2.2. Predictors of brain injury

2.2.1.1. Creatine phosphokinase isoenzyme BB

Creatine phosphokinase has three isoenzymes: the muscle type (CK-MM), the heart type (CK-MB) and the brain type (CK-BB). The function of creatine phosphokinase is to transfer an energy-bond between ATP, PCr and ADP (Bakay et al. 1986). The molecular mass of CK-BB is 40 to 53 kDa, and it is normally found in the serum in very low concentrations (the upper normal serum concentration is 3µg/L) (Phillips et al. 1980). CK-BB is not normally found in CSF, and it does not cross the intact blood-brain barrier (Maas 1977a, Maas 1977b). Somer introduced CK-BB as a potential marker of brain injury (Somer et al. 1975). In brain injury, an elevated concentration of CK-BB was measured from CSF and serum (Kaste et al. 1977). The CSF concentration of CK-BB in piglets correlated with increasing periods of HCA (Fessatidis et al. 1993b) and has been shown to be a valuable marker of neurological outcome after cardiac arrest (Roine et al. 1989). Lundar observed an association between high CK-BB activity in the CSF and adverse neurologic outcome after HCA (Lundar et al. 1983a).

However, CK-BB as a marker of cerebral injury has several weaknesses. A few years ago the radioimmunoassay method had a cross-reaction between CK-BB and CK-MB, but nowadays this problem has been solved (Bell et al. 1978). Another methodological weakness is the high extracerebral concentrations of CK-BB as compared, for example, with S100β protein, NSE or especially GFAP (Table 4). Hypothermia also increases the CK-BB concentration in serum, both with and without CPB (Vaagenes 1986, Vaagenes et al. 1987).

In summary, CK-BB is not a very specific marker of brain injury after cardiac surgery, new markers having been shown to be superior.

2.2.1.2. S100β protein

The S100 protein is a small dimeric cytosolic calcium-binding protein with a molecular weight of 22 kDa (Zimmer et al. 1995). It exists in various forms depending on its chain (α or β ) structure (Zimmer et al. 1995). The αα-form is found in striated muscles, heart and kidney (Haimoto & Kato 1988, Hasegawa et al. 1993). The isoforms αβ and β β are predominantly present in astroglial and Schwann cells, and are commonly referred to as the brain-specific S100β protein (Ingebrigtsen & Romner 2002); they are also found in the adipose tissue at lower concentrations and in other tissues (Table 4) (Haimoto et al. 1987, Johnsson 1996).

An increased serum and CSF concentration of S100β protein indicates both neuronal injury and increased permeability of the blood-brain barrier (Persson et al. 1987). Elevated serum S100β protein levels are found in the blood and CSF after cerebral stroke, subarachnoid hemorrhage, cranial trauma, coma after cardiac arrest and many other neurological disorders (Sindic et al. 1982, Persson et al. 1987, Missler et al. 1997, Grocott et al. 1998, Rosen et al. 1998, Raabe et al. 1999). Because of this, serum S100β protein has been suggested as a promising marker of brain injury in cardiac surgery (Johnsson 1996, Westaby et al. 1996). S100β protein is eliminated from the serum by the kidney and excreted in the urine. Its exact biologic half-life is not known, but recent studies suggest that it is well below 60 minutes (Jönsson et al. 2000).

The first study on S100β protein measurement after cardiac surgery was published in 1992 (Sellman et al. 1992), and showed no measurable S100β protein concentration in the CSF after CPB. A certain association between serum S100β protein levels and postoperative cerebral complication was published in 1995 (Johnsson et al. 1995). Since a positive correlation between CPB perfusion duration and S100β protein was observed, serum S100β protein level was considered to be indirect proof of brain injury (Westaby et al. 1996). Increased serum levels of S100β protein were demonstrated to be significantly associated with cerebral embolic event (Grocott et al. 1998) and adverse early postoperative neuropsychological outcome by numerous studies (Westaby et al. 1996, Blomquist et al. 1997, Kilminster et al. 1999, Georgiadis et al. 2000, Herrmann et al. 2000b). The association between increased serum S100β protein concentrations and long-term neuropsychological outcome is, on the other hand, less clear (Herrmann et al. 2000b, Westaby et al. 2000).

During operations under HCA, a positive correlation has been demonstrated between S100β protein serum levels and the duration of HCA and CPB respectively (Lindberg et al. 1998, Bhattacharya et al. 1999, Wong et al. 1999).

