The Accuracy of a Near-Infrared Spectroscopy Cerebral Oximetry Device and Its Potential Value for Estimating Jugular Venous Oxygen Saturation

Abstract

Background

An intriguing potential clinical use of cerebral oximeter measurements (S_ctO₂) is the ability to noninvasively estimate jugular bulb venous oxygen saturation (S_jvO₂). Our purpose in this study was to determine the accuracy of the FORE-SIGHT® (CAS Medical Systems; Branford, CT), which is calibrated to a weighted average of 70% (S_jvO₂) and 30% arterial saturation, for Food and Drug Administration pre-market approval 510 (k) certification by adapting an industry standard protocol, ISO 9919:2005 [www.ISO.org] (used for pulse oximeters) and to evaluate the use of S_ctO₂ and S_pO₂ measurements to noninvasively estimate jugular venous oxygen saturation (S_nvO₂).

Methods

Paired blood gas samples from the radial artery and the jugular venous bulb were collected from 20 healthy volunteers undergoing progressive oxygen desaturation from 100 to 70%. The blood sample pairs were analyzed via co-oximetry and used to calculate the approximate mixed vascular cerebral blood oxygen saturation, or reference S_ctO₂ values (refS_ctO₂), during increasing hypoxia. These reference values were compared to bilateral FORE-SIGHT S_ctO₂ values recorded simultaneously with the blood gas draws to determine its accuracy. Bilateral S_ctO₂ and S_pO₂ measurements were then used to calculate S_nvO₂ values which were compared to S_jvO₂.

Results

Two hundred forty-six arterial and 253 venous samples from 18 subjects were used in the analysis. The ipsilateral FORE-SIGHT S_ctO₂ values showed a tolerance interval (TI) of [−10.72 10.90] Lin’s concordance correlation coefficient (CCC) with standard error (SE) of 0.83 ± 0.073 with the refS_ctO₂ values calculated using arterial and venous blood gases. The combined data had a CCC of 0. 81 + 0.059 with TI of [−9.22 9.40] with overall bias was 0.09% and amplitude of the root mean square of error after it was corrected with random effects analysis was 2.92%. The bias and variability values between the ipsilateral and the contralateral FORE-SIGHT S_ctO₂ measurements varied from person to person. The S_nvO₂ calculated from the ipsilateral S_ctO₂ and S_pO₂ data showed a CCC + SE of 0.79 ± 0.088, TI = [−14.93 15.33], slope of 0.98, Y-Intercept of 1.14%) with S_jvO₂ values with a bias of 0.20% and an Arms of 4.08%. The S_nvO₂ values calculated independently from contralateral forehead FORE-SIGHT S_ctO₂ values were not as correlated with the S_jvO₂ values (contralateral side CCC + SE = 0.72 ± 0.118, TI = [−14.86 15.20], slope of 0.66 and y-intercept of 20.36%).

Conclusions

The FORE-SIGHT cerebral oximeter was able to estimate oxygen saturation within the tissues of the frontal lobe under conditions of normocapnia and varying degrees of hypoxia (with 95% confidence interval of [−5.60 5.78] with ipsilateral blood ample data). These findings from healthy volunteers also suggest that the use of the calculated S_nvO₂ derived from S_ctO₂ and S_pO₂ values may be a reasonable noninvasive method of estimating S_jvO₂ and therefore global cerebral oxygen consumption in the clinical setting. Further laboratory and clinical research is required to define the clinical utility of near-infrared spectroscopy determination of S_ctO₂ and S_nvO₂ in the operating room setting.

Introduction

Cerebral oximetry using near-infrared spectroscopy (NIRS) is a continuous, noninvasive, optical-based method of measurement used to estimate cerebral tissue oxygen saturation (S_ctO₂). NIRS devices, including both cerebral and conventional pulse oximeters, calculate oxy- and deoxyhemoglobin concentration by measuring the absorbance of light at specific wavelengths.¹ However, cerebral oximeters do not preferentially measure oxygen saturation in pulsatile blood flow. Instead, cerebral NIRS devices estimate S_ctO₂ by measuring oxy- and deoxyhemoglobin in arterioles, capillaries and venules in intracerebral tissue. The device interrogates a region of cerebral tissue approximately 1.5 cm below the sensor (half the distance between the transmitter and the receiver) and provides a weighted measure of hemoglobin changes in the arterial, capillary, and venous compartments, which makes this a regional cerebral tissue saturation monitor. As opposed to arterial saturation determined from conventional pulse oximetry (a measurement made from the changes in absorbance due to changes in concentration at the apex and the nadir of pulsatile flow between the transmitter and the receiver), cerebral oximetry may allow for more clinically nuanced information about cerebral oxygen supply and demand, which can be gleaned by examining the venous component of the cerebral blood flow (CBF) by mathematically manipulating S_ctO₂ with S_pO₂.

