and 0.34, or a liberal transfusion group, with haematocrit thresholds between 0.30 and 0.46.2 In each group, transfusion thresholds decreased with improving clinical status. The authors reported a decrease in transfusions in the restricted group: from 5.2±4.5 (mean±SD) in the liberal group to 3.3±2.9 in the restricted group. The second study3 randomly assigned 451 preterm infants below 1000 g to either a restricted transfusion group, with thresholds between 68 and 115 g/l (consistent with haematocrit between 0.23 and 0.35),4 or a liberal transfusion group, with thresholds between 77 and 135 g/l (consistent with haematocrit between 0.26 and 0.41). Transfusion thresholds decreased with improving clinical status and advancing postnatal age. The authors reported little evidence of clinical benefit (death or major morbidity) for either approach, with the low threshold group receiving fewer transfusions (mean 4.9 vs 5.7).3 However, both studies produced conflicting results with regards to certain major neurological adverse events.4 In the first study, adopting relatively high cut-off points for haematocrit levels in the control group, the incidence of cerebral haemorrhage and cystic periventricular leucomalacia increased significantly in the restrictedly transfused group.2 In the other study no such differences were found, but adverse outcomes (death or major morbidity) were as frequent as 70% in both groups, and the difference regarding haemoglobin levels between the liberal and restricted groups was minimal.3 Previously, it was demonstrated that apparently stable anaemic preterm infants may be in a clinically unrecognised high cardiac output state.5 In this state, cerebral oxygen delivery may be at risk. There are some indications that insufficient oxygen delivery to the brain might be a mechanism involved in cerebral haemorrhage and periventricular leucomalacia.2 6
What is already known on this topic
▶. There is considerable controversy in neonatology as when to transfuse preterm infants with anaemia with RBC.
▶. Apparently stable anaemic preterm infants may be in a clinically unrecognised high cardiac output state, in which cerebral oxygen delivery may be at risk. Insufficient oxygen delivery to the brain might be a mechanism involved in cerebral haemorrhage and periventricular leucomalacia in the preterm infant.
What this study adds
▶. If haemoglobin levels are below 6.0 mmol/l (97 g/l), rcSO2 is low and FTOE is high in preterm infants. Following a RBC transfusion, cerebral oxygenation (rcSO2 and FTOE) improves quickly.
▶. Our findings indicate that cerebral oxygenation in preterm infants may be at risk when haemoglobin levels decrease under 6.0 mmol/l (97 g/l).
It is difficult to measure cerebral oxygenation non-invasively. A new, non-invasive method that could be useful is near infrared spectroscopy (NIRS).7 NIRS measures regional tissue oxygen saturation.8,–,11 This measure is thought to reflect the oxygen saturation in a mixed vascular bed dominated especially by venules. The fractional tissue oxygen extraction (FTOE) is calculated on the basis of the values for the regional tissue oxygen saturation and the transcutaneous arterial oxygen saturation (tcSaO2).9 11,–,14 FTOE reflects the balance between cerebral oxygen supply and cerebral oxygen consumption, and may thus serve as an indicator of cerebral hypoxia and ischaemia.15 16 In the case of anaemia, the supply of oxygen may be impaired, resulting in a lower tissue oxygen saturation and a higher FTOE.
The values for the regional cerebral tissue oxygen saturation (rcSO2) and FTOE in preterm infants with anaemia are largely unknown. The aim of this study was therefore to determine whether rcSO2 and FTOE were associated with the haemoglobin level, and if so, at which level. Second, we were interested to see if a RBC transfusion influenced the values for rcSO2 and FTOE during the 24-h period thereafter. We hypothesised that as a result of reduced cerebral oxygen supply, rcSO2 will be lower and FTOE will be higher in infants before RBC transfusion, and that the transfusion will lead to a higher rcSO2 and lower FTOE.
For the purpose of this study we consecutively selected 33 preterm infants, who had been admitted to the neonatal intensive care unit of the University Medical Center Groningen between December 2006 and August 2007. The selection criteria were a gestational age of less than 35 weeks and need for a RBC transfusion, according to our local guidelines (table 1). We excluded infants with major chromosomal or congenital abnormalities. The cut-off levels for haemoglobin in our guidelines range from 5.0 to 8.0 mmol/l (81–129 g/l). They are close to the cut-off levels reported in the restricted groups of previous randomised controlled studies: Bell et al2 reported cut-off levels between 4.6 and 7.0 mmol/l (74–113 g/l), and Kirpalani et al3 reported cut-off levels between 4.8 and 7.2 mmol/l (77–116 g/l).
