Near-infrared spectroscopy as an index of brain and tissue oxygenation

Abstract

Continuous real-time monitoring of the adequacy of cerebral perfusion can provide important therapeutic information in a variety of clinical settings. The current clinical availability of several non-invasive near-infrared spectroscopy (NIRS)-based cerebral oximetry devices represents a potentially important development for the detection of cerebral ischaemia. In addition, a number of preliminary studies have reported on the application of cerebral oximetry sensors to other tissue beds including splanchnic, renal, and spinal cord. This review provides a synopsis of the mode of operation, current limitations and confounders, clinical applications, and potential future uses of such NIRS devices.

Reflectance near-infrared spectroscopy

Jobsis³⁴ first reported in 1977 that the relatively high degree of transparency of myocardial and brain tissue in the near-infrared (NIR) range enabled real-time non-invasive detection of tissue oxygen saturation using transillumination spectroscopy. By 1985, Ferrari and colleagues¹⁹ reported some of the first human cerebral oximetry studies using near-infrared spectroscopy (NIRS). After United States Food and Drug Administration (FDA) approval, in May 1993, the first commercial cerebral oximetry device, INVOS 3100^®, was marketed (Somanetics Corporation, Troy, MI, USA). Subsequently, after FDA approval, CAS Medical Systems (Branford, CN, USA) and Nonin Medical Inc. (Minneapolis, MN, USA) also began marketing NIRS cerebral oximetry devices.

NIR light can be used to measure regional cerebral tissue oxygen saturation (rSO₂). This technique uses principles of optical spectrophotometry that make use of the fact that biological material, including the skull, is relatively transparent in the NIR range. However, because of the poor signal-to-noise ratio as a result of the low intensity of transmitted light, most commercially available devices use reflectance-mode NIRS in which receiving optodes are placed ipsilateral to the transmitter and exploit the fact that photons transmitted through a sphere will traverse an elliptical path in which the mean depth of penetration is proportional to the transmitter and receiver optode separation. Fundamental challenges posed in utilizing transcranial reflectance NIRS to measure cerebral tissue oxygen saturation include the potential requirement for knowledge of the photon pathlength, the presence of non-haeme chromophores, and variable light absorption by overlying extracerebral tissue.

Tissue oxygen saturation

Measurement of tissue oxygen saturation and tissue haemoglobin content is determined by the difference in intensity between a transmitted and received light delivered at specific wavelengths as described by the Beer–Lambert law (see below). A decrement in transmitted light intensity is equivalent to the quantity of the substance and the amount of light absorbed by a unit quantity of that substance, defined as the extinction coefficient (ε), a factor that varies with the substance and the incident-light wavelength. The depth of penetration is proportional to the mean pathlength of photons through tissue.

Transmission of light at a given wavelength through tissue depends on a combination of reflectance, scattering, and absorptive effects. Reflectance is a function of the angle of the light beam and the regularity of the tissue surface. This decreases with increasing wavelength, thus favouring transmission of NIR vs visible light. Scattering is a function of tissue composition and number of tissue interfaces while absorption is determined by the molecular properties of substances within the light path. Above 1300 nm, water (H₂O) absorbs all photons over a pathlength of a few millimetres with a secondary peak between 950 and 1050 nm, whereas below 700 nm, increasing light scattering and more intense absorption bands of haemoglobin prevent effective transmission. In the 700–1300 nm range, NIR light penetrates biological tissue several centimetres.⁴⁸

Within the NIR range, the primary light-absorbing molecules in tissue are metal complex chromophores: haemoglobin, bilirubin, and cytochrome. The absorption spectra of deoxyhaemoglobin (Hb) ranges from 650 to 1000 nm, oxyhaemoglobin (HbO₂) shows a broad peak between 700 and 1150, and cytochrome oxidase aa3 (Caa3) has a broad peak at 820–840 nm (Fig. 1).³⁴ The wavelengths of NIR light used in commercial devices are selected to be sensitive to these biologically important chromophores and generally utilize wavelengths between 700 and 850 nm where the absorption spectra of Hb and HbO₂ are maximally separated and there is minimal overlap with H₂O. The isobestic point (wavelength at which oxy- and deoxyhaemoglobin species have the same molar absorptivity) for Hb/HbO₂ is 810 nm. As discussed below, the isobestic absorption spectra can be utilized to measure total tissue haemoglobin concentration.

