Prognostic value of somatosensory-evoked potentials in neurology: A critical review in hypoxic encephalopathy
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.181555
Source of Support: None, Conflict of Interest: None
Prediction of prognosis in comatose patients surviving a cardiac arrest is still one of the intractable problems in critical care neurology because of lack of fool-proof ways to assess the outcome. Of all these measures, somatosensory-evoked potential (SSEP) has been perhaps the most evaluated and heavily relied-upon tool over the past several decades for assessing coma. Recent studies have given rise to concerns regarding the “absoluteness” of SSEP signals for the prognostic evaluation of coma. In this critical review, we searched the literature to focus on studies conducted so far on the prognostic evaluation of postanoxic coma using SSEPs. All those studies published on the use of SSEP as a prognostication tool in postanoxic coma were reviewed. A narrative review was created that included the strengths as well as limitations of the use of SSEP in postanoxic coma. The use of SSEP in coma has been universal for the purpose of prognostication. However, it has its own advantages as well as limitations. The limitations include challenges in performing and getting SSEP signals during coma as well as the challenges involved in reading and interpreting the signals. The recent usage of therapeutic hypothermia has become another factor that often interferes with the SSEP recording. Finally, based on these study results, some recommendations are generated for the effective use of SSEPs in comatose patients for further prognostication. We advocate that SSEP should be an integral component for the assessment of postanoxic comatose patients due to its several advantages over other assessment tools. However, SSEP recordings should follow certain standards. One should be aware that its interpretation may be biased by several factors. The bias created by the concept of “self-fulfilling hypothesis” should always be borne in mind before discontinuation of life support systems in terminal patients.
Keywords: Coma; prognostication; somatosensory-evoked potential
Coma remains to be one of the most challenging conditions for management as well as prognostication for the treating neurologist. Prediction of prognosis in comatose patients surviving a cardiac arrest is still one of the intractable problems in critical care neurology because of lack of fool-proof ways to assess the outcome. Currently, the methods being used for predicting a poor outcome within 72 h of the cardiac arrest include ocular reflexes, electroencephalogram (EEG) reactivity, somatosensory-evoked potentials (SSEPs), and serum biomarkers. Of all these measures, SSEP has been perhaps the most evaluated and heavily relied-upon tool for assessing the prognosis of patients in coma. SSEP measures the conduction time across the spinal cord and central conduction pathways through the dorsal roots, ascending pathways, and thalamus to the primary sensory cortex.
The SSEP signals typically consist of small (<10–50 μV) electrical signals that can be recorded noninvasively from the skull. These signals are originating from the activity of somatosensory cortex after a set of electrical stimuli are administered to one of the peripheral nerves. These signals have been named according to the time-latency at which they appear from the application of the stimulation; for example, the negative deflection arising at around 20 ms has been named as N20 and so on [Figure 1].
SSEPs constitute an integral component of several neurological evaluations. The most important of all of its applications in neurology has been the prognostic evaluation of coma. SSEPs have especially been used for prognostic predictions in posttraumatic and anoxic–ischemic coma. SSEPs have gained an important position in the assessment of coma because of their reportedly high specificity and their advantageous position over other assessment methods (such as clinical signs, motor responses, and pupillary light reactions). However, concerns have arisen in recent studies regarding the “absoluteness” of SSEP signals for the prognostic evaluation of coma., If SSEP is not used with adequate sophistication and knowledge, the interpretation of the originating signals might not be as reliable as had once been thought. In a critical condition like coma, this could result in a premature/inappropriate withdrawal of the life support systems, thereby leading to the death of the individual and thus, perpetuating the so-called “self-fulfilling hypothesis.”, In this critical review, we reviewed the literature on the prognostic evaluation of postanoxic coma using SSEPs. The advantages as well as limitations of SSEPs in postanoxic coma following a cardiac arrest are studied. Based on the review, recommendations are being proposed for the recording as well as interpreting of the SSEP signals in postanoxic coma. We strongly suggest that the interpreting neurologist should have a multi-dimensional understanding of SSEPs. Caution must be exercised at several levels before coming to a decision regarding the continuation/withdrawal of life support based on the SSEP findings.
