Article Text
PDF
Abstract
Cardiopulmonary resuscitation, basic life support and early defibrillation are leading to more survivors of out-of-hospital cardiac arrest reaching hospital. Once stabilised on an intensive care unit, it can be difficult to predict the neurological outcome using clinical criteria alone, particularly with modern management using sedation, neuromuscular blockade and hypothermia. If we are to prevent ongoing futile life support, it is important to try to identify the majority of patients who, despite best efforts, will not make a meaningful recovery. Somatosensory evoked potentials are widely available electrophysiological tests that can provide an objective biomarker of a poor neurological outcome and assist in predicting the prognosis.
- EVOKED POTENTIALS, SOMATOSENSORY
- EEG
Statistics from Altmetric.com
Introduction
Hypoxic–ischaemic brain injury typically follows cardiac arrest or profound hypotension. Its neurological outcome depends upon several factors, including the severity of the initial insult, effectiveness of immediate resuscitation and transfer, and the post-resuscitation management on the intensive care unit. About 60 000 out-of-hospital cardiac arrests occur every year in the UK; emergency medical services care for about 30 000 of these. Predictors of survival include the time to emergency response, witnessed arrest, effective bystander cardiopulmonary resuscitation, initial shockable rhythm, early defibrillation, and most significantly, prehospital return of spontaneous resuscitation.1 While hypoxic–ischaemic encephalopathy generally has a poor neurological prognosis, there are a significant number of survivors with no or minimal cognitive impairment. The neurologist is often asked to assess patients in the intensive care unit. We clearly need an accurate predictor to avoid withdrawal of treatment from those who might make a good neurological recovery, but in others to prevent ongoing futile life support.2 In 2006, the American Academy of Neurology published guidelines on prognostication after cardiopulmonary resuscitation following cardiac arrest, identifying three methods reliably to assist prediction of poor outcome: clinical examination findings (motor response to pain of no better than extension, myoclonic status epilepticus, absent pupillary and corneal reflexes), biochemical markers (serum neurone-specific enolase >33 μg/L) and electrophysiological studies (bilateral absence of the N20 wave of the somatosensory evoked potential).3
However, early prognostication is challenging, not least because sedation, ventilation, hypothermia and neuromuscular blockade potentially confound neurological assessment. Indeed, the American Academy of Neurology guidelines were published before any systematic review of the effects of these (particularly hypothermia) on prognosis. Therapeutic hypothermia can improve neurological outcome in comatose patients after cardiac arrest4 and is widely used. Therapeutic hypothermia may also depress neurological function and delay sedative clearance; thus, in cooled patients, predictions based on clinical examination alone may be overly pessimistic. Several studies on the effects of mild hypothermia have confirmed this. First, Al Thenayan et al5 reported two patients whose motor responses were no better than extension on day 3, but who subsequently regained motor responses and awareness. Rossetti et al6 prospectively studied 111 patients and found more false positive mortality predictions than the American Academy of Neurology guidelines would suggest: these included incomplete brainstem reflex recovery (4% vs 0%), absent motor response to pain (24% vs 0%) and myoclonic status epilepticus (7% vs 0%). Bisschops et al7 retrospectively studied 103 patients and identified 4 of 36 with a favourable outcome who had at least one of the three clinical signs at day 3. Samaniego et al8 identified several patients whose clinical examination proved unreliable and recommended that irreversible management decisions should not be based on a single prognostic variable. Somatosensory evoked potentials are an obvious objective choice to improve reliable prognostication since they change little with mild hypothermia, neuromuscular blockade and sedatives.9 Here we review somatosensory evoked potentials and their clinical uses, and particularly their role in prognostication after hypoxic–ischaemic encephalopathy, and include our own findings after therapeutic hypothermia.