However, an increased serum level of S100β protein is observed in almost all patients undergoing CPB, and there was a variation in S100β protein levels at different sampling times (Jönsson et al. 1998). Serum levels and the degree of variation decreased gradually with time after termination of extracorporeal circulation, and an increase in S100β protein during CPB as a marker of brain injury was questioned (Jönsson et al. 1998). In fact, it has been shown that S100β protein is released from mediastinal adipose tissue during cardiac operations (Jönsson et al. 1999), and an increased concentration of S100β protein was detected in cardiotomy suction blood.

The conclusion based on increased serum S100β protein levels after CPB as a marker of disturbance of the blood-brain barrier during CPB could also be erroneous (Bokesch 1999, Lloyd et al. 2000), because CPB did not actually increase the serum S100β protein (Anderson et al. 2001).

The timing of increased S100β protein in the serum has been pointed out as being of key prognostic importance: when the serum level of S100β protein was elevated 48 hours after coronary artery bypass surgery, it had a negative predictive value for medium-term survival (Jönsson et al. 2001) and long-term survival (Johnsson et al. 2003) and correlated positively with the size of the infarcted brain tissue (Jönsson et al. 2001). In fact, it has been shown that seven hours after the end of CPB, the S100β protein released from the fat cells is not detected in the blood, and an increased S100β protein level correlates with decreased memory function (Svenmarker et al. 2002).

In summary, the S100β protein is a potential marker of brain injury. Serum analysis of S100β protein is flamed by contamination during CPB, which causes a certain limitation in its accuracy. However, at 7, and surely 48, hours after surgery it is a very good marker of brain injury and a predictor of outcome.

2.2.1.3. Neuron-specific enolase

Enolases are a family of ubiquitous glycolytic enzymes occurring as series of dimeric isoenzymes including three subunits, the α, β and γ chains (Cooper 1994). The isoforms γ γ and αγ are restricted to neurons, where they act as glycolytic enzymes in cytoplasm, and are named neuron-specific enolase (NSE). NSE is located in the cytoplasm of neurons, and has a molecular weight of 78 kDa (Ingebrigtsen & Romner 2002). The biologic half-life of NSE in serum is 24 hours (Ishiguro et al. 1983). It was first used as a tumor marker for small-cell lung cancer, neuroblastoma, and other malignancies of neuroendocrine origin (Cooper 1994). Later, it was introduced as a marker of brain damage (Steinberg et al. 1984, Sellman et al. 1992).

Serum NSE levels have been shown to correlate with infarct volume (Missler et al. 1997) and functional impairment after stroke (Wunderlich et al. 1999). In cardiac surgery, postoperative serum concentrations of NSE have a predictive value with respect to early neuropsychological and neuropsychiatric outcome after cardiac surgery (Herrmann et al. 2000b). In patients with traumatic head injuries, a serum NSE concentration of more than 10µg/L is considered pathologic (Raabe et al. 1998). A recent study suggested that NSE is the most useful marker of brain injury, and the most appropriate timing for NSE blood sampling is 36 hours after coronary artery bypass grafting (Rasmussen et al. 2002). However, this study included only 15 patients, and a significant correlation with the outcome was observed only in one of the several sampling points after surgery. This finding is thus far from being conclusive.

The measurement of NSE in the serum is also associated with a number of methodological weaknesses. NSE protein is present at relatively high levels in erythrocytes, thus even mild haemolysis is enough to increase significantly the serum levels of NSE (Brown et al. 1980). Furthermore, high extracerebral concentrations of NSE can also be detected in other tissues (Table 4). Certainly, this is a major problem when enclosing this marker in a situation likely to be associated with hemolysis, such as CPB (Pierangeli et al. 2001). Because of these problems, serum NSE has failed to be a specific and efficient marker of brain injury during and shortly after CPB (Georgiadis et al. 2000, Johnsson et al. 2000) and during HCA (Hovels-Gurich et al. 2001).

Certain problems with serum analysis have lead to attempts to use CSF NSE as a marker of brain injury. A significant increase in CSF NSE levels has also been demonstrated in patients suffering traumatic brain injury (Ross et al. 1996) and after CPB (Sellman et al. 1992). A correlation between CSF NSE levels and the Glasgow coma scale score has been demonstrated after traumatic brain injury (Ross et al. 1996). The methodological weakness of CSF NSE is that the measurement shows high sensitivity to blood contamination during sampling, because NSE is present in blood cells in significant concentrations (Schmitt et al. 1998). The predictive value of CSF NSE levels in the detection of brain injury after CPB is estimated to be limited, and even elevated CSF NSE levels might be related to blood-brain barrier disturbances (Schmitt et al. 1998).

