Clinical review: Neuromonitoring - an update

Critically ill patients are frequently at risk of neurological dysfunction as a result of primary neurological conditions or secondary insults. Determining which aspects of brain function are affected and how best to manage the neurological dysfunction can often be difficult and is complicated by the limited information that can be gained from clinical examination in such patients and the effects of therapies, notably sedation, on neurological function. Methods to measure and monitor brain function have evolved considerably in recent years and now play an important role in the evaluation and management of patients with brain injury. Importantly, no single technique is ideal for all patients and different variables will need to be monitored in different patients; in many patients, a combination of monitoring techniques will be needed. Although clinical studies support the physiologic feasibility and biologic plausibility of management based on information from various monitors, data supporting this concept from randomized trials are still required.

pressure triggers an active vasodilatory response that keeps cerebral blood fl ow (CBF) constant over a wide range of mean arterial pressures (MAPs), thereby prevent ing brain hypoperfusion (cerebral pressure autoregula tion) [7]. In brain-injured patients, cerebrovascular reactivity may be impaired, and a decrease in MAP or cerebral perfusion pressure (CPP) may thus translate into reduced CBF and secondary ischemia [8].
Cerebral oxygen utilization is proportional to the product of CBF and the arterio-venous diff erence in oxygen content [9]. Mechanisms other than perfusionlimited ischemia (microvascular collapse, endothelial swelling, perivascular edema) may increase the diff usion gradient for oxygen between venous blood and brain tissue (PvO 2 -PbtO 2 ), reduce cellular oxygen delivery and attenuate the ability of the brain to increase oxygen extraction fraction in response to reduced CBF [10].

Neurological examination of the ICU patient
Clinical neurological examination is a fundamental compo nent of neuromonitoring and should take into account the eff ects of sedative drugs, which may markedly infl uence neurological responses. Th e degree of awareness should fi rst be assessed. Clinical assessment is based mainly on evaluating eye and motor responses to verbal orders and noxious stimuli. Th e depth of coma can be evaluated by the Glasgow Coma Scale (GCS) [11] or the Full Outline of UnResponsiveness (FOUR) score [12], which includes assessment of the pupillary light and corneal refl exes and of the breathing pattern. Delirium is a fl uctuating state characterized by altered attention, spatio-temporal disorientation, disorganized thinking and alteration of awareness. Delirious patients can be hypo active, hyperactive or both alternately. Th e Con fusion Assessment Method for the ICU (CAM-ICU) [13] or the ICU Delirium Screening Checklist (ICU DSC) [14] can be used to assess the presence and degree of delirium.
Neurological examination should then assess for neck stiff ness, motor responses, plantar and deep tendon refl exes and cranial nerve functions. Th e presence of focal neurological signs should prompt brain imaging, which should also be considered, along with an electroencephalogram (EEG), if there is no obvious cause for delirium or coma.

