Arterial and Venous Contributions to Near-infrared Cerebral Oximetry

After obtaining approval from the Institutional Review Board at the Children’s Hospital of Philadelphia and informed parental consent, we studied 20 children undergoing cardiac catheterization for diagnostic or therapeutic purposes. Inclusion criteria included age less than 8 yr and a diagnosis of congenital heart disease requiring cardiac catheterization with general anesthesia and mechanical ventilation. Exclusion criteria included known structural neurologic or craniofacial disease and anemia (hemoglobin < 8 g/dl). Children were enrolled from the Day Medicine Clinic or inpatient care units and were studied from August 1997 to March 1999.

Anesthetic management was conducted according to our institutional practice. Premedication included pentobarbital (4 mg/kg orally) for age 6 months to 1 yr and pentobarbital and meperidine (3 mg/kg orally) for age 1–8 yr. The anesthetic was either (1) propofol (3 mg/kg then 200 μg · kg⁻¹· min⁻¹), (2) propofol–fentanyl (propofol 2 mg/kg then 125 μg · kg⁻· min⁻, fentanyl 5 μg/kg then 2 μg · kg⁻¹· h⁻¹), (3) fentanyl–midazolam (fentanyl as before, midazolam 0.2 mg/kg then 0.2 mg · kg⁻¹· h⁻¹), or (4) isoflurane (1% expired)–fentanyl (1 μg · kg⁻¹· h⁻¹). All children received mechanical ventilation through an endotracheal tube. Tracheal intubation was facilitated by intravenous pancuronium (0.1 mg/kg). Lidocaine 1% was infiltrated into the catheterization site. End-expired carbon dioxide, pulse oximetry (N-200; Nellcor Puritan Bennet, Pleasanton, CA), electrocardiogram (Hewlett-Packard, Andover, MA), arterial pressure (Dinamap; Critikon, Tampa, FL), and rectal temperature were monitored. No blood products, vasodilators, or vasoconstrictors were administered during the study.

At present, cerebral oximetry instrumentation is based on continuous-wave, time-domain, or frequency-domain technologies. In continuous-wave technology, the instrument emits light at constant intensity (I^o) and detects this light whose intensity is attenuated (I) after passing through the head (fig. 1A). Sco²is determined from the intensity changes through an expression based on the principles of the Beer Law: where L represents the path length of light through the tissue, C the concentration of the compounds (usually oxyhemoglobin, deoxyhemoglobin, and water), ε the extinction coefficient of the compounds, and n the number of compounds and wavelengths of the light being used. Because path length through the tissue is uncertain, continuous-wave instruments are semiquantitative, limited to monitoring changes in Sco²over time (baseline Sco²cannot be determined). In frequency-domain technology (fig. 1B), the instrument emits light whose intensity is oscillated at high frequency. The light passing through the tissue is amplitude demodulated and phase shifted relative to the emitted light. Sco²is determined from amplitude and phase, which correspond to intensity changes and path length, respectively. Time-domain instruments use a pulse light source, require expensive hardware, and are not commercially viable in the foreseeable future. Continuous-wave and frequency-domain instruments are viable, although only frequency-domain instruments are absolutely quantitative.

In our study, Sco²was measured with a prototype fdNIRS (PMD, NIM Incorporated, Philadelphia, PA). 12Briefly, the instrument uses laser diodes at measuring wavelengths of 754 nm, 780 nm, and 816 nm, with a reference wavelength at 780 nm that is not directed through the sample. The laser light intensities are oscillated at 200 MHz. The instrument uses heterodyne frequency-domain technology to monitor phase shifts at the three measuring wavelengths relative to the internal reference. The instrument is calibrated before use against an external standard. Fiberoptic bundles mounted in soft rubber housing (optical probe) deliver the laser light to and from the subject’s head. The distance separating the emitter and detector fiberoptic is 3 or 4 cm. A computer captures the phase signals at 2/s and signal averages them over 15 s. Sco²is calculated from amplitude and phase signals according to an algorithm. 12Instrument precision and bias relative to blood oximetry is 6% and −2%, respectively, from 0% to 100% saturation. 12 

Cerebral oximetry views a banana-shaped tissue volume between the emitter and detector located approximately 2 cm deep to the surface. 9,10In young children, the thin extracerebral tissues do not contaminate the measurement of Sco²from the surface of the head. 10,12,13With the probe on the forehead, the cerebral oximeter appears to monitor both gray and white matter in the frontal neocortex.