In summary, the value of the NSE as a marker of brain injury after cardiac surgery with CPB is highly questionable.

2.2.1.4. Glial Tissue-Specific Protein

Glial fibrillary acidic protein (GFAP) is an intermediate filament protein expressed almost exclusively in the astrocytes, where it represents the major part of the cytoskeleton. GFAP is a monomeric molecule with a molecular mass ranging from 40 to 53 kDa (Missler et al. 1999). Increased concentrations GFAP in the CSF have been shown in normal-pressure hydrocephalus (Albrechtsen et al. 1985), dementia (Teunissen et al. 2002) and stroke (Aurell et al. 1991). GFAP is a very sensitive and specific marker of rapid astrocytic response to injury and diseases (Eng & Ghirnikar 1994). The measurement is not sensitive to the effect of haemolysis, the concentration of GFAP is stable for at least three freezing and thawing cycles and normal freezer storage (van Geel et al. 2002). Hermann and colleagues provided the first systematic clinical evaluation of a strong association between serum levels of GFAP and severity of stroke (Herrmann et al. 2000a). Serum GFAP protein seems to be a promising marker for brain injury, and it seems to fulfill the demand for a highly specific marker for brain injury (Herrmann et al. 2000a). However, further investigations are required to evaluate its prognostic accuracy in patients undergoing cardiac surgery.

2.2.1.5. Other markers of brain injury

A number of molecules and proteins have been suggested as markers of neuronal damage. Adenylate kinase is an intracellular cytoplasmic enzyme that is present in neurons as well as in other cells such the erythrocytes. In the case of ischemic brain damage, the CSF concentration of adenylate kinase is increased significantly (Ronquist & Frithz 1982), and its increase has also been observed in the case of neurological dysfunction after CPB (Åberg et al. 1984). However, its association with neurological outcome is rather weak (Johnsson 1996). The problem is the high concentration of adenylate kinase in the blood, and contamination from blood is possible unless specific precautions are taken (Bakay et al. 1986).

ICAM-5 (telencephalin) is only expressed in the somatodendritic membranes of telencephalic neurons. 48 hours after hypoxic-ischemic injury, the level of ICAM-5 was elevated in serum, and it was suggested as a potential marker for somatodendritic neuronal damage (Guo et al. 2000).

CSF cleaved tau proteins are structural microtubule binding proteins primarily localized in the axonal compartment of neurons. Elevated CSF cleaved tau proteins have been demonstrated in patients suffering from traumatic head injury and in multiple-sclerosis patients (Zemlan et al. 1999). CSF tau protein levels have been shown to predict the increase of ICP and the clinical outcome after traumatic head injury (Zemlan et al. 2002). Cleaved tau proteins can also be measured from blood, and a preliminary study showed increased C-tau levels in patients with unfavorable outcome after severe head injury (Chatfield et al. 2002). Further studies are required to explore whether C-tau could be used as a marker for brain ischemic injury (Chatfield et al. 2002).

Myelin basic protein (MBP) is detectable in developing oligodendroglia, and it is bound to the extracellular membranes of central, and to a lesser extent peripheral, myelin (de Vries et al. 2001). However, as a marker of brain injury, NSE is superior compared to myelin basic protein (Garcia-Alix et al. 1994).

Endothelin-1 is identified as a most potent vasoconstrictor peptide with 21-amino acids. It is present in endothelial cells, were it also exists in two different isopeptides, endothelin-2 and endothelin-3 (Lampl et al. 1997). Endothelin-1 has been shown to participate in astrocyte activation and oxidative stress after trauma (Beuth et al. 2001). It is elevated in the CSF after stroke (Lampl et al. 1997), and after traumatic head injury both in the CSF and blood (Beuth et al. 2001). A correlation between the CSF concentrations of endothelin-1, the volume of the lesion, and the Matthew Scale score has been shown after stroke (Lampl et al. 1997). After traumatic head injury, a correlation between the concentration of CSF endothelin and the Glasgow coma scale has been observed (Beuth et al. 2001). Studies have shown that an increase in endothelin-1 may exacerbate brain injury associated with head injury or stroke (Sato & Noble 1998, Park & Thornhill 2000). Further studies are required to evaluate the role of CSF endothelin as a predictor of outcome in ischemic brain injury.

