Chapter 16: Evoked Potentials in the Operating Room and ICU

During operative procedures requiring anesthesia resulting in depressed consciousness surgeons have limited means to assess the integrity of the nervous system using clinical examination techniques alone. Monitoring techniques during surgery or interventions (such as interventional neuroradiologic procedures) may allow documentation of acute, but still reversible, changes in neurologic function. Additionally, these techniques can be used intraoperatively to assist in identifying important neural structures. Techniques commonly used for intraoperative monitoring include electroencephalography (EEG) and evoked potentials; the later will be discussed in this chapter. Evoked potentials include somatosensory evoked potentials (SSEPs), brainstem auditory evoked potentials (BAEPs), visual evoked potentials (VEPs), and motor evoked potentials (MEPs). Each of these specifically assesses different sensory or motor pathways and can be selected based on the individual clinical scenario. Wiedemayer et al 2002¹ reported that of 423 operations with intraoperative monitoring, surgical decisions were successfully modified in 5.2%. Using both SSEP and BAEP monitoring, the rates were: true-positive findings with intervention, 42 cases (9.9%); true-positive findings without intervention, 42 cases (9.9%), false-positive findings, 9 cases (2.1%), false-negative findings, 16 cases (3.8%), and true-negative findings, 314 cases (74.2%).

Evoked potentials (EPs) are electric potentials recorded from the nervous system following presentation of a stimulus, which can be auditory, visual, or electrical. EPs are orders of magnitude smaller in amplitude than EEG signals and require signal averaging and precise localization of the recording electrode to measure a response.² The recorded potential is time-locked to the stimulus and most of the noise occurs randomly, allowing the noise to be averaged out by collecting repeated responses. EPs can be recorded from cerebral cortex, brainstem, spinal cord, and peripheral nerves. Usually, the term evoked potential is reserved for responses involving either recordings from, or stimulation of, central nervous system structures. Thus, evoked compound motor action potentials (CMAPs) or sensory nerve action potentials (SNAPs) as used in nerve conduction studies (NCS) are generally not thought of as evoked potentials, although they do meet the above definition. Clinically, the recorded electric potentials are evaluated for morphology, latency, and amplitude. These values may then be compared to laboratory-established norms or compared to the contralateral side, in which case patients may serve as their own control.

Visual Evoked Potentials (VEPs)

These are generated by placing a recording electrode over or near to the visual cortex and applying visual stimuli such as a flashing light or a flickering checkerboard.³ Measured from the primary recording electrodes over the primary visual cortex, these stimuli typically produced a negative deflection at approximately 75 milliseconds (N75) and a positive deflection at approximately 100 milliseconds (P100).

Somatosensory Evoked Potentials (SSEPs)

This test assesses the integrity of the dorsal column-lemniscal system.⁴ This pathway projects via the dorsal column of the spinal cord to the cuneate nucleus in the lower brainstem, to the ventroposterior lateral thalamus, to the primary somatosensory cortex, and then to a wide network of cortical areas involved in somatosensory processing. Median and tibial nerves are most often stimulated in SSEP testing, although others (eg, ulnar, peroneal) may be used when appropriate.

The stimulus for SSEPs is a brief electric pulse delivered by a pair of electrodes placed on the skin above the nerve. To minimize artifact produced by electric stimulation, a ground electrode is placed between the stimulation site and the recording site. Both standard surface disc electrodes and needle electrodes can be used as recording electrodes. For upper limb SSEP recording, electrodes are placed at the clavicle between the heads of the sternocleidomastoid muscles (Erb point) and on the skin overlying cervical bodies 6-7. On the scalp, electrodes are placed at CP3 and CP4 of the International 10-20 System. For tibial SSEPs, electrodes are placed in the popliteal fossa and over the lumbar vertebra. At least two bipolar channels (eg, CPz-Fpz and CP3-CP4) should be used to record the cortical component. The SSEP waveform is obtained by averaging typically from 500 to 2000 stimuli; it is necessary to repeat at least two independent averages to demonstrate reproducibility. SSEPs are recorded using a broad bandpass filter with high-pass and low-pass filters set typically to 30 and 2000 Hz, respectively. Notch filters to eliminate power line noise (50 or 60 Hz) should be used with caution because of their tendency to create "ringing" oscillatory artifacts.