It is important to have a neurophysiological understanding of the SSEP changes observed in coma. SSEPs survey the integrity of the neuroaxis and are indicators of the illness progression and prognosis in both acute brain and spinal injuries. Based on electrophysiological studies, it has been hypothesized that the hypoxic–ischemic injury causes electrophysiological derangements resulting in an increased latency (due to decrease in the conduction velocity) and amplitude (due to dispersion of conduction). Two types of abnormalities have been proposed by these studies. The first type of electrophysiological abnormalities is the dissociation of excitation. All of the SSEP abnormalities observed in ischemic coma have been attributed to this abnormality. These changes include increased amplitude and latency, as well as changes in the conduction velocity and time–frequency distributions.,,, The characteristic changes in magnitude and latency associated with anoxic coma have been observed in several studies in the past decades ,, and will be dealt within the next section. The second kind of abnormality is that obtained due to disturbed connections induced by alterations in the neural signals and sources that result due to the ischemic injury. These kinds of alterations are almost always ignored and could also represent the noise in the central nervous system induced by the ischemic injury.
A plethora of human as well as animal study data has provided the evidence regarding the predictive ability of SSEP in postanoxic coma. The most universally accepted finding has been the unfavorable outcome (persistent vegetative state [PVS]/death) associated with bilateral absence of cortical responses [Figure 2].
Thus, in patients who remain comatose after a cardiac arrest, absent SSEPs have been shown to be highly specific in predicting a poor outcome. Several meta-analyses have been conducted for evaluating this predictive ability.,,, A summary of these meta-analyses is reported in [Table 1].
These meta-analyses have included several large-sample cohort studies as well as a multitude of small-sample studies., 12, ,,,, All these studies invariably have shown a high specificity of the absent N20 responses in predicting a poor outcome (PVS/death) in postanoxic comatose patients. This predictive ability seems to persist irrespective of the therapeutic interventions used, for example, hypothermia, or the type of anesthesia used., An important addition was the study conducted by Madl et al. This study evaluated the N70 SSEP waveform instead of the N20 waveform and still found a very high specificity (97%) in predicting a poor outcome. This was comparable with that obtained with the lack of N20 response. This one study showing a high specificity of the N70 waveform is still of limited value due to the absence of any other study corroborating this finding. Several limitations also exist in the recording as well as the interpreting of SSEP responses in coma. Therefore, this particular finding should be subjected to further well-structured studies before being introduced as a well-established procedure for the prognostication of postanoxic comatose patients in routine clinical practice.
As mentioned before, the major advantage of SSEP as a prognostication tool is its higher specificity as compared to other methods. This fact was very evident in the meta-analysis presented by Kamps et al., which compared the false-positive rates (FPRs) of SSEP with several other assessment methods such as the corneal reflexes, Glasgow coma scale (GCS), and response to painful stimuli. They observed that the FPRs of SSEP in detecting a poor outcome were the lowest (0.007) as compared to the corneal reflex (FPR = 0.02), pupillary response (FPR = 0.004), and GCS (FPR = 0.21), thus showing the highest specificity for SSEP signals. Similarly, Madl et al., found that the N70 SSEP responses had the highest specificity as compared to the other prognostication methods.
The second major advantage of SSEP leading to its superiority over the other methods is that its conduction time, and thus the signals generated are not affected by several factors including the level of consciousness and the anesthetics used. Propofol produces minimal to <10% suppression of the SSEP amplitude. It has been observed that benzodiazepines such as midazolam as well as the commonly used opioids have a minimal effect on the SSEP amplitude and latency. The only exception seems to be remifentanil. Thus, SSEPs are often useful in conditions in which the clinical examination is of limited value, like in sedated patients. Cortical N20 responses are also not influenced even by sedation levels that are sufficient to induce an isoelectric EEG, thus indicating their superiority over EEG reactivity.
The third major advantage seems to be that SSEPs can be conducted within 24 hours or even earlier after a cardiac arrest thus providing an earlier indication of a poorer prognosis, at which time other tests, especially a clinical examination would be unhelpful, especially for prognostication. However, this early recording is not recommendable for deciding the prognostic outcome due to its limitations, which we shall be discussing below.