The United States Food and Drug Administration requires cerebral oximeters meet parts of the International Organization for Standards (ISO) # 9919:2005 performance standards established for pulse oximeters when requesting pre-market 510(k) clearance. This ISO standard calls for a root-mean-square difference (measure of accuracy) of less than or equal to 4.0 % over a range of 70 to 100 % arterial blood oxygen saturation. While these standards are for pulse oximeters, they are generally adopted for the purpose of evaluating cerebral oximeters. The specifications also recommend stepped reductions in inspired oxygen concentration as a method to assess the accuracy of any type of oximeter used for patient monitoring.

There is no universally agreed upon “gold standard” measurement of S_ctO₂ in healthy volunteers. However, comparison of NIRS generated S_ctO₂-generated values with reference S_ctO₂ (refS_ctO₂) values derived from simultaneously sampling and measurement of radial artery oxygen saturation (S_aO₂) and jugular bulb venous oxygen saturation (S_jvO₂) via co-oximetry as performed by Henson et al² as the standard.

The purpose of this study was to: 1) determine the accuracy of the FORE-SIGHT cerebral oximeter S_ctO₂ values for Food and Drug Administration 510 (k) pre-market approval by comparing them to refS_ctO₂ values derived from simultaneous radial artery and jugular bulb venous samples obtained from healthy subjects during room air breathing, mild hypoxemia and hyperoxemia in the presence of normocapnia; and 2) assess the accuracy of noninvasive venous oxygen saturation (S_nvO₂) derived from S_pO₂ and S_ctO₂ values.

Methods

After IRB approval (August, 2004), 10 male and 10 female healthy volunteers aged 21 to 35 years with a body mass index of 18–30 kg·m⁻² were enrolled into the study after signing a written consent form. Because of the relatively low systemic saturation targets, the IRB mandated that we report every 3 subjects to the Data Safety and Monitoring Board for review.

An 18G peripheral IV, 20G radial arterial and a 5F venous jugular bulb cannulae were inserted in each of the 20 subjects under 2% lidocaine HCl local anesthesia. The jugular bulb cannula³ was initially inserted under ultrasound guidance (Sonosite Titan, Sonosite, Seattle, WA) and advanced cephalad until the subject reported minor ear discomfort.⁴ A lateral skull radiograph confirmed cannula tip placement cranial to a line extending from the atlanto-occipital joint space and caudal to the lower margin of the orbit, thus insuring the absence of extracranial venous blood contamination.⁵ Bilateral oximeter sensors were applied to the subject’s forehead, symmetrically along the midline to avoid the sagittal sinus and as far above the orbital ridge to maximize interrogation of brain as the hairline allowed, and connected to a dual-channel FORE-SIGHT cerebral oximeter. Continuous venous jugular bulb (S_jbO₂) and cardiac output in liters per minute were monitored with a Vigilance hemodynamic monitor (Edwards LifeScience, Irvine, CA), and finger arterial oxygen saturations with a Nellcor pulse oximeter (Model N595, Coviden, Boulder CO) respectively. Periodic noninvasive arterial blood pressure, continuous arterial blood pressure, heart rate and electrocardiogram were monitored with a PROPAQ (Model CS, Welch Allyn, Beaverton, OR). Subjects were positioned supine with the head of the bed elevated 35°.

The subject inspired gas mixtures delivered to a facemask from a sequential gas delivery system^6,7 (Respiract UDB 1000; Thornhill Research, Toronto, Canada) capable of producing and maintaining target end-tidal partial pressure of oxygen (PetO₂) and carbon dioxide (PetCO₂) values independent of one another. The system works by dividing the inspired breath into 2 components, fresh gas and the last end-tidal breath (which has already reached equilibrium with pulmonary venous gas pressures), such that only the small sample of fresh gas contributes to the gas exchange as long as the alveolar ventilation rate exceeds that of the fresh gas flow rate.⁸ End-tidal O₂ and CO₂ values were continuously measured using a Datex Capnomac monitor (Ultima I V90 EN, GE Healthcare, Waukesha, WI) via a sampling port located on the mask, with a sampling line of approximately 2 meters. Each subject initially inspired 21% O₂ followed by 2 sequences of reduced oxygen breathing designed to produce a progressive and then an acute reduction in blood O₂ saturation. After the period of reduced oxygen breathing, the inspired O₂ concentration was returned to 21% and then increased to 50%. The subject was allowed a brief rest period between the 2 hypoxia sequences. The progressive sequence consisted of 10 steps of 4 minutes with inspired oxygen concentrations of 21, 12, 10, 9, 8, 9, 10, 12, 21, and 50%. The acute sequence consisted of 4 steps with inspired oxygen concentrations of 21, 8, 21, and 50%. The target PetCO₂ during each step of the 2 sequences was 40 mmHg (Figure 1).

Figure 1.

Example of hemodynamic and respiratory data obtained during the progressive hypoxia sequence in a single subject.

Each step was maintained for 4 minutes and paired arterial and jugular venous blood samples were obtained during the last 2 minutes of the step. The jugular venous samples were drawn at a rate of 2 mL/min and the arterial samples were taken at the midpoint of the jugular venous draws (Figure 2).^9,10

Figure 2.