For blood transfusions on our unit, a single donor approach was used. The RBC aliquots were stored for a maximum of 2 weeks, had a haematocrit of 0.60 and were preserved with saline, adenine, glucose and mannitol. The administered RBC transfusion volume was 15 ml/kg in 3 h. The guidelines were followed as closely as possible, but the final decision for a RBC transfusion was made by the attending neonatologist.
Informed parental consent was obtained in all cases. The study was approved by the local institutional review board.
Near infrared spectroscopy
We used the INVOS 4100 monitor (Somanetics Corporation, Troy, Michigan, USA) in combination with the paediatric somasensor to obtain rcSO2 values. This technology is based on the fact that biological tissues are relatively transparent to near infrared light (600–900 nm wavelength). The optical sensor measures the quantity of reflected light photons as a function of two wavelengths (730 and 805 nm), and determines the spectral absorption of the underlying tissue.14 17 NIRS differentiates oxygenated haemoglobin from deoxygenated haemoglobin, which each have distinct absorption spectra. The ratio of oxygenated haemoglobin to total haemoglobin reflects the regional oxygen saturation of tissue. The somasensor has two detectors, at 3 cm and at 4 cm distance from the near infrared optode. The detector placed at 3 cm from the optode receives light scattered predominantly from the scalp and skull. The detector placed at 4 cm receives light scattered from the scalp, skull and cerebral tissue. Therefore, by subtraction the two detectors measure the oxygen saturation of the underlying cerebral tissue. From previous studies it is estimated that the depth of the signal is at least between 15 and 20 mm, enough to reach the cortical grey and white matter of the infants.18 19
For this study, we placed the optical sensor on the left frontoparietal region of the infant's head and held it in place using elastic bandaging. Fifteen minutes were allowed for stabilisation of the measurement. For each RBC transfusion, we measured the rcSO2 three times for 1 h at a time. The first measurement was made immediately before (t=0), the second 1 h after the completion of the RBC transfusion (t=1) and the third 24 h thereafter (t=24). Simultaneously, we measured tcSaO2 by pulse oximetry. We calculated FTOE using the equation FTOE=(tcSaO2−rcSO2)/tcSaO2.11 13 14 The mean values for tcSaO2, rcSO2 and FTOE were calculated during the 1-h period of measurement, and this mean was used as a single value for further analysis.
Repeatability of the rcSO2 measurements, using the same device as we did, is reported to be stable, with limits of agreement less than 6%.20 In our hands, repeatability of the rcSO2 measurements after refixation of the optode, and allowing 10 min to stabilise rcSO2, was similar, with a mean difference of 4.5% (range 1–8%) between three consecutive measurements of the same patient (EA Verhagen and AF Bos, unpublished data). Previous studies have demonstrated that FTOE in preterm infants range from 0.15 to 0.40 during the first weeks after birth.11 14 20 21
In all infants, we collected haemoglobin and haematocrit values before RBC transfusion and 24 h after transfusion. We prospectively collected data with regard to the perinatal and neonatal characteristics that might influence haemodynamics. These included gestational age, birth weight, patency of the ductus arteriosus and ventilatory status. Simultaneously to rcSO2 and tcSaO2 measurements we recorded the infants' heart rate, respiratory rate, blood pressure and blood gas values.
SPSS 16.0 software for Windows was used for most statistical analyses. The mean and median values for rcSO2 and FTOE, along with the other variables, were calculated for the 1-h periods of measurement: t=0, t=1 and t=24. The Spearman rank order correlation test (two-tailed) was used to determine correlations between the haemoglobin concentration before transfusion and NIRS parameters, and between the absolute rise in haemoglobin following transfusion and NIRS parameters.