It is a commonly held belief that patients who have lost cortical SSEPs after a successful cardio-pulmonary resuscitation will never regain consciousness. In addition, SSEPs have an advantageous position over other tools. Taken together, these findings led to the recommendation in the 2006 American Academy of Neurology (AAN) practice parameters that “bilateral absence of cortical SSEP response with median nerve stimulation recorded on days 1–3 or later after resuscitation could accurately predict a poor outcome.” However, SSEPs are not without limitations. Recent studies have shown that SSEPs may not be unequivocal predictors of a poorer outcome, as has been considered so far. Thus, blindly following the absence of SSEP in predicting a negative outcome may lead to the dangers of the “self-fulfilling hypothesis” (in which a researcher's firm belief or expectation regarding the results expected from a technique erroneously leads to his/her conviction that those very results have been obtained) and an early discontinuation of life supports in these patients. For example, in a recent review, Sandroni et al., found that in 6/8 studies on SSEP, the investigated predictor was used for taking a decision to withdraw treatment, leading to the risk of taking an erroneous decision based on the “self-fulfilling prophecy.” These limitations exist both in the recording and interpretation of SSEPs. Therefore, the practice of withdrawing the life-sustaining and rehabilitative efforts only on the basis of the absence of the SSEP response may not be a well-substantiated step.
It has been repeatedly mentioned in the literature that the prerequisite for the absent SSEPs to be consistently predictive of a poor outcome is that the cortical responses have to be absent bilaterally in a “technically well-performed test.” However, a technically well-performed test would need to fulfill several criteria which are often too difficult to implement on a regular basis. Three of these technical challenges that need to be overcome while conducting the SSEP in comatose patients include the following factors.
The interference effects of noise
The most challenging aspect in the recording of SSEP seems to be in neutralizing various electrical noises interfering with the neuro-electrical signals originating in the cortex. It is also one of the major causes of disagreement over the signal interpretation across different studies. In fact, in the study by Zandbergen et al., the inter-observer agreement was moderate with the main source of disagreement being the noise. For assessing prognosis after cardiac arrest, only the short cortical latencies (N20, expected to appear 20 ms after median nerve stimulation) are used. SSEPs, being only several microvolts in amplitude, are prone to interfering with other electrical signals in this intensity range. Usually, three sources of noise have been identified that interfere with SSEP signals: (i) Biological noise from the subject's own body; (ii) electrical noise from instrumentation; and (iii) electrical noise from the surrounding environment. The signal-to-noise ratio (SNR) of SSEPs is especially low during the first 24 h after the patient sustains a cardiac arrest. It is practically impossible to completely eliminate all of these noises. However, some techniques have been applied for tackling this problem: (a) Giving of muscle relaxants to patients with too much muscular activity; (b) turning off whenever possible, the intensive care unit electrical equipment that can interfere with the electrical recordings; (c) providing a higher frequency of stimuli (up to 1000 or more); (d) increasing the stimulus intensity that also improves the SNR; and (e) using appropriate averaging techniques such as ensemble averaging that are usually used to improve the SNR by averaging waveforms in a large time window (20–30 trials); (f) recording of waveforms after 24 h of the onset of coma as the noise levels decrease after this time. The cortical responses can only be reliably interpreted when the peripheral and spinal responses are also present. If peripheral responses are not present, then the loss of signal on SSEP recording may be due to peripheral nerve damage.
The effects of therapeutic hypothermia on the N20 response
Therapeutic hypothermia (TH) is an important treatment-related variable in the cases that develop a cardiac arrest. Interest in the application of hypothermia in humans was recently revived after animal studies showed a positive outcome in rats treated with mild TH following reperfusion after a cardiac arrest. The mechanism of the effects of hypothermia on the brain seems to be more complicated than was earlier thought, which probably includes the reduction of both the early hyperemia and the delayed hypoperfusion that develops after a cardiac arrest. Studies indicate that the use of TH influences the N20 response. This has also been acknowledged by the current AAN practice parameters which state that bilateral absence of the N20 response between 24 and 72 h after a cardiac arrest can accurately predict a poor prognosis in patients who have “not undergone hypothermia.” However, the effects of hypothermia on SSEP seem to be more complex and confusing. Most studies have been very recent with only a few delving into this issue. Most of the data on SSEP as a prognostication tool have been acquired before the use of hypothermia and many more studies are required to find out exactly how TH affects the SSEP responses.
