Chapter 33. Intraoperative Neurologic Monitoring

1. Anesthetic strategies to enhance intraoperative monitoring of the nervous system include techniques that minimize interference with neurophysiologic monitoring as well as techniques that preserve neurocognitive function during the structure and function mapping in the awake patient.

2. The cellular basis of normal electroencephalography (EEG) reveals a variety of pathways to produce alterations of electrical and neurocognitive function.

3. Synchronous EEG is seen with sleep, sedation and anesthesia, and cerebral ischemia.

4. Processed EEG algorithms can aid the objective assessment of EEG changes. As long as there is an understanding of the EEG features analyzed by these algorithms, pitfalls leading to inaccurate assessment can be avoided.

5. Achieving reliability with evoked potential monitoring depends on minimizing anesthetic effect, maintaining a constant anesthetic level, and ensuring adequate nervous tissue perfusion.

6. Intraoperative wakefulness for cortical mapping has been achieved by a variety of techniques. For a successful procedure, all techniques must address maintenance of effective ventilation during craniotomy and a balance of clear sensorium and sufficient analgesia to enable effective patient participation during cortical mapping.

7. Subarachnoid block for placement of epidural stimulating electrodes allows for maintained patient perception of electrode stimulation while providing effective anesthesia for laminotomy.

Intraoperative neurologic monitoring is based on detecting changes in neurologic function that reflect injury to the nervous system. Neurologic function can be assessed intraoperatively either by repeated neurologic physical examinations or by inducing observable responses through electrical or magnetic stimulation of the nervous system. Because much of intraoperative neurophysiologic monitoring depends on some measurement of the electroencephalogram (EEG), this chapter starts with a brief overview of the cellular mechanisms responsible for the genesis of the EEG. Phenomena that modify electrical brain function and electrical activity are considered so readers can appreciate the possibilities and limitations of electrical monitoring in guiding anesthetic administration and safeguarding the integrity of the nervous system. The use of anesthetic techniques that minimize interference with monitored neurophysiologic function are presented as are techniques that allow for rapid emergence enabling intraoperative neurologic testing, immediate postoperative assessment, and continuous postoperative assessment in the intensive care unit.

Electroencephalography

Electrical activity in the brain, which is what an EEG records, was first measured in 1875 by Richard Caton, who noted the electrical oscillations on the exposed cortical surface of animals. The physiologic mechanisms and cortical morphology responsible for generating the EEG are presented by Martin,1 and some essential features are summarized here.

Neural Basis of Electroencephalography

The EEG is derived from postsynaptic potentials on pyramidal cells. There are 2 major classes of cortical neurons, pyramidal and nonpyramidal. Pyramidal cells derive their name from their distinctive shape and are notable as the only neuron that projects axons out of the cerebral cortex and locally as well. Their apical dendritic spines are oriented perpendicularly to the cortical surface and extend through the lamina of the cortex, enabling connection with all of the nonpyramidal cortical cells. These nonpyramidal cells serve to modulate pyramidal cell output through stimulation (glutamate alteration of postsynaptic potential) or inhibition (GABA [γ-aminobutyric acid] alteration of postsynaptic potentials) (Fig. 33-1 and Table 33-1).

Collective voltage of cortical neuron ensembles is measurable at the scalp. The parallel orientation of the pyramidal cells enables the constructive addition of polarity from each cell to be measured at the surface of the cerebral cortex (Fig. 33-2).

Because the postsynaptic potentials last for a relatively long period of time and are geometrically aligned, it is possible for the potentials to summate to a sufficiently large magnitude to be measurable by electrodes placed on the scalp. Action potentials, which are of a much larger magnitude than the postsynaptic potential, are too brief to enable summation and therefore do not contribute to the EEG. Similar mechanisms for the generation of the EEG are presented by Rampil,2 Lukatch and Greenwald,3 and McPherson.4 The EEG is an electrical potential versus time measurement that measures cortical voltages at the scalp resulting from the collective postsynaptic potentials of ensembles of cortical pyramidal cells.

Synchronous versus Desynchronous Electroencephalographic Patterns

This cellular mechanism forms the conceptual basis of 2 basic patterns of EEG, synchronous and desynchronous EEG. A synchronous EEG is composed of large-amplitude peaks with slow-frequency oscillations. Referring to the basic mechanism presented previously, a synchronous EEG results when an ensemble of many cortical dendrites are polarized in synchrony. The voltage amplitude is large because the individual depolarizations are additive. The thalamus serves as the orchestrator of pyramidal cell depolarization during EEG synchrony, and the frequency of thalamic stimulation determines the slow rate of cortical polarization or depolarization.5 A desynchronous EEG results when cortical dendrites are polarized by a less circumscribed group of afferent nerves. The consequence of diverse sources of polarizations is a faster frequency of oscillation. The polarizations are not additive and therefore are of lower amplitude. Drawing from this conceptual framework, certain patterns are evident in the normal waking and sleeping EEG (Fig. 33-3). A normal EEG from an awake, alert person is characterized by irregular EEG oscillations with a variable frequency greater than 12 cycles/s and relatively small amplitude. Low-amplitude, high-frequency EEG activity, the hallmark of desynchronous EEG, is the typical EEG pattern of the awake and dreaming brain (the consciously perceiving brain). Departures from normal waking consciousness during sleep are characterized by a slowing of the oscillation frequency and an increase in amplitude of the voltage oscillation. Neuronal discharge is no longer subject to the ambient environmental stimuli–responses characteristics of wakefulness. Orchestration of neuronal discharge is now directed by central pattern generators (presumably located in the thalamus).5 The constructive interference of these waves is evidenced by large-amplitude and slow-frequency EEG traces, the hallmark of the synchronized EEG of the sleeping brain.

Desynchronous Electroencephalography: Wakefulness and REM Sleep

One of the early findings in sleep research was that the EEG reverted from a synchronous pattern (low frequency, large amplitude) to a desynchronous pattern when a subject passed from non–rapid eye movement (NREM) sleep to rapid eye movement (REM) sleep. Because subjects were able to report dreams (cognitive activity) at the conclusion of these episodes and because awake patients also demonstrated desynchronous EEGs, the notion that EEG desynchrony was the electrical correlate of a cognitively functioning brain. Conversely, EEG synchrony was the hallmark of the unconscious brain. The transition from desynchronous EEG to synchronous EEG during sleep onset is characterized by the gradual appearance of slow oscillations in the EEG. The tendency for an EEG to develop a synchronous pattern is also known as slowing.