Example of the S_pO₂, FORE-SIGHT™ S_ctO₂, Vigilance S_jbO₂, and co-oximeter determined S_jvO₂ and S_aO₂ values during the progressive hypoxia sequence in a single subject.

If the subject did not reach the desired S_pO₂ minimum target value of 70% during either sequence, the O₂ content in the delivered gas mixture was reduced to 5% FiO₂ and additional blood samples were drawn at the new S_pO₂ plateau. Blood draws were omitted if the subject reached the desired S_pO₂ threshold of 70% in fewer steps or became excessively anxious. Blood samples were analyzed immediately by co-oximetry (Model IL282; Instrumentation Laboratories, Lexington, MA).

The NIRS-derived S_ctO₂ values were calibrated against S_aO₂ and S_jvO₂ blood samples using a 3-compartment cerebral blood volume (CBV) model (refS_ctO₂; equation 1) :¹¹

$$\text{ref}{\mathrm{S}}_{\mathrm{c}\mathrm{t}}{\mathrm{O}}_2=\mathrm{K}\mathrm{a}*{\mathrm{S}}_{\mathrm{a}}{\mathrm{O}}_2+\mathrm{K}\mathrm{v}*{\mathrm{S}}_{\mathrm{j}\mathrm{v}}{\mathrm{O}}_2+\mathrm{K}\mathrm{c}*{\mathrm{S}}_{\mathrm{c}}{\mathrm{O}}_2+\text{Error}$$ (1)

where the distribution coefficients of arterial (Ka), venous (Kv) and capillary blood (Kc) oxygen saturations are set at 0.3, 0.7, and 0.0 for the FORE-SIGHT® (CAS Medical Systems; Branford, CT), respectively. The ratio of the volume of arterial blood (30%) to venous blood (70%) in the cerebral tissue is based on the work of Ito et al.³⁶ The capillary blood compartment volume is considered negligible and so the capillary blood saturation (S_cO2) contribution to CBV is considered nil (Kc = 0.0). The error term is a generic term that sums the various sources of error that may influence the S_ctO₂ reading. Equation 1 was solved for jugular bulb venous oxygen saturation (S_jvO₂; equation 2).

$${\mathrm{S}}_{\mathrm{j}\mathrm{v}}{\mathrm{O}}_2=(\text{ref}{\mathrm{S}}_{\mathrm{c}\mathrm{t}}{\mathrm{O}}_2-\mathrm{K}\mathrm{a}*{\mathrm{S}}_{\mathrm{a}}{\mathrm{O}}_2+\text{Error})/\mathrm{K}\mathrm{v}$$ (2)

Substituting pulse oximetry (S_pO₂) values for co-oximeter determined S_aO₂ and NIRS S_ctO₂ values for co-oximeter determined refS_ctO₂ produced a noninvasive estimate of venous oxygen saturation and global cerebral oxygen consumption (S_nvO₂; equation 3):

$${\mathrm{S}}_{\mathrm{n}\mathrm{v}}{\mathrm{O}}_2=({\mathrm{S}}_{\mathrm{c}\mathrm{t}}{\mathrm{O}}_2-\mathrm{K}\mathrm{a}*{\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2+\text{Error})/\mathrm{K}\mathrm{v}$$ (3)

Furthermore, the relationship between the arterio-jugular oxygen difference (AVDO₂), cerebral metabolic rate for O₂ (CMRO₂) and CBF as a function of arterial blood oxygen-carrying capacity (C_aO₂) and venous blood oxygen-carrying capacity (C_jvO₂)¹² can be expressed in terms of S_jvO₂, S_aO₂, partial pressure of oxygen in the arteries (P_aO₂), and partial pressure of oxygen in the jugular vein (P_jvO₂), if the difference between the arterial and venous concentration of hemoglobin is negligible (equations 4 and 5).

$$\mathrm{C}\mathrm{M}\mathrm{R}{\mathrm{O}}_2/\mathrm{C}\mathrm{B}\mathrm{F}=({\mathrm{C}}_{\mathrm{a}}{\mathrm{O}}_2-{\mathrm{C}}_{\mathrm{j}\mathrm{v}}{\mathrm{O}}_2)$$ (4)

$$\mathrm{C}\mathrm{M}\mathrm{R}{\mathrm{O}}_2/\mathrm{C}\mathrm{B}\mathrm{F}=({\mathrm{S}}_{\mathrm{a}}{\mathrm{O}}_2-{\mathrm{S}}_{\mathrm{j}\mathrm{v}}{\mathrm{O}}_2)\times (1.34\times [\mathrm{t}\mathrm{H}\mathrm{b}])+0.003\times ({\mathrm{P}}_{\mathrm{a}}{\mathrm{O}}_2-{\mathrm{P}}_{\mathrm{j}\mathrm{v}}{\mathrm{O}}_2)$$ (5)