One unequivocal finding across all the studies has been that hypothermia increases the latency of the N20 waveform. The earlier studies unrelated to cardiac arrest had evaluated the effects of hypothermia on the SSEP N20 response. It had been observed that conduction velocity in the median nerve decreases by 2 m/s for every 1°C decrease in body temperature, thus resulting in an increased N20 latency., At temperatures of 32°C to 34°C, the N20 latency becomes prolonged. In fact, some studies have even reported a complete disappearance of the N20 waveform below body temperatures of 30°C.,, Similar findings have been reproduced in postcardiac arrest patients undergoing TH. Recently, Zanatta et al., studied the amplitude and latency of four different SSEP signals – N9, N13, P14/N18 interpeak, and N20/P25 retrospectively in 84 patients undergoing a cardio-pulmonary bypass during normothermic (36 ± 0.43°C) and mildly hypothermic (32°C ± 1.38°C) conditions. SSEPs were recorded in normothermia immediately after the induction of anesthesia, and in hypothermia, as the temperature reached its steady-state, and specifically when the nasopharyngeal temperature was equivalent to the rectal temperature (±0.5°C). They observed that hypothermia significantly increased the latency of all SSEPs (N9, N13, P14/N18, and N20/P25).
In a sharp contrast to the effects of hypothermia on the latency of SSEP signals, the effects of hypothermia on the amplitude of the waveform remain a much more confusing and unexplored area. In an earlier study by Hayes et al., whereas the normal controls showed a decreased amplitude of the tibial nerve SSEPs, the spinal cord injury patients showed an increase in the amplitudes, which was more than 100% in three patients. Recent studies actually report an increase in the N20 amplitudes during hypothermia. Animal studies have also shown that hypothermia induces a significant increase in the SSEP amplitude while increasing the SSEP latency. However, contrary to this finding, the study by Tiainen et al., demonstrated that the patients who underwent hypothermia and those continuing with normothermia did not have any significant differences in the N20 amplitudes. It was, however, not clear from the study if the SSEP recording was done during the induction of hypothermia or after the procedure had been performed. An absent N20 response during hypothermia seems to have the ability to predict a poor outcome although its specificity is still questionable. For example, in the same study by Tiainen et al., three patients who had an absent N20 response during hypothermia never regained consciousness, leading the authors to conclude that the prognostic ability of median nerve short-latency SSEP does not seem to be affected by TH. Similarly, Bouwes et al. reported that nine patients, who had an absent N20 during hypothermia, also had absent N20 after re-warming during normothermia, thereby giving a positive predictive value of 1. However, in the same study, one of the patients who had bilaterally absent N20 responses during hypothermia regained his responses after rewarming, thus creating a situation that was at variance with the conclusions of that study. Two of the similar exceptions were also observed by Leithner et al., where the authors observed either no response or a minimally detectable N20 response after 24 h of cardiac arrest and the recovery of these responses subsequently in patients who had received TH. However, it is not clear if these responses were observed during or after the hypothermia procedure. Irrespective of all the prevailing confusions, the above-mentioned data clearly indicate that the process of hypothermia affects the latency and possibly the amplitude of SSEP responses, especially of the N20 waveform. Thus, SSEP recording during the maintenance of TH for the purpose of prognostication should be avoided until larger studies prove the contrary fact that TH has no role in influencing the waveform. This is despite the fact that some studies do mention a high-degree specificity of absent N20 responses even in hypothermia.,
The N20 responses after the re-warming procedure seem to have a good predictive value in prognostication. This seems to be because the changes of SSEP and EEG induced by TH revert back to normal after re-warming of the patient., In the prospective cohort study of 77 patients, all 13 (17%) patients, who remained comatose after re-warming with bilaterally absent N20 responses, had a poor outcome (0% FPR). In two large prospective studies performed in patients treated with TH, bilateral absence of N20 at re-warming (that is, on day 2 or day 3) was found to be a reliable tool for predicting a poor outcome with a FPR of 0%. A pooled analysis of recent studies still gave a very low FPR of < 0.5%.,,,, Another study of 100 patients in whom the N20 response was assessed at least 24 h after the completion of re-warming and the discontinuation of sedation found that no patients with bilaterally absent N20 responses survived. This study was important because it compared several indicators such as the brainstem reflexes, myoclonus, absent motor response to pain, and SSEP signals; and SSEP was found to be one of the most reliable indicators for predicting the outcome after coma. In another multi-centric trial by the “Hypothermia after Cardiac arrest Study Group,” out of the 14 patients with an absent N20 response more than 72 h after a cardiac arrest, none regained consciousness.
