Effects of Inspired Hypoxic and Hypercapnic Gas Mixtures on Cerebral Oxygen Saturation in Neonates with Univentricular Heart Defects

After obtaining institutional review board approval (Children's Hospital of Philadelphia, Philadelphia, PA), informed parental consent was obtained. This study was conducted between June and October 1999 at the Children's Hospital of Philadelphia. Eligibility criteria included a diagnosis of univentricular heart defect with Sao²greater than 80% (evidence of Q̇p/Q̇s > 1.0). Exclusion criteria were postnatal age greater than 1 month, preoperative seizures, associated craniofacial anomalies that precluded near-infrared spectroscopy monitoring, and participation in an investigational drug study. The study was conducted in the operating room after induction of anesthesia and before start of stage 1 Norwood surgery.

Per institutional practice, prostaglandin infusion was discontinued immediately before transport to the operating room. No premedication was administered. After placement of electrocardiogram, pulse oximeter, and arterial pressure monitors, the subjects received fentanyl citrate (20 μg/kg bolus, then 1 μg · kg⁻¹· h⁻¹) and pancuronium bromide (0.2 mg/kg), were nasally intubated, and mechanically ventilated to normocapnia with 21% Fio²using a Servo SV 300 ventilator (Siemens-Elema, Solna, Sweden) without positive end-expiratory pressure. Heart rate, systolic arterial pressure, diastolic arterial pressure, and mean arterial pressure (MAP), Sao²and nasopharyngeal temperature were continuously monitored during the study.

To deliver 17% Fio²and 3% Fico², nitrogen and carbon dioxide, respectively, were added to the inspiratory limb of the ventilator circuit. Nitrogen (size H tank) flow was adjusted to obtain 17% Fio²as measured in the inspired limb (MaxO²oxygen analyzer; Ceramatec, Salt Lake City, UT). Carbon dioxide (size E tank) was adjusted to achieve a partial pressure of 20 mmHg in the inspired limb (equivalent to 2.8% CO², with a water vapor pressure of 47 mmHg at 37°C; measured using Nellcor Ultracap; Nellcor Inc., Pleasanton, CA). No changes were made to the minute ventilation during the study period.

The near-infrared cerebral oximeter used in this study (NIM Incorporated, Philadelphia, PA) was a prototype 3 wavelength frequency-domain device. 10This device uses laser diodes at measuring wavelengths of 754, 785, and 816 nm with an internal reference wavelength at 780 nm. The emitted light is sinusoidally oscillated at 200 MHz, and the phase-shift and intensity of the detected light relative to the emitted light were monitored by heterodyne frequency domain technology. Fiberoptic bundles mounted in soft rubber housing (optical probe) delivered the light to and from the head, and emitter and detector were separated by 3 or 4 cm. The probe, placed on the forehead below the hairline, monitors Sco²located in the frontal cerebrum; the scalp and skull do not contribute to the optical signal. 7,10The main unit housing the electronic hardware sends data to a computer for storage and analysis. Sco²is calculated from the phase shift signals. 10Instrument precision relative to cooximetry is 6% from 0 to 100%. Near-infrared spectroscopy Sco²represents predominantly oxygen saturation in the venous blood. 11 

The protocol consisted of two treatment periods with three baseline periods (before and after each treatment). Treatments were 17% Fio²and 3% Fico². Arterial blood gases, arterial pressure, saturation, and cerebral saturation were obtained at the end of each period. Arterial blood gases were drawn from an indwelling umbilical arterial catheter in all subjects. The sequence of treatment was assigned randomly from a previously generated chart. Each treatment was maintained for 10 min, followed by a 10–20-min baseline period.

Sample size calculation (n = 16) was based on a power of 0.8 and α= 0.01 to detect a 20% change in Sco²from baseline with treatment (3% Fico²or 17% Fio²). Data are presented as mean ± SD. Significance was set at 0.01 after Bonferroni correction for multiple comparisons.

The primary outcome measure was Sco². Secondary outcome measures included heart rate, MAP, Sao², p  H, arterial carbon dioxide tension, arterial oxygen tension, and base excess. To examine the effect of period (time effect) and the order of treatment (carryover effects), independent two-sample t  tests were conducted (SPSS Inc., Chicago, IL). In the absence of time and carryover effects, and differences between baseline and the two treatments were then examined by paired t  tests for parametric data (data were normally distributed). Sco²values during the last 1-min of each period (baseline, 3% Fico², 17% Fio²) were averaged by the computer to obtain the representative Sco²value for that period. Change in Sco²was the difference between treatment and the average of the baselines before and after the treatment.

Of the 16 patients enrolled, data from 15 were included in the analysis. One neonate was excluded for protocol violation (error in administering 3% Fico²). Table 1shows the demographic data of 15 subjects. Before the study, in the intensive care unit, all neonates had received prostaglandin infusion (0.025–0. 05 μg · kg⁻¹· min⁻¹), 10 had received dopamine (3–5 μg · kg⁻¹· min⁻¹), 13 had received digoxin and lasix, and 4 had inspired a hypoxic gas mixture (17–19% Fio²) by spontaneous ventilation in a hood. There were 12 full-term birth infants and 3 prematurely (< 37 weeks) born infants. None of the infants had neurologic abnormalities by history or physical examination.