Synchronous Electroencephalography: Natural Sleep, Drug-Induced Sedation, "Light" Anesthesia, and Cerebral Ischemia

All of these states are usually marked by an alteration of consciousness. But all of these states are not equivalent in terms of the degree of altered wakefulness or in terms of the likelihood of return to normal consciousness. However, for all of these states, a return to normal waking consciousness is associated with a return of a desynchronized EEG pattern. In summary, the synchronous EEG does not specifically identify the cause of the EEG synchrony (and its associated alteration of consciousness). Therefore, synchronous EEG will be less than specific in predicting the consequences.6

The following 4 points, although highly simplified, represent key concepts in EEG generation: (1) EEG signals recorded at the skull surface represent the summated postsynaptic potential of hundreds of thousands to millions of pyramidal neurons. (2) Although EEG rhythms recorded at the skull surface are generated by neocortical neurons, these rhythms may originate either in the neocortex itself, or they may be "imposed on" the neocortex by subcortical structures that "pace" the activity of neocortical neurons. (3) Multiple neurotransmitter systems can be involved in generating different types of EEG activity. (4) EEG activity within individual frequency bands (ie, delta, theta, alpha, beta, or gamma) likely represents more than 1 phenomenon."3

Studies Suggesting the Cellular Mechanisms for Electroencephalographic Synchrony: How Sleep, Anesthesia, and Cerebral Lesions Slow the Electroencephalogram

Amzica and Steriade5 have reviewed the evidence for the cellular mechanisms of slow wave generation during natural sleep. Their summary describes 3 different oscillations that "coalesce into the polymorphic wave of slow wave EEG sleep" (Fig. 33-4). Salient points from their review include

1. Cellular discharge data are obtained from cats anesthetized with ketamine and xylazine. Therefore, there is a leap of faith that the mechanism of slowing from sleep versus anesthesia is the same.

2. Human data during natural sleep are not obtained from single-cell but from scalp EEG. Therefore, there is an assumption that cellular data from anesthetized cats can be extrapolated to unmeasured cellular phenomena of sleeping humans because of a similarity of the EEGs from both groups.

3. Deafferented cortex (cortex disconnected from thalamic input by surgical prep or tumor) tends to reveal a slow wave pattern.

The EPITIDE (enhanced phasic inhibition, tonic inhibition, and depressed excitation) theory of anesthetic action on patterned brain activity proposes that anesthetic-induced prolongation of inhibitory currents may slow EEG activity by limiting neuronal discharge frequencies of EEG-generating neurons. Lukatch and Greenwald3 propose that the origin of the EEG rhythms arise in neocortical neurons and may also be "imposed" on neocortical neurons by subcortical structures that "pace" the cortex. Multiple neurotransmitter systems are involved, and EEG activity within individual frequency bands likely represents more than 1 phenomenon. The anesthetic effect on EEG is not a simple transition from the desynchronized EEG to the synchronous but rather involves an initial EEG activation at subanesthetic concentrations (perhaps an EEG correlate of the clinical excitement phase of subanesthetic levels during induction and emergence). At greater anesthetic exposure, the EEG evolves into a synchronous pattern of slowing followed at even greater exposure by isoelectric EEG with bursts of oscillation. It is hypothesized that at the network level, this cellular effect could produce a state in which low-frequency EEG oscillations (eg, delta activity) are supported by the network but higher-frequency oscillations are filtered out (much like the activity of a low-pass filter in an electronics circuit). Lukatch and Greenwald3 state that the theory relies on anesthetic effect on the cortical networks to explain the EEG effect. The theory does not address the role of subcortical networks of cortical slowing proposed for the mechanism of natural sleep-related EEG slowing. Further observations are required for confirmation.

Electroencephalographic changes induced by a variety of insults include EEG slowing, burst suppression, and isoelectric activity. Neuronal electrophysiologic changes responsible for these EEG changes were explored by Rabinovici et al7 This study used an in vitro rat brain slice model that enabled the measurement of single-cell discharge and simultaneous cortical field potential (analogous to EEG) resulting from exposure to the brain slice to hypoxia, ischemia, and hypoglycemia. The findings show that EEG slowing, burst suppression, and isoelectric EEG occur with varying patterns depending on the lesion and that some patterns of damage occur with recovery from the insult. This study did not specifically address the role of subcortical structures (ie, thalamus) in the generation of slowed EEG, suggesting that thalamic mechanisms may not be necessary for the genesis of a slowed EEG in response to cerebral insult.

The work of Amzica and Steriade5, Lukatch and Greenwald,3 and Rabinovici et al7 illustrates that ischemia, hypoxia, hypoglycemia, and anesthesia produce cortical EEG slowing by a variety cellular electrophysiologic mechanisms that may not be distinguishable from the macroscopic EEG.

Intraoperative Electroencephalography Uses

The foregoing discussion indicates that a variety of "natural," pharmacologic, and pathologic causes alter conscious brain function and the associated EEG. This overlap points to the necessity of obtaining confirmatory data to determine the cause and treatment of such EEG changes.

Ischemia Monitor

The EEG is not homogeneous throughout the skull (Fig. 33-5). The EEG activity varies considerably depending on the location of the electrode pair being monitored. Therefore, a standardized scheme of electrode application, the 10-20-20 system, has been developed to more reliably position EEG sensing electrodes using surface landmarks of the head for reference points. This reference system enables electrode to be placed at positions that are likely to detect EEG changes because of alteration of local perfusion evoked EEG changes because of stimulation of a particular extremity (Fig. 33-6).

Doubt has been cast8 on the standard 10-20-20 system to have sufficient spatial sensitivity to determine small areas of cerebral ischemia. The EEG, however, is symmetric about the midline with EEG activity on 1 hemisphere generating a "mirror image" of the same area on the other side of the midline. This feature adds a benefit of internal control when assessing changes in an area of EEG activity. If the mirror image area of 1 hemisphere does not mimic the change on the other side, then the change is focal and likely attributable to a change in the activity of the area under monitoring. Symmetric changes in the EEGs of both hemispheres' activity suggest a global source of altered neural activity and therefore more likely attributable to global changes (ie, blood pressure, oxygenation, anesthetic depth). The converse is also true. Focal changes in EEG activity over 1 portion of the brain may not be detected over another portion of the same brain. This fact has consequence when one tries to describe a global change in cerebral activity from 1 focal electrode pair. At least 2 electrode pairs from mirror image sites on either cortex are necessary to conclude that a particular EEG change reflects a "global" change in activity.