Fig 1

Absorption spectra for oxygenated haemoglobin (HbO₂), deoxygenated haemoglobin (Hb), Caa3, melanin, and water (H₂O) over wavelengths in NIR range. Note the relatively low peak for Caa3. Commercial cerebral NIRS devices currently utilize wavelengths in the 700–850 nm range to maximize separation between Hb and HbO₂. The presence of melanin as found in human hair can significantly attenuate Hb, HbO₂, and Caa3 signals.

As outlined previously, the absorption of NIR light in tissue is determined by the Beer–Lambert law relating pathlength of NIR light to the concentration and absorption spectra of tissue chromophores and is conventionally written as:

where ΔA is the amount of light attenuation, L the differential photon pathlength through tissue, and μ the absorption coefficient of chromophore X and can be expressed as [X]×ε, where [X] is the tissue concentration of chromophore X and ε the extinction coefficient of chromophore X, thus [X]=ΔA/L×ε, which, in theory, allows measurement of tissue oxygen saturation (SO₂).

Multiwavelength NIRS and absolute vs relative oxygen saturation

Since ΔA is measured directly and ε has been determined for various tissue chromophores, absolute chromophore concentration [X] is thus inversely proportional to the optical pathlength. However, photon pathlength cannot be measured directly due to reflection and refraction in the various tissue layers involved. Unless pathlength can be determined, only relative change in chromophore concentration can be assessed. Modelling and computer simulation can be used to estimate photon tissue pathlength. By using successive approximation, an analysis algorithm can be calibrated to provide a measure of absolute change of chromophore concentration, as utilized by some commercial devices.

In order to measure absolute tissue chromophore concentrations, a different approach is used based on radiative transport theory and using multiple NIRS wavelengths and frequency-domain NIRS (fdNIRS) or time-domain NIRS (tdNIRS) analyses to determine tissue absorption coefficients (μ). Theoretically, approaches such as fdNIRS or tdNIRS avoid the need for actual photon pathlength determination.^42,46 Fundamental to such techniques is that tissue absorption coefficients can be measured directly using multiwavelength NIRS. Since

tissue chromophore concentration can thus be measured absolutely, there is no requirement for determination of optical pathlength.⁴⁰ This approach has been shown to yield reasonable fidelity using an in vitro model of human skull and brain, but haemoglobin concentration <6 g dl⁻¹ yields errors of ∼15% and increasing skull thickness produces errors as high as 32%.⁴⁰ Accordingly, some correction for extracerebral tissue must still be made even with such ‘absolute’ measurements.

NIRS limitations and confounds

Extracerebral tissue

Transcutaneous NIRS is reflective of a heterogeneous tissue field containing arteries, veins, and capillary networks and also other non-vascular tissue. For NIRS of cerebral tissue, photons must penetrate several tissue layers including scalp, skull, and dura, which can contain various concentrations of blood and tissue-derived chromophores. Both computer simulation and experimental tissue models of transcranial NIR light transmission have demonstrated an elliptical photon distribution centred around the transmitter whose mean depth is proportional to the separation of the optodes by a factor of ∼1/3.²¹ Increasing transmitter/receptor distance increases depth of penetration and minimizes the effect of extracerebral tissue,²¹ but power must be limited to prevent direct thermal tissue damage. Since signal intensity is inversely proportional to the square of the pathlength, 5 cm separation appears to be the functional maximal optode spacing.⁵⁸ This provides a mean depth of NIR light penetration ∼1.7 cm giving increased weighting to cerebral vs extracerebral tissue.⁵⁸ As there is still significant attenuation from extracerebral tissue even with optimized transmitter/receiver separation, additional techniques must be utilized.

Spatial resolution

Since mean depth of photon penetration approximates 1/3 the transmitter/receiver separation, by utilizing two differentially spaced receiving optodes—one spaced more closely and the other spaced farther from the transmitter—a degree of spatial resolution can be achieved. Accordingly, the closer receiver (e.g. 3 cm separation) detects primarily superficial tissue, whereas the farther optode (e.g. 4 cm separation) reflects deeper tissue. Incorporation of a subtraction algorithm enables calculation of the difference between the two signals and thus a measure of deeper, cortical tissue saturation. Thus differential spacing of receiving optodes can provide spatial resolution to distinguish signals from cerebral vs extracerebral tissue. In certain models, this has been interpreted as demonstrating transcutaneous photon penetration to the level of the cerebral ventricles.⁵⁸ It has been estimated that ∼85% of cerebral regional oxygen saturation (rSO₂) is derived from cortical tissue with the remaining 15% derived from overlying extracerebral tissue.