After completion of the catheterization, the cardiologist advanced the catheter located in the systemic venous circulation into the left or right jugular vein until the tip was at the level of the jugular bulb, confirmed by fluoroscopy. The catheter located in the systemic arterial circulation was positioned with its tip in the aorta. The cerebral oximeter optical probe was applied to the forehead at midline below the hairline.

Subjects experienced three conditions selected in random order: baseline, 100% inspired oxygen, and hyperventilation. Baseline was normocapnia with inspired oxygen at the preoperative concentration (usually room air). One hundred percent inspired oxygen was also at normocapnia. Hyperventilation was achieved by increasing tidal volume or ventilatory rate to decrease end-tidal carbon dioxide by 10–15 mmHg, while returning inspired oxygen to the baseline concentration. The protocol required approximately 40 min, with each condition achieved in approximately 3 min, followed by 10-min steady state ventilation. Sco²at the condition was taken as the average value over the last minute, during which arterial and jugular bulb samples were drawn. Jugular bulb samples were drawn slowly. 17 

Demographic and physiologic data were recorded. Mean arterial and systemic venous pressure, Sao², Sjo², Sco², and arterial blood gases were recorded during each condition. A subject was defined as normoxic  if baseline Sao²was > 95%.

Data are presented as mean ± SD. Comparisons between conditions or groups were made by analysis of variance. When a significant overall F was found, pairwise multiple comparisons were made using Tukey’s test. Linear regression was used to determine the relation between Sco²and Sjo², Sao², and Swo², where Swo²was defined as Swo²=(0.25)Sao²+(0.75)Sjo²Swo²was used in other studies and was included in the current study for historical comparisons. 2,3,14The contribution of Sjo²and Sao²to Sco²for each subject–condition was determined by solving paired equations for α and β, Sco²=α Sao²+β Sjo²1 =α+β where α and β represent the fraction of arterial and venous blood in tissue and Sjo², Sao², and Sco²were the measured values for the subject–condition. Equations 3 and 4 assume all blood in tissue exists in either arterial or venous compartments. Arterial/venous ratios that were negative were not included in the analysis but were noted as “undeterminable.” A one sample t  test was used to validate the hypothesis that the α/β ratio is different from the 25/75 ratio. Pearson’s correlation coefficients were calculated between Sco²and demographic, physiologic, and arterial and jugular venous blood ratio variables. Multivariable regression was explored between Sco²and the variables having correlation coefficients with 0.01 level of significance to adjust for the multiple correlation tests. Bias and precision of fdNIRS Sco²relative to Sjo², Sao², and Swo²were calculated. 3 

Table 1lists the study subjects. The subjects’ age ranged from newborn to 6 yr, with 12 males and 8 females, 16 being white and 4 black. Fifteen subjects had complex congenital heart disease, among which 13 had single-ventricle disease. Ten subjects received medications for heart failure and 2 received prostaglandin infusions for ductus arteriosus-dependent lesions. Of the 20 subjects, 14 successfully completed the protocol, while 4 completed two conditions and 2 completed one condition. Reasons for incomplete data on all conditions included inadequate hyperventilation (n = 4) or malfunction of the jugular bulb catheter (n = 3) or cerebral oximeter (n = 3). Arterial/venous ratios were undeterminable in four instances (three subjects) because Sjo²was greater than Sco²(Sjo²vs.  Sco²: 85%vs.  73%, 84%vs.  58%, 96%vs.  88%, and 60%vs.  45%).

Figure 1displays a representative Sco²tracing (subject 9) during the study protocol. Table 2presents the physiologic data in all subjects and in the normoxic subjects. In all subjects, Sco², Sao², and Sjo²increased significantly from baseline to 100% O², whereas from baseline to hyperventilation, Sao²increased and Sjo²decreased, whereas Sco²did not change significantly. In normoxic subjects, Sco²decreased significantly with hyperventilation, whereas Sjo²tended to decrease (P = 0.15).

