Clinicians have sought methods to monitor tissue oxygenation for over a century.1 These efforts culminated in the development of the ear oximeter in the 1940s and then pulse oximetry in the 1970s. The observation that tissue including bone is transparent to light in the near-infrared spectrum (700 to 1,300 nm) led to the introduction of cerebral oximetry.2,3 A growing understanding of the importance of neurologic complications for patient outcomes after cardiac surgery has provided a strong impetus for implementing cerebral oximetry into clinical practice, albeit adoption in the United States remains low.4–7 Other applications continue to evolve, including monitoring during noncardiac surgery such as shoulder surgery in the beach chair position or during one-lung ventilation for thoracic surgery.8,9 The purpose of this clinically focused review is to summarize the applications of cerebral oximetry monitoring for adult patients undergoing cardiovascular surgery.
Principles of Near-infrared Spectroscopy Monitoring
In the United States, there are seven Food and Drug Administration (Silver Spring, Maryland)–approved cerebral oximetry monitoring systems including Invos 5100C and 7100 (Medtronic/Covidien, USA); ForeSight Elite (Casmed/Edwards Lifesciences, USA); Equanox three- and four-wavelength versions (Nonin Medical, USA); O3 Regional Oximetry (Massimo, USA), and the NIRO-200 NX system (Hamamatsu Photonics, Japan). All approved monitors use similar technology that is susceptible to artifacts and limitations as previously reviewed (fig. 1).2
Near-infrared spectroscopy measurement of tissue oxygen saturation is based on a modification of the Beer–Lambert law whereby the difference in the intensity of transmitted and received light is a function of the concentration of a light absorbing substance in a solution. Oxy- and deoxyhemoglobin have distinct peak absorption wavelengths in the near-infrared range. An isobestic wavelength at 810 nm provides a measurement of total hemoglobin concentration. Tissue saturation can be measured, thus, as the derived ratio of oxyhemoglobin to total hemoglobin concentrations. Other substances absorb near-infrared light including water, melanin, bilirubin, and cytochrome C, each with distinct but overlapping absorption spectra. Commercial cerebral oximetry monitors mostly restrict light to 700- to 850-nm wavelengths to focus on hemoglobin species.2
Cerebral oximetry monitoring is performed by attaching adhesive pads to the forehead that contain a near-infrared light source and sensors. Emitted photons must traverse various tissues including skin, muscle, bone, dura, and cerebral spinal fluid to reach the cerebral cortex. Although there is some scatter of photons at each tissue interface, their general path has been modeled to be elliptical with the depth of their penetration a function of the distance between the light source and sensors by a factor of roughly one third.2 A transmitter–sensor distance of 4 cm results in light penetration of tissue of approximately 1.3 cm. The use of two or more sensors at different distances from the light source allows for spatial resolution of reflected photons. For example, sensors placed 3 cm and 4 cm from the light source provide an estimate of light absorbed by superficial and deeper tissue, respectively. Subtraction algorithms evaluate the difference between deeper and superficial photon absorption for measurement of oxygen saturation.
It is estimated that 85% of cerebral oximetry measurements are derived from the superficial cerebrum and 15% from extracerebral tissue, including the scalp.2,10,11 This confounding effect of scalp light absorption might explain decrements in cerebral oxygen saturation observed during general anesthesia after the IV administration of the α-adrenergic agonist phenylephrine but not after administration of the indirect acting α- and β-adrenergic agonist ephedrine.12 An opposing argument is that an imbalance of cerebral oxygen supply versus demand occurs after α-adrenergic stimulation of receptors located on contractile pericytes in the microcirculation such that cerebral oximetry measurements are accurately reduced.13 The latter possibility was examined in a study of patients with brain tumors undergoing positron emission tomography where a continuous IV infusion of either phenylephrine or ephedrine was randomly administered.14 Cerebral metabolic rates for oxygen, cerebral blood flow, and cerebral oxygen saturation were not decreased by phenylephrine. Ephedrine, however, increased cerebral blood flow and cerebral oxygen saturation compared with phenylephrine. Thus, cerebral oximetry decrements after phenylephrine are not due to compromised cerebral oxygenation; its effects on cerebral oxygen saturation measurements are inconsistent.