Somatosensory evoked potentials
Somatosensory evoked potentials are the complex series of electrical potentials generated in sensory pathways at peripheral, spinal, subcortical and cortical levels of the nervous system (see figure 1). They can be elicited by electrical or other stimuli, including tactile, mechanical and thermal.10 Clinical somatosensory evoked potentials are elicited most effectively by electrical stimulation of a peripheral nerve, which preferentially activates the faster conducting fibres from proprioceptors and mechanoreceptors destined for the dorsal column–lemniscal (or ‘lemniscal’) pathway. Somatosensory evoked potentials have been used to study disorders of the brain, brainstem, spinal cord, sensory spinal nerve roots and peripheral nerves. Depending on the clinical question, neurophysiologists can use multiple bipolar or referential montage recording techniques. Clinical recordings are nearly always bilateral, usually non-invasive, rapidly acquired and portable, such that they can be recorded at the patient's bedside or in the operating theatre. ‘Short-latency’ somatosensory evoked potentials are defined as the portion of the waveform occurring within the first 25 ms (after stimulating the upper limbs) or 50 ms (for the lower limbs). They can localise impaired sensory conduction and may suggest axonal loss (attenuated or absent responses) or demyelination (prolonged or absent responses) but are neither disease specific nor pathognomonic of any particular process.11
Schematic of the somatosensory system. After Crucco et al10, reproduced with permission.
Stimulation and recording technique
Somatosensory evoked potentials are elicited by transcutaneous bipolar, monophasic square wave pulses of 0.1–0.2 ms duration from a constant current stimulator (table 1). Rather than the supramaximal stimulation technique of peripheral nerve conduction studies, a 2–3 times sensory threshold for a pure sensory nerve or ‘mild’ muscle twitch for a mixed nerve achieve optimal responses. Recording disc electrodes (with skin-electrode impedances ideally <5000 Ω) are placed over the peripheral, spinal and cortical sensory pathways. Cortical somatosensory evoked potentials are best recorded over the primary somatosensory cortex (S1, area 3b), but have a wide central projection. The analysis times for ‘normal’ ‘short-latency’ responses are 50 ms (upper limb) and 100 ms (lower limb) but may need extending in some clinical situations (eg, demyelinating disease) or to observe long-latency responses (>100 ms post stimulus). At least two repetitions are needed to assess reproducibility of responses.12 The elicited waveforms are labelled with the prefix N or P, indicating polarity, followed by the nominal post-stimulus latency (in milliseconds) of the healthy population (eg, N20 indicates a negative response over primary somatosensory cortex at ∼20 ms post stimulation). The peak latency, rather than its onset, along with amplitude of the peak to subsequent opposite polarity peak, and inter-peak intervals are measured (see figure 2 and table 2). Abnormality is determined by the absence of an obligate waveform or prolongation of a component beyond 2.5 or 3 SD of the normal range.
Median nerve somatosensory evoked potentials: suggested stimulus value (range) and recording variables
Median nerve somatosensory evoked potentials: suggested recording channels, mean latencies and upper limits of normal in young adults of body height 1.70±0.1 m
Median nerve somatosensory evoked potentials. Walsh et al,13 reproduced with permission.
Clinical uses
In general, somatosensory evoked potentials can detect peripheral or central nervous system disease; their clinical value is based upon their localising role, ease of use, reliability, reproducibility, sensitivity, and in certain circumstances, prognostic ability. They are most likely to be abnormal when there is concordant neurological abnormality. However, they may also detect subclinical lesions of the sensory pathways, particularly in demyelinating disease and in an unconscious patient.13 However reassuring a ‘normal’ somatosensory evoked potential might appear in medically unexplained sensory disturbance, it cannot exclude organic disease since somatosensory evoked potentials primarily index the large sensory fibres of the lemniscal pathways only.11 Somatosensory evoked potentials help most when testing a clear pathophysiological hypothesis. It is worth emphasising that their results are not pathognomonic for any specific disease and can be affected by peripheral neuropathy, which is common in the elderly and may be coincidental. In contrast to EEG, somatosensory evoked potentials have no real role in assessing cortical activity, depth of coma or anaesthesia, although they can help to distinguish between cerebral, uni-hemispheric and brainstem lesions.