Of the demographic and physiologic variables in all subjects, only Sao²and Sjo²correlated significantly with Sco²(r = 0.61, P < 0.01; r = 0.76, P < 0.001, respectively). Multivariable regression found a significant relation: Sco²= 0.46 Sao²+ 0.56 Sjo²− 17 (R = 0.71, P < 0.03). Other variables, including age, hemoglobin, type of cardiac lesion and anesthetic, and mean arterial and systemic venous pressure, did not significantly correlate with Sco²(r = 0.1–0.3, all P > 0.2).

Accordingly, we focused on the relation of Sco²to Sao², Sjo², and Swo². Sco²was linearly related to Sao²(fig. 3), Sjo²(fig. 4), and Swo²(fig. 5). Sao²ranged from 68% to 100%, Sjo²from 27% to 96%, Swo²from 37% to 98%, and Sco²29% to 92%. Cerebral oximetry bias and precision were, respectively, 1.4 and 11.4 relative to Sjo², −30 and 14.2 relative to Sao², and −3.2 and 5.4 relative to Swo². As noted by the biases and intercepts of the lines, Sco²was less than Sao², greater than Sjo², and close to Swo². FIGURE 

Because NIRS Sco²was closest to Swo², we examined in greater detail the arterial and venous contribution to cerebral oximetry (table 3). Cerebral venous blood (Sjo²) accounted for the majority (≈ 85%) of NIRS-monitored blood. Although the arterial/venous ratio did not vary significantly in individual subjects with changing conditions (all comparisons, P > 0.2), significant variability existed between subjects (P < 0.001). For example, in 10 subjects (4 were normoxic), NIRS monitored essentially cerebral venous blood (arterial/venous ratio ≈ 0/100), whereas in 2 subjects (1 was normoxic), it monitored a more equal mixture (arterial/venous ratio ≈ 40/60). In the remaining 8 subjects, the ratio fell between these values. No demographic or physiologic factor (e.g. , age, hemoglobin, arterial or systemic venous pressure, or Sao²) significantly correlated with the arterial/venous ratio (r = 0.05–0.3, all P > 0.2). The cerebral arterial/venous ratio in all subjects at baseline was significantly different from the 25:75 ratio (P = 0.03).

More than 20 yr ago, Jobsis 18demonstrated the feasibility of monitoring cerebral oxygenation noninvasively with near-infrared light. The method, which became known as NIRS, was qualitative because light scattering by tissue precluded knowledge of optical path length necessary to solve the Beer Law equation. Current commercially available NIRS monitors have the same problem, although they use empirical calibration or a differential path length factor to make them “semiquantitative,” describing relative changes in cerebral oxygenation from an unknown baseline. 1–8,14fdNIRS, a new class of instrumentation, uses frequency-domain technology to determine optical path length to make it absolutely quantitative. 13,15In a validated in vitro  brain model, fdNIRS was found accurate against blood oximetry. 13,16A validated in vivo  model to test NIRS accuracy does not exist. Previous human and animal studies 2,3,14used a weighted average of Sao²and Sjo²in a fixed ratio of 25:75 (Sao²:Sjo²) for comparison with NIRS Sco². However, this weighted average has not been validated for the cerebral circulation in vivo .

In the present study, we investigated the relation of fdNIRS Sco²to a number of physiologic variables in infants and young children, including Sao²and Sjo², to test the weighted average in vivo  model. Of the variables examined, only Sao²and Sjo²significantly correlated with Sco². NIRS Sco²fell between Sao²and Sjo², closer to Sjo²than Sao². The arterial/venous contribution to NIRS Sco²averaged 85% venous and 15% arterial, not differing differ significantly between normoxia, hypoxia, and hypocapnia. However, this arterial:venous ratio differed significantly among subjects and from the 25:75 arterial:venous ratio. Thus, our results do not validate this fixed, weighted average in vivo  model to test cerebral oximetry.