2.2.2.1. Electroencephalography

The electroencephalogram (EEG) is a well-established method for monitoring brain electrical function. EEG is a method for recording cerebral electrical potentials, including action potentials and postsynaptic potentials (Binnie & Prior 1994). The EEG is very sensitive in detecting regional synaptic depression accompanying cerebral ischemia and even hypotension without hypoxia (Gavilanes et al. 2001). An EEG recorded after cerebral ischemia has been shown to predict the extent of cerebral damage (Binnie & Prior 1994, Stecker et al. 2001). Many abnormalities in EEG are known to be associated with brain ischemic injury. The following have been shown to be associated with ischemic brain injury after cardiac surgery: increased slow activity (theta and delta); a decrease in fast activity (alpha); a decrease in EEG frequency in several channels, indicating general slowing of the EEG (Sainio 1974, Arroyo et al. 1993, Vanninen et al. 1998); EEG seizures after pediatric cardiac surgery using HCA; (Helmers et al. 1997, Rappaport et al. 1998) slow recovery of EEG power after experimental HCA (Mezrow et al. 1995).

Both after cardiac arrest and HCA, the EEG recovers from electrical silence through burst suppression to continuous EEG even when no cerebral injury occurs. A delay in such an EEG burst-suppression recovery is an indicator of brain damage (Binnie & Prior 1994, Stecker et al. 2001), and when EEG remained in burst suppression after recovering to normothermia, patients suffered severe postoperative neurological complications (Stecker et al. 2001). The EEG amplitude also rises after ischemia, but because of the large variation in amplitude, it is not a good marker of injury (Sainio 1974). A similar EEG frequency after ischemia as detected preoperatively has been shown to predict good neurological outcome (Arroyo et al. 1993), but analysis is quite sensible to artefacts. In pediatric cardiac surgery, postoperative clinical and EEG-detected seizures were associated with a consistent pattern of worse developmental and long-term neurological function and with cerebral damage as detected by MRI (Rappaport et al. 1998). Whether clinical or EEG-detected seizures are markers of brain injury or themselves contributors to development of injury remains unclear (Helmers et al. 1997). An EEG power amplitude of less than 500 µV² two hours after HCA strongly predicts clinically evident neurologic impairment (Mezrow et al. 1995). The impact of injured brain areas on EEG abnormalities such as recovery is not well known.

Neuromonitoring during cardiac or carotid surgery has been considered a sensitive method for identifying cerebral ischemia (Edmonds et al. 1996, Sebel 1998). Unfortunately, many non-injurious processes may produce the same EEG changes as occur during hypoperfusion or hypoxia (Austin et al. 1997). Because of the complexity of conventional EEG analysis and its susceptibility to electrical and mechanical artifact, its prognostic value has been questioned (Witoszka et al. 1973). However, after HCA, EEG seems to be of prognostic importance (Stecker et al. 2001). Quantitative EEG is a sensitive method in the detection of slight cerebral functional changes, and useful in identifying patients with suspected cerebrovascular problems before surgery (Toner et al. 1998). EEG monitoring during CPB has a positive effect on outcome when an interoperative approach to increase brain blood flow was adopted in the case of EEG abnormalities (Hansotia et al. 1975). On the other hand, after cardiac surgery, subclinical brain injury was observed in quantitative EEG in one fifth of the patients three months after surgery (Vanninen et al. 1998).

In summary, quantitative and classic EEG analysis is a sensitive marker of brain injury. Even partial ischemia is seen on EEG, and subclinical injury could be seen in EEG. Specificity of the EEG analysis could be variable, depending on the abnormal finding and the method of analysis. Furthermore, interpretation of quantitative EEG is complex, and may require special expertise (Toner et al. 1998).

2.2.2.2. Somatosensory evoked potentials

Somatosensory evoked potentials (SEP) are the electrophysiologic responses of the nervous system to sensory stimulation (Chiappa & Ropper 1982a,b). The evaluation of SEP is an established form of neuromonitoring. SEP reflects the functional integrity of specific sensory pathways. A loss of cortical SEP is a good indicator of cerebral hypoperfusion, and is able to identify patients at risk of stroke during carotid surgery (Schwartz et al. 1996, Beese et al. 1998). Loss of SEP response is an indicator of cerebral ischemia during hypothermic low-flow CPB (Wilson et al. 1988).