Cerebral arterial/venous blood partitioning

Cerebral NIRS devices measure mean tissue oxygen saturation and, as such, reflect haemoglobin saturation in venous, capillary, and arterial blood comprising the sampling volume. For cerebral cortex, average tissue haemoglobin is distributed in a proportion of ∼70% venous and 30% arterial,⁴⁷ based on correlations between position emission tomography (PET) and NIRS.⁵⁸ However, clinical studies have demonstrated that there can be considerable biological variation in individual cerebral arterial/venous (A/V) ratios between patients, further underscoring that the use of a fixed ratio can produce significant divergence from actual in vivo tissue oxygen saturation, thus confounding even ‘absolute’ measures of cerebral oxygenation, for example, fdNIRS or tdNIRS.⁸⁰ A further confound can be introduced if there is significant variation in haemoglobin concentration as a consequence of haemodilution which may give rise to changes in cerebral rSO₂ without attendant alterations in jugular venous oxygen saturation.⁸⁷ Whether this represents subclinical regional ischaemia, changes in photon pathlength, alterations in cerebral A/V partitioning, or other factors remain unclear.^60,87

In clinical practice, the use of cerebral NIRS as a trend monitor with interventions designed to preserve cerebral saturation values close to their individual baseline values has produced a significantly lower incidence of adverse clinical events in patients undergoing coronary artery bypass (CAB) surgery.⁵¹ A trend monitoring approach thus minimizes confounds introduced by biological variation in individual cerebral A/V ratios and outer layer tissue composition. These can produce an ‘offset’ in measured saturation values and result in inaccurate therapy if based on the assumption that a device is measuring ‘absolute’ in vivo cerebral oxygenation.

Extracerebral tissue

It is important to recognize that confounders such as extracerebral or subdural haematoma can change the proportion of cerebral to extracerebral haemoglobin and thus offset tissue oxygen saturation values by a variable amount. Using computed tomographic assessment of skull thickness (t-skull), cerebrospinal fluid area (a-CSF), and haemoglobin concentration, NIRO-100^® (Hamamatsu Photonics KH, Hamamatsu City, Japan) was compared with INVOS 4100^® (Somanetics Corporation) in a recent study of 103 cardiac surgical and neurosurgical patients.⁸⁸ This study demonstrated that INVOS rSO₂ values were potentially influenced by haemoglobin concentration, t-skull, and a-CSF. There was a potential confound in this evaluation as there was no assessment of superficial tissue attenuation of NIR light, for which INVOS uses a subtraction algorithm as compensation.⁸⁸ One implication of this is reflected in the potential for artifact and signal attenuation when extracerebral tissue is thickened or oedematous. Since haemoglobin represents the primary chromophore at these wavelengths, extracranial or subdural haemorrhage can artifactually influence measured cerebral saturation values. Based in part on PET studies, most clinical NIRS devices assume venous/arterial distribution in cortical tissue of ∼70/30%.³² Consequently, changes in rSO₂ largely reflect alterations in cerebral venous saturation and may also vary between patients.

Significant changes in extracerebral tissue saturation as induced by a scalp tourniquet have been shown to confound the ability to measure changes in cerebral rSO₂.²² However, in a non-tourniquet clinical study of oxygenation of blood drawn from both the facial vein and the jugular venous bulb, there was no correlation between cerebral rSO₂ and facial vein oxygenation, but there was a significant correlation between regional cerebral oxygenation and jugular venous bulb oxygenation.²⁴ The authors concluded that extracranial tissue oxygenation had a negligible influence on the values recorded using NIRS but noted that individual changes in jugular venous bulb oxygenation may be poorly reflected.²⁴

Non-haeme tissue chromophores

Because melanin pigmentation, as found in hair,⁶² can significantly attenuate light transmission and impede NIRS measurements, optimal placement of transmitting and receiving optodes is high on the frontal eminences, ∼2–3 cm above the orbital ridge to avoid frontal sinuses. Melanin in skin is confined to the epidermal layer at a depth of 50–100 µm and as such does not appear to produce significant attenuation of NIRS signal. However, conjugated bilirubin has an absorption peak at 730 nm, and is deposited throughout all tissue layers such that concern has been raised about the ability of NIRS to assess cerebral oxygenation in the presence of jaundice. In a study of 48 patients undergoing orthotopic liver transplantation, total plasma bilirubin was related to rSO₂ as determined from NIRS. During reperfusion of the grafted liver, rSO₂ increased by an average of 7%, and plasma bilirubin concentration did not influence the increase. Although bilirubin dampens the cerebral NIRS, even at high bilirubin values changes in cerebral perfusion can be discerned.⁴⁴ This further supports the approach of establishing a baseline value in each patient individually and observing for perturbations from that baseline rather than relying primarily upon a specific threshold value.

