Main

Cerebral hypoxia and ischemia are important causes of long-term morbidity among babies who have undergone neonatal intensive care. Because of this there has been much interest in studying the cerebral circulation during the neonatal period using a variety of methods. These studies have provided important insights into the mechanisms which control or influence CBF in health and disease. However, despite the information generated by these studies, they have had little impact on clinical management during neonatal intensive care. A possible reason for this is that most of the studies have concentrated on measuring CBF or changes in CBF. Although such measurements provide information about the rate of oxygen delivery, or changes in the rate of oxygen delivery, to the brain, they cannot give a complete picture of cerebral oxygenation because they do not take into account the cerebral oxygen requirement. Cerebral oxidative metabolism and energetics have been studied after perinatal asphyxia using magnetic resonance spectroscopy(1), but these techniques cannot be easily performed at the cotside during the course of intensive care. Attempts have also been made to study changes in cerebral oxygenation using NIRS to monitor changes in the cerebral concentration of cytochrome aa3(2). However, this technique has, as yet, found limited acceptance. There have been concerns about the differences in the algorithms used by different NIRS systems for measuring cytochrome aa3(3,4), and concerns have been raised that the changes observed after perinatal asphyxia may represent changes in the optical characteristics of the brain, rather than true changes in cytochrome aa3(5).

Clinical events and interventions during the course of neonatal intensive care have been shown to alter cerebral hemodynamics. As long as oxygen delivery remains sufficient for the brain's requirements, a reduction in perfusion is unlikely to be of any clinical relevance. In these circumstances, a reduction in the rate of CDo2 would not be accompanied by a reduction the rate of oxygen consumption as long as the level of cerebral activity did not alter significantly. A spontaneous reduction in CVo2 may be a consequence of reduced cerebral activity. However, reduced CVo2 due to a reduction in cerebral perfusion may indicate that oxygen delivery has been reduced to a level that does not meet cerebral oxygen demand. Such a reduction would represent true hypoxia and is likely to be of clinical relevance. The implication of this is that measurements of CVo2 may provide useful information about the adequacy of cerebral oxygenation.

Measurements of CVo2 seem to be of use when studying adults with cerebral ischemia. A range for CVo2 of 3.2 to 3.7 mL·100 g-1 min-1 is typically seen in resting healthy adults(612). Several studies in adults with cerebral ischemia have shown that rates of CVo2 of less than 1.3-1.5 mL·100 g-1 min-1 are strongly associated with permanent cerebral ischemic injury(13,14) and that areas of brain vulnerable to infarction are better identified by measurements of CVo2 than by measurements of CBF, cerebral blood volume, or cerebral oxygen extraction(14). The adult brain, however, is structurally and metabolically different from the neonatal brain as there are significant maturational changes within the brain during the last trimester. It is possible that the minimal level of CVo2 that applies to the adult brain may be inappropriate for the neonate. For example, Altman et al.(15) measured CVo2 in 11 babies using positron emission tomography and documented normal neurologic outcome in four babies with CVo2 values less than 1.3 mL·100 g-1 min-1. The aims of this study were to evaluate a noninvasive method for the assessment of CVo2 at the cotside using near infrared spectroscopy, and to investigate the relationship between gestational age and CVo2.

METHODS

CVo2 and CDo2 were calculated in micromoles·100 g-1 min-1 from: Equation 1,Equation 2 where CHbF is micromoles·100 g-1/min-1; Sao2 and CSvo2 are present.

To allow for comparison with previously published data, CVo2 was recalculated in milliliters·100 mL-1 min-1. Details of the derivation of Equations 1 and 2, and the method used to recalculate the CVo2 values in milliliters·100 mL-1 min-1 are explained in the "Appendix."

Sao2 was measured by pulse oximetry (Datex Satlite Trans, Datex, Finland). CSvo2 and CHbF were measured using NIRS (NIRO 500, Hamamatsu Photonics, UK).