In patients who have undergone HCA, delayed recovery of SEP has been shown to be a good marker of cerebral injury (Taylor et al. 1985), and intraoperative SEP alteration is associated with transient or permanent neurological sequences with high specificity (Ghariani et al. 1999). Acute, unilateral decreases in amplitude of the cortical potential are more useful than changes in latency in detecting intraoperative stroke (Stecker et al. 1996).

There is a problem inherent in the use of SEP amplitude asymmetry as a criterion for cerebral injury, since it is strongly biased towards detecting unilateral or asymmetric central nervous system lesions (Stecker et al. 1996). When SEP is used during CPB with mild hypothermia, the commonly used rule that a 50% decrease in evoked potential amplitude suggests a neurologic injury is too conservative, and it is not uncommon to see 90% decreases in amplitude in patients without strokes (Stecker et al. 1996).

In summary, SEP is a sensitive marker of stroke during CPB, but its specificity is not satisfactory.

2.2.3.1. Cerebral microdialysis

Evaluation of changes in the chemical microenvironment is important in order to understand the mechanisms underlying the development of brain ischemic injury. These changes of extracellular biochemical microenvironment in tissues can be monitored by a microdialysis tool in vivo (Benveniste et al. 1984, Tossman & Ungerstedt 1986). The method of microdialysis was introduced in the 60s, when push-pull cannulas, dialysis sacs, and dialyrodes were inserted into animal tissues to study biochemistry directly (Ungerstedt 1991). The first report on microdialysis technique dates back to 1966, when Bito described a fluid-filled semipermeable membrane implanted in dogs (Bito et al. 1966). Tossman and Ungerstedt simplified the technique, introducing the current microdialysis method (Tossman & Ungerstedt 1986).

Technique of microdialysis. In vivo microdialysis measures the chemical composition of the extracellular fluid (ECF). Microdialysis, as the name suggests, functions on the principle of the diffusion of water-soluble substances through the semipermeable dialysis membrane until equilibrium is attained (Fig. 2) (Benveniste 1989, Ungerstedt 1991). Depending on the permeability of the membrane, the molecular weight of these substances can be up to 100 kDA (Benveniste & Huttemeier 1990).

Nowadays there are a variety of different probes. The basic probes consist of a small polycarbonate tube or probe with a diameter of 0.2–0.6 mm and a length of dialysis area of 10 to 30 mm at the tip of the probe (Fig. 3). Substances from the interstitial fluid can diffuse through the membrane into the perfusion fluid (Ungerstedt 1991). Furthermore, the probe is connected to a microperfusion pump and constantly perfused with a physiological solution at a constant flow rate of 0.1 to 5 L/min, which is collected for later analysis (Di Chiara 1990). Depending on the availability of analysis methods, every soluble molecule in the interstitial space can, theoretically, be measured by microdialysis.

The recovery of a particular substance is defined as the ratio, expressed as a percentage (%), of the concentration in the dialysate to the concentration in the interstitial fluid. The perfusion flow rate is inversely related to the relative recovery (Fig. 4), restricting the size of the samples and the minimum time between samples. When recovery is less than 100%, the concentration in the dialysate depends both upon the supply of substances from the blood capillaries and how much the cells take up from the interstitial fluid. For example, a glucose supply to the microdialysis catheter can decrease due to a decrease in the capillary blood flow and/or to an increase in the cell uptake of glucose. (Ungerstedt 1991)

Monitoring brain metabolism and ischemia. Cerebral microdialysis has been used to study chemical changes in several cerebral pathological processes, for example Parkinson, epilepsy, malignant neoplasia, brain ischemia and stroke, traumatic brain injury and subarachnoid haemorrhage in an intensive care setting. During the last few years, this method has been developed into a promising tool for monitoring and targeting therapy of brain injury (Persson & Hillered 1992, Muller 2002). A methodological study of microdialysis showed the high quality of one of the most-used brain catheters, CMA70, and clinical analyzer CMA 600 (Hutchinson et al. 2000).

Normal and ischemic brain tissue concentrations of glucose, lactate and pyruvate, and the lactate/pyruvate and lactate/glucose ratios, are shown in Tables 5 and 6.