Intracranial pressure
Elevated intracranial pressure (ICP; >20 mmHg) is associated with increased mortality after acute traumatic brain injury (TBI) [15]. Th e most recent Brain Trauma Foundation guide lines [16] recommend (level II evidence) that ICP should be monitored in all salvageable patients with severe TBI (GCS score of 3 to 8 after resuscitation) and an abnormal head CT scan.
ICP monitoring is also used in non-traumatic neurological disorders, such as SAH, and to a lesser extent in brain tumors, infarctions, intracerebral hemorrhage, and infections. After SAH, external ventricular drainage is recommended in patients with higher World Federation of Neurological Societies (WFNS) scale scores or acute hydrocephalus [17], to monitor ICP and to drain cerebrospinal fl uid (CSF) for ICP control. In spontaneous intracerebral hemorrhage, ICP monitoring is seldom used, but recent guidelines [18] suggest that in patients with a GCS score of <8, those with clinical evidence of transtentorial hernia tion, or those with intraventricular hemorrhage or hydrocephalus, ICP monitoring and treatment may be considered (class IIb; level of evidence: C).
ICP is a function of intracranial content: some content is easy to detect and correct, such as an expanding contusion, while others are more diffi cult -for example, in cases of diff use cerebral swelling, which may indicate vasodilation or edema. Th e volume-pressure curve, which describes ICP changes to corresponding increases in volume, has an exponential shape: as long as compensatory mechanisms can be used (that is, displacement of CSF from the supratentorial space into the spinal sac), ICP will remain constant despite the volume increase. When compensatory capabilities are exceeded, ICP rises sharply. Attempts have been made to identify loss of compensation using ICP waveform analysis: the normal ICP waveform has three components (P1 > P2 > P3), but is altered when pressure increases, with an increase in the P2 component compared to P1 (Figure 2); the ICP pulse amplitude (diff erence between systolic and diastolic values) also tends to increase [19]. However, this feature depends on the CPP and on the physical properties of the measuring system, and so should be used with caution. Figure 1. Some of the pathological cascades contributing to aggravate neuronal injury after acute brain injury. Initial mechanisms, such as excitotoxicity and mitochondrial dysfunction, initiate damage in the very early minutes/hours; other factors, such as energy dysfunction, edema, and infl ammation come into play later in the process. CBF, cerebral blood fl ow.
Although non-invasive evaluation of ICP is possible using the transcranial Doppler (TCD)-derived pulsatility index [20] or optic nerve sonography [21], the only methods for continuous on-line monitoring of ICP remain invasive. Intra-ventricular devices have long been considered the 'gold standard'; however, intra-parenchymal pressure monitoring provides equivalent pressure measurements [22]. Intraventricular probes are particularly useful when CSF drainage is desirable.
ICP management should also take CPP into account ( Figure 3). Th erapies to reduce intracranial hypertension that also may reduce arterial pressure and CPP (for example, barbiturates) require careful titration to preserve adequate cerebral perfusion.

Cerebral blood fl ow
Direct, continuous measurement of regional CBF is now feasible using a thermal diff usion probe (TDP; Hemedex ® Cambridge, Massachusetts, USA) inserted into the brain parenchyma. Th e sensor at the tip provides a quantitative measure in the volume of tissue surrounding the sensor. Th e TDP technique showed good agreement with CBF measured by xenon-CT [23].
Correct and stable positioning of regional CBF probes with a skull bolt is crucial and a CT-scan should be performed to verify correct placement. Precise quantifi cation of regional CBF using a TDP is dependent on the presence of a stable temperature and may be altered in conditions of hyperthermia or rapid fl uctuation in a patient's temperature.
In patients with SAH, Muench and colleagues [24] used TDP monitoring to guide medical therapy for delayed cerebral ischemia, and demonstrated that increasing blood pressure with vasopressors was the only intervention that improved CBF and PbtO 2 ; hypervolemia and hemo dilu tion had only marginal eff ects at best. Hence, normo volemia and MAP/CPP augmentation are increasingly used instead of 'triple-H' therapy in these patients.
TDP monitoring may also guide CPP manage ment after TBI [25]. TDP data are currently limited to small singlecenter studies and larger trials are needed to confi rm the validity and use of this technique.

Transcranial Doppler
Direct measurement of blood fl ow in the brain arteries would be of value in all cases of brain damage. Using an ultrasound probe, fl ow velocity (rather than fl ow itself ) can be measured in the main arteries of the base of the brain. TCD combines ultrasound and the Doppler principle to represent erythrocyte fl ow in the basal cerebral arteries. In some 10% of patients, transtemporal insonation is not feasible because of anatomical barriers. When cerebrovascular resistance increases -for example, in vasospasm of the middle cerebral artery -systolic TCD velocity increases, whereas diastolic velocity decreases, leading to a clear increase in the pulsatility index (the ratio of the diff erence between systolic and diastolic fl ow to diastolic fl ow). For this reason, TCD is often used to monitor the time course of vasospasm after SAH [26]. However, correct interpretation of the pulsatility index is complex, because it depends not only on cerebrovascular resistance, but also on several systemic and cerebral variables [27].
TCD has also been used to assess cerebrovascular autoregulation in TBI and SAH patients [7]. Flow velocity responses to spontaneous or induced (with vasopressors) changes in CPP may quantify the status of static pressure autoregulation; autoregulation estimated by TCD in the middle cerebral artery correlated with autoregulation studied by positron emission tomography in TBI patients [28].
Th e main advantage of TCD is that it is non-invasive and can be carried out at the bedside. However, the quality of the TCD signal is operator-dependent and correct interpretation requires training. TCD signals may also vary over time with temperature, arterial carbon dioxide tension (PaCO 2 and, in some series, correlation with actual CBF measure ments was disappointing [29].

