Brain oxymetry in the operating room: current status and future directions with particular regard to cytochrome oxidase

Central nervous system and psychiatric disorders are more likely to occur following cardiovascular surgery when compared to other surgeries.¹ A large prospective study that evaluated 2108 patients undergoing coronary artery bypass graft (CABG) surgery at 24 institutions recorded adverse cerebral outcomes in 6.1% of patients overall,² 3.1% of these patients experienced focal injury, stupor, or coma, with an associated in-hospital mortality rate of 21%. Although the incidence of stroke appears to be less than 1% for patients under the age of 65 years, the cumulative stroke risk for patients over age 65 is 5%; in two separate studies, the stroke rate in patients over age 75 years was 7 to 9%.^3- 4 Because cerebral injury continues to be a major source of morbidity and mortality after cardiovascular surgery, protection of the brain is a primary concern during operative and perioperative periods.^5- 7 Incriminated as a major factor in this complication is the presence of a level of cerebral perfusion that is inadequate to meet the cerebral metabolic demands during cardiopulmonary bypass (CPB) and/or selective cerebral perfusion, even though the patient is under deep hypothermia. Consequently, it is not the level of cerebral blood flow per se during cardiovascular surgery that is likely to be important, but an appropriate balance between cerebral perfusion and cerebral oxygen consumption.

Near-infrared spectroscopy (NIRS) is now in widespread use in such clinical fields as anesthesiology,⁸ neurosurgery,⁹ and pediatrics.^10- 11 NIRS is a noninvasive technique that enables physicians to continuously monitor variation in the oxygenation of hemoglobin and the redox state of cytochrome oxidase in living tissue. The cytochrome oxidase signal from NIRS has a controversial history. A seminal work in the study of cytochrome oxidas by NIRS is that of Matcher et al. ,¹² which showed that algorithms from different groups produced markedly different results when analyzing the same dataset, thereby highlighting the unreliable nature of the cytochrome oxidase signal. In the present work, we review NIRS by discussing the usefulness and limitations of the technique with particular regard to cytochrome oxidase, using an animal model and actual clinical data.

As expected, the risk of stroke after cardiac surgery increases with the severity of carotid disease: a follow-up study of 4047 patients examined before CABG reported that the risk of stroke increased from 1.9 to 6.3% in the presence of $>50%$ carotid stenosis.¹³ It is therefore important to survey the existence and severity of carotid disease. The carotid compression study was done against healthy volunteers $(n=4)$ and patients with cerebral vascular disease $(n=7)$. Figure 1 shows the typical changes in cerebral oxygenation and blood flow of the middle cerebral artery in a healthy volunteer, as assessed by NIRS and transcranial Doppler (TCD), respectively. Following compression of one carotid artery at the monitoring site, oxygenated hemoglobin ([oxy-Hb]) decreased and deoxygenated hemoglobin ([deoxy-Hb]) increased (as shown by point 2 in Fig. 1). At this time, blood flow in the middle cerebral artery decreased and arterial pulsation was absent; however, it is interesting that at several seconds after the start of compression, [oxy-Hb] and [deoxy-Hb] began to increase and decrease, respectively, and blood flow in the middle cerebral artery resumed (as shown by point 3 in Fig. 1). When the carotid artery was compressed at the monitoring site in some other healthy volunteers, similar reactions (delayed improvement of cerebral oxygenation and blood flow) were seen with some individual differences. The reason for this finding is the circle of Willis, a vascular structure at the base of the brain. This arterial ring directly connects the left and right carotid arteries via the anterior cerebral and anterior communicating arteries, and serves as important collateral circulation when one carotid artery is occluded. What happens if this collateral circulation is blocked? In 5 of 7 patients with cerebral vascular disease, cerebral oxygenation decreased and the blood flow rate in the middle cerebral artery was reduced when one carotid artery was compressed; the ischemic and hypoxic condition of the brain did not improve during the compression period. Figure 2 shows changes in cerebral oxygenation and the blood flow rate of the middle cerebral artery in a patient who was preoperatively diagnosed with occlusion of the anterior communicating artery. When one carotid artery was compressed, cerebral oxygenation decreased, with a reduction in blood flow rate in the middle cerebral artery; the ischemic and hypoxic condition of the brain did not improve during compression period (as shown by point 2 in Fig. 2). TCD has the advantage of being quantitative and that it actually measures the restoration of flow rather than a surrogate of flow; however, TCD is often hampered by impenetrable ultrasound windows.¹⁴ Thus, a brief compression test of the carotid artery using both NIRS and TCD monitoring could enable a survey of cerebrovascular disease, especially in regard to the severity of carotid disease and dysfunction of the circle of Willis.