NIRS was used to measure changes in cerebral concentrations of Hbo2 and Hb and total hemoglobin. The NIRS optodes were placed more than 4 cm apart in a frontotemporal or frontoparietal position determined by the size of the subject's head. A differential path length factor of 4.99 and a tissue density of 1.05 were used to quantify the data in micromoles/100 100 g(16). NIRS data were collected every half second. Pulse oximetry was performed with the oximeter operating in beat-to-beat mode. NIRS and pulse oximetry data were collected into a computer file using data logging software designed for the NIRO 500 (Onmain version 1.32). Subsequent analysis was performed using a computer spreadsheet package(Quattro Pro for Windows version 5.00, Borland International).

CSvo2 was calculated from the NIRS data obtained during partial jugular venous compression using a previously validated method(17). Unilateral partial jugular venous compression was applied for less than 10 s. CSvo2 was calculated from the NIRS data obtained during the first 5 s of the compression. Sao2 was taken as the mean of the 10 pulse oximetry measurements recorded during the first 5 s of venous compression. The NIRS data collected during each jugular venous compression were examined and were accepted as suitable for the calculation of CSvo2 if there was a stable baseline before the occlusion, a rise in the cerebral concentrations of both Hbo2 and Hb during the compression and a return to the same baseline after release of the compression.

CHbF (micromoles·100 g-1 min-1) was measured by NIRS using a technique in which a bolus of Hbo2 is delivered to the brain(18,19). The fraction of inspired O2 was increased to induce a rapid rise in Sao2. CHbF measurements were calculated from the data collected during the first 6 s of the rise in Sao2 and cerebral Hbo2 concentration. The NIRS and Spo2 data collected during each measurement of CHbF were also examined before the calculation of CHbF to determine their acceptability for calculation of CHbF. Spo2 data were accepted as suitable if Spo2 was less than 92% with a steady baseline before the measurement and there was a rise in Spo2 of greater than 5% over 6 s. NIRS data were accepted as suitable if there was a steady baseline for both Hbo2 and Hb before the measurement and the rise in Hbo2 was mirrored by a fall in Hb so that HbT remained constant through the study. CBF(milliliters·100 g-1 min-1) was calculated from the NIRS data by dividing the CHbF value by the arterial hemoglobin concentration(grams/mL) measured in each subject during each study.

Twenty babies were studied. A total of 48 CVo2 estimates were made with a median (range) of 2(15) estimates per subject. To assess the relationship with gestational age, the median value for CVo2, CDo2, and CSvo2 was taken for each subject. Correlations between gestational age and CVo2, CSvo2, and CDo2 were assessed by calculating Spearman's rank correlation coefficient (ρ). Correlations between CVo2 and both Paco2 and postnatal age were calculated using analysis of covariance to assess correlation within subjects(20), and a weighted correlation coefficient was calculated to assess correlation between subjects(21).

RESULTS

There were 9 male and 11 female infants studied. All were receiving intensive care. The median (range) gestation and birth weights were 27(2441) wk and 1.1 (0.5-3.6) kg, respectively. Eight babies were sedated with morphine (200 µg/kg i.v. six hourly), two babies were receiving pancuronium, and one baby was receiving phenobarbitone (2.5 mg/kg 12 hourly) for previous seizures. None was having clinically apparent seizures at the time of the studies. Three babies had abnormal cranial ultrasound appearances at the time of the studies, including one who had a parenchymal hemorrhage. Thirty-five of the 48 studies were performed in the first 2 d of life, five more studies were performed between d 3 and 7, and the other eight were performed before d 21.