Microcirculation
Microcirculatory alterations play a key role in the pathogenesis of organ dysfunction. Brain endothelial cells help regulate vessel diameter and permeability [30], contri buting to coupling of fl ow to metabolism [31]. Unfor tu nately, it is still not possible to visualize the brain microcirculation in clinical practice without direct expo sure of the cerebral cortex after craniectomy and data on microvascular abnormalities in humans are limited.
Direct evaluation of the brain microcirculation can be performed using sidestream dark fi eld imaging and intra vital microscopy [32]. Th ese methods allow continuous and in vivo observation of the microcirculation at high resolution as well as evaluation of vessel size, number, density and fl ow. Although intravital microscopy is still considered the gold standard, it is not suitable for human use. Th e sidestream dark fi eld technique can be applied on exposed cortical areas during craniectomy but care must be taken to limit pressure artifacts and minimize the risk of infection [33]. In patients with stroke from middle cerebral artery occlusion who underwent decom pressive craniectomy, brain microcirculatory density and fl ow were signifi cantly altered when compared to control patients [34]. Similarly, altered microvascular reactivity has been demonstrated in patients with SAH [35].

Brain tissue oxygen monitoring
Several techniques can be used to measure brain oxygena tion, the most common in the ICU being jugular venous bulb oximetry and direct PbtO 2 measurement. Near infra-red spectroscopy has also been used for this purpose.

Jugular venous bulb oximetry
Th is technique requires placement of a catheter into the jugular bulb. Th e arterio-jugular diff erence in oxygen content (AJDO 2 ) is proportional to CBF and inversely proportional to oxygen consumption (cerebral metabolic rate for oxygen, CMRO 2 ) and is often used as a global measure of adequate perfusion. Proper positioning of the probe in the jugular bulb is crucial, since blood draining from extra-cerebral structures, such as the neck and face, can contaminate the lower portions of the jugular vein.
Under most conditions in which arterial hemoglobin saturation is constant, just the jugular oxygen saturation (SjO 2 ) is measured. Th is measurement can be performed by intermittent sampling, or continuously using fi beroptic probes. Normal values for SjO 2 in patients without brain damage are about 57% (95% confi dence interval 52 to 62%) [36]. In the early hours after trauma, low SjO 2 values have been detected, especially in the most severe cases [37]. Severe and frequent SjO 2 desaturations are associated with worse outcome in TBI patients [38].
Because of the anatomy of the cerebral circulation, SjO 2 is an average indicator and a large volume of brain must be underperfused for an abnormality to be detected. Two-thirds of the blood fl owing in each jugular vein comes from the ipsilateral and one-third from the contralateral hemisphere [39]; focal areas of mismatch between fl ow/metabolism can, therefore, be missed by this global measurement [40]. Additionally, SjO 2 values in one jugular vein may diff er from values in the other [41]. Despite these limitations, intermittent SjO 2 sampling is a cheap and relatively easy tool for estimating adequacy of fl ow and metabolism, particularly when CBF is manipulated -for example, when using hyperventilation.