Next, we categorised the infants into three groups, on the basis of their initial haemoglobin concentration, following the threshold levels of our guidelines (table 1): group A, haemoglobin level before transfusion <6.0 mmol/l (<97 g/l); group B, haemoglobin level before transfusion 6.1–7.0 mmol/l (98–113 g/l); group C, haemoglobin level before transfusion 7.1–8.0 mmol/l (114–129 g/l). In order to detect differences in time, and between the three categories of haemoglobin level before transfusion, we built a multilevel model in which NIRS measurements (level 1) were nested within subjects (level 2), thereby taking dependency between measurements into account. We used this model to test differences between means. To test for differences between an estimated mean and the intercept one uses a t test.22 To test for differences between two estimated means one tests the contrast of the sum of the parameters from which each estimate is derived, using a χ2 test with 1 degree of freedom. Results were reported including 95% CI. We used MLwiN 2.11 (University of Bristol, Bristol, UK) for these multilevel statistical analyses.
To investigate whether other clinical conditions confounded our results, we tested the correlations between NIRS parameters, blood pressure and partial pressure carbon dioxide (Pco2) with the Spearman rank correlation test. We used the Mann–Whitney U test to test whether patency of the ductus arteriosus and ventilatory status at the time of measurement confounded the NIRS parameters before and after transfusion. We also tested whether patency of the ductus arteriosus and ventilatory status were evenly distributed among the three categories of haemoglobin level before transfusion, using the χ2-for-trend test. A p value less than 0.05 was considered significant.
Thirty-three infants were included in this study. Their gestational age ranged from 25 to 34 weeks (median 27.3 weeks) and their birth weights ranged from 605 to 2080 g (median 1010 g). In 14 infants a RBC transfusion was given twice. The study sample therefore consisted of 47 instances of RBC transfusion, which was applied between 1 and 93 days after birth (median 17 days), at a postmenstrual age between 25.9 and 39.0 weeks (median 30.1 weeks). The haemoglobin values before transfusion ranged from 3.7 to 7.9 mmol/l (60–128 g/l), with the median 6.9 mmol/l (111 g/l). Haematocrit values ranged from 0.19 to 0.36 (median 0.31). In 11 transfusions, the infants did not require respiratory support, in 18 transfusions, the infants were on continuous positive airway pressure and in 18 transfusions the infants were ventilated. In 17 transfusions, the infants required additional oxygen, but in only three infants was the inspiratory fraction of oxygen more than 0.30.
The course of tcSaO2, rcSO2 and FTOE in relationship to haemoglobin
Before transfusion median tcSaO2 was 93% (range 85–99), median rcSO2 was 60% (range 28–82%) and median FTOE was 0.35 (range 0.07–0.70) (table 2). RcSO2 and FTOE correlated strongly with the haemoglobin levels before the transfusion (Spearman's r=0.414, p=0.004 and r=−0.462, p=0.001, respectively; figure 1). TcSaO2 did not correlate with the haemoglobin levels. RcSO2 and FTOE values after the transfusion did also not correlate with haemoglobin levels after the transfusion.
We categorised the infants receiving RBC transfusions into three groups, following the threshold levels of our guidelines. When taking an arbitrary cut-off point for FTOE at 0.40 (figure 1B), six of nine infants in group A (67%), five of 21 infants in group B (24%) and three of 17 infants in group C (18%) had higher FTOE values (χ2-for-trend test 5.5, p=0.019).
The courses of rcSO2 and FTOE following RBC transfusion in the three groups are shown in figures 2 and 3. Differences were tested by a multilevel model in which rcSO2 and FTOE measurements (level 1) were nested within subjects (level 2). Following the transfusion, rcSO2 increased (figure 2A) and FTOE decreased (figure 2B) in each group. These changes were already present at 1 h after the transfusion (average increase in rcSO2 9%, 95% CI 7 to 11, p<0.001; average decrease in FTOE −0.10, 95% CI −0.12 to −0.08, p<0.001). This differed significantly between group A compared with group B, with an average difference in rcSO2 of 8.4% (95% CI 2.0 to 14.8) p=0.01, and near-to-significant in FTOE of −0.06 (95% CI −0.13 to 0.00) p=0.06, but not between group A compared with group C (for rcSO2, p=0.17; for FTOE, p=0.31) and between group B and group C (for rcSO2, p=0.15; for FTOE, p=0.31). There was no significant subsequent change after the next 23 h (average change 1–24 h in rcSO2 −1.1%, 95% CI −3.6 to 1.4, p=0.37, and in FTOE=−0.02, 95% CI −0.04 to 0.01, p=0.26) and this did not differ between haemoglobin groups for rcSO2 and FTOE, respectively (group A versus B, p=0.75 and p=0.70; group A versus C, p=0.53 and p=0.54; and group B versus C, p=0.68 and p=0.76).