The ischemic effect of carotid cross-clamping was reported by Chiappa et al in 19709 (Table 33-2). There was a variety of postclamping patterns. The changes were not global but reflected local changes in activity, underscoring the importance of a symmetric array of electrodes to detect changes over 1 hemisphere versus the other. EEG changes were sometimes observed over the cerebral hemisphere opposite to the clamped side, suggesting an interruption of collateral circulation.

A Cochrane meta-analysis10 reviewed 2 prospective randomized studies (composed of 590 patients) that compared routine selective shunting strategy with a no shunt strategy for carotid endarterectomy (CEA). The analysis reviewed a third study that explored the potential benefit of intraoperative EEG monitoring to identify CEA patients who would benefit from shunting. The authors concluded: "There is still insufficient evidence from randomized controlled trials to support or refute the use of routine or selective shunting during CEA. Furthermore, there is little evidence to support the use of one form of monitoring over another in selecting patients requiring a shunt."10

As regards the method of monitoring in selective shunting, until the efficacy of shunting has been demonstrated, further trials of the method of monitoring are probably not merited."10 This Cochrane analysis does not address the use of other potential strategies for brain protection during endarterectomy such as hypothermia, burst suppression, or induced hypertension or techniques performed on an awake patient (either endovascular carotid angioplasty or transcutaneous endarterectomy performed under local anesthesia).

After one has made the decision to monitor EEG for cerebral ischemic changes during any surgery, one is then confronted with the problem of determining if a change in EEG is attributable to focal hypoperfusion, generalized hypoperfusion (from low systemic blood pressure), or anesthetic effect. These possibilities guide anesthesiologists toward a strategy of anesthetic management during procedures that may result in ischemia.

1. Rely on anesthetic agents that have the least effect on EEG activity (short-acting narcotics, benzodiazepine) (Table 33-3).

2. Maintain a constant anesthetic level throughout procedure, especially at critical periods.

3. Be prepared to raise and lower blood pressure with nonanesthetic agents (inotropes, chronotropes, vasoactive drugs, and volume).

4. Be prepared to use alternative ischemia monitors (ie, somatosensory evoked potential [SSEP] if burst suppression is to be used for "cerebral protection").

5. Perform the carotid surgery with the patient awake.

Using Electroencephalography to Monitor Burst Suppression

The human data regarding burst suppression as a guide to administration of anesthetics for cerebral protection are mixed. Patients in whom focal iatrogenic ischemia is induced during middle cerebral artery (MCA) aneurysm clip ligation have a significant advantage over those receiving isoflurane when they are given pentobarbital as the primary neuroprotective agent or when they receive propofol or etomidate titrated to achieve EEG burst suppression.11 Melgar et al12 propose the use of EEG to identify CEA patients who develop ischemic changes following carotid cross clamp that are refractory to induced hypertension. To these patients, these authors administer etomidate to achieve burst suppression and proceed with endarterectomy without shunting. Bush et al13 conclude: "Percutaneous carotid stenting with neuroprotection provides comparable clinical success to CEA performed under local anesthetic."

However, a multicenter study of the protective effects of propofol-induced burst suppression during cardiac valve replacement showed no effect.14 Burst suppression achieved through various means (barbiturate or isoflurane) resulted in different effects on cerebral blood flow and calculated cerebral requirement for oxygen,15 demonstrating that the balance of oxygen delivery to consumption is altered differently by these 2 agents at equivalent degrees of burst suppression. Therefore, the use of burst suppression as a guide to anesthetic administration during procedures that place the cortex at risk must take account of the anesthetic agent used for the burst suppression, the surgical procedure for which it is intended, and the other hemodynamic and ventilatory parameters effecting oxygen delivery to the cortex.

Why Process an Electroencephalogram

The characteristics of the EEG associated with consciousness and unconsciousness are EEG synchrony and desynchrony. Synchronous EEG is obtained in natural sleep, sedation, anesthesia, and cerebral ischemia. Desynchronous EEG is observed during wakefulness and REM sleep (associated with dreaming). The degree of synchrony and desynchrony of the EEG varies. It is difficult for the human observer to reliably measure this feature by an inspection of the unprocessed EEG. Therefore, methods to objectively quantify the frequency and amplitude characteristics of the EEG have been developed.2,16 Two methods currently in clinical use are the bispectral index (BIS) and entropy.17,18

The Processed Electroencephalogram

EEG voltage oscillation is sufficiently irregular so that the voltage measured at one instant does not enable an exact prediction of voltage for the next instant. However, EEG voltage oscillation is not completely random. Knowing the voltage at a particular instant allows for prediction of the next voltage with some probability of certainty. The EEG signal is stochastic (from the Greek, stokhos, meaning "to aim"), meaning that it may be analyzed statistically but may not be predicted precisely. The stochastic nature of the EEG requires that the signal be described in terms of probability. The Fourier transformation of the EEG has been used to provide this summary and is the initial step used in both the bispectral and the entropy methods of EEG signal processing.

Bispectral Array

The three-pronged strategy underlying the analysis of "raw" EEG to derive a BIS is described by Sigl and Chamoun16 and Rampil2 (Fig. 33-7). This strategy includes (1) a determination of the slow wave content of the signal (beta ratio), (2) a determination of the bicoherence of all frequency pairs derived from a Fourier transformation (synch slow), and (3) quantifying the amount of burst suppression present in the EEG (burst suppression ratio). These 3 features are combined in an unspecified, weighted fashion to derive the BIS.

Because the synchronous activity of EEG (slow, large-amplitude oscillations) occurs during progression from wakefulness to natural sleep, does the BIS identify this as well? Sleigh answered this question in a study of human volunteers undergoing natural sleep (Fig. 33-8).19 The results, summarized in the accompanying figure and table, show that the BIS algorithm assigns low BIS numbers to EEG from naturally sleeping volunteers. A return of desynchronous EEG (ie, dreaming) is also identified by an increase in BIS number. The BIS shows a low number during physiologic sleep based on the prevalence of slow EEG activity during that time. The BIS assumption is that this slow EEG activity also corresponds to a deep state of anesthesia. The BIS algorithm does not uniquely identify an anesthetized EEG.