Direct PbtO 2 measurement
Direct PbtO 2 monitors are the most common technique used in the ICU to assess cerebral oxygenation. Probe positioning is crucial: usually inserted in the white matter, readings are dependent, in part, on proximity to intracranial pathology [42]. For example, if close to a contusion, values may be reduced and this should be considered when making management decisions. When the PbtO 2 probe is in what appears to be normal white matter on CT, then the reading, although local in nature, provides a reasonable estimate of global brain oxygenation.
PbtO 2 is not a 'surrogate' for ischemia or CBF, because it varies not only with CBF (and factors that regulate itfor example, CO 2 and MAP) but also with changes in arterial oxygen tension (PaO 2 ) among many other factors. PbtO 2 is, therefore, more a marker of the balance between regional oxygen supply and cellular oxygen consumption [9].
Monitoring PbtO 2 has been validated against fi beroptic SjO 2 monitoring, xenon-enhanced CT scanning, and SPECT. Th reshold values vary slightly depending on what type of PbtO 2 monitor is used but values <20 mmHg are considered worth treating and values <15 mmHg indicate brain hypoxia or ischemia [43,44]. Decreases in PbtO 2 have been associated with independent chemical markers of brain ischemia in microdialysis studies [45]. Th e number, duration, and intensity of brain hypoxic episodes (PbtO 2 <15 mmHg) and any PbtO 2 values ≤5 mmHg are associated with poor outcome after TBI [46,47]. Episodes of brain hypoxia are common and may occur even when ICP and CPP are normal [48]. Th e exact relationship with outcome may, however, vary according to whether the probe is positioned in normal white matter, the penumbra or in a contusion [42,49]. Monitoring of PbtO 2 can be used in combination with other intra-parenchymal monitors, mainly ICP and cerebral microdialysis. Clinical studies suggest that therapy based on information from both an ICP and a PbtO 2 monitor may be associated with better outcomes than that based on ICP monitoring alone [50].

Near infra-red spectroscopy
Oxyhemoglobin, deoxyhemoglobin, and oxidized cytochrome oxidase absorb specifi c portions of the light spectra. When a tissue layer is illuminated with a source of light in the near infra-red wavelength, the attenuation of the light signal is correlated to the relative proportions of oxyhemoglobin and deoxyhemoglobin (HbO 2 /Hb) and oxidized cytochrome oxidase in the tissue. Using appropriate calculations, an estimate of tissue oxygenation can, therefore, be made non-invasively and continuously.
Near infra-red spectroscopy has been used in TBI [51] and SAH [52], with ambiguous results. Th e technique has major limitations in adults: baseline normal values vary widely, extracranial contamination is a problem, and poor correlation with independent measurements of cerebral oxygenation has been reported frequently.

Microdialysis
Cerebral microdialysis (CMD) involves the insertion of a catheter tipped with a semi-permeable membrane (usually with a 20 kDa cutoff ) in the brain parenchyma. Th e CMD catheter is constantly perfused, thereby allowing regular sampling of the patient's brain extracellular fl uid [53]. CMD sampling is limited to the interstitial tissue area around the catheter, thus measuring regional brain metabolism. In clinical practice, a pattern of elevated lactate/pyruvate ratio and low glucose is considered as a warning sign for cerebral ischemia/hypoxia [54,55]. High lactate/pyruvate ratios in normoxic conditions have been interpreted as markers of hyper glycolysis. Elevated glutamate is a marker of cellular dysfunction, for example, delayed cerebral ischemia in high-risk SAH patients [56]. Absolute CMD values are important, but trends over time may provide more useful information.
Recently, CMD has been used to identify the 'optimal' threshold for blood glucose during insulin therapy [57], and when used in combination with PbtO 2 , CMD monitor ing has potential clinical utility to target adequate values of MAP/CPP [58,59] and blood hemo globin concentration [60].

Biomarkers
Neurological biomarkers, quantitative indicators of brain dysfunction or damage, may be obtained via sampling of biological tissues (blood, CSF, brain interstitial fl uid), electrophysiological recordings, or neuroimaging. For soluble biomarkers, modeling of extracerebral concen tration kinetics is complex and must account for passage across anatomical barriers into the bloodstream or CSF; serum levels of a marker may therefore correlate less closely with brain levels than with alterations in bloodbrain barrier integrity. An emerging body of work suggests that multi-marker panels may enhance sensitivity and specifi city for acute neurological injury [61,62]. Th ere is also considerable interest in genomic, proteomic and lipidomic profi les.
Brain biomarkers are categorized according to their source as primarily neuronal, astroglial, or microglial. Here we briefl y review the key biomarkers that have been tested clinically in acute neurological injury.

s-100β
Increased levels of this astroglial protein are associated with injury severity and are predictive of long-term death or severe disability in severe TBI [63,64], and following cardiac arrest [65] or aneurysmal SAH [66]. Although highly sensitive, the specifi city of this marker for acute brain injury is generally low [67].