Next, we analysed whether there were any differences between the three groups at each time point (t=0, t=1 and t=24). At t=0, rcSO2 was lower in group A compared with B and C (average difference 12%, 95% CI 8 to 17, p<0.001), whereas there was no significant difference between groups B and C (B on average 2.8% higher (95% CI −1.9 to 7.6) p=0.24). At t=1, rcSO2 was near-to-significantly lower in group A compared with groups B and C (average difference 5.4%, 95% CI −0.1 to 10.9, p=0.054), but not at t=24, p=0.15. At all three time points, rcSO2 values did not differ between groups B and C, p=0.24, p=0.70 and p=0.95, respectively. FTOE values were higher in group A compared with groups B and C at all three time points (average differences: t=0, 0.13 (95% CI 0.08 to 0.17) p<0.001; t=1, 0.08 (95% CI 0.02 to 0.13) p=0.009; t=24, 0.06 (95% CI 0.00 to 0.11) p=0.04). There were no significant differences at any time point between groups B and C (p=0.19, p=0.83 and p=0.58, respectively).
Twenty-four hours after the RBC transfusion, rcSO2 had increased from a weighted mean 61–72% and FTOE had decreased from a weighted mean of 0.34–0.23. At t=24 FTOE was still above 0.40 in two infants, both were from group A. Both infants were born at 25 weeks, with a birth weight of 730 and 900 g, respectively. One of them was transfused at postnatal day 30, with haemoglobin 5.7 mmol/l (92 g/l) before transfusion, and haemoglobin 6.9 mmol/l (111 g/l) 24 h later, the other was transfused at postnatal day 90, with haemoglobin 5.3 mmol/l (85 g/l) before transfusion and haemoglobin 7.8 mmol/l (126 g/l) 24 h later. In the first infant, there were no clinical factors identified that could account for the higher FTOE, in the second infant a concurrent septicaemia with a coagulase-negative staphylococcus species was diagnosed on the same day as the transfusion.
The change in FTOE following a RBC transfusion correlated strongly with the absolute rise in haemoglobin level. This correlation was already present 1 h after RBC transfusion (Spearman's r=−0.320, p=0.039), but was strongest at 24 h after RBC transfusion (Spearman's r=−0.466, p=0.002).
The relationship between rcSO2 and FTOE and the clinical variables
As several clinical conditions may influence cerebral haemodynamics and oxygenation, we investigated whether these conditions confounded the differences found in rcSO2 and FTOE. We checked blood pressure, Pco2, patency of the ductus arteriosus and ventilatory status during the period of measurement. The clinical characteristics of the groups A, B and C are shown in table 3. We found no differences between the three groups in any of the clinical variables. All intraventricular haemorrhages were already diagnosed before the occurrence of anaemia. The ductus arteriosus was patent in seven children at the time of transfusion, and closed at the time of the other 40 measurements. RcSO2 and FTOE before and following transfusion were not associated with the patency of the ductus arteriosus, or with ventilatory status. Mean blood pressure and Pco2 did not correlate with the simultaneously measured rcSO2 and FTOE. In the absence of significant associations of potential confounders, we refrained from multivariate regression analyses.
The present study indicated that rcSO2 decreased and FTOE increased in preterm infants who were to receive a RBC transfusion according to our rather restrictive transfusion guidelines.4 Following a RBC transfusion, cerebral oxygenation improved quickly. Further improvement during the following 24 h did not occur. However, in infants with low haemoglobin levels before transfusion (haemoglobin <6.0 mmol/l, haemoglobin <97 g/l), cerebral FTOE was still higher 24 h after the transfusion, compared with the infants who had a higher haemoglobin level (>6.0 mmol/l, >97 g/l) before the transfusion. This was present despite haemoglobin levels that were no different between groups after the transfusion. From our findings we speculate that cerebral oxygenation in preterm infants may be at risk when haemoglobin levels decrease under 6.0 mmol/l (97 g/l).