The monitoring of brain oxygen status using NIRS was first applied to clinical practice by Jöbsis in 1977.¹⁵ Published studies indicat that NIRS could be used as a continuous and noninvasive method of observing changes in the cerebral oxygenation state during hypoxia and ischemia, because changes occur in the optical properties under these conditions; however, the interpretation of NIRS data, especially the cytochrome oxidase signal, remains controversial.^16- 18 Mitochondrial cytochrome oxidase consists of two hemoproteins (heme a and heme a3) and two copper proteins ($CuA$ and $CuB$). A wide absorption spectrum is seen in the infrared region, with a peak at $830nm$ for $CuA$; detection by NIRS of changes in its absorption enables the noninvasive and real-time assessment of the redox state of cytochrome oxidase. A possible source of error that might interfere with accurate measurements of the redox state of cytochrome oxidase is the derives overlapping of the absorption spectra for hemoglobin and cytochrome oxidase in the near-infrared region, with the absorption coefficient for hemoglobin being an order of magnitude greater than that for cytochrome oxidase.¹⁹ Several studies have shown that while the absorption coefficient of cytochrome oxidase was originally considered a constant, it appears to vary markedly in isolated mitochondria and in the blood-free perfused rat brain model, depending on the energy conditions.^20- 21 In other words, cytochrome oxidase cannot be accurately measured using a formula in which a constant is used for the absorption coefficient of cytochrome oxidase.

Clinical data shown in the present work were calculated using a formula derived by researchers at Hokkaido University (Tamura group).²² One important assumption of our algorithm is that in the near-infrared region (700 to $900nm$), the absorbance changes in the range shorter than $780nm$ are mainly attribute to Hb, whereas those at wavelengths longer than $780nm$ are attributed to both Hb and oxidized cytochrome oxidase. Thus, the changes in hemoglobin oxygenation and the redox state of cytochrome oxidase were calculated by the following algorithm:Display Formula

We applied our algorithm to an animal model²⁷ almost identical to that described by Sakamoto et al. ²⁶ After the institution of CPB, an aerobic (100% $O2$)-to-anaerobic (100% $N2$) transition was induced using a membrane oxygenator to determine the maximum changes in the full-scale value of the redox state of cytochrome oxidase. After stabilization under 100% $O2$, the animal received $200mg∕kg$ of sodium cyanide intravenously to inhibit electron transport from cytochrome oxidase to the oxygen molecule. Hematocrit was then reduced under CPB from 35 to 5% by hemodilution using Ringer’s solution. Figure 3 shows that after the infusion of sodium cyanide, [oxy-Hb] and [total-Hb] were significantly decreased during hemodilution, but that the cytochrome oxidase signal showed little change and that the variance of cytochrome oxidase during hemodilution was less than 4% of full scale (100% $O2$-to-100% $N2$ transition) under CPB in the dog model. The relative concentration of [total-Hb] (as measured by NIRS) showed a strong positive correlation with hematocrit values in the blood, but there was no correlation between the cytochrome oxidase signal and hematocrit values during hemodilution (Fig. 4).²⁷ Using our algorithm, cytochrome oxidase signals are completely unaffected, even when hematocrit values change from 35 to 5%. This may resolve the problems associated with cytochrome oxidase measurement that arise by blood dilution associated with cardiovascular surgery.

Because cytochrome oxidase is the terminal enzyme in the respiratory chain in mitochondria, its redox state is directly related to the intracellular oxygenation state and to the energy state of the tissue. Cooper et al. ²⁸ showed that when blood is withdrawn stepwise in rats, [oxy-Hb] falls linearly with decreases in the rate of oxygen delivery to the brain, whereas the redox state of cytochrome oxidase is unchanged until the oxygen delivery rate is reduced to below approximately one-half of the normal physiological level. However, other studies showed that cytochrome oxidase can be oxidized by increasing oxygen availability to the brain tissue under conditions of hyperoxia or hypercapnia, and have argued that cytochrome oxidase might be partially reduced.^29- 31 The question of whether cytochrome oxidase is partially reduced in normoxia remains unanswered.