Glial fi brillary acidic protein
Glial fi brillary acidic protein is associated with intermediate fi laments in astrocytes. Increased serum levels are observed following moderate and severe TBI [68], ischemic stroke [69], SAH [70], and cardiac arrest [71]. Serum glial fi brillary acidic protein levels more accurately discriminate TBI outcomes than do levels of s-100β [72].

Neuron-specifi c enolase
Neuron-specifi c enolase is an enzyme that is found in neurons, cells of neuroendocrine origin, erythrocytes and platelets. Increased serum neuron-specifi c enolase has been observed in patients with ischemic stroke, TBI, SAH and after cardiac arrest [63,65,73]. Neuron-specifi c enolase can help discriminate between favorable and unfavorable outcomes following severe TBI and cardiac arrest [63,65].

Tau
Tau is a neuronal protein that stabilizes microtubules, thereby promoting axonal transport. Signifi cant elevations in tau have been observed in the CSF [74] and brain interstitial fl uid [75] of patients with severe TBI. CSF tau protein is also elevated in patients with SAH, and is associated with neurological severity and outcome [76].

Amyloid β
Amyloid β is a peptide cleaved from neuronal and glial amyloid precursor protein. CSF amyloid β fragment levels are lower in TBI and in SAH patients than in control subjects, and lower levels are predictive of shortterm functional outcome [77]. Recently, amyloid β has been measured in brain interstitium [75,78].

Neurofi lament heavy chain
Elevated levels of NfH (neurofi lament heavy chain), a neuronal protein that is integral to the axonal cytoskeleton, have been reported in the brain interstitial fl uid of patients with severe TBI and are associated with physiological abnormalities and mortality [79]. CSF NfH is increased following SAH and hypertensive intracerebral hemorrhage [80,81]. Serum concentrations of NfH do not reliably identify brain injury, perhaps because the large size of this molecule impedes passage through the blood-brain barrier.

Ubiquitin carboxy-terminal hydrolase L1
CSF levels of UCH-L1 (Ubiquitin carboxy-terminal hydro lase L1) are signifi cantly elevated in severe TBI patients and are associated with lower levels of conscious ness and worse six-month functional outcomes [82]. Serum levels of UCH-L1 are associated with TBI severity and survival [83].

Alpha II-spectrin
Alpha II-spectrin is a fundamental component of the neuronal cytoskeleton, which in the setting of injury is degraded into fragments referred to as spectrin breakdown products. CSF concentrations of these are elevated after TBI and correlate with the severity of injury and with six-month outcome [84]. Spectrin breakdown products are also detectable in serum and have diagnostic and prognostic value in children with TBI [85].

Electro-physiological measurements Electroencephalogram
Th e classical indication for EEG is to detect and manage seizures, including status epilepticus. In critically ill patients, seizures are often non-convulsive and may aggravate brain injury [86]. Intermittent EEG is less sensitive than continuous EEG for detecting nonconvulsive status epilepticus [87]. Approximately 50% of non-convulsive seizures are recorded within the fi rst 60 minutes of EEG recording, but in some patients up to 48 hours of monitoring may be required [88]. Over the past decade, a more advanced form of EEG, quantitative EEG (qEEG), has been developed, in which the raw EEG signal is converted into a digital form using fast Fourier transformation (compressed spectral array) (Figure 4).
Several groups have used qEEG to detect seizures and other causes of brain dysfunction, such as ischemia. In SAH patients, the percent alpha trend variability (PAV) decreased acutely with the onset of vasospasm-induced decreases in CBF and returned to normal with treatment and resolution of vasospasm [89]. More recently, Rathakrishnan and colleagues [90] reported that changes in α power could help to distinguish delayed cerebral ischemia from vasospasm after SAH. Th us, there appears to be good evidence for the use of qEEG to detect cerebral ischemia.
qEEG may also be used to monitor the depth of sedation in the ICU. Th e most commonly used modality is the 'bispectral index' (BIS) [91], an algorithm-derived number that approximates the degree of sedationinduced EEG activity when patients are under general anesthesia. It ranges from 0 (brain dead) to 100 (normal brain function); a value of <60 indicates general anesthesia, and <30 indicates burst suppression.
Selected characteristics of raw EEG, such as the background rhythm and Synek scale, spontaneous variability and responsiveness stimulation, have been used to assist in estimating prognosis [92]. qEEG is more objective because it uses derived measures, and a larger amount of data that have been trended over time. Th e lack of PAV over time during the initial three days after severe TBI was associated with clinical outcome at 1 month [93] and 6 months [94]. PAV trends are automatically created by modern EEG software, and can be useful in tracking improvement or lack of improvement in patients in real time ( Figure 5). Caution is required when using PAV and other qEEG parameters, however, because they may be aff ected by deep sedation.