Cerebral oximetry has been validated in humans after intra-arterial injection of the near-infrared absorbing dye idocyanine green, with positron emission tomography scanning, and by comparison with jugular mixed venous blood.15–17 The measurements do not require pulsatile flow as with pulse oximetry. Rather, it averages the oxygenation of venous, capillary, and arterial blood. Cerebral oximetry requires estimating the distribution of blood between the arterial and venous vasculature. Manufacturers use fixed ratio assumptions in their algorithms that range from 25 to 30% for arterial and 70 to 75% for venous cortical blood volume. These estimates, though, are not constant and vary between individuals and clinical scenarios. The accuracy of cerebral oximetry fixed arterial to venous blood ratio assumptions was investigated in 23 volunteers during isocapnic hypoxemia.18 All five cerebral oximetry monitors tested accurately detected hypoxemia induced by randomly varying the fractional inspired oxygen concentration from 100% to 70%. The average bias (difference between cerebral oxygen saturation reading and manufacturer weighted jugular venous and arterial blood oxygen saturation) with declining oxygen varied markedly between manufacturer and with arterial oxygen saturation. The ForeSight Elite, NIRO-200NX, and EQUANOX-3 wavelength had greater positive mean bias at low arterial oxygen saturation. A positive bias represents cerebral oximetry overestimation of the weighted average of arterial and jugular oxygen saturation. Bias was negatively influenced by darker skin pigment and female sex. The amount of bias during hypoxia was reduced when the difference between measured arterial and jugular venous oxygen saturation was used in the calculation rather than a fixed weighted average. Thus, the proportion of arterial and venous blood in the brain is not fixed but dynamically varies.
Cerebral Oxygen Desaturation
There is no consensus on what decrement of cerebral oxygen saturation from baseline represents a clinically important threat to cerebral oxygenation. Nonetheless, a widely used paradigm for defining a “desaturation” is a greater than 20% oxygen saturation reduction from baseline or an absolute value of less than 50%.2 These definitions originate from studies of patients undergoing carotid endarterectomy where cerebral oxygen saturation was evaluated during documented episodes of cerebral ischemia.19–22 We prospectively applied this definition in a observational study of 235 patients at eight centers to determine the frequency of cerebral oxygen desaturation during cardiopulmonary bypass.23 We observed desaturations in 61% of patients. A corrective intervention algorithm was effective in restoring cerebral oxygenation in 92% of persistent episodes. In a randomized, masked study involving 201 patients undergoing cardiac surgery at eight different centers, cerebral desaturation defined as a decrement greater than 10% from baseline occurred in 63% of patients and 40% had decrements greater than 20% of baseline.24 Due to wide interindividual steady state variability and dynamic error in the measurements, cerebral oximetry does not provide an absolute cerebral oxygen saturation measurement but rather should be interpreted as a trend monitor.18,25 Thus, determining an absolute threshold for defining “desaturation” is difficult. As a result, the use of cerebral oxygen saturation cutoffs to define desaturation in clinical practice may have less relevance than in clinical research.
There is a variety of proposed applications of cerebral oximetry monitoring for patients undergoing cardiac surgery, including for preoperative risk stratification.26 Case reports have described its value for detecting malposition of aortic and venous cannulae, particularly during aortic arch surgery, or to confirm selective antegrade cerebral perfusion.27 The highest interest for cerebral oximetry understandably has been to reduce the risk of neurologic complications. Other applications include monitoring of somatic tissue oxygenation, monitoring of patients during extracorporeal membrane oxygenation, and cerebral blood flow autoregulation monitoring.
Avoiding Neurologic Complications
Neurologic complications after cardiac surgery have a spectrum of manifestations that include delirium, delayed neurocognitive recovery (i.e., postoperative cognitive dysfunction), and clinical or subclinical stroke.4–6 Of these, the mechanism(s) of stroke is best understood with the majority of episodes due to cerebral embolism. Cerebral hypoperfusion, though, can directly cause ischemic injury, particularly in the setting of cerebral arterial stenosis or interruption of cerebral blood flow. It may further exacerbate ischemic injury from embolism by reducing collateral flow and/or reducing microembolism washout.28 The value of cerebral oximetry for detecting cerebral hypoperfusion has been evaluated in patients undergoing carotid endarterectomy. In a study of 323 patients, cerebral oximetry had a sensitivity of 68% and specificity was 94% for detecting cerebral ischemia during carotid artery cross-clamping compared with electroencephalography or somatosensory evoked potentials.21 It must be acknowledged that the different brain regions monitored may have varying collateral blood supply.
In contrast to stroke, the etiology of neurocognitive dysfunction is unclear, but it is likely the result of the interaction of patient-related factors with any combination of cerebral microembolism, hypoperfusion, central nervous system inflammation, or other variables.4–6 In a recent randomized trial, we found a 28% relative reduction in the risk for postoperative delirium in patients having blood pressure targets during cardiopulmonary bypass set above the lower limit of cerebral autoregulation compared with usual care.29 This suggests that optimizing cerebral perfusion may allow for reducing the risk for postoperative delirium.