In our previous study, under conditions of hyperoxia and/or hypercapnia in-vivo rat models, cytochrome oxidase did not cause further oxidation, whereas [oxy-Hb] was increased.²² When the electron flow rate through the respiratory chain is decreased by blocking of the glucose metabolism in the blood-free perfused rat brain model, the redox state of $CuA$ was slightly hyperoxidized, although not significantly so.²¹ Different values of a half-maximal reduction (P50) of the cytochrome oxidase for coupled mitochondria have been quoted by different authors: Sugano et al. ³² quoted 0.03 to $0.27μM$, depending on the metabolic activity; Gnaiger³³ quoted $0.24uM$; and Wilson et al. ³⁴ quoted approximately $0.7μM$. Hoshi, Hazeki, and Tamura²⁰ demonstrated that the oxygen dependence of the redox state of heme $a+a3$ depends on the mitochondrial energy state as well as the respiratory rate, but that $CuA$ is independent of these factors (the P50 of heme $a+a3$ was 0.078 and $0.16μM$ in states 4 and 3, respectively; the P50 of $CuA$ was $0.075μM$). These calibration studies showed that the P50 of cytochrome oxidase is much higher than that of Hb (P50 of Hb is $42μM$), by approximately three orders of magnitude. Based on the results obtained from both animal^21- 22 and clinical investigations,³⁵ cerebral cytochrome oxidase, especially $CuA$, may be fully oxidized under normal physiological conditions. Thus, the start of reduction can be used as an alarm, indicating that the brain has reached critical metabolic and functional levels.

Figure 5 shows changes in intracerebral oxygenation in a patient scheduled to undergo aortic arch replacement due to dissecting aortic aneurysm.³⁵ When extracorporeal circulation was initiated by supplying blood from the femoral artery, intracerebral cytochrome oxidase reduction was detected. As brain cells were thought to be in a critical state, arterial cannulae were quickly inserted into both right and left common carotid arteries and separate cerebral perfusion (SCP) was initiated; the redox state of cytochrome oxidase then rapidly returned to preoperative levels, and pupil dilation disappeared. Intraoperative findings revealed a flap at the base of the right brachiocephalic artery, and when blood was supplied from the femoral artery, the right common carotid artery became completely occluded.³⁵ As we know from numerous clinical experiments, hypothermia is beneficial in achieving cerebral protection during cerebral hypoxia and/or hypoperfusion; indeed, increases in PCr and ATP in the brain are observed during cooling.^36- 37 When our algorithm was used in the present study to measure the redox state of cytochrome oxidase, under deep hypothermia (nasopharyngeal temperature: $15°C$), the redox state showed oxidation after the initiation of SCP rather than the preoperative level (Fig. 5)³⁵; however, some groups reported that while the relative cerebral hemoglobin saturation increased during hypothermia, cytochrome oxidase showed a marked reduction.^17,38- 39 The previous studies suggested several potential explanations for this phenomenon, such as the increased affinity of hemoglobin for oxygen and the decreased efficiency of cytochrome oxidase in transferring electrons to oxygen, among others.³⁸ In this respect, the redox behavior of cytochrome oxidase in the present study appears to reveal oxidation rather than reduction under deep hypothermia, which was inconsistent with the conclusions of their previous study.^17,38- 39 Thus, further investigations are required to re-evaluate the redox behavior of cytochrome oxidase under hypothermia.