Electrocorticography and spreading depression
Electrodes placed on the surface of the human cortex (electrocorticography), usually in proximity to damaged areas, can detect spontaneous waves of depolarization, which appear as spreading depression. Evidence is accumulating that these spreading depolarization waves cause an imbalance in energy consumption and delivery and may lead to secondary brain injury [95,96]. Spreading depression has been identifi ed in TBI [97] and in patients with SAH [98] in whom the number of spreading depressions correlated signifi cantly with the development of delayed cerebral ischemia even in the absence of vasospasm. Electrocorticography requires a craniotomy, but recent studies have suggested there may be correlates with human scalp direct and alternating current electroencephalography [99].

Evoked potentials
Somatosensory evoked potentials (SSEPs) are measured on the scalp as evoked EEG responses to an electrical stimulus applied typically to the median or tibial nerves. SSEPs are less aff ected by pharmacological agents or hypothermia than is the EEG. Th e main variable used for prognosis is the cortical response, which usually occurs at 20 msec after the stimulus, and hence is called the N20 peak. After cardiac arrest, bilateral absence of the N20 SSEP is associated with persistent vegetative state or death in all patients [100,101]. SSEPs can be measured during periods of sedation and even hypothermia, but most clinicians wait for return of normothermia. Th ere is some concern about inter-and intra-observer variability in interpretation of SSEPs [102]. Th e use of so-called 'cognitive' potentials, such as N70, has been suggested to improve the predictive value of poor prognosis of this test but this approach is not widely used [103].

Computed tomography
CT is the imaging modality of choice in the initial evaluation of patients with TBI or when acute hemorrhage is suspected. It helps to rule out surgical masses, and to identify early signs of intracranial hypertension. Since hemorrhagic lesions or edema may evolve over the fi rst hours after injury, a CT scan must be repeated whenever there is clinical deterioration even if the initial imaging was apparently normal. Th e severity of TBI lesions may be classifi ed using the Marshall [104] or Rotterdam [105] scales. New features, such as angioCT and CT perfusion, add important information to noncontrast CT and are increasingly used in stroke and SAH evaluation.

Magnetic resonance imaging
Magnetic resonance imaging (MRI) has great spatial resolution and, in patients with TBI, can identify pathologic abnormalities that are undetected or poorly characterized with CT, such as traumatic axonal injury. Acute ischemic stroke can also be detected earlier using MRI than CT. MRI is multi-parametric and can provide anatomical detail and quantitative information on brain physiology and metabolism, also allowing neuronal activa tion to be mapped. MRI has the advantage over other radiological methods in that it does not require ionizing radiation. Challenges for MRI scanning of critically ill patients include: monitoring and resusci ta tion devices that are incompatible with the magnetic fi eld, the need for sedation or even neuromuscular blockade to prevent movement artifacts, and the risks inherent to transport outside of the ICU environment.
Specifi c MRI sequences can enhance the diagnostic and prognostic evaluation of patients with acute brain injury.

Diff usion weighted imaging
Reductions in the apparent diff usion coeffi cient (ADC), a marker of water diff usion, are associated with acute tissue infarction in stroke and in cardiac arrest. Reduced ADC values have been noted in major gray matter and white matter structures after cardiac arrest and are predictive of six-month outcome [106].