Evidence on whether interventions for cerebral oxygen desaturations mitigate risk for neurologic outcomes in the past has been limited to case reports and observational studies.27 There are now several randomized, controlled algorithm-based interventional trials for cerebral oxygen desaturations in patients undergoing cardiac surgery. Two recent meta-analyses of these trials have been published and are summarized in table 1.30,31 Variability between meta-analyses is likely explained by differing criteria for study inclusion and approach to data analysis. These meta-analyses demonstrate that currently there is insufficient evidence to support or refute whether interventions for cerebral oxygen desaturations during surgery improve neurologic outcomes due to the many limitations of the existing studies, especially the low number of studied patients. In an eight-center pilot study, it was estimated that 3,080 evaluable patients are likely needed to assess differences in the rate of mortality, 4,638 patients for stroke, and 1,610 for delirium. Importantly, a large percentage of strokes after cardiac surgery occur postoperatively and thus are not likely to be influenced by intraoperative interventions alone.32 Further, the mechanism of delayed cognitive recovery is not known and perhaps not prevented solely by avoiding compromised cerebral oxygenation only during surgery. We have found that hypotension after surgery in the intensive care unit is common and associated with brain injury biomarker release.33 Thus, a strategy of ensuring cerebral perfusion in both the operating room and the intensive care unit may be necessary. Finally, monitoring of any physiologic parameter can only provide information to clinicians. It is the corrective interventions based on the information provided by the monitor that can potentially impact patient outcomes.34
Noncerebral Tissue Oxygenation Monitoring
Near-infrared spectroscopy monitoring of somatic tissue oxygenation strongly correlates with cardiac output in experimental models of hemorrhagic shock and resuscitation.1,35 This type of monitoring has been proposed for tracking somatic tissue oxygenation in patients with extremity compartment syndromes, in patients with peripheral vascular disease undergoing revascularization, and to monitor oxygenation of free flaps.1 Case reports and observational studies have reported the use of somatic tissue oxygen saturation monitoring of the thenar eminence, forearm, and lower leg in patients undergoing cardiac surgery. A single-center observational study of 121 patients found that somatic tissue oxygen saturation (measured over the flank) less than 65% or a greater than 20% decline from baseline predicted acute kidney injury after surgery.36 A limitation of renal monitoring in adults is the variability in skin and subcutaneous tissue depth making it unknown whether the kidney or even large flank muscles are indeed interrogated with near-infrared light that may penetrate only approximately 1.3 cm.
Mechanical Circulatory Support
The use of extracorporeal membrane oxygenation for treating cardiac and/or pulmonary failure continues to expand. Although lifesaving in many situations, its use is associated with complications including neurologic complications such as stroke and seizures.37,38 Differential hypoxia syndrome (“Harlequin syndrome”) where Pao2 is less in the upper compared with the lower body can occur with peripheral venous–arterial extracorporeal membrane oxygenation. This complication occurs when left ventricular ejection of blood with a low Pao2 mixes in the aorta with blood coming from the extracorporeal membrane oxygenation circuit having a higher Pao2. Differential hypoxia syndrome is associated with acute cerebral dysfunction that can predispose to stroke or seizures. Diagnosis is challenging when critical illness or sedation renders patients incapable of participating in neurologic examination. Attenuated or absent arterial pulsatility with extracorporeal membrane oxygenation may interfere with upper and lower extremity pulse oximetry that may compromise its use to detect this syndrome. For this reason, some centers monitor cerebral oximetry, which is not reliant on arterial pulsations. In a series of 56 patients on veno-arterial extracorporeal membrane oxygenation, 32% developed acute cerebral dysfunction.37 Cerebral oxygen desaturations occurred in 74% of the patients. Cerebral oxygen desaturation and high lactate concentrations were independently associated with mortality. Early detection of differential hypoxia syndrome allows for adjustments of extracorporeal membrane oxygenation drainage/flow to reduce the content of ejected left ventricular blood with a low Pao2.
A pitfall of peripheral veno-arterial extracorporeal membrane oxygenation is the risk for lower-limb ischemia with femoral cannulation.39 The traditional approach to diagnosis of limb ischemia is intermittent clinical examination and Doppler pulse examination that can delay detection. Somatic tissue oxygenation monitoring with near-infrared spectroscopy can be used to ensure extremity perfusion distal to the site of vascular access.40 Placing sensors bilaterally on the feet or superficial lower leg muscles allows for comparing tissue oxygenation between the extremities. This method can be used as well with vascular access with large cannula as for percutaneous left ventricular assist devices. Early detection of vascular compromise allows for vascular cannula modification and/or insertion of an arterial shunt to avoid distal extremity ischemia.