Phase coupling, which is measured by the bicoherence portion of the algorithm, may be the least useful component of the algorithm.20 In a study of processed EEG obtained from human subjects undergoing induction of anesthesia, Miller et al19 conclude that although the power content of slow-frequency EEG increases as subjects progress from wakefulness to anesthesia, the degree of phase coupling remains unchanged (Fig. 33-9).

These results show that the bispectral analysis did not give any more information than power spectral-based analysis and that most of the changes in the bispectral values result from decrease in the relative high-frequency content of the EEG caused by anesthesia.20

Entropy Monitors

Entropy is a concept related to the amount of "disorder" within a system18 and has been applied to thermodynamics and information theory by Shannon and Weaver21 and to power spectrums by Johnson and Shore22 in 1984. There are a variety of ways to compute the entropy of a signal in the time domain or frequency domain. The algorithm implemented by Datex-Ohmeda uses a combination of time and frequency domain approaches (Fig. 33-10). This approach allows for the contributions to entropy from any frequency to be explicitly separated. The computations are constructed in such a way that the length of the time window for each individual frequency is individually chosen in order to optimize response times. To summarize the algorithm, the raw EEG (signal) is transformed to a power spectrum by fast Fourier transformation (Fig. 33-10, step 1). A Shannon function is performed on each integer frequency to determine the contribution of each frequency to the overall entropy (Fig. 33-10, step 2). The total amount of entropy is determined by summing all the entropy contributions from each integer frequency (Fig. 33-10, step 3). Entropy ranges from 0 (order) to 1 (complete randomness).

State Entropy and Response Entropy

State entropy (SE) is computed over the frequency range from 0.8 to 32 Hz. It includes the EEG-dominant part of the spectrum and therefore primarily reflects the cortical state of the patient. The time windows for SE are chosen optimally for each particular frequency component and range from 6 to 15 seconds. Response entropy (RE) is computed over a frequency range from 0.8 to 47 Hz. It includes both the EEG-dominant and electromyography (EMG)-dominant part of the spectrum. The time windows for RE are chosen optimally for each frequency.18 White et al's study23 of 30 patients undergoing laparoscopy compared the entropy monitor and BIS and showed that both the BIS and RE parameters increased after reversal of cisatracurium paralysis. However, the study was not designed to determine the ability of the RE monitor to function as an early predictor of emergence in a muscle-relaxed patient.

Wheeler et al24 studied the RE in patients who were induced and paralyzed and found that the RE increased in a portion of patients undergoing arterial catheter placement or head pin placement before recovery from paralysis. They concluded that "increased RE during painful stimulation was not dependent on recovery from paralysis but was seen more often in patients anesthetized with 0.8% compared with 1.4% isoflurane. This suggests that RE reflects FEMG and may be useful to identify inadequate anesthesia and patient arousal during painful stimuli."24

Entropy: Clinical Studies Anesthesia versus Sedation

White et al's study23 concluded "the changes in SE and RE values followed a similar pattern to the BIS values during the perioperative period. Analogous to the BIS, the entropy indices display a high degree of sensitivity and specificity in assessing consciousness during the induction of and emergence from anesthesia and were able to detect changes associated with administration of IV (propofol) and volatile (desflurane) anesthetics during the maintenance period. Finally, the Entropy module experienced less interference with the displayed indices during use of the electrocautery unit than the BIS monitor."

How Does an Entropy Monitor Handle Burst Suppression

Periods of zero EEG voltage have zero entropy because the EEG value is constant. Burst suppression patterns contain variable amounts of entropy because a variable amount of the signal is composed of bursts of oscillations. Therefore, the entropy algorithm has been successful in identifying and quantifying the degree of burst suppression.

Conclusion: Bispectral and Entropy Processing of the Raw Electroencephalogram

Whatever method is used, bispectral analysis or entropy, the strategy still seems to address the issue of identifying a synchronous versus a desynchronous EEG. The entropy methods use electrical frequencies in the EMG range in an attempt to use muscle activity as a gauge to anesthetic depth. Those strategies are useful if one holds a number of dictums in mind:

1. There are multiple ways to generate a synchronous EEG, including ischemia, natural sleep, and anesthetic. Therefore, a synchronous EEG may not be predictive of the patient's response to stimulation nor for the time for change from a synchronous to desynchronous EEG.

2. Although a synchronous EEG may imply a nonperceiving brain, it does not guarantee that state and it does not define absolute cutoffs for achieving a nonperceiving state.

3. The use of synchrony or desynchrony evaluation of EEG is effectively done when one looks for a change in the EEG pattern in response to stimulus. An absence of change during noxious stimulation may add confidence that cerebral arousal did not occur (retrospective), but it does not add confidence that noxious-induced activation will not occur (prospective). However, if noxious event-related changes in synchrony to desynchrony do occur, the patient might be at risk for a more overt response to noxious stimulus (in the form of gross movement or recollection of the event).

Intraoperative evoked potential monitors have received the greatest clinical usage in monitoring the integrity of the spinal cord during surgery to correct spinal column deformities, remove spinal cord tumors and vascular malformations, and correct vascular lesions that jeopardize the spinal cord vascular supply. The decision to use this monitoring is based on the risk and the magnitude of possible deficit versus the predictive value (positive or negative) of the monitor in a given operative setting. Except for a few exceptions, (ie, facial nerve monitoring during acoustic neuroma resection),25 there are no mandatory uses of this monitoring even though a body of evidence supports its use in a variety of circumstances, including the following: (1) peripheral nerve or plexus surgery, (2) spinal cord surgery (deformity correction, traumatic spinal fracture repair, tumor removal), (3) brainstem surgery (posterior fossa tumor removal), (4) cerebrovascular surgery (CEA, aneurysm repair), or (5) identification of the sensory portion of the sensorimotor cortex (central sulcus identification or cortical mapping).26

How to Evoke a Cortical Response: Anatomy

Site of Stimulation

SSEPs are elicited by stimulation of a peripheral nerve at a distal site, typically the median or ulnar nerves at the wrist for acquiring SSEPs from the upper extremities and the posterior tibial nerve at the ankle or the peroneal nerve at the fibular head for acquiring lower extremity SSEPs. Stimulation is typically delivered by adhesive electrodes or fine-needle electrodes.