Timing of the test administration
Recently, the issue of the appropriate timing for performing SSEP has been addressed. The earlier studies did not hesitate to conduct SSEP recordings within the first 24 h. In fact, a recent meta-analysis by Lee et al., demonstrated that the predictive ability of SSEP was superior to clinical signs only within 24 h of the onset of coma. Even consensus guidelines have recommended that SSEPs must be recorded within 24 h or even earlier. A closer look at some of the recent studies shows that this finding may not be as straightforward as it appears to be, especially in the era of TH. If we go by the parameter of SNR, then it seems logical to assume that the first 24 h should be avoided for the recording of SSEPs due to a very low SNR ratio in this period. In addition, in the present day scenario, because TH is initiated as early as possible, this period should be avoided for conducting SSEPs. Perhaps, these and several other factors contribute to the poor reliability of SSEP responses during the first 24 h of cardiac arrest observed across several studies. In one of the early studies, Robinson et al., observed that some false-positives were identified when SSEPs were performed too early (within 1 day) after the anoxic injury; thus, this test should not be performed earlier than 24 h after the cardiac arrest. Recently, Abend and Licht have reviewed the literature on the pediatric population having a hypoxic–ischemic encephalopathy and found that the SSEP responses were of prognostic value only if they were performed at least 24 h after the inciting event. The reason why absent responses during the early period after a cardiac arrest should not be taken with absolute certainty has been discussed by Leithner et al., who proposed that recovery of consciousness and cognitive functions is possible in spite of absent or minimally present N20 responses more than 24 h after the cardiac arrest, although in a very small proportion of patients. He observed that the N20 responses may actually show recovery when sufficient time is allowed to relapse beyond this time window before conducting the SSEPs.
Lack of sensitivity and the limitation in detecting a good outcome
While the absence of P20 waveform has been considered as being highly specific in detecting a poor prognosis, the presence of a N20 response peak (on one or both sides) cannot be interpreted as a definite favorable predictor because many of these patients will not regain consciousness. In fact, almost one-half of the patients showing the presence of the N20 waveform will end up having a poor outcome. Expressed more directly, unfortunately, only a small proportion of patients with a poor outcome after resuscitation have absent SSEPs. Therefore, many of the patients who have SSEPs present may also have a poor outcome, which results in a low sensitivity value for the prognostic capabilities of the test.
In the study by Tiainen et al., three patients (one in the hypothermia group and two in the normothermia group) had bilateral N20 responses, but did not awaken. Thus, the sensitivity was 75% (95% confidence interval, 30–95%) in the hypothermia group and 80% (95% confidence interval, 49–94%) in the normothermia group. Therefore, the presence of N20 response cannot be considered as an indicator of a good outcome.
Exceptions to specificity
The high specificity of the SSEP N20 responses in predicting a poor outcome has been maintained over several well-designed prospective cohort studies. However, several exceptions have surfaced recently in the literature which should also be taken into consideration while deciding on continuing/discontinuing the life support systems in postanoxic comatose conditions. In one such case, a 16-year-old boy, who had a cardiac arrest and was not treated with TH, had a good outcome despite the repeated absence of cortical SSEP responses (at days 3 and 9). In another study of 185 patients, one patient who had absent N20 responses did survive subsequently, and ultimately achieved a good recovery. Another patient in the same study had severe bilateral reduction of N20 amplitudes, but also survived and achieved a normal cognitive functioning. Owing to these isolated cases, the authors concluded that a bilaterally absent N20 response at 72 h may not predict a poor prognosis with absolute certainty. This finding was important because in both these patients, the absent N20 responses were observed after re-warming and after >24 h, thereby eliminating time-based and hypothermia-related effects.
In addition to these isolated cases, at least in one prospective study, the use of SSEP was found to be questionable for predicting the outcome in comatose patients. Oddo and Rossetti conducted a multivariable ordinal logistic regression and operator characteristic curves on 134 consecutive adults treated with TH after cardiac arrest. Variables related to the cardiac arrest, clinical examination, EEG reactivity, SSEPs, and serum neuron-specific enolase were included in the model to predict the clinical outcome at 3 months. In the 72 patients who had a poor outcome, the multivariable ordinal logistic regression identified the absence of EEG reactivity, incomplete recovery of brainstem reflexes in normothermia, and neuron-specific enolase levels being higher than 33 μg/L, but not SSEPs, as independent predictors of a poor outcome. In addition, the combination of clinical examination, EEG reactivity and neuron-specific enolase yielded the best predictive performance, with a 100% positive predictive value. Addition of the SSEP results to this model did not improve the prognostic accuracy.