A normal median SSEP waveform is shown in Figure 16-1 and normal values in Table 16-1.⁵ The purpose of obtaining peripheral potentials (ie, N9, Erb point) is to differentiate peripheral causes of conduction delays as seen in peripheral neuropathy from central ones. The P14 is generated in the caudal medial lemniscus within the lower medulla and the N20 reflects activation of the primary somatosensory receiving area, located in the posterior bank of the rolandic fissure in Brodmann area 3b. Middle and late latency potentials are less frequently used. A delayed or absent signal at the Erb point may suggest a brachial plexus injury. A delayed or absent potential at the base of the brain may suggest pathology in the upper cervical cord or brainstem. If the typical brainstem potentials (ie, N14) are present, but cortical potentials like the N20 are not seen, this suggests disruption to the thalamicocortical projections, which may be related to pathology such as tumors, anoxic injury, or ischemia.

Motor Evoked Potentials (MEPs)

Other evoked potentials typically used in the operating room are MEPs (Figure 16-2). These EPs are generated using magnetic stimulation at or close to the primary motor cortex and have the recording electrode placed in the relevant muscle.

BAEPs are produced by an audible clicking stimulus. The first 10 milliseconds on the response elicited from the audio stimulus represents the electric transmission of that signal through the brainstem and thus is known as the BAEPs.³ Recording electrodes are between the Cz (according to the international 10-20 system) and the ipsilateral ear. The normal BAEP typically shows five to six peaks labeled with corresponding Roman numerals (Figure 16-3). Although generators of individual peaks are still somewhat controversial, commonly identified generators include the distal auditory nerve (wave I), auditory nerve as it exits the porus acoustics or the cochlear nucleus (wave II), the cochlear nucleus or ipsilateral superior olivary nucleus (wave III), the superior olive or nucleus or axons of the lateral lemniscus (wave IV), and the inferior colliculus and ventral lateral lemniscus (wave V). As waves II and IV are less reliably recorded across individuals, clinical interpretation is based primarily on assessment of waves I, III, and V.

Absence or delays in wave V with a normal wave I latency can be seen with conduction abnormalities central to the distal portion of cranial nerve VIII, while absence of wave I with preserved wave V may reflect a problem with the peripheral hearing apparatus or with the auditory nerve, but commonly reflects technical difficulty of recording wave I. Functional disruption of the brainstem may cause loss of waves II to V with perseveration of wave I. Unilaterally abnormal BAEPs most often reflect ipsilateral brainstem damage. More rostral structures, including generators within the primary auditory cortex and surrounding areas, are responsible for longer latency components⁶; however, similar to other EPs, the later components of the BAEPs are substantially affected by arousal state and are less commonly used in the OR or ICU setting.⁷

In the OR, EPs are used for two related purposes: (1) to minimize postoperative injury and (2) to guide treatment. Monitoring may detect functional impairment of neuronal function before permanent deficits occur. Intraoperatively, neuronal function is at risk from a number of mechanisms, including direct surgical transection, transmission of heat from surgical devices, stress, compression, and compromise of blood supply. In addition to EPs, recording needles may be placed in muscles of pertinent myotomes to look for spontaneous electromyographic (EMG) activity caused by stimulation of the relevant nerve. The surgeon may also have a specialized dissecting device that also serves as a stimulating electrode. EMG activity in the muscles of the face may, for example, signal the surgeon that he/she is dissecting too close to the facial nerve while trying to remove a pontomedullary tumor. A surgeon may also apply an electric stimulus to a screw placed during back surgery. The corresponding EMG activity detected in the relevant muscle may give the surgeon an idea of how close the screw is to the nerve root.

ORs are electrically noisy environments with abundant artifacts from use of surgical equipment such as a Bovie device and also ungrounded electric equipment. Consideration must also be given to the type of stimulation and recording electrodes that are used. It is also important to note the placement of the ground electrode on the patient. Stray current from various surgical devices may travel along the surface of the skin and cause burns at the site of electrodes. Specific consideration should be given to the type of electrodes used. Surface electrodes are appropriate for awake patients, but they may be easily dislodged and replacement may be very challenging in a prepped and draped patient. Needle electrodes are less likely to be dislodged and have a very low infection risk. Neurologic injury may occur from positioning and baseline studies, post-induction but prior to the final positioning, and may be helpful to differentiate EP changes present at baseline from those due to positioning or the surgical intervention.