During the study, 8 of 15 subjects received hypoxic mixture first followed by 3% Fico². Because neither the order of treatments (crossover effect, P = 0.16) or time (period effect, P = 0.45) was statistically significant, data are presented by condition (i.e. , Fico², Fio², baseline) rather than by treatment order (baseline, 3% Fico², baseline, 17% Fio², baseline; or baseline, 17% Fio², baseline, 3% Fico², baseline). Mean data from the 15 subjects are shown in table 2, and the change from baseline (δ) in the MAP and Sco²for each subject is shown in table 3.

Heart rate, systolic arterial pressure, and Sao²did not change significantly with either Fico²or Fio²(table 2). With 3% Fico², diastolic arterial pressure and MAP increased significantly by 12 and 10%, respectively. During 17% Fio², diastolic arterial pressure and MAP did not change significantly (P = 0.02 and 0.11, respectively). With 3% Fico², p  H decreased and arterial carbon dioxide tension increased as intended, and arterial oxygen tension increased (P < 0.01); although the base excess increased, it was not significant (P = 0.09). With 17% Fio², p  H increased and arterial oxygen and carbon dioxide tensions decreased (although no alterations were made to the minute ventilation, adding nitrogen increased the minute ventilation;P < 0.01); the decrease in base excess was not significant (P = 0.3). Nasopharyngeal temperature remained unchanged during the study (36 ± 0.4 vs.  35.98 ± 0.4°C, start vs.  end), as did hematocrit (42 ± 6 vs.  42.6 ± 6%, start vs.  end).

Figure 1shows a typical tracing of Sco²during the study in one subject, and table 4shows the results of the 15 subjects. Sco²increased significantly with 3% Fico²(P < 0.01), whereas it did not change with 17% Fio²(P = 0.85). In 8 of 15 subjects, 17% Fio²decreased Sco², the largest decrease being 6.5%. By comparison ,  3% Fico²increased Sco²in all subjects, the largest increase being 26%. At baseline, in two subjects Sco²was less than 40% (37 and 38.5%, respectively). In one of these subjects, Sco²decreased to 32% with 17% Fio², whereas Sco²increased to 48% with 3% Fico². In the other subject, Sco²did not change with 17% Fio², but increased 10% during 3% Fico²administration.

Increase in Sco²began 2 ± 0.8 min after administration of 3% Fico², and the increase was linear at a rate of 0.075 ± 0.05%· mmHg CO²⁻¹· min⁻¹(r²= 0.68). With the discontinuation of Fico², Sco²returned to baseline by 8 ± 1.5 min. The relation between change in Sco²and MAP during Fico²was not significant (r²= 0.27, P = 0.048). The Sco²response to 3% Fico²in the premature (10 ± 3%) and full-term birth (13 ± 6) infants was similar.

There were no complications from the study.

In this group of neonates with SV heart defects, we found that Sco²and arterial pressure increased with 3% Fico², whereas these parameters did not change significantly with 17% Fio². These observations suggest that 3% Fico²provides better preoperative hemodynamics and cerebral oxygenation than 17% Fio²in this patient population.

Previous studies conducted in immature animal models of SV heart defects reported increases in pulmonary vascular resistance and a decrease in pulmonary to systemic blood flow ratios (Q̇p:Q̇s) with 10% Fio²and 5% Fico². 12However, no cerebral oxygenation or vital organ blood flow data exist in either animal model or human neonates. In anesthetized, mechanically ventilated human neonates with SV heart defects, Tabbutt et al.  13compared the impact of 17% Fio²and 3% Fico²on systemic oxygen delivery after the same protocol as our study. 13They found that both treatments decreased Sao²and Q̇p:Q̇s similarly, although 3% Fico²increased systemic oxygen delivery, whereas 17% Fio²had no significant impact. These data suggest that 3% Fico²provides better systemic hemodynamics and oxygenation than 17% Fio². Our observations support these conclusions. However, lower inspired oxygen concentrations, as in the animal models, might provide a similar hemodynamic result as 3% Fico², with or without changes in cerebral oximetry.

Cerebral oximetry is an emerging technology to noninvasively monitor brain tissue oxygenation at the bedside. It is particularly applicable to the critically ill neonate and infant population, where the thin extra-cranial tissues do not interfere with brain monitoring, and in whom diagnosis of cerebral hypoxia–ischemia is otherwise problematic. At present, the instrumentation is based on continuous-wave or frequency-domain technologies. Continuous-wave devices have been commercially available for several years. They can monitor changes in Sco²over time but cannot determine baseline levels accurately. Frequency-domain devices (used in this study), a new technology using cellular phone technology, can accurately determine baseline Sco²as well as changes over time. It should be commercially available in the near future. However, the clinical utility of cerebral oximetry remains to be determined. Our study serves as an example of how the technology might be used clinically.