The brain depends almost exclusively on the aerobic consumption of glucose for energy production (Goodman et al. 1999). However, information on brain ECF glucose levels under normal conditions in response to brain ischemia/hypoxia is scant (Valtysson et al. 1998). The brain glucose concentration reflects the balance of the supply from the blood and utilization by cells (Fellows et al. 1992). The results of the rat ischemia model support the contention that brain glucose may be a valuable marker of severe ischemia and may help to differentiate between partial and complete ischemia (Valtysson et al. 1998). In the case of complete ischemia, glucose is depleted almost totally from the brain tissue (Ljunggren et al. 1974a, Rimpiläinen et al. 2001), whereas with a lesser degree of ischemia it is decreased, but still found in the ECF (Tables 1 and 6) (Schulz et al. 2000).

Several studies have shown that high-energy metabolites ATP and PCr, along with glucose and glycogen stores, are consumed during the first few minutes during normothermic ischemia, and at this time lactate has reached its maximal concentration and glucose has disappeared from the brain tissue (Ljunggren et al. 1974a). Lactate increases during total or, in particular, partial ischemia (Schulz et al. 2000). However, the absence of a rapid increase of ECF lactate represents an important finding for distinguishing between intracellular lactate and ECF lactate (Persson et al. 1996). Studies have shown that lactate increased mainly during recirculation rather than during complete ischemia, and during repolarisation rather than depolarization (Persson et al. 1996). It should be recalled that the lactate production depends on substrate availability, and in the case of a fast lack of glucose and pyruvate, lactate production is consequently limited (Fig. 5) (Persson et al. 1996).

Pyruvate metabolism resembles glucose metabolism. During severe ischemia its brain concentration decreases, and in the case of complete ischemia, pyruvate is used to produce lactate (Table 6) (Schulz et al. 2000). After an ischemic insult, an increase of pyruvate levels is a better marker of reperfusion than glucose levels (Persson & Hillered 1992).

The lactate/pyruvate ratio is a well-known marker of cell ischemia (Hillered et al. 1990), and a much more reliable marker of cerebral ischemia compared to lactate or pyruvate alone (Valtysson et al. 1998). The use of a ratio has the further advantage of abolishing the influence of changes in recovery over the dialysis membrane (Persson & Hillered 1992). When mitochondrial function is impaired, as during anoxia or severe ischemia, the intracellular NADH/NAD⁺ ratio increases together with accumulated [H⁺] and drives the lactate dehydrogenase reaction towards lactate (Valtysson et al. 1998). Pyruvate may also be consumed when α-ketoglutarate and alanine is formed from glutamate and pyruvate by alanine aminotransferase (Siesjö 1978). Changes in the brain lactate/pyruvate ratio appear to closely reflect the intracellular redox state (Siesjö 1978).

The lactate/glucose ratio is also a reliable marker of ischemia (Zauner et al. 1997, Goodman et al. 1999). When cerebral oxygenation is partially reduced, lactate accumulates in the extracellular space. When severe enough, such episodes are associated with depletion of glucose from the extracellular space. This state may lead to uncompensated anaerobic glycolysis in which neurons and astrocytes compete for the extracellular glucose in a desperate bid for a trickle of adenosine triphosphate production. In such a severe metabolic state, an increase in lactate/glucose ratio is observed, and is associated with a poor clinical outcome Table 6 (Zauner et al. 1997, Goodman et al. 1999).

Other markers of ischemia and cell damage. Glutamate is released from neurons during ischemia and accumulates in the interstial space. It is responsible for initiation of a pathological influx of calcium leading to cell damage. Glutamate concentrations have been shown to predict the postoperative outcome of patients with subarachnoidal hemorrhage (Persson et al. 1996). Glutamate has been shown to increase in ECF during ischemia and also in secondary ischemia (Hillered et al. 1990, Boris-Moller & Wieloch 1998). It is an indirect marker of cell damage, but it is sometimes difficult to interpret its changes due to the fact that the neuronally released glutamate is mixed with the large metabolic pool of glutamate (Lipton & Rosenberg 1994). In experimental models of HCA, brain glutamate levels increased after prolonged HCA, but without a clear association to postoperative outcome (Rimpiläinen et al. 2001, Rimpiläinen et al. 2002). Difficulties in detecting a narrow peak level of glutamate, as well as the commonly observed variability in individual increase of glutamate, may explain the lack of correlation with the postoperative outcome.