Cerebral injury continues to be a major source of morbidity and mortality after thoracic aortic surgery; therefore, protection of the brain is a primary concern during the operative and perioperative periods.^5- 7 The presence of a level of cerebral perfusion that is inadequate to meet the cerebral metabolic demands during CPB and/or selective cerebral perfusion, even though the patient is under deep hypothermia, has been incriminated as a major factor in this complication. To evaluate whether cerebral Hb oxygenation and the redox state of cytochrome oxidase during surgery, as measured by NIRS, are accurate predictors of postoperative cerebral outcome, we retrospectively examined the relationship between these two parameters in 66 patients who underwent thoracic aortic surgery.⁴⁰

In all cases, the various changes occurred in cerebral Hb oxygenation ([oxy-Hb], [deoxy-Hb], and [total-Hb]) at the initiation of CPB, and during the periods of hypothermia and rewarming. We classified the changes in the cerebral oxygen index ([oxy-Hb] minus [deoxy-Hb]) in the 66 cases at the completion of surgery as either a marked and prolonged increase (type X), a subsequent return to baseline (type Y), or a marked and prolonged reduction (type Z). When we retrospectively assessed the redox behavior of cytochrome oxidase during the operation in detail, we realized that there were three different types of cytochrome oxidase behavior. In terms of time course, we classed these as either no change (type A), a temporary, marked reduction with a subsequent return to baseline (type B), or a marked and prolonged reduction (type C) (Fig. 6). Figures 7- 9 show the typical type of cytochrome oxidase behavior. The characteristics of the type-A pattern are that the redox state of cytochrome oxidase remains at its initial level (no change) throughout the operation, even though marked changes occur in [oxy-Hb], [deoxy-Hb], and [total-Hb] at the initiation of CPB and/or during the periods of hypothermia and rewarming; 34 of the 66 cases (51.5%) were of this type (Fig. 7). In the type-B pattern, a transient reduction of cytochrome oxidase is observed, but recovery to baseline occurs by the end of the operation; 29 of the 66 cases (43.9%) showed this pattern (Fig. 8). Figure 9 shows the changes in cerebral oxygenation in a case representative of type C; only 3 of the 66 cases (4.5%) were of this type. The characteristics of type C are that the redox behavior of cytochrome oxidase shows a marked and prolonged reduction both during and following the operation.⁴⁰

In the present study, 9 of the 66 patients (13.6%) suffered from postoperative brain injury (Table 1). This resulted in severe coma (four cases), hemiparesis (three cases), convulsion (one case), and sight deficit (one case). We tested for significant differences between the occurrence of brain injury and the pattern of changes in the signals of hemoglobin and cytochrome oxidase using the chi-square test for independence ($l×m$ contingency table). There was no significant relationship between the occurrence of brain injury and the cerebral oxygen index type ([oxy-Hb] minus [deoxy-Hb]). In contrast, the relationship between the occurrence of brain injury and the cytochrome oxidase behavior type was highly significant (Table 2).⁴⁰

Although it appears that the cerebral oxygen index ([oxy-Hb] minus [deoxy-Hb]) by NIRS can measure regional changes in cerebral oxygen metabolism caused by blood dilution, cooling, rewarming, and changes in $FIO2$, there was no significant relationship in the present study between the occurrence of severe brain injury and the magnitude of the cerebral oxygen index at the end of the operation. We cannot explain exactly why the cerebral oxygen index ([oxy-Hb] minus [deoxy-Hb]) by NIRS could not detect the occurrence of a brain injury. However, one of the explanations for them is shown in the following paragraphs.

Figure 10 shows changes in the cerebral oxygenation of patients who had undergone cardiopulmonary resuscitation (CPR) three days prior. When pupil dilation occurred, indicating inadequate blood supply to the midbrain, the value of jugular venous oxygen saturation $(SjvO2)$ was extreme high, and cerebral hemoglobin oxygenation gradually increased; however, the redox state of cytochrome oxidase continued to decline, and the patient was diagnosed as brain dead two days later. Hyperoxia of the internal jugular venous blood has been reported in patients who suffered coma or brain death; this was thought to be because the external carotid regions had received “luxury perfusion,” and the absence of filling of intracranial arteries was commonly observed in the patients with brain death, resulting in hyperoxia of the venous blood returning from the extracerebral tissues.⁴¹ Kyttä et al. ⁴² have reported that high values of $SjvO2$ in patients who suffered brain death does not always indicate decreases in oxygen extraction in the intracranial brain tissue, and the degree of cerebral tissue hypoxia cannot be judged by near-infrared measurement of hemoglobin oxygenation alone. In the present work, we have shown that there was no significant relationship between the occurrence of severe brain injury and the magnitude of the cerebral oxygen index ([oxy-Hb] minus [deoxy-Hb]) (Table 2), which was almost the same as that in the case of brain dead patients observed by Kyttä et al. However, our data demonstrate that the redox behavior of cytochrome oxidase (as measured by NIRS) can provide direct, real-time information regarding crises in cerebral oxygen metabolism associated with relatively widespread hypoxia and/or ischemia, and that acquiring such information enables us to both assess the likely degree of cerebral damage and predict the postoperative cerebral outcome (Table 2).⁴⁰