Prognosis in hypoxic–ischaemic encephalopathy
Trojaborg and Jørgensen14 showed that the cranial nerve areflexia seen in deeply comatose patients with isoelectric EEGs is associated with absent median nerve somatosensory evoked potentials. Indeed, Goldie et al15 regarded absent brainstem auditory and somatosensory evoked potentials as evidence of brain death. Walser et al16 found at postmortem that amplitude reduction of cortical potentials yielded an estimate of cortical damage in hypoxia. Zegers de Beyl and Brunko17 prospectively studied 50 adults after cardiopulmonary resuscitation, within 8 h of coma onset, and found absent cortical somatosensory evoked potentials in 30 patients who died or remained in a vegetative state; none recovered consciousness even after ‘surviving’ for weeks. Throughout the 1980s and 1990s, single-centre observational outcome studies confirmed this, even enhancing the predictive accuracy by recording long-latency somatosensory evoked potentials,18 and by multivariate analysis of combined clinical and electrophysiological variables.19 However, short-latency somatosensory evoked potentials alone at 24 h predicted the eventual individual outcome significantly more accurately than clinical findings alone.20 Systematic reviews in patients older than 10 years showed that after hypoxic–ischaemic encephalopathy, and only in this condition, absent cortical somatosensory evoked potentials powerfully predict the prognosis, with a near 100% specificity for poor outcome (defined as death, persistent vegetative state or unconsciousness at 1 month),21 including after hypothermia.22
Following a prospective multicentre cohort study, there is now near universal agreement that, provided peripheral nerve and cervical cord function is preserved (ie, brachial plexus N9 and cervical N13 responses present), the bilateral absence of the N20 somatosensory evoked potential component reliably predicts non-awakening from hypoxic–ischaemic encephalopathy (ie, death, the vegetative or minimally conscious states).23 The main determinant of neurological outcome in hypoxic–ischaemic encephalopathy is preservation of cortical function, of which the N20 is an objective biomarker. Widespread ischaemia or cortical necrosis is required to obliterate the cortical somatosensory evoked response.24 The bilateral absence of N20 components received a level B recommendation (ie, ‘probably effective’) for accurately predicting poor outcome in the landmark Practice Parameter of a Quality Standards Subcommittee of the American Academy of Neurology.3 This conclusion was subsequently endorsed by an independent group of leading European authorities.25 Conversely, the presence of ‘normal’ cortical responses does not confer a favourable outcome, with a pooled sensitivity of only 39–46% for awakening when N20 is present bilaterally.3 ,25
False positive prediction of poor outcome is a major concern in prognostication as it may lead to a ‘self-fulfilling prophecy’ of death, after withdrawing life-sustaining treatment. Considering the very many patients studied in coma with somatosensory evoked potentials, there are relatively few case reports in the English language of awakening with bilaterally absent somatosensory evoked potentials. Table 3 lists (not exhaustively) most of these reports from some of the more prominent authors with experience in this field. Typically, but not exclusively, false positive predictions have occurred in children under 10 years, after traumatic brain injury, or in studies performed in the first 24–48 h of coma onset.26 ,27 Sedation does not usually markedly affect somatosensory evoked responses in subjects with ‘normal brain function’ after traumatic brain injury. There has been recent concern that hypothermia may be a confounding factor,3 ,28 although a subsequent meta-analysis found that somatosensory evoked potentials show comparable reliability in patients treated with or not treated with hypothermia, with an identical false positive rate of only 0.7%.29 This is important because a single-centre study showed that physicians may use somatosensory evoked potentials to estimate prognosis and thus the withdrawal of life support.