Equation 6 shows the noninvasive version of equation 5 where S_pO₂ and S_nvO₂ have been substituted for S_aO2 and S_jvO₂ respectively to determine the relationship between CMRO₂ and CBF:

$$\mathrm{C}\mathrm{M}\mathrm{R}{\mathrm{O}}_2/\mathrm{C}\mathrm{B}\mathrm{F}=({\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2-{\mathrm{S}}_{\mathrm{n}\mathrm{v}}{\mathrm{O}}_2)\times (1.34\times [\mathrm{t}\mathrm{H}\mathrm{b}])+0.003\times ({\mathrm{P}}_{\mathrm{a}}{\mathrm{O}}_2-{\mathrm{P}}_{\mathrm{j}\mathrm{v}}{\mathrm{O}}_2)$$ (6)

CMRO₂ / CBF is approximately proportional to the difference between S_nvO₂ and S_pO₂ if the effects of dissolved oxygen (0.003 × (P_aO₂ – P_jvO₂)) are negligible. In the clinical setting, this approximation using noninvasive parameters can be important because the decrease in S_ctO₂ can indicate either a decrease in CBF, or an increase in CMRO₂, or both.

Data Analysis

A priori statistical power analysis determined that a minimum of 100 samples would be required to obtain a 99% confidence interval ((1 − □) of a width of ± 4% (each side) around the estimate of the mean with a standard deviation of 5. In the present study, we had 18 subjects with 2 sensors each, and each sensor yielded an average of 13 data points per sensor per subject, which netted an average of 27 data points total per subject. The ipsilateral and contralateral forehead S_ctO₂ values, filtered within the device with a moving average filter with a window of 6 seconds, recorded at each step of the progressive and acute sequences at the midpoint of the predetermined 30 second jugular venous blood draw to avoid extracranial contamination, as reported by Matta and Lam,⁹ were compared to each other, and the refS_ctO₂ value, and used to compute S_nvO₂ which was then compared to S_jvO₂. A paired Student’s t-test was used to compare the ipsilateral and contralateral sensor S_ctO₂ measurements for each subject. The correlation and the Lin’s concordance correlation coefficient (CCC)^13,14 (and the 95% confidence interval denoted as standard error (SE)¹⁵) between S_ctO₂ and refS_ctO₂ values were examined by constructing linear regression curves for each subject; for each forehead sensor independently and for both sensors combined. The agreement between ipsilateral and contralateral S_ctO₂ and refS_ctO₂ values across all subjects and all samples was explored by constructing the Bland and Altman¹⁶ plot and modified for random effect analysis for repeated measures by Myles and Cui.¹⁷ The degree of error, defined as the difference between the S_ctO₂ and the refS_ctO₂ values, was assessed in terms of parametric analysis of the errors, including bias (mean of the difference between FORE-SIGHT S_ctO₂ and refS_ctO₂), standard deviation (SD), confidence intervals (bias ±1.95 SD), tolerance intervals (for 99% of future measurements to be within 95% confidence), and first order linear regression,¹⁸ adjusted for random effects analysis as explained by Myles and Cui¹⁷ and adjustments for repeated measures as previously described.¹³ The analysis was repeated for the error between the calculated S_nvO₂ values and the co-oximeter-determined venous oxygen saturation values (S_jvO₂). Additional nonparametric analysis was performed in the form of a cumulative distribution function (CDF) because the error is not exactly normally distributed.¹⁹

Results

Of the 20 subjects enrolled in the study 18 were included in the analysis. The prototype FORE-SIGHT was unable to resolve NIRS measurements from 1 subject due to unexpected attenuation of the light through the tissue and blood gas equipment failure prevented blood gas analysis in another. The non-Gaussian distribution of subjects’ demographic information is summarized in Tables 1 and 2.

Table 1.

Demographics

	Age*	Weight*	BMI*
Female	26.0 [22.0, 33.4]	65.4 [49.8, 75.6]	22.2 [20.1, 26.5]
Male	23.3 [21.4, 35.2]	76.9 [63.4, 99.1]	23.6 [21.8, 29.6]
Overall	25.1 [21.4, 35.16]	70.1 [49.8, 99.1]	22.8 [20.1, 29.6]

^*Median and Range

Table 2.

Demographics

	African America n	Asian	Caucasia n	Hispanic	Total
Female	0	2	7	0	9
Male	2	1	5	1	9
Total	2	3	12	1	18

Sixteen paired arterial and venous blood samples were collected during the progressive phase followed by the acute phase of the experiment from 2 subjects, 15 from 7 subjects, 14 from 6 subjects, 13 from 2 subjects and 7 from 1 subject. Variations in the number of samples collected were due to having to reduce or increase the inspired gas exposure duration to achieve the target S_pO₂ of 70% or the necessity to terminate the hypoxia exposure based on subject tolerance. Of the 254 total paired samples drawn, 8 arterial samples and 1 venous sample could not be analyzed due to clotting of the sample. Therefore 246 arterial and 253 venous samples were analyzed and included in the analysis. The lowest recorded S_aO₂ during the course of the study, across all subjects, was 64%. The P_etCO₂ was independently targeted to 40 mmHg for the entire experiment, which resulted in mean arterial P_aCO₂ during the 2 phase exposure (example in Figure 1) was 38.8 ± 2.2 mmHg (mean ± SD). The average of the ipsilateral and contralateral S_ctO₂ values recorded at baseline during room air breathing was 74.6 ± 3.8% (mean ± SD) with a range of 66.6 to 83.3%.