Figure 11 shows a valuable clinical case. An 18-year-old man was involved in a motorcycle accident and underwent emergency surgery for aortic dissection. During surgery [Fig. 11], the patient suffered massive hemorrhage, life-threatening arrhythmia, cardiac arrest, pupil dilation, and a marked reduction in levels of cytochrome oxidase. Open-heart cardiac massage was immediately performed. The heart began to beat again $30min$ later, and cytochrome oxidase reoxidation was seen; however, cytochrome oxidase reduction was again detected $1h$ later. While monitoring cytochrome oxidase reduction, therapy was actively performed to protect the brain, including mild hypothermia $(32°C)$ and the administration of various drugs (magnesium, steroids, barbiturates, mannitol, glyceol, etc.). Complete oxidation of cytochrome oxidase was achieved several hours after the operation [Fig. 11]. The patient regained consciousness without brain injury. These findings suggest that in cases of severe brain injury (systemic brain ischemia, hypoxia, or cerebral edema), the brain can be resuscitated using active brain protective therapy while monitoring cytochrome oxidase redox. This shows that NIRS represents not only a monitoring technique for detecting brain injury, but also an indicator for assessing therapeutic effects in brain injury.

Lifshitz, Janmey, and McIntosh⁴³ suggest that mitochondria are maximally swollen at $3h$ after brain injury, and scattering decreases and transmission increases may incorporate components of swelling, internal structure modification, and changes in concentration of mitochondria, which can occur simultaneously. Time-resolved spectroscopy (TRS) is a new cerebral monitoring method that measures mean optical path length, and the absorption coefficient $(μa)$ and scattering coefficient $(μs′)$.⁴⁴ Since the therapies targeted toward mitochondria have proven beneficial, both histologically and behaviorally, in the recovery from ischemia⁴⁵ and trauma,⁴⁶ it might be important to monitor not only the redox state of cytochrome oxidase in the mitochondria by a conventional NIRS instrument, but also the scattering properties, which are influenced by swelling and structure modification in the mitochondria, by a TRS instrument.

NIRS has been clinically available for more than a decade, and initially has been given attention as a useful perioperative cerebral monitoring technique for noninvasively detecting intracerebral oxygenation. What are the applications of the currently available forehead-mounted NIRS devices? Central nervous system and psychiatric disorders are more likely to occur following cardiovascular surgery in comparison with other surgeries. Al-Rawi, Smielewski, and Kirkpatrick⁴⁷ showed that hemoglobin concentration in the brain calculated by using the modified Beer-Lambert law was significantly affected by extracranial blood flow changes, while tissues oxygen index (TOI) with the use of spatially resolved reflectance spectroscopy reflected intracranial oxygenation (not extracranial changes) with a high degree of sensitivity and specificity. Based on postmortem studies and correlative analysis of intraoperative events with neurologic outcomes, three primary mechanisms appear responsible for brain injury in otherwise uncomplicated cardiac operations: cerebral emboli, cerebral hypoperfusion, and cerebral hypoxia. The causes of these disorders include: 1. macroemboli (i.e., calcific and atheromatous debris) and microemboli (i.e., microgaseous bubbles, platelets, white blood cells, and fibrin agglutination); 2. decreased cerebral blood flow (pump oxygenator-related low flow, low perfusion pressure, hypocapnia, positional abnormality of blood supply/removal of cannulae, carotid artery disease, and cerebrovascular disorders); and 3. cerebral hypoxia induced by hemodynamic disturbance following respiratory failure. Unfortunately, NIRS cannot detect changes in remote areas from the measurement site, making it difficult to detect local brain injuries such as brain embolism using this method. However, forehead NIRS adequately detects changes affecting the entire brain, such as cerebral hypoperfusion or cerebral hypoxia. In particular, it readily enables the detection of medical accidents such as problems with extracorporeal circulation, occlusion of a supply tube used in separate cerebral perfusion, abnormal oxygen supply from an oxygenator, arterial or venous cannulae anomalies related to the use of an artificial pump, artificial respirator malfunctions, and abnormal respiration or circulation. Furthermore, the NIRS equipment setup requires little time, which is important in emergencies. Because cerebral flow correlates to cardiac output in shock,⁴⁸ NIRS should be used during shock or resuscitation as a device for monitoring respiration and circulation.