For each CVo2 estimation, CSvo2, Sao2, and CHbF were measured repeatedly over a 30-min period. A median (range) of 5(215) attempted measurements of CSvo2 and 4(28) attempted measurements of CHbF were made during each study. Not all attempted NIRS measurements met the predetermined success criteria. The median (range) success rates for CSvo2 and CHbF measurements in each subject were 82% (20-100%) and 75% (22-100%), respectively. The median number (range) of successful measurements of CSvo2 was therefore 4(210), and the median number(range) of successful CHbF measurements was 3(25) in each study. Mean values of CSvo2, Sao2, and CHbF obtained during each study were used to calculate the CVo2 estimate for that study.

The reproducibility of the estimate was assessed by calculating the coefficient of variation of each component measurement during each estimation. The median (interquartile range) coefficients of variation for CSvo2, Sao2, and CHbF were 4.9% (3-7.6%), 1.1% (0.5-2.8%), and 20.6% (10.5-28%), respectively.

Table 1 shows the median (range) values for CHbF, CSvo2, Sao2, CBF, CDo2, and CVo2 from all 48 studies. There was a significant positive correlation between CVo2 and gestational age (n = 20, ρ = 0.55,p = 0.014) as shown in Figure 1. There was also a significant positive correlation between CDo2 and gestational(n = 20, ρ = 0.56, p = 0.012) as shown in Figure 2, but there was no significant correlation between CSvo2 and gestational age (n = 20, ρ =-0.041, p = 0.86).

Table 1 Median and range of each variable measured and the derived values for CDo2 and CVo2
Figure 1
figure 1

Plot of the relationship between median CVo2 from each individual and gestational age.

Figure 2
figure 2

Plot of the relationship between median CDo2 from each individual and gestational age.

There was no significant correlation between CVo2 and Paco2 either intersubject (r = 0.25, p > 0.1) or intrasubject(r = 0.5, p = 0.35). There was also no apparent correlation between CVo2 and postnatal age either intersubject(r = 0.1, p > 0.1) or intrasubject (r = 0.2,p = 0.33). There was no significant difference in CVo2 between the eight patients who were receiving morphine sedation and the 11 who were not receiving any sedation (p = 0.32, Mann-Whitney U test). The subject who was receiving phenobarbitone had a median value for CVo2 of 0.66 mL·100 g-1 min-1 and this is within the range of values seen in the other subjects. The values for CVo2 obtained in the three subjects with abnormal cerebral ultrasound appearances were also within the range of values seen in the other subjects and there was no significant difference between CVo2 in these three individuals and the rest of the subjects (p = 0.71, Mann-Whitney U test).

DISCUSSION

Previous studies of CVo2 in neonates have relied on invasive measurements with direct cannulation of the jugular bulb(22,23) or exposure to ionizing radiation during xenon-133 measurements of CBF(24) or positron emission tomography(15). We have described a noninvasive optical method for the estimation of CVo2 in neonates that can be performed at the cotside. The highest rates we observed were nine times greater than the lowest rates. We observed an increase in CVo2 with increasing gestational age, although there was only full-term baby in this study. There was no suggestion from this study that CVo2 was influenced by Paco2 or postnatal age, although the main aim of the study was not to test those hypotheses and it is possible that a study designed to address those specific questions may find otherwise. CVo2 was not obviously different in babies with abnormal cranial ultrasound scans but the number of babies with abnormal scans was small and it is likely that a larger study would be more conclusive. Similarly, CVo2 was not obviously different in babies who were receiving sedative drugs although the numbers of such babies in this study were too small for us to reach any definite conclusions about this.

The accuracy of the CVo2 estimate will be affected by a combination of the inaccuracies in each of the component measurements of Sao2, CSvo2, and CHbF. Each of these three separate measurements has been validated previously. Nearly all studies evaluating measurements of Sao2 by pulse oximetry in neonates have found pulse oximetry to be accurate to within ±2%(25). We have validated the method for measurement of CSvo2 by comparison with co-oximetry of blood from the jugular bulb and found close agreement, with a mean difference between NIRS and co-oximetry measurements of 1.5%(17). Measurements of CBF in neonates using NIRS have been validated by comparison with xenon-133 measurements(26,27). The measurement of CHbF is the measurement which caused the greatest inaccuracy in the estimate. The measurement shows the poorest reproducibility with a median coefficient of variation of 20.6%. This is similar to that noted in other studies which have used this particular NIRS method to measure blood flow(26,28,29).