The significance of the difference between ipsilateral and contralateral S_ctO₂ varied from subject to subject, with overall significance of P = 0.61 (Table 3). The ipsilateral S_ctO₂ correlated better with a CCC ± SE of 0.83 ± 0.073 and a tolerance interval (TI) of [−10.72 10.90] versus the contralateral side of 0.78 ± 0.098 with TI of [−10.59 10.74], with the combined data showing a CCC of 0. 81 ± 0.059 with TI of [−9.22 9.40] (Figure 3, Table 4).

Table 3.

Paired Student T-Test for the ipsilateral and contralateral values of S_ctO₂ and S_nvO₂ per subject. P value below 0.05 is considered significant.

Subject	Student T- Test P Value (SctO2)	Student T- Test P Value (SnvO2)	(SctO2) Ipsilateral / Contralateral dichotomy significant?	(SnvO2) Ipsilateral / Contralateral dichotomy significant?
S02	2.86 e-9	2.86 e-9	Yes	Yes
S03	4.85 e-6	4.85 e-6	Yes	Yes
S04	2.45 e-9	2.45 e-9	Yes	Yes
S05	2.08 e-8	2.08 e-8	Yes	Yes
S06	1.70 e-4	1.70 e-4	Yes	Yes
S07	9.64 e-1	9.00 e-1	No	No
S08	3.67 e-1	3.64 e-1	No	No
S09	6.78 e-2	1.18 e-1	No	No
S10	4.46 e-3	1.88 e-3	Yes	Yes
S11	1.51 e-1	1.99 e-1	No	No
S13	3.42 e-12	3.42 e-12	Yes	Yes
S14	1.69 e-1	2.98 e-1	No	No
S15	8.33 e-3	7.17 e-3	Yes	Yes
S16	1.33 e-2	2.46 e-2	Yes	Yes
S17	3.99 e-5	9.10 e-7	Yes	Yes
S18	1.29 e-2	1.78 e-2	Yes	Yes
S19	1.68 e-1	1.68 e-1	No	No
S20	3.98 e-1	8.98 e-1	No	No
Overall	7.23 e-1	6.93 e-1	No	No

Figure 3.

Plot of FORE-SIGHT™ S_ctO₂ ipsilateral and contralateral combined sensor values versus calculated refS_ctO₂ values.

Table 4.

Parametric analysis and Random Effects Analysis correction for the ipsilateral, contralateral and combined S_ctO₂ values compared to refS_ctO₂ values.

Parametric Analysis with Random Effects Analysis Correction:	SctO2 Ipsilateral	SctO2 Contralateral	SctO2 Combined
Bias	0.09	0.08	0.09
Standard Deviation	2.92	2.88	2.86
95% Confidence Interval	[−5.60 5.78]	[−5.53 5.69]	[−5.48 5.57]
95% Tolerance Interval	[−10.72 10.90]	[−10.59 10.74]	[−9.22 9.40]
Precision	0.20	1.14	0.46
Pearson Correlation	0.85	0.78	0.81
Lin’s Concordance Correlation Coefficient ± Standard Error	0.83 ± 0.073	0.78 ± 0.098	0.81 ± 0.059
Slope	1.05	0.74	0.89
Intercept	−3.08	17.96	7.44

The complete parametric analysis of the errors for the ipsilateral, contralateral, and the combined data are in Table 4. The ipsilateral side was on the same side as the jugular bulb catheter, and the data reflect a more accurate measurement from the sensor located on the same side as the catheter.

The calculated proportional error for all ipsilateral and contralateral S_ctO₂ values is shown on the random effect analysis-corrected Bland-Altman plot (Figure 4). There was a proportional error of less than 0.04% per decade (for every 10 % change in the (FORE-SIGHT S_ctO₂ + refS_ctO₂) / 2 value; X axis) and the error decreased as saturation values decreased.

Figure 4.

Bland-Altman plot of the difference between FORE-SIGHT S_ctO₂ and refS_ctO₂ values.

The relationship between blood gas-determined venous jugular bulb S_jvO₂ and the S_nvO₂ values calculated using the combined ipsilateral and contralateral FORESIGHT readings was examined using linear regression analysis (Figure 5) and a Bland-Altman plot (Figure 6)

Figure 5.

Plot of the calculated noninvasive venous oxygen saturation (S_nvO₂) versus the blood gas determined jugular bulb venous oxygen saturation (S_jvO₂) values.

Figure 6.

Bland-Altman plot of the calculated noninvasive venous oxygen saturation (S_nvO₂) versus the blood gas determined jugular bulb venous oxygen saturation (S_jvO₂) values.