Non-metabolizing tissue

Tissue oxygen saturation in non-metabolizing tissue can be high or low, and can be near normal in dead or non-metabolizing brain because of sequestered cerebral venous blood in capillaries and venous capacitance vessels.¹⁷ Schwarz and colleagues⁷¹ examined rSO₂ in 18 adult human cadavers and found values in one-third of the subjects that exceeded the lowest values that they had previously recorded in normal subjects raising concern regarding the validity of the rSO₂ measurement. However, Maeda and colleagues⁴⁵ examined cerebral venous oxygen saturation during 214 autopsies and found the values to range from 0.3% to 95.1% apparently as a consequence of total haemoglobin content, cause of death, and cadaver storage conditions. Accordingly, rSO₂ or other measures of cerebral oxygen saturation can appear discordantly high, which rather than indicative of error may reflect the pathophysiology of non-metabolizing yet non-perfused tissue.⁷¹ In clinical practice, it is thus the detection of ‘context-sensitive change’ in cerebral NIRS (e.g. during cooling or rewarming) that is of fundamental importance rather than an ‘absolute’ value.

Clinical applications

A number of the limitations as discussed above have raised questions regarding the clinical utility of cerebral oximetry monitoring.^60,89 However, a number of clinical studies and case reports have demonstrated that despite such limitations, the ability of cerebral oximetry monitoring to detect otherwise clinically silent episodes of cerebral ischaemia in a variety of clinical settings renders it an important safeguard for cerebral function. In a recent study in patients with subarachnoid haemorrhage, episodes of angiographic cerebral vasospasm were strongly associated with reduction in trend ipsilateral NIRS signal.⁵ Furthermore, the degree of spasm (especially more than 75% vessel diameter reduction) was associated with a greater reduction in same-side NIRS signal demonstrating real-time detection of intracerebral ischaemia.

The proper management of brain oxygenation is one of the principal endpoints of all anaesthesia procedures, but the brain remains one of the least monitored organs during clinical anaesthesiology. There are some medical procedures where iatrogenic brain ischaemia is present, including carotid endarterectomy (CEA) in patients with high-grade carotid artery stenosis, temporary clipping in brain aneurysm surgery, hypothermic circulatory arrest for aortic arch procedures, and others in which the pathology itself generates brain ischaemia, such as traumatic brain injury and stroke. One of the most common limitations seen in studies assessing the impact of cerebral oximetry monitoring has been the absence of a defined protocol based on physiologically derived interventions to treat decreases in rSO₂. In order to provide a pathophysiological rationale for interventions and to facilitate clinical strategies designed to improve cerebral rSO₂, an intervention algorithm has been devised (Fig. 2),¹⁶ and has proven effective in improving outcomes in at least two separate randomized prospective clinical trials.^10,51

Fig 2

Proposed algorithm in the use of brain oximetry. CT, computed tomography; ICHT, intra-cranial hypertension; MAP, mean arterial pressure; MRI, magnetic resonance imaging. Reprinted from Denault and colleagues,¹⁶ with permission from SAGE Publications Inc.