Xiao et al. ⁴⁹ reported that NIRS can be applied to monitor cerebral blood volume, Hb oxygenation state, cytochrome oxidase redox state, and water content following cardiac arrest in rats. The data obtained in the present study show that NIRS represents not only a monitoring technique for detecting brain injury, but also an indicator for assessing therapeutic effects following brain injury [Figs. 11]. Because the ultimate goal of resuscitation is social reintegration, noninvasive cerebral monitoring such as NIRS should be performed as soon as possible in brain resuscitation.

Questions have been raised regarding brain oxygenation by NIRS because intraoperative NIRS monitoring is known to have shown no abnormalities in patients who subsequently developed marked postoperative brain injury such as stroke.⁴⁰ Many clinicians are therefore skeptical about using NIRS for detecting brain injury. The current commercially available NIRS device is designed to be placed on the forehead; however, it is impossible to detect regional cerebral infarction in areas located far from the monitoring site (i.e., in the occipital lobe, temporal lobe, and basal ganglia), because the light cannot reach these areas. Ideally, if a probe could be placed in six locations (on the left and right of the forehead, temporal lobe, and occipital lobe), then most of the brain (except for the basal ganglia) could be covered, and detection sensitivity would be increased several fold.

In the clinical setting, NIRS signals obtained by blocking the internal carotid artery are clearly greater than those obtained by blocking the external carotid artery.⁴⁷ Therefore, a certain percentage of light signals detected by NIRS are assumed to originate from brain tissue; however, many clinicians question the penetration depth of NIRS. A recent study regarding the contribution of different parts of the head to the near-infrared signal revealed that most of the signal attributed to the brain arises from the upper 1 to $2mm$ of the cortical surface, for a source–detector separation of $30mm$.⁵⁰ These results indicate that the cerebral hemoglobin signal is highly sensitive to hemodynamic changes in the scalp because of the marked contribution of extracerebral tissue; however, changes in the redox state of cytochrome oxidase will be affected primarily by mitochondrial oxygen concentration, and may be less prone to extracerebral contamination because cytochrome oxidase is not present in red blood cells and is present in a lower concentration in the skin compared to that in brain tissue even though under “luxury perfusion,” resulting in hyperoxia of the venous blood returning from the extracerebral tissues (Fig. 10).

Numerous systematic review studies tend to report positive results regarding the clinical efficacy of NIRS for cerebral monitoring in cardiac surgery,^51- 54 for bedside cerebral blood flow measurements, and for mapping structure and function.^55- 57 However, most of these studies have major methodological limitations and low levels of evidence. Therefore, further investigations, such as multicenter randomized controlled trials, are clearly required to determine the clinical significance of cerebral monitoring (cerebral hemoglobin oxygenation and redox state of cytochrome oxidase) by NIRS.

We apply NIRS to both animal and clinical investigations. Based on these results, we discuss the significance of the measurement of cerebral hemoglobin oxygenation and cerebral cytochrome oxidase in vivo and in the clinical setting. From our data, we conclude that the redox behavior of cytochrome oxidase during surgery is likely to be a good predictor of the postoperative cerebral outcome. Our data indicate that although changes in local regions located far from the monitoring site may escape detection, the oxygen sufficiency or deficiency of brain tissue on a more global scale can be evaluated by near-infrared measurement. We are now sufficiently confident of its reliability to alter our strategy if the NIRS data indicate that the patient is experiencing deficient cerebral oxygenation; however, further investigation is required to establish its clinical efficacy and justify its routine use during operative and perioperative periods.

The authors gratefully acknowledge Yoko Hoshi (Department of Integrated Neuroscience, Tokyo Institute of Psychiatry) for helpful discussions.