The ipsilateral, contralateral and combined data for S_nvO₂ and the data obtained from the jugular bulb catheter (S_jbO₂) are compared to the S_jvO₂ values (Table 5). The data show that the S_nvO₂ calculated from the same side as the jugular bulb catheter (CCC ± SE = 0.79 ± 0.088, TI = [−14.93 15.33]) was more accurate than the contralateral sensor (CCC ± SE = 0.72 ± 0.118, TI = [−14.86 15.20]). The jugular bulb catheter (Vigilance) fared only slightly better with CCC + SE = 0.82 ± 0.078 with a TI of [−12.60 9.66].

Table 5.

Comparison of the calculated S_nvO₂ values derived from the ipsilateral, contralateral and combined measurements and the Vigilance (S_jbO₂) measurements to the venous jugular bulb oxygen saturation (S_jvO₂) values from blood samples.

Parametric Analysis with Random Effects Analysis Correction:	SnvO2 Ipsilateral	SnvO2 Contralate ral	SnvO2 Combined	SjbO2 Vigilance
Bias	0.20	0.17	0.19	−1.47
Standard Deviation	4.08	4.06	4.01	3.00
95% Confidence Interval	[−7.80 8.20]	[−7.78 8.12]	[−7.68 8.05]	[−2.88 8.89]
95% Tolerance Interval	[−14.93 15.33]	[−14.86 15.20]	[−12.89 13.26]	[−12.60 9.66]
Precision	0.17	1.92	0.99	0.34
Pearson Correlation	0.81	0.73	0.77	0.85
Lin’s Concordance Correlation Coefficient ± Standard Error	0.79 ± 0.088	0.72 ± 0.118	0.76 ± 0.071	0.82 ± 0.078
Slope	0.98	0.66	0.82	0.93
Intercept	1.14	20.36	10.75	2.56

All values were compared against jugular bulb venous (S_jvO₂) blood sample values determined by co-oximetry.

The CDF plot of the errors in S_ctO₂ (Figure 7 and Figure 8) clearly shows subject 2 to be an outlier, thus the parametric analysis will be partially biased away from the norm. It also shows that approximately 90% of the error is within 8.5% for the ipsilateral, and 10% for the contralateral. Similarly, the CDF plot of the S_nvO₂ shows once again that subject 2 is an outlier. It also shows that 90% of the S_ctO₂ errors on the ipsilateral are within 12% and 14% on the contralateral.

Figure 7.

Cumulative Density Function Plot of the absolute value of the error between S_ctO₂ and refS_ctO₂.

Figure 8.

Cumulative Density Function Plot of the absolute value of the error between (S_nvO₂) and the blood gas determined jugular bulb venous oxygen saturation (S_jvO₂) values.

Discussion

We found a strong positive correlation between the ipsilateral forehead S_ctO₂ measurements and the refS_ctO₂ values calculated using arterial and jugular bulb oxygen saturations and the FORE-SIGHT S_ctO₂ values. The ipsilateral sensor performed better than the contralateral, and the comparison between the 2 sides on a subject-by-subject basis showed that ipsilateral-contralateral dependency varied widely. Differences between the refS_ctO₂ and FORE-SIGHT S_ctO₂ values may reflect interindividual variability²⁰ and/or differences in the regional blood flow of the cerebral tissue, as well as contamination from extracranial tissue²¹. Henson et al. sought to use S_ctO₂ as a predictor of S_jvO₂, and found that the Somanetics 3100 cerebral oximeter device correlated well with venous oxygen saturation.²

The ± 10% tolerance interval of future readings may seem excessively large but given the sigmoid nature of the oxygen-hemoglobin dissociation curve²², this may not be clinically significant. As an example, a reading of 75% S_ctO₂ may be between 65% or 85%, but a value within this range does not suggest an imminent clinical threat to the patient. An S_ctO₂ of 55% indicates a venous saturation between 50% and 21%, which is the arbitrary threshold for neurocognitive impairment as set and published by Yao et al²³, given that the arterial saturation is usually near 100%. Several studies have reported neurocognitive deficits and prolonged hospital stay when the INVOS™ (Covidien (formerly Somanetics), Mansfield, MA) rSO₂ decreases below 50% for a prolonged period, 40% for more than 10 minutes, or nadir below 35% where the corresponding cerebral venous saturations decreased below 35%, 20%, and 13% accordingly.^24,25,26 There is no clear S_ctO₂ threshold for clinical intervention so the S_ctO₂ may serve as a graded indicator of risk organized in decades. The normal range is considered 65% and above, elevated risk between 55% and 65%, and high risk at 55% and below (55% S_ctO₂ corresponds to 35% venous saturation, assuming 100% S_aO₂). Because arterial blood oxygen saturations are very well managed in modern anesthesia practice, estimation of the noninvasive estimate of venous oxygen saturation (S_nvO₂) as a surrogate of S_jvO₂ may have some merit and indicate clinically relevant changes in the ratio between CMRO₂ and CBF. A crude summary of physiologic conditions and their possible contributors are tabulated in Table 6.

Table 6.

Possible causes of inflections in the S_nvO₂ due to physiological state change in the cerebral tissue.