Degradation of membrane phospholipids is a well-known phenomenon in acute brain injury (Hillered et al. 1998). Glycerol is an integral component of the cell membrane. Loss of energy during brain ischemia leads to an influx of calcium, which triggers events for membrane phospholipid degradation to glycerol (Fig. 6) (Kristian & Siesjö 1998). During and after the ischemic condition, the release and production of glycerol is unbalanced. Glycerol concentrations rise during and after cerebral ischemia, and glycerol has been shown to be a sensitive and reliable marker of cell damage in experimental cerebral ischemia (Hillered et al. 1998, Frykholm et al. 2001). When compared with glutamate, brain glycerol has a wider peak after HCA, and its decrease takes several hours, making the measurement much easier.

2.2.3.2. Tissue monitoring probes

Direct brain tissue oxygen monitoring in patients with traumatic head injury was first reported by Maas (Maas et al. 1993). The placement of intracranial chemistry probes can be accomplished with techniques similar to those used for standard ICP probes (McKinley et al. 1996). The microsensor probe is 0.8 mm in diameter, and the sensing area is 18 mm² (Fig. 7).

Later, a probe by which it is possible to monitor cerebral tissue oxygen, carbon dioxide, pH and temperature was introduced (Zauner et al. 1995). A fiber-optic probe with sensors for monitoring the partial pressure of tissue oxygen (Pti0₂), carbon dioxide (PtiCO₂) and pH has been proved to provide results comparable to those achieved by other tissue monitoring methods (McKinley et al. 1996).

Brain tissue oxygenation. The normal level of brain PtiO₂ is 33 to 36 mmHg (Hoffman et al. 1996a). Monitoring of PtiO₂ by tissue probe is safe, reliable and sensitive in detecting changes in brain tissue oxygenation with results comparable with jugular vein oxygen saturation measurement (Kiening et al. 1996). A brain PtiO₂ below 8 mmHg for longer than 30 minutes is associated with increased extracellular glutamate and cerebral infarction (Kett-White et al. 2002). The quality of the oxygenation measurement is excellent, provided the first hour after insertion (adaptation time) is excluded (Dings et al. 1998).

Brain tissue carbon dioxide pressure (PtiCO₂). The normal level of brain PtiCO₂ is about 49 mmHg (Hoffman et al. 1996a). There are conditions associated with an increased or decreased PtiCO₂ such as increased cerebral vascular constriction (Carmona Suazo et al. 2000). The increased brain tissue PtiCO₂ observed in patients with a compromised cerebrovasculature is consistent with ischemic tissue PtiCO₂ (Hoffman et al. 1996a). However, as a marker of ischemia, PtiO₂ is faster and more reliable than PtiCO₂ (Hoffman et al. 1996b). The decrease of PtiCO₂ after hypothermic or normothermic ischemia is a marker of reperfusion, and faster than PtiO₂ or pH (Hoffman et al. 1996b).

Brain tissue pH. Continuous monitoring of pH and PtiCO₂ allows the precise monitoring of the acid-base status of the brain tissue. The normal level of brain pH is about 7.25 in humans (Hoffman et al. 1999). It has been shown by this method that a pH of less than 7.0 for more than 20 minutes is associated with severe brain injury (Hoffman et al. 1996a). Interestingly, monitoring of brain tissue pH provides indirect evidence of ICP (Hoffman et al. 1996a).

In summary, a combination of brain tissue pH, PtiCO₂ and PtiO₂ measurement by means of a minimally invasive tissue monitoring probe gives valuable and reliable information on tissue oxygenation, carbon dioxide pressure and acid-base balance.

2.2.3.3. Near-infrared spectroscopy

Near-infrared spectroscopy (NIRS) is a non-invasive optical monitoring technique providing information on vascular oxygenation by showing changes in tissue oxyhemoglobin, deoxyhemoglobin and total hemoglobin. It also provides information on cellular oxygenation by detecting changes in oxidized cytochrome a, a3 (CytOx), the last enzyme of the respiratory chain (Jöbsis 1977, Wray et al. 1988). The basis of this monitoring method is near-infrared light, which, in the spectral range of 700 to 1000 nm, is absorbed less than visible light and can penetrate much further, up to a depth of 6 cm into the tissues (Wray et al. 1988). These optical methods allow transmission spectroscopy to be performed in vivo.