Subcortical components (ie, waves I to V of the BAEP) are relatively unaffected by most medications commonly used in the OR and the ICU, whereas cortical ones are more susceptible to medication effects, especially at high doses. Sedative medications, including benzodiazepines, propofol, barbiturates, nitrous oxide, and halogenated inhalational agents, all depress the cortical components of evoked potentials in a dose-related manner.^8-10 Interpeak changes may be more stable (ie, I to V interpeak latencies for BAEPs) and comparing left- and right-sided recordings may be useful, particularly for SSEPs. Neuromuscular blockade should not affect SSEP,² but will affect MEPs.

EPs and particularly SSEPs are frequently used to predict outcome after cardiac arrest and have been recommended for prognostication. These recommendations are based on one prospective and a number of retrospective studies.^11-20 The prospective study of 407 cardiac arrest patients showed that bilaterally absent N20s were seen in 45% of patients who were comatose at 72 hours, and all of these had poor outcome.²⁰ In a meta-analysis of 4500 postanoxic patients, bilaterally absent N20s within the first week had a 100% specificity to predict poor outcome.¹⁹ Combining SSEPs with other predictors of outcome after cardiac arrest such as EEG or serum markers of neuronal injury improves prognostic accuracy.^16-18 While some predictions can be made for poor recovery, it is much more difficult to predict good outcome.

Two randomized control studies have shown that therapeutic hypothermia post–cardiac arrest related to ventricular fibrillation and ventricular tachycardia improves both survival²¹ and neurologic outcome.^21,22 In a prospective study of 111 cardiac arrest patients who underwent therapeutic hypothermia,²³ none of the patients who had absent SSEPs 24 hours after weaning of sedation had a favorable neurologic outcome. Another study,²⁴ which was retrospective, looked at 185 therapeutic hypothermia patients post–cardiac arrest. This study found that of the 36 patients who had bilaterally absent SSEPs, only one patient had a good recovery. The mean time to SSEP study was 3 days. In a small control study by Tiainen et al²⁵ of 60 cardiac arrest patients, 30 of whom underwent therapeutic hypothermia, no patient in either the hypothermia group or the normothermia group with absent SSEPs awoke. These SSEPs were preformed 24 hours post–cardiac arrest, and so the patients in the hypothermia group had a mean temperature of 33°C. Despite these studies, larger prospective studies need to be performed to determine the utility of SSEPs in prognosticating after cardiac arrest.

BAEPs have rarely been studied systematically in cardiac arrest patients. In a small cohort of 13 patients, middle latency auditory evoked responses (MLAEPs) were absent in all patients who died or remained in a persistent vegetative state.¹⁴ In the study by Tiainen et al²⁵ mentioned above, BAEPs were found not to be useful for prognostication after cardiac arrest.

One study¹⁴ examined the value of using median nerve SSEP, BAEP, and MLAEP in 131 ICU patients who were comatose as a result of anoxic ischemic injury, TBI, complication of neurosurgery, or encephalitis. These investigators used a grading system to categorize electrophysiologic studies and test the accuracy to prognosticate outcome (Table 16-2). This grading system is not widely used clinically.

Electrophysiologic studies were preformed between days 1 and 46 after onset of coma.¹⁴ Among patients with anoxic brain injury, 38% recovered with SSEP grade 1 classification. No patient recovered from coma with grade 2 or 3 in the anoxic ischemic group. In the TBI patients, 81% with grade 1 median SSEP recovered from coma, whereas 86% recovered with a grade 2 SSEP. No one in the TBI group recovered with grade 3. In the poststroke group, 60% recovered from coma with grade 1 median SSEP, 69% recovered with grade 2, and again no one recovered from coma with grade 4 SSEP. Of note, no patient in the anoxic group with an N20-P24 amplitude of less than 1.2 μV recovered. This was not noted in the TBI patients or the poststroke patients. In summary, absent primary cortical median nerve SSEP responses were very accurate in predicting poor outcome after anoxic brain injury.¹⁴ Many studies have not taken dynamic changes of the EP findings into account, but some evidence suggests that secondary loss of initially present cortical responses carries a poor prognosis.¹⁴

ERPs are long-latency potentials also visualized through signal-averaging techniques of the EEG²⁶ and are thought to reflect more complex cognitive processing of stimuli. Examples of ERPs include P300, N100, and MMN (mismatch negativity). The N100 is thought to represent attention,²⁷ the P300 is elicited by a rare task related stimulus,²⁶ while the MMN is elicited by an "oddball" sound in a sequence of sounds.²⁶ A meta-analysis compared the predictive ability of late-stage EPs for awakening from coma due to ischemia, hemorrhage, trauma, anoxic injury, metabolic etiologies, and postoperative causes (Table 16-3).²⁸ The presence of a N100 had a sensitivity of 71% (CI 66-76) and specificity of 57% (CI 51-63) for good outcome across the several studies. The MMN had a sensitivity of 38% (CI 32-43) and specificity of 91% (CI 85-95). The P300 had a sensitivity 62% (CI 53-69) and specificity of 77% (CI 70-82).