Pathway of the Response through the Spinal Cord

The SSEP signal enters the spinal cord through dorsal nerve roots and ascends the spinal cord via multiple pathways.27,28 The posterior column spinal pathways, which decussate at the cervicomedullary junction, primarily mediate the SSEPs. Other pathways such as the dorsal spinocerebellar tracts and the anterolateral columns, which decussate near the nerve root entry level, may contribute to the early SSEP responses that are used for monitoring purposes.26

Implications of Vascular Supply to the Spinal Cord

The blood supply for nourishing the posterior column pathways that mediate SSEPs is generally thought to be the posterior spinal arteries (Fig. 33-11). The anterior spinal artery is generally believed to provide the primary blood supply to the anterior and anterolateral portions of the spinal cord, which make up the remaining two-thirds of the spinal cord. Motor pathway function is mediated by spinal cord pathways, which receive their blood supply from the anterior spinal artery. Therefore, loss of motor function because of compromise of the blood supply to the anterior spinal artery may be associated with little or no loss of the sensory function that is mediated by the dorsal column pathways (anterior cord syndrome).

Global Cerebral Blood Flow

When cerebral perfusion decreases to about 18 cc/min/100 g of tissue, electrical activity of the brain decreases and SSEPs begin to diminish in amplitude. When perfusion decreases to 15 cc/min/100 g of tissue, electrical activity of the brain decreases still further and SSEPs are generally not recordable. Further decreases in blood flow to the brain, particularly if they are sustained, will result in cellular damage and irreversible changes in electrical activity.26 These responses depend on the blood supply to the brain and brainstem and the specific arterial branches that provide this supply. Perforating branches of the basilar artery and the vertebral artery supply the brainstem. The middle cerebral artery provides the blood supply to the area of the cortex, which mediates the upper-extremity SSEPs, whereas the anterior cerebral artery provides the blood supply to the area of the brain which mediates the lower-extremity SSEPs.

What Is the Response, and How Is It Quantified

Banoub et al29 conclude: "The single cortical sensory evoked response has a low amplitude (1-2 microV) compared with the much larger electroencephalogram waves (50-100 microV). Therefore, the EP (evoked potential) wave has to be extracted from concurrent spontaneous electroencephalogram activity by repetitive stimulation and computer-signal averaging techniques" (Fig. 33-12). They also write: "The EP waveform consists of a series of peaks and valleys presented as a graph of voltage over time and described in terms of amplitude, latency, and morphology. Amplitude is commonly measured as the waves' peak-to-peak voltage difference. Latency is the time from stimulus to the peak of the response. Interpeak latency is the interval between the peaks of interest"29(Fig. 33-13).

Anesthetic Effect on Somatosensory Evoked Potential: Implication of Synapses between Peripheral Nerve Sites, Brainstem, and Cortex

General anesthetics do not affect ascending SSEP responses up to the level of the medullary nuclei (nucleus cuneatus and nucleus gracilis). These early subcortical responses are predominantly a reflection of the integrity of spinal cord white matter and provide little direct information about the condition of spinal cord gray matter. The short latency evoked responses from these "subcortical" structures are useful to determine the integrity of the monitoring system (ie, Is the stimulator working? Is electrode impedance sufficiently low to enable peripheral nerve stimulation?) and to distinguish "real problems" in spinal cord integrity from "false-positive problems" caused by anesthetic-induced decrement of cortical SSEPs. Anesthetic effects on sensory responses are more pronounced in regions where synaptic transmission is prominent. Therefore, the effects are more pronounced in the EEG and cortically generated SSEP peaks. Responses of the brainstem, spinal cord, and peripheral nerve are markedly less affected because fewer synapses occur in these pathways. Anesthetic effects are clearly dose related. However, many agents have a disproportionate effect at low doses. Therefore, during periods of acute neural risk, a steady state of anesthesia is important.30

Choosing an Anesthetic Plan for Somatosensory Evoked Potential Monitoring29

1. Criteria for significant changes are difficult to establish and therefore are empiric. A significant change in SSEP, reflecting loss of integrity of peripheral nerve or spinal cord, is usually taken as a 50% decrease in peak amplitude and a 10% increase in peak latency provided these changes are not caused by anesthetic or temperature.

2. "General anesthesia has an inhibitory effect on neurotransmission and therefore on the EP. The effect of anesthetics is greater on synaptic transmission than on axonal conduction. For this reason, responses recorded from polysynaptic pathways (eg, cortical recordings) are affected by anesthesia to a much greater extent than those recorded from oligosynaptic pathways (eg, spinal cord and subcortical recordings).

3. "All volatile anesthetics produce a dose-dependent increase in SSEP latency and a decrease in amplitude. All volatile anesthetics, even at concentrations above 1.0 MAC [minimum alveolar concentration], only minimally affect the subcortical waveform, resulting in high recordability and reliability."29

4. The effect of volatile anesthetics on cortical SSEP amplitude is compounded by nitrous oxide.

5. Intravenous anesthetics generally affect SSEPs less than inhaled anesthetics (Table 33-3).

6. Etomidate and ketamine increase SSEP amplitude.

7. Propofol, midazolam. and barbiturates have a moderate depressant effect on SSEP amplitude. In a normothermic individual, SSEP monitoring can detect cerebral ischemia during barbiturate anesthesia at doses that induce burst suppression.30-32

8. Most authors report minimal to no effect of opioids on SSEP amplitude.

9. Clonidine can be used as an anesthetic adjuvant without compromising SSEP monitoring. Dexmedetomidine affects SSEP amplitude minimally at sedative doses. During isoflurane anesthesia, dexmedetomidine blunts isoflurane's effect on SSEP amplitude.

Application of Evoked Potentials

SSEP monitoring is generally used to assess intraoperative spinal cord, brainstem, or regional cortical function. In the spinal cord, SSEP monitoring is used during spinal cord manipulation (eg, spinal distraction and rod placement, intramedullary tumor and vascular malformation removal). In the brainstem, SSEPs are useful for monitoring lesions at the cervicomedullary junction, particularly below the entrance of the eighth cranial nerve to the brainstem, such as clivus cordoma resection. In the cortex, SSEPs are used for monitoring function in the anterior cerebral artery or middle cerebral artery distributions during procedures that place those areas at risk for ischemic damage (eg, vascular surgery or tumor resection).