The increased cerebral oxygen extraction can be explained by a decreased oxygen transport capacity, as a result of the lower haemoglobin level. It is possible for preterm infants to compensate for a low haemoglobin level by increasing cardiac output.5 23 Along with the higher cardiac output, cerebral blood flow will also increase,24,–,26 but this increase might be absent or limited in cases of mild anaemia.27 28 In fact, our data suggest that the increase in cerebral blood flow is insufficient to compensate for the decreased oxygen transport capacity. As the FTOE reflects the balance between oxygen supply and oxygen consumption, this means that oxygen delivery to the brain may be at risk in cases of anaemia. It is interesting to note that changes were detectable 1 h after transfusion, and did not increase further at 24 h after the transfusion. This adds support to the hypothesis that increasing the fraction of haemoglobin A and therefore tissue oxygen delivery might be an important factor in the benefits of neonatal blood transfusion.
Our data support the notion that adherence to a too restrictive RBC transfusion policy in preterm infants may be harmful. When compared with studies that have investigated liberal versus restrictive guidelines, our transfusion guidelines are rather restrictive.2 3 In the randomised study that found significant differences in neurodevelopmental outcome,2 the liberal guidelines in the sickest infants allowed the haematocrit level to be as low as 40% (in our study it was approximately 36%) and in the more stable infants 30% (we had 26%).4
A few other studies have used NIRS techniques to determine the course of oxygenation parameters following RBC transfusions in preterm infants.26 27 29,–,31 Some of those studies were performed several years ago, when haemoglobin levels were not allowed to get as low as in the present study. Despite this, the results of those studies also indicated that oxygen extraction in peripheral and cerebral tissue improved following transfusion.26 27 29,–,31 Nevertheless, the studies did not provide a cut-off point for haemoglobin at which cerebral oxygen delivery may be at risk.
Our study may have implications for clinical practice. NIRS can play a role in determining whether cerebral oxygenation is at risk in cases in which the haemoglobin level approaches the level for RBC transfusion according to one's guidelines. Those infants with a high cerebral FTOE might benefit from an earlier RBC transfusion than strictly required according to one's guidelines. This has been studied previously measuring peripheral tissue oxygen extraction of the forearm with NIRS, and taking a cut-off point of 0.47 for giving a blood transfusion.30 These peripheral fractional oxygen extraction measurements, however, failed to identify many of the infants felt by the clinicians to require blood transfusions.30 It might be that cerebral FTOE, rather than peripheral fractional oxygen extraction, is a more sensitive predictor of the need for a transfusion. This requires further study.
There are several limitations to our study. Being a single centre study, our results may not be applicable to the general population. Our study group is relatively small, and it may be that clinical variables that we are not aware of also contributed to the variation in rcSO2 and FTOE. We did not measure cerebral blood flow. It is difficult to obtain reliable indices of cerebral blood flow, and we refrained from measuring blood flow velocity by Doppler ultrasound. Doppler flow measurements in, for example, the middle cerebral artery could have given an extra dimension to this study, as high values would underscore the presence of haemodynamic significant anaemia, and also indicate the presence of a ductal steal in the case of patent ductus arteriosus.
The values we found for rcSO2 and FTOE showed a wide range. This finding is confirmed by various other studies6 11 14 32 and points to a large interindividual variation. We stress the fact that we did not identify any clinical variables other than haemoglobin related to this variation. A final limitation is that we did not include long-term follow-up in the present study. The prevalence of cerebral haemorrhages in our group was low and there were no children with periventricular leucomalacia. However, our study is definitely underpowered to find clinically relevant diverse outcomes in these groups as a result of the haemoglobin level before transfusion.
In conclusion, the present study indicates that cerebral oxygenation in preterm infants may be at risk when haemoglobin levels decrease below 6 mmol/l (97 g/l). Following a RBC transfusion, cerebral oxygenation improved quickly (within 1 h). Further improvement during the 24 h following did not occur. However, in infants with low haemoglobin levels before transfusion (haemoglobin <6.0 mmol/l, haemoglobin <97 g/l), cerebral FTOE was still higher at 24 h after the transfusion, compared with the infants who had higher haemoglobin levels (haemoglobin >6.0 mmol/l, haemoglobin >97 g/l) before the transfusion. Our findings may have implications for the treatment of anaemia with RBC transfusions.
This study was part of the research programme of the postgraduate school for Behavioural and Cognitive Neurosciences, University of Groningen. The authors are grateful to K Van Braeckel for statistical advice.
Competing interests None.
Patient consent Obtained from the parents.
Ethics approval This study was conducted with the approval of the local ethical review board.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.