The technique for CHbF measurement is not applicable to all sick neonates. To make the measurement it is necessary to induce a rapid rise in Sao2 by manipulation of the inspired oxygen concentration. This is not possible in babies with severe respiratory disease, or in babies who remain well oxygenated with low inspired oxygen concentrations. This is a major limitation of the technique. Alternative NIRS techniques for blood flow measurement, such as injecting indocyanine green(30) will be more widely applicable and may show better repeatability.

The quantification of NIRS data for the calculation of hemoglobin flow or blood flow requires accurate knowledge of the optical path length. This is calculated by multiplying the distance between the two NIRS optodes by a"differential path length factor" to account for scattering of light within the tissues. The differential path length factor of 4.99 used in this study was obtained from measurements made in neonates using phase resolved spectroscopy(16). However, other studies have measured the differential path length factor from the time of flight of picosecond pulses of light and have produced different, although similar, values(31,32). It is likely that the optical characteristics of each baby's brain will be different and whichever differential path length factor is used, it can be considered only as an estimate. This problem could be resolved by performing NIRS with real time measurements of optical path length which may be possible as the technology develops.

Ignoring dissolved oxygen undoubtedly leads to an underestimation of CVo2. Across the range of hemoglobin concentrations and oxygen tensions seen in the subjects of this study, less than 1% of the oxygen carried in the blood would be dissolved rather than bound to hemoglobin, and the underestimate will therefore be small.

The measurement of CSvo2 by NIRS with jugular venous compression is easily performed on neonates undergoing intensive care. It causes no apparent distress to the babies and no change in the heart rate or the arterial saturation as measured by pulse oximetry (our unpublished observations). Bilateral jugular venous occlusion has been used to measure CBF in these babies using strain gauge plethysmography and electrical impedance(3335). An increase in the risk of intracranial hemorrhage has not been reported with these methods. We have observed that the change in cerebral blood volume which occurs during unilateral jugular compression is less than that which occurs during an unsynchronized expiration by a mechanically ventilated infant. Abolition of such unsynchronized expirations by neuromuscular paralysis during mechanical ventilation has not been shown to reduce the risk of intraventricular hemorrhage. It is highly unlikely that the measurement of CSvo2 by NIRS with unilateral jugular venous compression will increase the risk of intraventricular hemorrhage.

The values for CVo2 obtained in this study are comparable with the few published measurements of CVo2 in neonates obtained by more invasive methods. Garfunkel et al.(22) measured CVo2 in 33 children including three term neonates with severe brain abnormalities. CVo2 in these babies was between 1.1 and 2.1 mL·100 g-1 min-1. Frewen et al.(23) measured CBF, Cao2, and CSvo2 from the internal jugular vein in a group of nine term babies with hypoxic ischemic encephalopathy. Although they did not report CVo2 values for individual subjects, CVo2 calculated from the mean values for Cao2, CSvo2, and CBF gave values of 1.7 mL·100 g-1 min-1 in normal survivors and 1.3 mL·100 g-1 min-1 in babies who died or survived with severe brain damage. The first published measurements of CVo2 in preterm neonates were those reported by Skov et al.(24) who used xenon-133 to measure CBF, and NIRS with head tipping to measure CSvo2. That group reported a value of 1.4 mL·100 g-1 min-1 in 10 asphyxiated term babies and a lower value of 1.0 mL·100 g-1 min-1 in nine preterm babies with hyaline membrane disease. Altman et al.(15) measured CVo2 using positron emission tomography and reported CVo2 to be between 0.4 and 1.3 mL·100 g-1 min-1 in six term babies, four of whom had hypoxic ischemic encephalopathy, and values of 0.06-0.54 mL·100 g-1 min-1 in five preterm babies.