Indication:	SaO2	CMRO 2	CPP	CBF	Notes
Low SnvO2			Low	Low	Low CPP may be due to Low MAP or high ICP, or both
Low SnvO2		High			Some regions of the brain may be more metabolically active, which may indicate a low SnvO2 relative to systemic SvO2
Low SnvO2			High	Low	Possibly vasoconstriction due to low PaCO2 and vasopressors
Low SnvO2	Low				Low arterial blood saturation
High SnvO2				High	Possibly vasodilatation due to high PaCO2, drugs that relax smooth muscle, and diuretics
High SnvO2			High	High	Possibly vasodilatation due to high PaCO2, vasopressors, high Cardiac Output, and / or high Mean Arterial Pressure (MAP)
High SnvO2				High	Possibly vasodilatation due to high PaCO2 and vasopressors
High SnvO2		Low			Low brain activity

All blank spaces denote normal ranges.

The Bland-Altman plot demonstrated a modest degree of proportional error, which declined as the blood oxygen saturation value decreased; a positive characteristic in the setting of clinical hypoxia. In addition, the data points were equally scattered above and below the zero line suggesting there is no consistent bias of 1 measurement approach (FORE-SIGHT S_ctO₂) versus the refS_ctO₂. S_nvO₂ calculated using the FORE-SIGHT S_ctO₂ and S_pO₂ values was strongly correlated with blood gas determined S_jvO₂.

The wide variability of the average FORE-SIGHT S_ctO₂ values recorded at baseline during room air breathing is generally consistent with what has been reported clinically.^27,28,29 The significance of the cerebral oximeter can be further demonstrated by substituting equation 3 (assuming blood distribution to be 70% venous (Kv = 0.7), 30% arterial (Ka = 0.3), and negligible amount in the capillaries (Kc = 0) and ERROR assumed to be zero for the ideal case as suggested by Ito et al.¹¹) into equation 6 and solving the ratio of CMRO₂ to CBF as a function of S_ctO₂ (equation 7).

$$\mathrm{C}\mathrm{M}\mathrm{R}{\mathrm{O}}_2/\mathrm{C}\mathrm{B}\mathrm{F}=({\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2-{\mathrm{S}}_{\mathrm{c}\mathrm{t}}{\mathrm{O}}_2)\times \mathrm{K}1+\mathrm{K}2$$ (7)

• Where K1 = (1.34 × [tHb]) /0.7

• And

• K2 = 0.003 × (P_aO₂ − P_jbO₂).

Solving equation 7 for S_ctO₂ then becomes (equation 8):

$${\mathrm{S}}_{\mathrm{c}\mathrm{t}}{\mathrm{O}}_2={\mathrm{S}}_{\mathrm{p}}{\mathrm{O}}_2+(\mathrm{K}2-\mathrm{C}\mathrm{M}\mathrm{R}{\mathrm{O}}_2/\mathrm{C}\mathrm{B}\mathrm{F})/\mathrm{K}1$$ (8)

From this equation, it can be deduced that the S_ctO₂ has a proportional relationship to S_pO₂ and CMRO₂ and inversely proportional with CBF. Thus, the variation in baseline S_ctO₂ can either be directly due to interpersonal differences in brain activity and CBF. Any preexisting conditions affecting CBF such as sympathetic activation due to anxiety and reduced CBF due to vasoconstriction (abnormal PaCO2 due to hyperventilation, vasoconstrictors onboard, etc.), or atherosclerosis can affect the baseline S_ctO₂.

In the current study, the effect of CO₂ on CBF, and therefore oxygen delivery, was controlled by maintaining the PetCO₂ at 40 mmHg with a scheme described by Slessarev et al.²⁰ However, inspection of the individual subject’s linear regression plots show a slight inflection of the plots at lower saturations, which may indicate that the reference saturation and the NIRS saturation measurements may not be linear at lower saturation levels due to changing CMRO₂ and other confounding effects. The end-tidal gas-forcing device software assumes constant oxygen consumption rates that may have introduced an effort-dependent error. Because Seifert and Secher have shown that sympathetic activity can possibly influence CBF and CMRO₂ during cycling exercise,³⁰ the increased respiratory effort with the diaphragm muscle during this study could have affected a change in CBF and CMRO₂.

The morphology of the Bland-Altman plot [which resembles a “less than” symbol (<), with smaller spread in the 50% range than the 70% range], also suggests an error that is proportional to the magnitude of the S_ctO₂ reading and corroborates the possibility that the S_ctO₂ may be slightly nonlinear. In addition, hypoxemia produces changes in CBF³¹ and the full extent of the changes in the venous distribution and mixing of the various tributaries that drain into the jugular venous returns are not entirely known. Due to this uncertain venous mixing, the venous sample taken at the jugular bulb may not have reflected the true venous saturation at the interrogated tissue.