Case reports of recovery in patients with bilaterally absent N20 somatosensory evoked potentials responses
Our practice
At present, it therefore seems prudent to perform prognostic somatosensory evoked potential studies only in people after hypoxic–ischaemic encephalopathy who are aged over 18 years, at least 48 h or more after coma onset, who are normothermic, and with as little sedation as humanely possible. In our clinical practice, we perform somatosensory evoked potentials at least 72 h after their cardiac arrest, when a patient has been re-warmed following mild therapeutic hypothermia for 24 h (32–34°C), and their sedation safely reduced. For ease of interpretation, we regularly use neuromuscular blocking agents to reduce bioelectric artefacts (figure 3). We have assessed 30 consecutive patients (19 men, mean age 56 years, range 27–84) treated with therapeutic hypothermia after cardiopulmonary resuscitation for out-of-hospital cardiac arrest. Bilateral median somatosensory evoked potentials are recorded along with a standard EEG after day 3 post-cardiac arrest in the intensive care unit, with disposable scalp silver/silver chloride electrodes. We record a standard 20-min EEG using the Micromed system (Micromed, Woking UK) to determine if the patient is in myoclonic or non-convulsive status epilepticus, which is then treated accordingly with antiepileptic medication. The EEG abnormality is graded according to a standard scale (after Synek),30 endorsed by European experts,25 and reactivity to external auditory and noxious stimuli determined. We then stimulate both median nerves at the wrists with 0.1 ms square wave pulses at 0.5Hz and record two sets of 250 averaged somatosensory evoked potential trials, with a filter band pass of 3 Hz to 3 kHz, and active recording electrodes at Erb's point (N9), Cv5 (N13) and C3′/C4′ (N20) using a Medelec Synergy machine (Optima Medical, London, UK). The treating physician selected all our patients for electrophysiological assessment on the basis that the neurological outcome was uncertain, as they were still in coma (ie, Glasgow coma scale score ≤8) and that there was total or partial brainstem areflexia (ie, absent pupillary light reflexes and/or corneal ‘reflexes’). This multimodal approach, as outlined in figure 4, is recommended by both Rossetti et al6 and Samaniego et al8 along with their co-workers.
Somatosensory evoked potentials (EP): the top panel shows recording in a cardiac arrest patient without neuromuscular blockade; the bottom panel shows clearly absent cortical potentials (lower traces) in the same patient after neuromuscular blockade.
Prognostic algorithm adopted in Bristol, UK.
Our findings
The 30-day outcome post-cardiac arrest was determined for our patients from their medical records: only six (20%) survived to hospital discharge with either mild or no neurological sequelae, while most (24 patients (80%)) died soon after cardiac arrest and withdrawal of life-sustaining treatment (4–21 days). The electrophysiological recordings were dichotomised into either ‘malignant or uncertain’ EEG patterns and ‘absent or present’ N20 responses. A malignant EEG pattern and bilaterally absent N20 portending a poor neurological outcome (see table 4). If the EEG was an uncertain pattern and the N20 present, neurological outcome could not be determined with any degree of confidence (ie, indeterminate). In 13 patients, the N20 was absent bilaterally, while one patient had asymmetric responses (ie, one N20 absent and the other side present); all of these patients died. The N20 responses were present bilaterally in 16 patients but only 6 survived (38%). Overall, this gave a sensitivity of only 54% (95% CI 27 to 81), but specificity of 100% and a false positive rate of 0%. The EEG alone predicted outcome less well, in that 25 patients had a malignant pattern but 4 survived, while 5 had an uncertain pattern, 2 of whom survived, giving a sensitivity of 88% (95% CI 74 to 102), but specificity of only 33% and a false positive rate of 16%.
Dichotomised somatosensory evoked potential and EEG findings against patient outcome
Discussion
We adopted this pragmatic combined electroclinical algorithm for neurological prognostication out of a clinical need, the lack of definitive scientific evidence and concerns that clinical signs alone may be unreliable.5–8 It is difficult to discount the self-fulfilling prophecy in our patient population, although we had no patient survive after hypoxic–ischaemic encephalopathy who had bilaterally absent N20 responses. We feel that the bilateral absence of N20 component of the somatosensory evoked potential therefore remains an accurate predictor of poor outcome following hypoxic–ischaemic encephalopathy, as long as safeguards are in place and an appreciation of when somatosensory evoked potentials are invalid.26 Timing of the somatosensory evoked potential recording may be important, and authorities have recommended that somatosensory evoked potentials should not be performed within 24 h of cardiac arrest27 as there are reports of initially absent N20 responses after anoxia that subsequently recover by 72 h.31 For this reason, and so that our patients could be re-warmed, we therefore adopted a policy of recording after 3 days post-cardiac arrest. Some data suggest that therapeutic hypothermia does not affect the efficacy of somatosensory evoked potential findings,9 ,29 ,32 although our conservative approach was influenced by the very fact that it has been questioned.28 ,33 In fact, the evidence suggests that, if anything, the amplitude of cortical somatosensory evoked potentials increases with moderate hypothermia, although somatosensory evoked potentials disappear below 30°C.