Cerebral Autoregulation Monitoring
The brain has several homeostatic processes to maintain a constant supply of oxygenated blood to meet its high metabolic demand. Cerebral blood flow-pressure autoregulation is one such process whereby cerebral blood flow is maintained across a range of blood pressures, ensuring a steady supply of oxygenated blood. When blood pressure is below or above the constraints of autoregulation, cerebral blood flow is pressure-passive, predisposing to cerebral ischemia or cerebral hyperemia, respectively. Monitoring of cerebral autoregulation at the bedside is possible using a variety of methods that mathematically model cerebral blood flow or cerebral blood volume in response to cerebral perfusion pressure (or arterial pressure in the absence of elevated intracranial pressure). Cerebral blood flow as the output signal can be monitored with middle cerebral artery transcranial Doppler blood flow velocity measurements. Mean velocity index is calculated as the Pearson correlation coefficient between low-frequency (less than 0.05 Hz) changes in cerebral blood flow velocity and perfusion pressure. This index is near zero when cerebral blood flow is autoregulated since flow is constant despite changes in perfusion pressure. In contrast, when perfusion pressure is below or above the limits of autoregulation, mean velocity index approaches 1 since changes in pressure alter cerebral blood flow. Intracranial pressure (when clinically monitored) can serve as a surrogate for cerebral blood volume and modeled against cerebral perfusion pressure for continuous autoregulation monitoring.41 Similar to mean velocity index, pressure reactivity index is calculated as the Pearson correlation coefficient between low frequency (less than 0.05 Hz) changes in intracranial pressure and perfusion pressure. Arteriolar dilation or constriction in response to changes in perfusion pressure collectively effect cerebral blood volume, resulting in changes in intracranial pressure. Functional autoregulation is associated with a pressure reactivity index near zero, but it approaches 1 when autoregulation is impaired.
Cerebral oxygen saturation represents the balance of cerebral oxygen supply (i.e., cerebral blood flow, hemoglobin concentration, oxygen saturation) versus oxygen metabolic demand. At low frequencies in the range associated with autoregulation, cerebral oxygen saturation has been shown experimentally to be an acceptable surrogate of cerebral blood flow, providing a clinically feasible method for autoregulation monitoring, as we have described.42 Calculating the Pearson correlation coefficient between low frequency (less than 0.05 Hz) changes in cerebral oxygen saturation and perfusion pressure yields cerebral oximetry index. Similar to mean velocity index, cerebral oximetry index near zero indicates independence of cerebral blood flow from perfusion pressure, while when pressure is below or above the limits of autoregulation, cerebral oximetry index approaches 1.42 Our group has validated the cerebral oximetry index experimentally and in adults during cardiopulmonary bypass.43
The pressure reactivity index has been used in patients with traumatic brain injury for monitoring autoregulation where it has been shown to predict poor outcome.41,44 Intracranial pressure monitoring is not typically performed in patients undergoing cardiac surgery, but a similar index can be derived noninvasively from the total hemoglobin concentration measured with near-infrared spectroscopy at the isobestic wavelength.45 Vasodilation and vasoconstriction, respectively, reciprocally increase or decrease the total hemoglobin concentration due to collective changes in cerebral vessel capacity. The variable hemoglobin volume index represents low-frequency correlation between total hemoglobin concentration and perfusion pressure (fig. 1).
Our group has performed clinical studies of adult patients undergoing cardiac surgery using cerebral oximetry index. In retrospective analysis, we found that the product of the magnitude and duration that mean arterial pressure was below the lower limit of autoregulation during cardiopulmonary bypass was independently associated with risk for acute kidney injury as well as for major morbidity and mortality after cardiac surgery.46,47 Importantly, simply raising mean arterial pressures targets during cardiopulmonary bypass might result in arterial pressure above the upper limit of autoregulation and risk for cerebral hyperperfusion. We have found that the product of the magnitude that mean arterial pressure is above the upper limit of autoregulation is associated with risk for delirium.48 Our group thus has argued in favor of personalizing blood pressure management during surgery with cerebral autoregulation monitoring.49
Near-infrared spectroscopy provides a clinically feasible method for monitoring cerebral oxygen saturation. Limitations with the technology and interindividual variations in these measurements must be acknowledged such that cerebral oximetry should be interpreted as a trend monitor. The insufficient evidence currently available from randomized controlled trials does not prove or disprove that optimizing cerebral oxygen saturation during cardiac surgery can improve patient outcomes. New applications of cerebral oximetry including for patients undergoing extracorporeal membrane oxygenation and as a bedside monitor of cerebral autoregulation might expand its clinical utility.
The study was funded in part by a grant from the National Heart, Lung, and Blood Institute (Rockville, Maryland; NIH RO1HL092259; Dr. Hogue, principal investigator).
Dr. Hogue aas received payment for advisory board membership from Medtronic, Inc. (Minneapolis, Minnesota) and Edwards Lifesciences (Irvine, California). He serves on a Data Safety Monitoring Committee for Merck, Inc. (Kenilworth, New Jersey). The remaining authors declare no competing interests.