Cardiac surgery

Coronary artery bypass surgery

There have been a number of case–control and retrospective studies of cerebral oximetry in cardiac surgical procedures that have shown improvements in outcome associated with cerebral oximetry monitoring and correlations between cerebral desaturation and adverse outcomes.¹⁸ However, to date there have been relatively few randomized, prospective clinical trials. In a study utilizing cerebral oximetry in 265 patients undergoing primary CAB surgery and randomized to active monitoring and a series of interventions designed to improve rSO₂ or to a control group in which blinded monitoring was used, a significant association was found between prolonged cerebral desaturation and early cognitive decline, and also a three-fold increased risk of prolonged hospital stay.⁷² However, cerebral desaturation rates were similar between the groups and ascribed to poor compliance with the treatment protocol, resulting in no difference in the incidence of cognitive dysfunction between the groups. In a prospective, randomized blinded study in 200 patients undergoing coronary artery grafting, active treatment of declining cerebral rSO₂ values prevented prolonged cerebral desaturations and was associated with a shorter intensive care unit length of stay and a significantly reduced incidence of major organ morbidity or mortality.^51,52 The intervention protocol undertaken to return rSO₂ to baseline resulted in a rapid improvement in rSO₂ in 84% of cases and did not add undue risk to the patient, including no increase in allogeneic blood transfusions.^16,51 There were also numerically fewer clinical strokes in the monitored patients consistent with previous studies. For example, a significant reduction in perioperative stroke rate, from 2.0% to 0.97%, was observed in patients in whom INVOS rSO₂ cerebral oximetry was used to optimize and maintain intraoperative cerebral oxygenation in comparison with an untreated comparator group operated upon in the preceding 18 month interval.²³

Deep hypothermic circulatory arrest

Moderate (25–30°C) and deep (<25°C) hypothermia remain a mainstay for cerebral and systemic protection during complex aortic arch repair, since surgical access can require interruption of systemic perfusion for relatively protracted periods. As there is relatively little ability to monitor cerebral function during such times since EEG becomes progressively attenuated below 25°C, cerebral NIRS has been advocated as a means of monitoring and detecting onset of cerebral ischaemia during deep hypothermic circulatory arrest.^38,39 In a study of 46 consecutive patients in whom selective anterograde cerebral perfusion (SACP) was established by perfusion of the right subclavian artery (with or without left carotid artery perfusion) or by separate concomitant perfusion of the innominate and the left carotid arteries, bilateral regional cerebral tissue oxygen saturation index was monitored by INVOS 4100 NIRS.⁵⁹ Six patients died in hospital, and six patients (13%) experienced a perioperative stroke in all of whom rSO₂ values were significantly lower during SACP and in whom rSO₂ tended to be lower in the affected hemisphere. During selective antegrade cerebral perfusion, regional cerebral tissue oxygen saturation decreasing to between 76% and 86% of baseline had a sensitivity of up to 83% and a specificity of up to 94% in identifying individuals with stroke. It was concluded that monitoring of regional cerebral tissue oxygen saturation using NIRS during SACP allows detection of clinically important cerebral desaturations and can help predict perioperative neurological sequelae.⁵⁹

There have also been a number of case reports of aortic arch surgery in which cerebral oximetry has been shown to detect cerebral hypoperfusion from a variety of factors including ascending aortic dissection with occlusion of carotid lumen,³³ intraoperative thrombosis of a common carotid graft,⁶⁹ kinking or obstruction of perfusion cannula during SACP,⁶⁷ or due to diminished Blalock–Taussig shunt flow after paediatric cardiac surgery.⁶⁵

Carotid endarterectomy

During CEA, temporary cross-clamping of the internal carotid artery (ICA) is an integral part of the surgery and can produce brain ischaemia in patients with poor collateral flow. The perioperative stroke rate after CEA can be as high as 5%,^2,66 a situation that renders intraoperative brain monitoring of special interest.⁵³ Monitoring devices such as transcranial Doppler (TCD), EEG, and somatosensory evoked potentials (SSEP) have been used successfully for a number of years but have logistic limitations and disadvantages.^3,12,15,83 In up to 20% of patients, TCD cannot be performed due to relative absence of transcranial window, while SSEP and EEG measurements are influenced by anaesthetic agents and electrocautery, and involve a high level of technical complexity. In at least one large clinical study of 314 patients undergoing awake CEA, EEG identified cerebral ischaemia in only 59% of patients needing shunt placement, with a false-positive rate of 1.0% and a false-negative rate of 41%, concluding that both stump pressure (SP) and EEG as a guide to shunt placement have poor sensitivity.²⁷ Although measurement of SP after common carotid clamping is also used as a method to assess adequacy of collateral flow through the Circle of Willis, it is not widely used in the clinical setting as it is affected by numerous factors including arterial pressure, Pa_co2, and type of anaesthetic agent,^31,50 and has the significant disadvantage of being a one-time discontinuous measurement, thus rendering it incapable of detecting ischaemia developing later during performance of the arterectomy.