The specificity has been observed to decrease further when multiple observers/raters were involved. In the study by Pfeifer et al., when four neurologists were involved in rating, the specificity value for a poor neurological outcome of the SSEP pattern D achieved a specificity of 93.5%, and that of E, a specificity of 98.4%. Surprisingly, in this study, the mean correct prediction by the experts was 81.8% in the resuscitated patients with a good neurological outcome; whereas, in those with a poor neurological outcome, the correct prediction was achieved in only 63% (60.5–66%) patients.
Poor inter-rater agreeability
One of the less discussed limitations of SSEP is its moderate reproducibility of interpretation. In one study, SSEP recordings from 56 patients with hypoxic ischemic encephalopathy were interpreted independently by five experienced clinical neurophysiologists. The inter-observer agreement was moderate (kappa 0.52, 95% confidence interval = 0.20–0.65) with the main source of disagreement related to the noise levels.
In another study, five neurologists agreed only moderately about the presence of the N20 peak detected after cardiac arrest in their patients. In a recent study, only well-preserved SSEP patterns had “very good” expert agreement (kappa-coefficient (κ): 0.88). For patterns predicting a bad outcome, kappa was only 0.76 (”good”). With a specificity of 93.5%, the pattern of bilaterally absent N20 predicted a poor outcome less accurately than was previously estimated. This unexpected suboptimal reliability indicates that SSEPs cannot unequivocally and accurately predict a bad outcome after cardiac arrest.
Pfeifer et al., investigated the inter-observer variability in interpretation of the median nerve SSEPs with regard to the neurological prognosis in survivors of cardiac arrest. Four experienced neurologists analyzed 163 median nerve SSEPs (133 from cardiac arrest survivors and 30 from healthy volunteers) on the basis of a predefined classification of SSEPs into five patterns (A–E). The experts were blinded to whether the SSEP finding was from a cardiac arrest survivor or a healthy volunteer. They were also unaware of the neurological outcome for the resuscitated patients. Three categories were defined for decision making: (i) “A good neurological outcome” represented by patterns A–C, (ii) “a poor neurological outcome” (patterns D and E), and (iii) “not evaluable.” The kappa-coefficient representing the agreement level of experts for all decision-making classifications was 0.75; for patients with a poor outcome it was 0.76, and for those with a good outcome, 0.88. The strongest correlation with a poor outcome was found for pattern E, that is, bilateral absence of the N20 peak. Thus, it seems that a proportion of patients who will be labeled as having a good/poor prognosis by one expert will be considered in the opposite category by another expert.
In the preceding sections, we have discussed the advantages as well as limitations in using SSEP for prognostication of postanoxic coma. It is important to mention that the diagnosis of brain death should be primarily based on clinical findings augmented by neurophysiological assessments. Therefore, the SSEP findings should not be taken in isolation, but instead should be a part of a multi-domain assessment. Broadly, five different recommendations may be offered for the conduction and interpretation of the SSEP recordings in coma. Prediction of a good prognosis should be avoided based on the intactness of SSEP wave forms. Finally, the most important recommendation in view of the exceptions to the N20 rule is that life support should not be withdrawn only on the basis of absence of N20 waveforms, and several other factors should be taken into consideration [Table 2].
Based on the literature review, it is obvious that SSEP has been used and will continue to be used for assessment of outcome in postanoxic coma. The most common and well established correlation has been observed between bilaterally absent N20 responses and a poor outcome. We advocate that SSEP should be an integral component for the assessment of patients in postanoxic coma due to its several advantages over other assessment tools. However, the recording of SSEP should be a part of multi-factorial assessment model and its should not form the single test for predicting the prognosis. Most importantly, clinical assessment should be an integral part of the assessment for diagnosing brain death. SSEP recordings should follow certain parameters and interpretation should be done with several factors in mind. Finally, the dangers of the bias created by “self-fulfilling hypothesis” should be borne in mind before deciding on discontinuation of life support.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2]
[Table 1], [Table 2]