We dichotomised our somatosensory evoked potential responses into ‘absent or present’ in order to simplify prognostication for the treating physicians, although the presence of the N20 appears insensitive to eventual neurological outcome. However, previous authors have pointed out that reduced cortical amplitude is also associated with a ‘bad outcome’,34 and this concurs with our own impression. There have also been concerns about interobserver variation in the interpretation of somatosensory evoked potentials after hypoxic–ischaemic encephalopathy,35 when the recording noise level prevents absence (or presence) of a low-voltage N20 being confidently determined. Both authors agreed that the responses in our patients were either absent or present, and the routine use of neuromuscular blockade to reduce noise may have facilitated this observation (see figure 3). We have seen that a malignant EEG pattern is not invariably associated with death or survival in the vegetative state, as found by others.36 Four of our patients who had malignant EEG patterns, two with burst suppression and two with generalised periodic epileptiform discharges, survived with minimal neurological sequelae. It is possible that these EEG patterns could have been pharmacologically induced and/or due to neurometabolic changes interacting synergistically with the after-effects of cooling, all of which the EEG is quite sensitive to. For this reason, we primarily view the EEG as a diagnostic tool for detecting ictal activity, rather than a prognostic one, although all six of our patients with ‘flat EEGs or electro-cerebral silence’ died (ie, Synek Grade 5: isoelectric EEG).30 These electrophysiological findings are in accordance with the systematic review and meta-analysis of Sandroni et al.37
Recently, there have been endorsements of multimodal approaches to prognostication after cardiac arrest, in patients either treated or not treated with hypothermia, by recognised authorities in the field38 and the European Resuscitation Council and the European Society of Intensive Care Medicine.39 Since a multicentre trial after cardiac arrest recently established that mild hypothermia at 33°C confers no benefit over ‘targeted temperature management’ of 36°C,40 hypothermia may become an outmoded treatment. Nonetheless, its introduction has refocused prognostication methods in the modern era. It is beyond the scope of this review to discuss the potential prognostic role of neuroimaging (CT and MRI techniques) or blood levels of biomarkers (S-100β protein and neurone-specific enolase), although no consistent threshold for a 0% false positive rate has been identified for neurone-specific enolase.37 While multimodal prognostication is logical, at the current time it is unclear whether a mathematical model can integrate both dichotomous and ordinal variables, and what weighting should be allocated to each variable. Furthermore, there are no internal or external validation studies available.39 Although it is recognised that no one single measurement will predict neurological prognosis with absolute certainty,33 somatosensory evoked potentials have been shown to be the single most reliable laboratory test for predicting poor outcome in hypoxic–ischaemic encephalopathy patients, treated either with or without hypothermia.3 ,37 Somatosensory evoked potentials can be performed safely at a patient's bedside in the intensive care unit and are now quite widely available in the UK. Our pragmatic stepwise combined electroclinical algorithm with binary probabilities (ie, poor vs indeterminate neurological outcome), in carefully selected patients, appears to be safe and effective. Indeed, this simple method may be as reliable as more complex scores derived with additional weakly predictive variables and interactive terms.
Key points
Hypoxic–ischaemic encephalopathy after cardiac arrest carries a high morbidity, but full neurological recovery is possible.
With current management techniques, prognostication based upon clinical and electroencephalographic features alone is prone to be overly pessimistic.
The bilateral absence of somatosensory evoked potentials is a reliable biomarker of poor neurological prognosis, with attention to timing and technique.
Ideally prognostication should be based upon a multimodality assessment, although quite how this can be applied in current clinical practice is yet to be determined.
Acknowledgments
We thank our clinical physiologists at the Bristol Royal Infirmary for the electrophysiological data collection, and Dr Jasmeet Soar FRCA FFICM FRCP, who commented on the manuscript.
References
Footnotes
Contributors NK formulated the article, analysed the data and drafted the manuscript. AO also analysed the data and revised the manuscript.
Competing interests None declared.
Provenance and peer review Commissioned; externally peer reviewed. This paper was reviewed by Eelco Wijdicks, Minnesota, USA and David Greer, Connecticut, USA.
Data sharing statement All the original EEG and somatosensory evoked potential recordings are available for independent review by a consultant clinical neurophysiologist through the corresponding author (NK) on paper or digital media.
Linked Articles
- Editors' commentary