Various studies have shown that cerebral oximetry monitoring can be a valuable tool for detection of cerebral ischaemia during CEA.^{8,9,30,50,77,85,86} In comparison with other modalities, non-invasive NIRS devices are easy and simple to use and provide continuous measurement of frontal cortex oxygen saturation. A major thrust of most studies utilizing intraoperative NIRS during CEA has been defining the sensitivity and specificity of changes in cerebral rSO₂ as correlated with either clinical signs of cerebral ischaemia or other neuromonitoring modalities. Among the earliest of these was a study from 1998 in which there was a positive correlation between TCD and NIRS comparing the percentage change in middle cerebral artery flow velocity vs change in rSO₂.³⁷ In 99 patients undergoing awake CEA with cervical plexus anaesthesia, regression analysis was used to evaluate the specificity and sensitivity of various rSO₂ cut-off points to detect neurologically defined intraoperative brain ischaemia.⁶⁸ A sensitivity of 80% with a specificity of 82% was found using a cut-off point of 20% relative decrease in rSO₂, with a false-positive and false-negative rate of 67% and 2.6%, respectively. Similar results were found in another study in which brain ischaemia with possible neurological compromise could result when cerebral rSO₂ was <54–56% during carotid cross-clamping. A reduction in rSO₂ of 16–18% during CEA was a predictor of neurological compromise.^30,68

A large cohort of NIRS data from 594 CEA performed under general anaesthesia was studied to determine the sensitivity, specificity, and predictive values of various rSO₂ cut-off points to predict the need for shunting or resulting in neurological complications.⁴⁹ The previously described 20% reduction by Samra and colleagues⁶⁸ was found to have a low sensitivity of 30% but a very high specificity (98%), with positive and negative predictive values of 37% and 98%, respectively. Accordingly, a cut-off point utilizing a 12% decrease in rSO₂ was identified as optimal, having a sensitivity of 75% and a specificity of 77% with a positive predictive value of 37% and a negative value of 98%.⁴⁹ Subsequently, in 50 patients having CEA under cervical plexus block, an independent neurologist evaluated clinical and EEG signs of ischaemia during continuous NIRS monitoring.⁶⁴ Ten per cent of patients experienced clinical and EEG brain ischaemia requiring shunt placement and in these patients, the reduction in NIRS averaged 17% whereas the NIRS reduction in those with no clinical or EEG ischaemia was 8%, a difference consistent with the 12% threshold as determined by Mille and colleagues.⁴⁹ Concern has been raised that in comparison with TCD, decreases in rSO₂ >13% during CEA, while sensitive, are less specific, have a false-positive rate of 17%, and can lead to unnecessary shunt placement.²⁵ However, in this study,²⁵ TCD was technically inadequate in four of 59 patients, underscoring the compromise between sensitivity, specificity, and reliability of these various monitoring modalities for intraoperative detection of cerebral ischaemia.

Overall, these studies indicate that utilizing a decrease in cerebral rSO₂ of >12% is a reliable, sensitive, and relatively specific threshold for brain ischaemia secondary to ICA clamping and necessitates shunt placement or other pharmacological or physiological intervention. A caveat is necessary, however, based on a recent report.²⁰ In this series, multi-modality neuromonitoring of 323 CEA procedures under general anaesthesia showed significant discrepancies in 24 patients (7.4%), of whom 16 showed no significant EEG/SSEP changes but profound changes occurred in rSO₂ and no shunt was placed, whereas in seven patients, there was no change in rSO₂ but a profound change in EEG/SSEP and shunts were placed. These authors reported that the sensitivity of rSO₂ compared with EEG/SSEP was 68%, and the specificity was 94% yielding a positive-predictive value of 47% and a negative-predictive value of 98%,²⁰ essentially similar to data from Mille and colleagues.⁴⁹

Post-CEA hyperperfusion syndrome

Postoperative neurological complication after CEA can be related to rebound increases in cerebral blood flow (CBF) after surgical repair of carotid stenosis. Impaired autoregulation as a consequence of chronic brain ischaemia with a rapid restoration of regional perfusion can generate a hyperperfusion syndrome characterized by headache, brain oedema, seizures, and in severe cases intracerebral haemorrhage.⁵⁷ A significant correlation between rSO₂ values immediately after declamping and changes in CBF was found with a sensitivity and specificity for detecting patients at risk of developing hyperperfusion syndrome of 100% and 86.4%, respectively, using a cut-off value of 5%.^56,57 With this 5% cut-off point, cerebral oximetry demonstrated a positive predictive value of 50% and a negative predictive value of 100%.