In this study refS_ctO₂ values were calculated for the purposes of evaluating the accuracy of the FORE-SIGHT device using distribution coefficients (0.30*S_aO₂ + 0.70 * S_jvO₂) derived from Ito et al.³² However, Pollard et al.³³ have estimated S_ctO₂ using a slightly different CBV distribution (0.25*S_aO₂ + 0.75 * S_jvO₂) as has McCormick et al. (0.18*S_aO₂ + 0.82* S_jvO₂ and 0.28*S_aO₂ + 0.72* S_jvO₂).³⁴ Not only does a device manufacturer’s choice of distribution coefficients affect the S_ctO₂ reported value, but use of a fixed set of distribution coefficients can also be a source of error because the CBV distribution and CBF both change due to various activities and body positions.³³ Functional magnetic resonance imaging studies by Ito et al³⁵ have shown that an elevated CMRO₂ during neural activation causes a decrease in venous blood oxygen saturation. Regional differences in cerebral activity may also cause cerebral blood distribution to change. Ito et al. used controlled visual stimulation to demonstrate significant differences in CBF and CBV during rest and periods of stimulation, which was consistent with positron emission tomography studies estimating CBF and CBV under hypercapnic conditions.³⁶ However, Ito et al. did not precisely control PaCO₂ and their findings indicate a wide range of possible distribution of the CBV under these conditions. They also found that CBV changed significantly owing to changes in P_aCO₂, and arterial compartment volume changed significantly in the direction of CBV changes. However, no significant change was observed in venous or capillary blood volume. This indicates that changes in CBV during hypercapnia and hypocapnia are produced by changes in arterial blood volume and not by changes in venous and capillary blood volume.³⁷ The appropriate distribution coefficients for CBV within arterial and venous compartments remain a somewhat contentious issue but it is generally accepted that any increase in the concentration of CO₂ within the arterial compartment and sympathetic activation due to mental activity can trigger vasodilatation and thus affect changes in the CBV.

Our results were obtained by controlling the PetCO₂ in order to achieve a P_aCO₂ of approximately 40 mmHg (i.e., 38.8 ± 2.2 mmHg), which allowed a more consistent descent down the oxyhemoglobin dissociation curve. The dissociation curve from 20% to 80% is approximately linear and a small shift in the curve to the ipsilateral or contralateral may produce an unacceptable source of error when validating the calibration of a cerebral oximeter. Ito et al. found that the arterial fraction of human CBV during hypercapnia and hypocapnia measured by positron emission tomography changed such that the 25% arterial fraction assumed by the INVOS cerebral oximeter and the 30% arterial fraction assumed by the FORE-SIGHT cerebral oximeter for calculation of S_ctO₂ were no longer valid under these conditions.^37,38 The importance of maintaining normocapnia when evaluating the accuracy of a NIRS cerebral oximeter device which uses a fixed arterial to venous blood volume ratio is readily apparent from Ito et al.’s findings. Unfortunately, hypocapnia and hypercapnia in the clinical settings add to the uncertainty and error of these devices that have been calibrated under normocapnic conditions, regardless of whether a device is a “trend” or “absolute” monitor.

The gas delivery system required the gas blender to be calibrated at the beginning of the study to the subject’s rate of CO₂ production (dVCO₂/dt) and rate of O₂ consumption (dVO₂/dt). Any physical movement³⁹ and / or any changes in the breathing effort alter the dVCO₂/dt and dVO₂/dt.⁴⁰ The increased ventilatory drive at lower S_aO₂ increases respiratory effort thereby increasing both dVCO₂/dt and dVO₂/dt. The nature of the end-tidal-forcing device and technique used in this experiment made it impossible to adjust for these changes during the experiment and the initial set points for dVCO₂/dt and dVO₂/dt were no longer applicable at lower S_aO₂ and therefore resulted in some drift of the end-tidal target gas concentrations. Therefore vasodilatation and vasoconstriction due to hypercapnia and hypocapnia could have altered the intracranial blood volume distribution, and in turn the S_ctO₂.

Aside from the sources of error noted, the NIRS device and its interaction with interindividual anatomical differences may also introduce error.⁴¹ The sensors are designed to be placed on the forehead equilaterally. If there is a positional bias towards one side or the other, the sagittal sinus, a prominent anatomical feature, may influence the accuracy of the sensor readings.^42,43 Because the device is also designed to reject common mode noise, exposure to ambient light should not be an issue. However, an excessive amount of ambient light can saturate the sensors and introduce error in the reading. Individual skull thickness can also affect scattering of the NIRS optical signal and influence device readings.^44,45 The present study included 2 African-Americans, 3 Asians, and 1 Hispanic male. The varying levels of skin pigmentation may have also manifested as artificially lower readings at lower saturations as Feiner et al. and Bickler et al. have reported.^46,47

Conclusions

We found that the FORE-SIGHT NIRS device was able to estimate oxygen saturation (with 95% confidence interval of [−5.60 5.78] with ipsilateral blood sample data) within the tissues of the frontal lobe under conditions of normocapnia and varying degrees of hypoxia as suggested by the ISO 9919:2005 standard. The study findings also suggest that the use of the calculated venous oxygen saturation (S_nvO₂) may be a reasonable noninvasive method of estimating jugular bulb venous oxygen saturation and therefore global cerebral oxygen consumption in the clinical setting. Further laboratory and clinical research is required to define the value of NIRS determination of S_ctO₂ and S_nvO₂ in the operating room and intensive care unit setting.