Cerebral oximetry and pulse oximetry differ in several respects. Although both use near-infrared light intensity signals, pulse oximetry monitors the pulsatile signal component reflecting the arterial circulation. Cerebral oximetry monitors the nonpulsatile signal component reflecting the gas-exchanging tissue circulation (capillaries, venules, arterioles), of which approximately 85% of the signal appears to originate from small venules. 11Pulse oximetry often fails with poor perfusion as the pulsatile signal diminishes. Cerebral oximetry is not susceptible to this failure, although it is subject to motion artifact like pulse oximetry. The critical cerebral oxygen saturation that results in brain damage remains uncertain. Studies in animal models suggest that the risk of brain damage increases as Sco²decreases to less than 40%, because electroencephalogram activity begins to slow, adenosine triphosphate decreases, and cytochrome aa³becomes reduced; these physiologic changes inevitably lead to neuronal necrosis. 14In our study, Sco²was less than this 40% threshold in two subjects at baseline (incidence, 2 of 15 [13%]), which decreased further with 17% Fio²in one subject, whereas Sco²increased above the threshold in both subjects with 3% Fico². Based on this limited experience, 17% Fio²might increase the risk of cerebral hypoxia–ischemia before surgery in this patient population.

Cerebral oxygen saturation reflects a balance between cerebral oxygen delivery and cerebral oxygen consumption. Cerebral oxygen delivery is a product of cerebral blood flow (CBF) and arterial oxygen content. The latter depends on Sao²and hemoglobin. We did not measure hemoglobin, but during the study period there were no volume shifts; hence, use of hematocrit should be appropriate. During the study, Sao²and hematocrit remained constant; thus, a change Sco²resulted from changes in CBF or cerebral oxygen consumption. Hypercapnia and hypoxia in the magnitude used in our study do not alter cerebral oxygen consumption, 15nor do they alter the proportion of arterial to venous blood in the brain. 11Thus, changes in Sco²most likely reflect changes in CBF.

Treatment with 3% Fico²increased Sco². The decrease in p  H with hypercapnia would lead to a right shift in the oxygen dissociation curve (Bohr effect), which would lead to a decrease in Sco²and not an increase. There was a small (3 mmHg) but significant increase in arterial oxygen tension with 3% Fico²treatment; however, this cannot explain the 12% increase in Sco²we observed. Hence, it is most likely that hypercapnic cerebral vasodilatation led to an increase in CBF with an increase in Sco².

Healthy adult volunteers breathing hypoxic mixtures (7–11% Fio²) during isocapnic conditions showed decreases in Sco². 16We had therefore expected treatment with 17% Fio²to lead to a decrease in Sco². However, the lack of change in Sco²reflects the fact that cerebral oxygen delivery was unchanged. Hypoxia can lead to cerebral vasodilatation, but 17% Fio²was an insufficient stimulus to evoke this. For example, in adult volunteers, CBF did not change with 18% Fio², and it increased only 5% in response to 16% Fio². 17 

Cerebral blood flow may also increase if the MAP increases above the limits of autoregulation. The upper lower limit of autoregulation is uncertain in human neonates but is approximately 90 mmHg in neonatal animals. 18,19In our study, mean arterial pressure increased 10% in response to 3% Fico²but was still within the limits of autoregulation. Hence, an increase in CBF, and hence Sco², on this basis was unlikely.

There are several limitations in our study that might not allow generalization of our findings to care of neonates with SV in the intensive care unit. Our subjects were anesthetized and mechanically ventilated. Their baseline and posttreatment Sao²and Sco²were greater than that of nonanesthetized, spontaneously breathing subjects. It is possible that the response to hypoxic or hypercapnic gas mixtures differs in nonanesthetized, spontaneously breathing subjects with lower Sao²and Sco². Despite treatment with 3% Fico²and 17% Fio², the Sao²remained greater than 90% in our subjects ,  indicating continued excessive pulmonary blood flow. Use of higher Fico²or lower Fio²might have reduced Sao²further. Time constraints did not permit us to test this hypothesis. Time constraints also limited each baseline and treatment period to 10–20 min. Previous work has shown the CBF response to hypercapnia and hypoxia to be completed by 5 and 6 min, respectively, with return to baseline by 1–2 min. 17,20Hence, treatment duration in our study should have been adequate to observe a response. However, chronic exposure (hours) might desensitize the cerebral response to hypoxia and hypercapnia.

In conclusion, we demonstrated that controlled ventilation with 3% Fico²increased the cerebral oxygen saturation as well as arterial pressure, whereas controlled ventilation with 17% Fio²maintained arterial pressure but did not change Sco²in most subjects, suggesting these treatments provide hemodynamic stability but affect cerebral oxygenation differently. The impact of 17% Fio²on cerebral oxygen saturation during spontaneous ventilation, prolonged administration of 17% Fio²and 3% Fico², as well as higher Fico²remain to be studied.