The use of NIRS has also been explored in head injury patients; however, the results have been equivocal. A poor correlation with ICP and jugular oximetry indicates a low sensitivity of cerebral oximetry after acute brain injury.⁷ However, good sensitivity of the cerebral oximetry for detection of intracranial haematomas correlating with computed tomography or MRI has been reported.³⁵ The use of cerebral oximetry in traumatic head injury remains an area of interest.²⁶

Paediatrics

In complex settings such as paediatric cardiac surgery, paediatric neurosurgery, and paediatric and neonatal intensive care, NIRS is being increasingly used to monitor and detect episodes of cerebral ischaemia both intraoperatively when combined with bispectral index monitoring,²⁹ and after operation where decreased cerebral rSO₂ within 48 h of surgery has been associated with adverse outcomes after the Norwood procedures.⁶¹ Premature and low birth weight infants are at significant risk for apnoea due to brain immaturity, intraventricular/intraparenchymal haemorrhage (IVH), and periventricular leukomalacia, the common pathophysiological pathway for all involving cerebral ischaemia.

Cerebral oximetry has been used to evaluate variations in the cerebral circulation in 11 preterm infants presenting with 145 apnoeic episodes.⁸⁴ Standard monitoring including Sp_o2, heart rate, ventilatory frequency, and arterial pressure was compared against the change in NIRS-derived cerebral blood volume and cerebral oxygenation during apnoeic episodes. A significant change in cerebral circulation was found during the apnoeic episode, such that when the Sp_o2 dropped below 85%, total cerebral haemoglobin increased and rSO₂ decreased.

In neonatal birth asphyxia, mild brain cooling has been utilized in an attempt to minimize subsequent cerebral hyperaemia and IVH. In a recent study, cerebral oximetry and EEG were used to document changes in cerebral perfusion during mild systemic cooling.¹ Cerebral NIRS identified a reduction in cerebral blood volume (CBV) during hypothermia that recovered during the rewarming period, whereas brain oxygenation remained stable. As brain cooling is thought to reduce delayed hyperaemia and to help maintain neuronal metabolism after cerebral insults, cerebral oximetry monitoring may be useful during hypothermia treatments in order to monitor changes in CBV and brain oxygenation as possible indicators of the efficacy of such treatment.

In many settings, mixed venous oxygen saturation (Sv_o2) is used to monitor the adequacy of cardiac output and as a surrogate for cerebral oxygenation during paediatric cardiovascular surgery and neonatal and paediatric intensive care.^75,81 Tortoriello and colleagues⁷⁵ validated the use of NIRS in estimating Sv_o2 in 20 paediatric cardiac surgery patients and demonstrated a positive correlation between rSO₂ and Sv_o2.⁷⁵ In a larger study of 155 critically ill neonates and infants, cerebral tissue oxygenation index (cTOI—defined as the ratio of oxygenated to total haemoglobin) correlated with arterial oxygen saturation, arteriovenous oxygen extraction, and central venous oxygen saturation.⁸¹ A significant correlation between cerebral rSO₂ and superior vena cava oxygen saturation during inhalation of either room air or oxygen 100% was reported in 29 postoperative paediatric heart transplant patients undergoing myocardial biopsy;⁶ rSO₂ was also the best predictor of pulmonary artery saturations. Intraoperative cerebral rSO₂ studied in comparison with Sv_o2 in 20 paediatric cardiac surgical patients <10 kg body weight showed that cerebral rSO₂ was more sensitive for cerebral desaturation and is thus an early and sensitive monitor of adequacy of brain perfusion because Sv_o2 primarily represents lower torso oxygenation status.⁶³ For paediatric patients in whom haemodynamic monitoring is necessarily limited, monitoring the adequacy of systemic perfusion using cerebral oximetry appears to be an appropriate surrogate.

Tissue perfusion

There is increasing interest in the utilization of cerebral oximetry sensors to monitor adequacy of tissue perfusion when placed on somatic sites in both adult and paediatric patients.^{4,41,70,76,78} In settings including volume rescuscitation in traumatic shock,^13,73,79 dehydrated paediatric patients,²⁸ and as an estimate of splanchnic perfusion after paediatric cardiac surgery,³⁶ somatic NIRS has been found to correlate with other indices of tissue perfusion. Lower extremity rSO₂ was used to confirm the development of compartment syndrome after surgical cutdown for vascular access,⁷⁴ whereas others have used NIRS to assess the effect of various anaesthetic agents on skeletal microcirculation.¹⁴