Fallon et al first reported on its use in patients undergoing cardiac surgery for investigation of cerebral hemodynamics during CPB (Fallon et al. 1993). NIRS measurements with magnetic resonance imaging (MRI) during HCA showed high correlation between the CytOx value and PCr levels and histological brain injury after HCA (Shin"oka et al. 2000). The oxygenated hemoglobin signal nadir time in NIRS is a useful predictor of safe duration of circulatory arrest. Interestingly, when the nadir time was below 25 minutes during HCA of 15 or 25 °C, there was no behavioral or histological evidence of brain injury (Sakamoto et al. 2001b). NIRS has also been used to verify the safe level of brain oxygenation during total arch replacement employing selective brain perfusion (Yamashita et al. 2001).

The usefulness of these measurements has been questioned because the measurements are relative to baseline, and CytOx provides a small signal and therefore is vulnerable to artifact (Matsumoto et al. 1996, Nollert et al. 1998). Another pitfall of this method is the fact that hematocrit interferes with the cytochrome signal (Kurth & Uher 1997, Sakamoto et al. 2001a).

Clinical experience with cerebral oximetry after stroke and cardiac arrest have shown some methodological characteristics. Oximetry by NIRS reflects the balance between regional oxygen supply and demand. In dead or infarcted nonmetabolizing brain, saturation may be near normal because of sequestered cerebral venous blood in capillaries and venous capacitance vessels, and because of contribution from overlying tissue (Nemoto et al. 2000). Indeed, NIRS has failed to be an indicator of cerebral ischemia during carotid surgery (Beese et al. 1998).

In summary, NIRS is a non-invasive, feasible and safe method for measuring cerebral oxygenation independent of brain function, cerebral blood flow and metabolism. It has all the advantages to make it the gold standard for real-time brain monitoring during cardiac surgery (Nollert et al. 2000). NIRS measurements, especially the CytOx signal, correlate well with high-energy phosphates and have a high sensitivity for predicting histological brain damage after HCA. However, the CytOx signal also has several limitations, such as a small signal and high artifact effect, depending on the hematocrit value (Nollert et al. 2000).

2.2.3.4. Intracranial pressure

An increase in ICP occurs after brain infections, head injury, ischemic injury and intracranial bleeding. It has been shown to be a major factor contributing to the severity of brain ischemic injury in neurosurgical patients (Juul et al. 2000, Schneweis et al. 2001), as it is associated with derangements in blood flow supply to the brain (Hekmatpanah 1970, Nagai et al. 1997). Such disturbances in cerebral blood flow begin at the microcirculatory level with the collapse of capillaries, which is associated with sloughing of red blood cells and formation of microemboli. A further increase in ICP may also involve larger intracranial arteries and veins, thus worsening the blood supply to the brain (Hekmatpanah 1970). Nakai employed a microvascular laser-Doppler flow-meter to measure the blood flow into the middle cerebral artery, and showed that with the increase in ICP, the flow patterns appeared in the following order: normal flow, sharp wave, systolic flow, systolic spike and no flow (Nagai et al. 1997). ICP threshold levels and cerebral perfusion pressure (CPP) predicting adverse outcome have been studied in patients with severe head injury (Chambers et al. 2001). CPP threshold levels for adverse outcome were 45mmHg in children and 55 mmHg in adults, and for ICP, 35 mmHg in both adults and children (Chambers et al. 2001). A reduction of ICP to less than 20 mmHg is considered a main therapeutic target in patients with severe head injury (Juul et al. 2000).

There is evidence that a significant increase in ICP also occurs during and after CPB (Lundar et al. 1983b, Lundar et al. 1985, Johnston et al. 1991, McDaniel et al. 1994). The etiology behind such an ICP increase during CPB remains unknown. Interestingly, a study employing MRI has shown an increased content of cerebral water after CPB, which was not found in patients undergoing off-pump coronary artery bypass surgery (Anderson et al. 1999). Obstruction of CSF venous flow has been suggested as a contributor to increased ICP (Philpott et al. 1998). Increased postoperative ICP has been observed to be associated with lower recovery of EEG and adverse neurological outcome after experimental HCA (Hagl et al. 2002).

Only a few studies have been performed to investigate the role of ICP in predicting the outcome after cardiac surgery. Anyway, there is evidence that a modest increase of ICP may also put the patient at high risk of cerebral hypoperfusion, especially in the case of decreased mean arterial pressure (Philpott et al. 1998). Thus, the quality of perioperative and postoperative monitoring of cerebral function would be greatly improved if intracranial compliance and pressure (ICP) could be continuously monitored. In cardiac surgery, a new non-invasive method for monitoring intracranial compliance allows on-line ICP monitoring and could be performed with reliable results (Michaeli & Rappaport 2002, Paulat et al. 2002).