The brainstem auditory evoked response (BAER) is derived in similar fashion to the SSEP, but in this application, the stimulus is an audible click delivered to the tympanic membrane through earphones. The evoked stimulus traverses the auditory nerve and brainstem tracts and arrives at the auditory cortex. The response of the EEG is summed in similar fashion to the SSEP, and a characteristic pattern of evoked peaks is generated (Fig. 33-14) corresponding to synapses that occur between eighth nerve and cortex. BAERs are typically used during surgeries involving the eighth nerve (especially removal of acoustic neuromas) or procedures involving the brainstem or posterior cranial fossa to ensure integrity of neural structures in this area.

The use of mechanically evoked motor potentials has taken the form of identifying muscles that may be rendered paralyzed during the dissection of a tethered spinal cord or meningomyelocele. Other uses are the monitoring of facial nerve function during head and neck surgeries that place the facial nerve at risk. Another method is to determine that the application of a pedicle screw has not come in close contact to the nerve root contained in a neural foramen. Motor evoked potentials (MEPs), stimulated over the motor cortex and measured at the peripheral muscle, enable a monitor of motor tract integrity and are thought to provide a more comprehensive scrutiny of the spinal cord when combined with SSEP monitoring.

Specific Cases and Use of Electrophysiologic Monitoring

The variety of surgical procedures and monitoring modalities that have been used productively are listed in Table 33-4.26

1. Before starting the case, consult with both the surgeon and the neurophysiologist. What is the procedure? What is the tissue at risk (eg, tissue perfusion, mechanical injury)? How is it at risk? When is it at risk (beginning, middle, end, postoperative)? What modality will be used to monitor tissue at risk? Are baseline measurements required preoperatively? Is there anything to be done with the anesthetic plan that enhances the predictive value of the monitor? Usually this takes the form of limiting potent agent and nitrous oxide, but above all keeping these exposure levels as constant as is consistent with an effective anesthetic. Is the monitoring enhanced through use of muscle relaxants or does the use of relaxants impede the detection of an evoked motor response? Monitoring techniques that require an unparalyzed but immobile patient rely on rapidly reversible deep narcotics and/or neuroleptic adjuncts such as dexmedetomidine. Are there alternatives in case of monitor failure (ie, planned intraoperative emergence to perform awake neurologic exam)?

2. There must be constant communication between anesthesiologist, neurophysiologist, and surgeon as to any significant changes in vital signs or administered anesthetic. The neurophysiologist must relay assessment as to the adequacy of monitoring and any change in the monitoring parameters so the cause-and-effect relationship can be identified.

3. Tailor the anesthetic plan to maximize a reliably timed and clear wake-up to enable rapid postoperative assessment of tissue at risk, which enables timely corrective action.

What Is the Effect of Anesthetic Agents on Spontaneous and Evoked EEG?

Banoub et al29 present is a succinct summary of anesthetic effect on SSEP.29 To summarize

1. All potent inhaled agents prolong SSEP latency in a dose-dependent fashion.

2. All potent inhaled agents diminish SSEP amplitudes in a dose-dependent fashion.

3. Sole use of 60% nitrous oxide decreases SSEP amplitude but does not effect latency

4. Additive use of nitrous oxide with potent agent further decreases SSEP amplitude but does not increase latency more than that seen with the potent agent alone.

Troubleshooting a Change in Somatosensory Evoked Potential

Because evoked potential monitoring is highly sensitive and relatively less specific, one can anticipate a fair frequency of "false-positive" changes in the amplitude or latency of evoked responses during surgical procedures. All 3 participants in the monitoring process (neurophysiologist, surgeon, and anesthesiologist) should have a consistent strategy for dealing with these events to determine whether the SSEP is a "false-positive" result (eg, SSEP changes caused by monitor artifact or anesthetic effect impeding the ability to monitor the nervous system). Some preliminary points to consider are the use of preprocedural monitoring in an already damaged nerve tract. If preanesthetic testing results are abnormal, then the likelihood of improvement with anesthesia and surgery is low and monitoring of that tract should not be attempted. The neurophysiologist will be adept at determining false-positive signs to faulty electrode application by way of impedance checks and by determining if the elicited responses are altered in redundant locations along the peripheral nerve–spinal cord–brainstem–cortex circuit. Global changes in SSEP will be explored to determine whether there is a problem with generalized electrical interference from electrocautery or other electrical monitors. While these causes are being eliminated, the anesthesiologist can ensure that the 2 main features of anesthesia-controllable effects—anesthetic depth and tissue perfusion—are optimal. If all of these maneuvers do not improve the SSEP, then plan for rapid wake-up and testing. This part of the plan involves patient preparation for responding to commands to move appropriate extremities.26

Auditory Evoked Responses

The genesis of auditory evoked responses (AEPs) shares similarities with the SSEP. They are electrical cortical potentials, evoked through repetitive stimulation of a peripheral nerve, and identified and measured by summating numerous, brief (millisecond) stimulus synchronized EEG epochs recorded from electrodes placed on the vertex and ear lobe. Cranial nerve VIII is stimulated by presenting audible clicks to the cochlea by earphones placed within the auditory canal. The AEPs are grouped based on their respective latencies. Brainstem auditory evoked potentials (BAEP), also called auditory brain stem responses (ABR), are composed of the short-latency waves (∼2-10 milliseconds) and are numbered I through VII. The purported neural structures33 that serve as the substrate for these short-latency waves are noted in Figure 33-14. Procedures that have been monitored effectively include resection of acoustic neuromas, decompression of the facial nerve for hemifacial spasm (which carries a large risk of hearing loss), posterior fossa work in which the brainstem is at risk, clipping of basilar artery aneurysms, and sectioning of cranial nerve VIII for intractable tinnitus.