Data on the arterial/venous ratio for the cerebral circulation is limited. 19–21The circulation is anatomically divided into five groups. The first group comprises large conducting arteries, which for the brain include the Circle of Willis and middle cerebral artery. The second group includes distal arterial branches and arterioles that function to regulate blood flow. The third group, consisting of end arterioles, capillaries, and postcapillary venules, serves gas exchange. The fourth and fifth groups include venules and large collecting veins, respectively. Each group’s volume contribution has been calculated from in situ  measurements of number, length, and radius of the vessels in one adult dog brain and one bat wing. 19,20In adult dog brain, the first through the fifth groups contain 10%, 17%, 28%, 20%, and 25% of the blood volume, respectively. 19In bat wing, the corresponding figures are 10%, 5%, 39%, 25%, and 21%, with group 3 being 1% end arteriole/capillary and 38% venular. 20From this data, we calculate the arterial/venous ratio to be 14/86 in bat wing and 28/72 in dog brain, assuming the end arteriole/capillary to venous percentage in group 3 is the same for dog brain as for bat wing. The commonly used 25/75 ratio, 2,3,14originates from Mchedlishvili’s 21calculations based on cerebrovascular resistance measured by another investigator (Tkachenko BI: Venous blood circulation. Meditsina, Leningrad, 1979, undocumented). To our knowledge, no published data exist about variance in the ratio among subjects or physiologic conditions. However, in our studies of the pial circulation in piglets using intravital microscopy, we have observed considerable variation in the relative numbers of arterioles and venules among animals (unpublished observations). Other investigators have reported vessel groups 1–4 to dilate and constrict during hypoxia and hypocapnia. 22,23Thus, it is likely biologic variation exists in arterial/venous ratios and that the ratio does not change substantially during hypoxia or hypocapnia.

Several biologic and instrument-related factors might have influenced our calculation of the cerebral arterial/venous ratio. Biologic factors relate to the assumptions of negligible extracerebral blood in the jugular bulb and of negligible capillary blood volume. If extracerebral blood in the jugular bulb were not negligible, it would increase Sjo²because oxygen extraction by extracranial tissue is minimal, 24with the result that our calculation of the ratio would be falsely high. Although extracerebral contamination of the jugular bulb is negligible in normal adult humans, 25it may not have been so in some of our subjects, given the vascular malformations present in complex congenital heart disease. Perhaps the four instances of Sjo²exceeding Sco²represented an example of this extracerebral contamination. If capillary blood volume were not negligible, it would make our calculation of the ratio falsely high or low, depending on the net change in the arterial and venous coefficients (equations 3 and 4). Because chronic hypoxia increases brain microvessel density, 26it is possible that capillary blood volume may not have been negligible in some of our subjects. However, similar cerebral arterial/venous ratios in the normoxic subjects, in whom these assumptions are likely valid, argue against these biologic factors biasing our calculation of the ratio, although they could increase its variance.

Instrument-related factors include measurement precision and the field of view. Our optical probe monitors the frontal neocortex. 12,27Ideally, the cerebral arterial/venous ratio is calculated from venous saturation in this tissue field rather than Sjo². Variations in regional cerebral oxygen extraction have been measured and may have contributed to the variance in arterial/venous ratios among our subjects. 24,28Measurement precision of the cerebral oximeter (6%) and blood oximeter (3%) could contribute to the variance in our calculation of the arterial/venous ratio. 12A carry-through error analysis of equation 2 using these precision values reveals that approximately one half of the variance could arise from this instrument factor. The other one half of the variance in the cerebral arterial/venous ratio would have to originate from true biologic variability among subjects. This biologic variability did not seem to originate from physiologic differences among subject because factors that regulate cerebral blood volume, such as central venous and arterial pressure, hemoglobin concentration, and arterial saturation, 23were not associated with the cerebral arterial/venous ratio.

There has been some question about the circulation monitored by cerebral oximetry. NIRS signal changes with Trendelenberg positioning clearly demonstrate a component of the venous circulation. 29Our findings and those of Brun et al.  30also show an arterial contribution to NIRS. In vitro  work illustrates the NIRS signal to originate mainly from small blood vessels. 31The NIRS signal changes during complete ischemia clearly demonstrate that it monitors gas-exchanging vessels. 5Gas exchange has been found to take place in arterioles and venules as well as in capillaries. 32These gas-exchanging vessels contain the majority of blood in the circulation (e.g. , vessel groups 3 and 4). 19–21Together, the body of evidence points to cerebral oximetry monitoring a mixed vascular bed dominated by gas-exchanging vessels, especially venules.

Our findings have implications in the evaluation of cerebral oximetry for use in clinical medicine. Because of biologic variation, use of a fixed arterial/venous ratio is not a good method to validate cerebral oximetry. This requires a direct evaluation of its measurement with a clinical outcome. Cerebral oximetry might be evaluated as a diagnostic or management device for cerebral hypoxia–ischemia in infants and children. 5–7,33,34