Most of the CVo2 estimates obtained in this study were considerably lower than the value of 1.3 mL·100 g-1 min-1, which has been reported as the lower limit for brain viability in adults, and lower values were seen in babies with a lower gestational age. The reason for the neonatal values of CVo2 being lower than adult values is likely to be due to the changes in the structural complexity and functional activity of the brain which occur across the range of gestational ages studied. There is an enormous change in the architecture of the cerebrum, with an increase in the number of neurones and in the number of synapses(36). This increase in the structural complexity is accompanied by an increase in functional activity as evidenced by electroencephalography(37). It is not surprising then that the oxygen demand of the brain increases with increasing brain maturity during this period. Measurements of CVo2 in fetal sheep have shown a similar rise with increasing maturity(38). This is the first study to show such a relationship in the human in vivo, although Himwich et al.(39) described a similar relationship from in vitro work using minced brain samples obtained from recently dead fetuses and babies.

The demand for other metabolic fuels also increases during the process of cerebral maturation. In a recent positron emission tomography study, Kinnala et al.(40) demonstrated an increase in cerebral glucose consumption with increasing postconceptional age. In that study the mean value for the rate of cerebral glucose consumption between 32 and 34 wk was 5.5 µmol·100 g-1 min-1. In our study, the mean CVo2 for the four infants between 32 and 34 wk was 31.47µmol·100 g-1 min-1. The ratio of oxygen to glucose consumption calculated from these two values is 5.7 µmol of oxygen for each µmol of glucose. This is similar to the ratio reported in the adult human brain(4144). This calculation lends further support to our estimates of CVo2 being of the correct magnitude.

Examination of the published data suggests that CVo2 continues to increase through childhood. Figure 3 shows CVo2 measurements obtained from several studies in various different groups of older children(15,2224,4548). The details of these studies are shown in Table 2. Most of the children in these studies were anesthetized or sedated, and many were undergoing intensive care; there are few published data from healthy, stable children. The value of 5.17 mL·100 g-1 min-1 obtained by Kennedy and Sokoloff(48) from a group of children with a median age of 6.1 y is higher than the values of 3.2-3.7 mL·100 g-1 min-1 typically seen in healthy adults(612). The values obtained by Kennedy and Sokoloff were from invasive methods, including cannulation of the jugular bulb, in unsedated children in whom cooperation was achieved by allowing them to watch "moving pictures." The anxiety caused by the procedures or the increased arousal from watching the moving pictures may have artificially elevated the CVo2 in these children. However, Chugani et al.(49) have demonstrated that cerebral glucose consumption increases through early childhood to a maximum value at about 8-9 y of age and subsequently falls slowly with the approach of adulthood. From this observation it may be expected that CVo2 would show a similar pattern of change with age through childhood, and the high values reported by Kennedy and Sokoloff may be representative of this. Later in adult life there is evidence that CVo2 decreases with aging after the age of 20 y(50,51).

Figure 3
figure 3

Plot of mean CVo2 against mean age from studies measuring CVo2 in children. The open circle represents this study, closed circles represent data from the studies in Table 2.

Table 2 Details of studies reporting measurements of CVo2 in children

There was an increase in CDo2 with increasing gestational age, but no change in CSvo2. This suggested that the increased oxygen demands of the maturing brain were met by increasing the rate of oxygen delivery, rather than by increasing the proportion of available oxygen which is extracted from the circulation.

We have shown the feasibility of estimating CVo2 in sick neonates at the cotside using noninvasive, optical methods. The values obtained were similar to those obtained, by more invasive techniques, in other studies and are in agreement with values which we would expect from the known rate of cerebral glucose consumption in neonates. There is an increase in CVo2 with increasing brain maturity and it is likely that this increase continues through childhood. The increasing oxygen demands of the brain with maturation are met by an increase in CDo2.