Midlatency AEPs (MLAEPs) have a poststimulus latency on the order of 20 to 100 milliseconds and are generated from the medial geniculate and primary auditory cortex.34 Although the amplitude is correlated with increasing indicators of anesthetic depth, this is not a perfect monitoring modality because of poor agreement among experts as to what determines a good peak necessary for analysis.35 Also, by determining the PK33 of the midlatency BAER, Alpiger et al37 showed that end-expiratory sevoflurane concentration was a better predictor and may turn out to be more useful in the clinical setting. Long-latency AEPs (also called the auditory late response ALR) have a latency of 50 milliseconds 38 and are generated from the frontal cortex and association areas.34

The early-latency BAEPs appear to be exquisitely sensitive monitors for pathologic events during surgery. Because anesthetics and mild hypothermia have minimal effects, they are specific monitors as well. The MLAEPs show promise as evoked responses for monitoring awareness or depth of anesthesia. When the concentration of anesthetics is increased, the amplitudes of the MLAEPs' peaks are decreased and their latencies are elongated. Further testing will determine if these AEP-based monitors may be superior to processed EEG in detecting the transition from unconsciousness to consciousness.34

Motor Evoked Potentials

A MEP is an electrical potential evoked from a muscle (myogenic response) or from a motor nerve (neurogenic response) by stimulating the motor cortex, spinal cord, or peripheral nerve. The source of stimulation can be either electrical or magnetic. MEP monitoring is primarily used to ensure integrity of the motor tracts of the spinal cord during spinal cord or spinal column surgery. MEPs are also used to determine the positioning of epidural motor strip electrodes for chronic pain treatment.39-42

Why Obtain a Motor Evoked Potential

The review by Sala et al43 contains an extensive list for surgeries that may be monitored with MEPs:

1. Monitor the anterior cord for spinal tumor removal, spinal vascular surgery (ie, arteriovenous malformation), and spinal column distraction (scoliosis surgery).

2. Identify motor strip for analgesic electrode placement.39-42,44

3. Monitor the brainstem and thalamus for ischemia during basilar artery aneurysm surgery.45

4. Facial nerve EMG during posterior fossa surgery.25,46

5. Upper arm EMG for surgery within the brachial plexus.47-49

6. Lower extremity EMG for myelomeningocele repair and release of tethered cord.43

7. EMG during acetabular surgery in which epidural was used for the surgery with use of passive EMG.50

Why Spinal Cord Stimulation Differs from Transcranial Stimulation

Toleikis et al51and Rose52 show that neurogenic monitoring after direct spinal cord stimulation may not specifically evaluate anterior spinal cord function. Measured impulses over the peripheral motor nerve may be attributable to antidromic stimulation of α1 spindle afferents that stimulate motor neurons. These authors believe that similar "contamination" may occur with transcranial stimulation. They offer a "collision" method of stimulation to circumvent this problem.51,52

Motor Evoked Potential Monitoring during Anesthesia

The effect of a variety of anesthetic agents, including alfentanil, sufentanil, fentanyl, remifentanil, thiopental, midazolam, etomidate, ketamine, and propofol, on neurogenic and myogenic MEPs was evaluated in a study combining clinical data obtained in 40 patients and experimental investigations conducted in 140 animals. Opioids, propofol, and thiopental suppressed myogenic but not neurogenic MEPs in a dose-dependent fashion; remifentanil exerted the least suppressive effects. Etomidate and midazolam did not suppress myogenic MEP even at plasma concentrations sufficient for anesthesia. Ketamine induced moderate reduction in compound muscle action potential (CMAP) amplitudes only at high doses. Remifentanil and propofol administered via TCI (target controlled infusion) systems allowed recording of myogenic potentials within a defined target plasma concentration range.53

The effect of various anesthetics on transcranial MEPs was reported in a review of 20 published studies comprising 537 patients: "Although propofol does demonstrate a dose-dependent reduction in MEP amplitude without effect on latency, it has repeatedly been shown to produce a more stable neurophysiologic environment for monitoring, when compared with inhalational anesthetics."54 Dexmedetomidine has been used as a supplement to total intravenous anesthetic (TIVA) to reduce the dose of propofol without evidence of detriment. Neuromuscular blockade is known to suppress MEP signal recording. Partial paralysis has been used on rare occasions but is generally too unpredictable to use regularly.

The anesthetics decrement of transcranial MEPs is likely caused by inhibition of the motor neuron because motor potentials evoked from direct stimulation of the spinal cord are maintained even at 1.5 MAC.55,56 Intraoperative transcranial motor stimulation can be evoked through the use of a TIVA with propofol and remifentanil. Multiple stimulation techniques may improve the quality of weak responses.56-59

Direct comparison of an inhaled anesthetic with desflurane and nitrous oxide with a TIVA technique showed successful motor evoked monitoring with both techniques.60 Tanaka et al61 propose a method of compensation of transcranial MEP that is an easy and accurate method for removing the effects of muscle relaxants.

The advent of MEP monitoring represents a landmark in this recent progress. MEP monitoring is the most appropriate technique to assess the functional integrity of descending motor pathways in the brainstem and foremost in the spinal cord. Mapping of the corticospinal tract at the level of the cerebral peduncle as well as mapping of the VII, IX–X, and XII cranial nerve motor nuclei on the floor of the fourth ventricle is of great value with which to identify "safe entry zones" into the brainstem62 (Figs. 33-15, 33-16, and 33-17).

Electromyography

Elicited responses to motor nerve stimulation are measured as CMAPs from intramuscular needle electrodes or surface-adherent electrodes placed over the muscle of interest. Stimulation of these responses can be either passive or active. Passive stimulation is used to alert surgeons that a dissection has trespassed on a peripheral nerve by mechanically stimulating a motor fiber resulting in a muscle twitch. Active stimulation is used to electrically stimulate tissue of interest to determine which muscles are innervated by the nerve of interest. This strategy is used for brachial plexus exploration or for identifying nerve roots during meningomyelocele repair or release of tethered spinal cord. The National Institutes of Health Consensus Conference on Acoustic Neuroma recommended that facial nerve monitoring46 should be used in all patients undergoing surgical resection of an acoustic neuroma.25

Intraoperative Wake-Up Test

Historically, before the use of electrophysiologic spinal cord monitoring, the functional integrity of the complete motor system (upper and lower motor neurons and peripheral musculature) was confirmed after spinal distraction for scoliosis by means of intraoperative emergence from anesthesia and demonstration of voluntary leg movement. This technique, the Stagnara wake-up test (named after one of its originators), was introduced in 1973 by Vauzelle et al63and is described by Owens.64 Careful preoperative patient preparation is necessary to determine that the patient understands what is required and is able to comply. Intraoperatively muscle relaxation and anesthesia are reversed while the narcotic level is maintained. The patient emerges to a level of consciousness at which point he or she is asked to follow commands. The commands include hand grip and movement of the feet. Use of the upper extremities ensures that the patient is sufficiently awake to follow commands. If quadriplegia is a possibility, the patient can be asked to open the mouth or grimace to determine that he or she is sufficiently awake.