NIRS has been reported as a monitor for non-cerebral tissue oxygenation with the objective of comparing liver tissue oxygenation (TOI[liver]) with Sv_o2 and intestinal perfusion measured by gastric mucosa pH (pHi) in 20 paediatric patients undergoing craniofacial surgery with expected major blood loss.⁸² Although only a moderate positive correlation was demonstrated between TOI[liver] and Sv_o2 and gastric pHi, intra-individual TOI[liver] values, however, demonstrated close correlation with Sv_o2 values but a varying correlation with gastric pHi values. These investigators concluded that while TOI[liver] provided a better trend monitor of central venous oxygen saturation than pHi, overall, because of its limited sensitivity and specificity to indicate deterioration of Sv_o2, TOI[liver] was not felt to provide additional practical information for clinical management in this setting.⁸²

More recently, correlations between renal rSO₂, abdominal (splanchnic) rSO₂, and gastric tonometry, central mixed venous saturation, and blood lactate were examined in 20 postoperative neonates with congenital heart disease within 48 h of surgery.³⁶ There was a strong correlation between abdominal rSO₂ and pHi and also between abdominal rSO₂ and Sv_o2 and a significant negative correlation between the abdominal rSO₂ and serum lactate. The investigators concluded that abdominal site rSO₂, measured in infants with either single or biventricular physiology, exhibits a strong correlation with gastric pHi and also with serum lactate and Sv_o2 and that rSO₂ measurements over the anterior abdominal wall correlate more strongly than flank rSO₂ with regard to systemic indices of oxygenation and perfusion. Abdominal NIRS monitoring thus appears to be a valid modality providing real-time, continuous, and non-invasive measurement of splanchnic rSO₂ in infants after cardiac surgery for congenital heart disease.³⁶

The relationship between cerebral and somatic rSO₂ measured in cerebral, splanchnic, renal, and muscle has been compared with blood lactate levels measured in 23 children after repair of congenital heart disease.¹¹ Cerebral rSO₂ had the strongest inverse correlation with lactate level followed by splanchnic, renal, and muscle rSO₂, and an averaged cerebral and renal rSO₂ ≤65% predicted a lactate level ≥3.0 mmol litre⁻¹ with a sensitivity of 95% and a specificity of 83%. Overall, it was felt that averaged cerebral and renal rSO₂ <65% as measured by NIRS predicts increased lactate in acyanotic children after congenital heart surgery and may facilitate the identification of global hypoperfusion caused by low cardiac output syndrome in this population.¹¹

Somatic NIRS is also being investigated as an indicator of need for transfusion in trauma patients thought to be at high risk for haemorrhagic shock.⁷³ A minimum somatic rSO₂ <70% correlated with the need for blood transfusion with a sensitivity of 88% and a specificity of 78%, whereas the need for blood transfusion within 24 h of arrival was not predicted by hypotension, tachycardia, arterial lactate, base deficit, or haemoglobin. The authors concluded that somatic rSO₂ may represent an important screening tool for identifying trauma patients who require blood transfusion.⁷³

Other

Since cerebral dysautoregulation can occur in head injury and in a variety of other conditions, the potential for cerebral NIRS to provide a reliable bedside non-invasive assessment of cerebral autoregulation is being actively investigated in a variety of clinical settings and may provide a further refinement in the assessment of risk of cerebral ischaemia.⁵⁵ Cerebral oximetry sensors have also been demonstrated to detect progressive spinal cord ischaemia after sequential intercostal artery ligation in a large animal swine study.⁴³ A recent preliminary clinical report has correlated changes in spinal cord perfusion during lumbar CSF drainage with changes in rSO₂ from cerebral oximetry sensors located over the lumbar spine area in a patient undergoing thoracic endovascular thoracoabdominal stenting.⁵⁴ Overall, these studies suggest an increasing role for cerebral and somatic oxygen saturation monitoring that, despite limitations, provides the only indication of compromised brain and tissue perfusion in a number of clinical settings. Against the ease of use and continuous nature of such NIRS monitoring must be considered the relative sensitivity and specificity of such devices vs other monitoring modalities.

Funding

Supported by the Department of Anesthesia and Perioperative Medicine, University of Western Ontario.

Author notes

†

Declaration of interest. J.M.M. has received honoraria/lecture/travel fees from neuromonitoring companies including Somanetics and Nonin Medical, but has no stock equity or other such financial interests.