Diff usion tensor imaging
Diff usion tensor imaging evaluates the directional preference (anisotropy) of water diff usion and allows quantitative assessments of white matter damage in patients with TBI. Th e magnitude of fractional anisotropy change correlates with clinical TBI severity in the acute setting and with long-term outcomes [107,108].

Magnetic resonance spectroscopy
Magnetic resonance spectroscopy measures tissue levels of selected neurometabolites and can provide important insight into TBI pathophysiology and natural history. It has been demonstrated that N-acetyl aspartate, a marker of neuronal integrity, is decreased in TBI, often in brain tissue that appears normal on cranial CT or on conventional MRI. Brain N-acetyl aspartate levels are associated with worse clinical condition and are predictive of functional outcome following TBI [108].

Susceptibility weighted imaging
Susceptibility weighted imaging is highly sensitive to micro scopic hemorrhagic lesions associated with TBI, which are believed to represent a subtype of diff use axonal injury. In children with diff use axonal injury, the number and volume of hemorrhagic lesions seen with susceptibility weighted imaging correlates with neurological severity on initial presentation and with long-term cognitive impairment and functional disability [109].

Functional MRI
Functional MRI demonstrates that cortical-subcortical networks governing motor activity are signifi cantly impaired following TBI [110].

Integration of variables
A single monitoring technique may not fully describe the complex pathophysiological changes in the brain and the concept of multimodality monitoring, the simultaneous digital recording of multiple parameters of brain function, has been introduced [111]. Multicenter collaborations, such as BrainIT (Brain monitoring with Informa tion Technology collaborative network) [112], have demonstrated that recording of many physiological variables across multiple patients is feasible and can provide new clinical insights [113]. Advanced statistical and mathe matical tools can be applied to the large volumes of data obtained with the aim of identifying patterns of brain injury and providing clinicians with easier identifi cation of specifi c targets [114]. Challenges in data integration and synchronization remain, but, ultimately, this concept should provide real-time, user-friendly advanced data analysis that, when applied at the bedside, could improve treatment.

Future trials
Neurological monitoring has developed rapidly over the past 25 years with widespread application of various monitors, but, although clinical studies suggest the physiological feasibility of this concept, there are still no published data from randomized trials to support that targeting any variables improves clinical outcome.
Neurocritical care research over the past 20 years has focused on the possibility that early pharmacological inter vention might improve outcome, which has defl ected attention from trials designed to answer simple questions about the everyday management of acute brain injury patients, such as optimum hemoglobin transfusion thresholds, hyperventilation, and arterial pressure manage ment after hemorrhagic stroke, for example. More extensive availability and use of neuromonitoring technology should facilitate the development of large clinical trials networks to provide the infrastructure necessary to design and conduct studies to resolve some of these issues.

Conclusion
Monitoring of brain function in critically ill patients must begin with a careful clinical evaluation. Monitoring systems typically look at single variables and there is a need to integrate/combine systems in order to obtain a full and accurate assessment of the patient's condition so as to be able to select the most appropriate therapy. Importantly, neuromonitoring is a dynamic process and not a single measurement; the ability to follow changes over time is vital to assess response to therapy and predict prognosis. No monitor will, by itself, improve patient outcomes but wise interpretation of the data and integration of the information obtained into an individualized therapeutic plan should help optimize care of critically ill patients with brain injury.

Key messages
• Monitoring of brain function should be considered in all comatose patients in the ICU. • Brain monitoring is centered on a careful clinical examination, although this can be limited in comatose and/or sedated patients. • Th e goals of neuromonitoring are to: identify worsening neurological function and secondary insults that may benefi t from specifi c treatments; improve pathophysiological understanding of cerebral disease in critical illness; provide physiological data to guide and individualize therapy; assist in prognostication. • At present there is no 'ideal' single brain monitor; a combination of monitoring techniques may provide better insight into brain function than a single monitor used alone. • Trends over time and threshold values are both important when assessing brain function. • Clinical studies suggest the physiological feasibility and biological plausibility of management based on information from various monitors but data supporting this concept from randomized trials are still required.