MMN showed a relatively high specificity for recovery of wakefulness, particularly in anoxic injury.²⁹ Fischer et al found the MMN to be the most powerful prognostic indicator for awakening from a coma and used it as the initial criterion to prepare a decision tree that they proposed for post–cardiac arrest prognostication.²⁹ This is not widely used.

The bulk of our experience of using EP in the hypothermic patient stems from intraoperative cases, particularly during cardiac and neurovascular surgery, but this needs to be considered for ICU patients given the increasingly widespread use of therapeutic hypothermia. In a cohort of nine aortic surgery patients, subcortical SSEP components (P13 and N14) became more consistently recognized at profound hypothermia (less than 20°C) when compared to normothermia; however, cortical components disappeared.³⁰ A linear correlation between decreasing temperature and increasing cortical and peripheral SSEP latencies (N10, P14, and N19) was seen in another cohort of cardiac surgery patients,³¹ while the amplitude was poorly correlated with the temperature. Whereas cortical potentials disappeared between 20°C and 25°C, peripheral components remained stable.

There is a linear increase of BAEP latencies and interpeak intervals by approximately 7% per drop of each degree celsius.³² At 26°C, BAEP normal values are approximately doubled. Longer latency components may disappear with moderate degrees of hypothermia, but ultimately all components will disappear below temperatures of 14°C or 20°C.^33,34

In a recent study of patients with cardiac arrest undergoing therapeutic hypothermia (goal 33°C), SSEPs and BAEPs were used to study patients between 24 and 28 hours after the arrest.²⁵ The investigators confirmed previously established EP prognosticators (ie, poor outcome with bilaterally absent N20 in the SSEP) in the setting of induced hypothermia. BAEPs had no additional value in outcome prediction.

Many studies have investigated the utility of EPs after TBI and demonstrated the ability to predict short-term mortality rate and long-term outcome. Most studies agree that bilaterally absent cortical SSEP responses in the setting of intact peripheral and spinal potentials universally indicate poor outcome.^35-38 Those with normal SSEPs (N = 553) have a favorable outcome (normal or moderate disability; Figure 16-5) in 71% of cases compared to unfavorable outcome (severe disability, vegetative state, or death) in 99% of those with bilaterally absent SSEPs (N = 777).³⁹ The positive predictive value and sensitivity are 71% and 59%, respectively, for normal SSEP, predicting favorable outcome; and 99% and 46.2%, respectively, for bilaterally absent SSEP, predicting unfavorable outcome. SSEPs perform favorably when compared to other predictors of outcome after severe TBI, including the GCS, pupillary or motor responses, CT, and EEG.⁴⁰ The predictive accuracy of abnormal, but present, cortical responses such as in our patient or unilateral abnormalities is more ambiguous.

Serial EPs may detect recovery^41,42 (see Figure 16-5) or secondary injury (ie, blossoming hemorrhages or ICP crises) after TBI.

BAEP abnormalities are more frequently seen in TBI patients with poor outcome.^43,44 Applying the above introduced grading system⁴⁴ to TBI patients, 75% of those with grade 1 or 2 BAEPs recovered from coma compared to 50% of those with grade 3 or 4. MLAEPs performed similarly.

SSEPs do not provide the same predictive accuracy of poor outcome in coma from other neurologic injuries and are currently not widely used for these patients. However, BAEP and SSEP abnormalities are seen in SAH with poor functional outcome.^45,46 Bilaterally absent cortical potentials carry a similarly dismal prognosis as in other etiologies discussed above.⁴⁵

Many practice parameter guidelines accept EPs as one option of a confirmatory test.⁴⁷ The American Academy of Neurology (AAN) recently updated guidelines on the determination of brain death.⁴⁸ SSEPs were not noted to be one of the typical ancillary tests that are used. Patients evolving to brain death will lose subcortical SSEPs and all BAEP responses when investigated with serial examinations.^49,50 Loss of the median nerve SSEP noncephalic P14 and of its cephalic referenced reflection N14 as well as the N18 is seen in brain death.^50,51