The Stagnara wake-up test has a number of limitations. It is a test of gross motor function and cannot precisely assess specific muscle group or nerve root function. It is not a test of sensory function; therefore, a patient could have a significant sensory deficit in the presence of grossly normal motor tract function. The test is time consuming, requires the reversal of anesthetic agents, and poses some risks to the patient, including aggressive patient movement resulting in trauma or extubation. Test administration can be difficult in patients with reduced capacities (eg, mental retardation, deafness, inadequate language skills, insufficient emotional capacity). Repeated administration is difficult and not often undertaken. In most cases, the test is administered at the completion of all corrective maneuvers. This single administration can reduce its sensitivity to the onset of a neurologic injury and the subsequent efficacy of intervention. However, the wake-up test can be useful to further evaluate the veracity of a change in neurophysiologic monitoring parameters in a patient under general anesthesia.

Awake Craniotomy

The "silent cortex" is so named because lesions of these cortical regions are not evident by simple neurologic testing. In distinction, lesions in the "eloquent cortex" are identified by alterations in normal neurologic function, including speech and motor and sensory function. Intraoperative monitoring of eloquent cortical function can be effectively performed through use of cortical mapping. Mapping is performed intraoperatively by transiently inducing functional ablation of small portions of the cortex with stimulating electrodes. Correlation of structure and function requires an awake patient who is capable of performing the neurologic function under surveillance (ie, speech or directed movement) so a cortical stimulation of a particular portion of cortex can be linked with an interruption of that function (ie, aphasia or uncontrolled contraction of a muscle group). Anesthetic techniques enabling this testing vary from an "awake craniotomy," which relies on local anesthetic blocks of scalp, periosteum, and dura supplemented by a "neurolept anesthetic."65 More recently, general anesthetic for craniotomy followed by an intraoperative emergence for cortical testing followed by reinduction of general anesthesia (the so called asleep–awake–asleep technique) has gained wide acceptance. Challenges of any method include the issue of airway control (ensuring sufficient airway patency and ventilatory drive to maintain acceptable oxygenation and carbon dioxide level) as well as minimizing coughing and the Valsalva maneuver in a patient with an open cranium. These challenges are met through use of various airway adjuncts, including nasal trumpets and laryngeal mask airways. Significant reflux risk can be addressed by endotracheal intubation followed by extubation at the time of emergence for intraoperative testing. Another challenge is maintaining sufficient balanced analgesia and anxiolysis during the awake portion of the case to enable patient participation. A variety of intravenous anesthetic techniques have been used and tend to rely on quickly titratable and reversible narcotic (eg, remifentanil) and anesthesia provided by an infusion of propofol or infusion of dexmedetomidine.66 A drawback to propofol is the respiratory depression associated with concomitant narcotic use. A relatively brief exposure to a propofol infusion is associated with a brief predictable emergence time. However, because of context sensitivity, a more prolonged exposure is associated with longer less predictable emergence. Dexmedetomidine infusion used with a narcotic infusion provides dose-related deep sedation as well as maintenance of ventilatory drive during craniotomy. However, airway patency is not ensured at the higher doses, and emergence times are not as brisk. However, a low dose of α2 agonist may be beneficial as a "rescue" medication for patients after emerging who demonstrate significant anxiety and hypertension that may preclude cortical testing.

Spinal Anesthesia for Epidural Placement of Spinal Cord–Stimulating Electrodes

The accurate placement of spinal cord–stimulating electrodes to treat chronic pain requires patient responsiveness to determine that the patient perceives the stimulating current in the same region as the perceived pain. Because the electrode placement may also require a laminotomy to introduce the electrode into the epidural space, the anesthetic challenge is to provide sufficient anesthesia for laminotomy and electrode placement while the patient is in a lateral or prone position. A combination of local anesthetic and conscious sedation is feasible. However, this technique runs the risk of airway loss. Additionally, prolonged deep sedation can impair patient response, which may hinder optimal electrode placement. A technique has recently been described that involves the use of subarachnoid bupivacaine. The spinal block is performed in the prone position, and hypobaric bupivacaine is introduced. Through blockade of the spinal nerve roots, there is complete dermatomal anesthesia, which enables a painless laminotomy to be performed. However, because spinal anesthetic does not completely block sensory tracts within the spinal cord, there is a maintenance of patient perception of stimulation from the epidural electrodes.67

Surgery on and around the nervous system is safer when the functional integrity of the nervous system can be ensured. Anesthetic agents, by definition and design, decrease neurologic function. Therefore, monitoring nervous system function during anesthesia can be challenging. The techniques presented in this chapter offer a standardized approach to assessing intraoperative neurologic function. Much work needs to be done to improve the impact of anesthetics on intraoperative neurologic monitoring, and these techniques will surely evolve.

• Lind G, Meyerson BA, Winter J, Linderoth B. Implantation of laminotomy electrodes for spinal cord stimulation in spinal anesthesia with intraoperative dorsal column activation. Neurosurgery. 2003;53:1150-1153; discussion 1153-1154.
• Martin JH. the collective electrical behavior of cortical neurons: the electroencephalogram and the mechanisms of epilepsy. In: Kandel ER, Schwartz JH, Jessel TM, eds. Principles of Neural Science. 3rd ed. New York, NY: Elsevier; 1991:777-791.
• Owen JH. The application of intraoperative monitoring during surgery for spinal deformity. Spine. 1999;24:2649-2662.
• Rampil IJ. A primer for EEG signal processing in anesthesia. Anesthesiology. 1998;89:980-1002.
• Sala F, Krzan MJ, Deletis V. Intraoperative neurophysiological monitoring in pediatric neurosurgery: why, when, how? Childs Nerv Syst. 2002;18:264-287.
• Toleikis JR. Intraoperative monitoring using somatosensory evoked potentials. A position statement by the American Society of Neurophysiological Monitoring. J Clin Monit Comput. 2005;19:241-258.