A structured approach to neurologic prognostication in clinical cardiac arrest trials

The majority of patients who are admitted to the ICU following cardiac arrest are unconscious [24]. Patients who improve their level of consciousness after withdrawal of sedative and analgesic substances usually have a good outcome [25]. For those who remain in coma, the prognosis becomes gradually worse with increasing time from the insult [26, 27]. Repeated clinical neurologic examinations in combination with electrophysiologic investigations (electroencephalography (EEG) and somatosensory-evoked potentials (SSEP)) constitute the foundation of neurological prognostication. This may be further supported by neuroradiologic examinations (computed tomography and magnetic resonance imaging) and biomarkers (neuron specific enolase in particular), but the evidence for these methods is less solid [9, 15, 16]. Since the specificity of clinical as well as neurophysiologic findings increases with time, a well-founded judgment of prognosis can usually be made at 72 hours after cardiac arrest in a patient who has not been treated with hypothermia [15, 26].

Hypothermia treatment makes the clinical neurological examination less reliable [17, 20, 22] and this may be, at least partly, related to an increased use [22] and decreased clearance [28, 29] of sedative medication.

SSEP is less influenced by sedative medication than EEG and a bilateral loss of the cortical N20-potential at 24 hours or more after cardiac arrest, predicts a poor outcome with high accuracy [23, 30]. The high specificity of SSEP appears to be retained for hypothermia-treated patients if the examination is performed after rewarming, but sporadic false predictions have been reported during hypothermia [19], and even after rewarming [31]. Furthermore, interobserver variability in the interpretation of SSEP has been reported [32].

Severely pathological EEG-patterns including burst-suppression, generalized status epilepticus and α-coma are associated with a poor prognosis after cardiac arrest. The EEG-pattern is sensitive to sedative medication and false predictions may occur [15]. For example, the occurrence of postanoxic status epilepticus has been reported in some patients with good outcome following cardiac arrest and induced hypothermia [33].

A status myoclonus, defined as generalized myoclonic seizures for more than 30 minutes and usually engaging facial and axial limb musculature, has been considered a reliable predictor of a poor prognosis if it occurs within 24 hours after cardiac arrest of a primary cardiac origin [34]. However, survival with good outcome despite early status myoclonus has been reported in hypothermia-treated patients [35].

A small fraction of cardiac arrest patients develops a total brain infarction with massive edema leading to herniation and complete loss of brain stem function [12, 22]. Recently, a case report has questioned the accuracy of clinical tests to diagnose brain death in cardiac arrest survivors treated with hypothermia [36].

The Brain Resuscitation after Cardiac Arrest Trial (BRCT) 1 [37] and BRCT II [38] studies were conducted prior to the introduction of hypothermia and included comatose adult cardiac arrest patients. The BRCT I studied the effect of thiopentone and BRCT II the calcium entry-blocker lidoflazine. Neither study describes any rules for treatment decisions in patients who remained in coma. In 1994, Edgren et al. used the 262 control patients from the BRCT I study to investigate whether it was possible to reliably predict a permanent vegetative state few days after cardiac arrest. Variation in treatment of patients who remained in coma existed in this group, as the authors report: “For ethical and economic reasons it was not possible to require indefinite intensive care in the protocol, and local variations in decision making were permitted. Although most patients were given unlimited therapy, some in the Scandinavian centers were changed to intermediate care as early as 2–5 days after cardiac arrest.” No further details on treatment limitation or WLST were reported [26].

In 2002, the Hypothermia after Cardiac Arrest (HACA) [7] group and Bernard et al. [8] reported on the positive effects of treatment with hypothermia after cardiac arrest. After enrolment of 275 patients, the HACA study stopped inclusion due to a lower than expected inclusion rate. Of the included patients, 132 died during follow-up, but there is no information on how decisions on treatment limitations or withdrawal of care were made.

In the Bernard study, active life support was withdrawn from most patients who remained deeply comatose at 72 hours and this may actually have decreased the treatment effect since hypothermia makes the clinical examination less reliable at this point [17, 20, 22] and care may therefore have been withdrawn prematurely in cooled patients with a potential for late recovery [39]. On the other hand, the authors declare that “patients with an uncertain prognosis underwent tracheostomy” and this possibility to allow patients a prolonged observation period may have introduced a bias in favor of hypothermic treatment since the treating physicians were not blinded for temperature. The same holds true for decisions on WLST in both the Bernard and the HACA study. Of the 77 included patients in the Bernard study, 45 died during follow-up and the cause of death was severe neurologic injury and withdrawal of active therapy in the majority (34/45), which is similar to more detailed studies on the cause of death after CA [12, 13]. The authors do not report which diagnostic methods were used in the study or how the results of investigations were weighed in treatment decisions. However, they state in the discussion that “because it was not feasible to blind clinicians to the patients’ treatment-group assignments, there is a possibility that bias affected patient care and outcome state” [8].

Another recent study, comparing hemofiltration with or without additional cooling versus control (no hemofiltration, no cooling), presents no information on how prognostication was performed or if decisions on treatment limitations and WLST were made [40].

Finally, a pilot study on the effect of high dosed erythropoietin was presented, but nothing is reported about treatment decisions and prognostication. Moreover, it is not evident from the protocol of the ensuing phase III study (http://​www.​controlled-trials.​com/​mrct/​trial/​706537) what diagnostics are used for prognostication and whether there are rules for WLST.

In an attempt to explore the optimal target temperature management strategy for comatose survivors after cardiac arrest, the TTM-trial was launched in November 2010. In the trial, two strict target temperatures of 33°C and 36°C for 24 hours are compared. Primary outcome is survival until the end of trial [41]. The sample size of 950 adult patients with cardiac arrest of presumed cardiac origin was reached in January 2013, and the trial will be finished in mid 2013 after the stipulated 180-day evaluation. A standardized and transparent protocol for prognostication and WLST was regarded as a key component in the trial design, in order to reduce the risk of self-fulfilling prophecy and consequent bias.

A manual is available and investigators have been instructed concerning principles of prognostication at investigator meetings. The person who performs the prognostication is blinded to the intervention and all patients are regarded as if they were treated with hypothermia to 33°C. As a general principle, all patients in the trial are actively treated until a minimum of 72 hours after the end of the intervention period. Neurological prognostication is performed at this time-point or later for all patients who remain unconscious.

An earlier prognostication is allowed if (1) the patient becomes brain dead, (2) has an early myoclonus status or (3) if there are strong ethical reasons to withdraw intensive care. In the study protocol it is defined that: “the neurological evaluation will be based on a clinical neurological examination (including GCS motor score, pupillary and corneal reflexes), median nerve SSEP and EEG”. In all unconscious patients, a conventional EEG is performed at 12–36 hours after rewarming. SSEP is performed at 48–72 hours after rewarming at centers where this technique is available. In addition, information from neuroimaging (MRI and CT) may be used but cannot constitute the sole reason for withdrawal of intensive care. Biochemical markers for brain damage are not used for prognostication, instead serial blood samples are collected and stored in a central biobank for later analyses. The physician performing prognostication should make one of the following recommendations:

Findings allowing for discontinuation of life support have been specified in the protocol (Table 1). In the case-record-form, findings from all examinations, recommendations and decisions are recorded.

Since we believe that the process of neurological prognostication may have an effect on the results of the study, we decided to blind the assessor for treatment temperature. However, the decision to continue or withdraw intensive care is multi-disciplinary and blinding is difficult to maintain throughout the prognostication process. However, as all recommendations and subsequent decisions are recorded, any systematic difference between the treatment arms will be detectable.

By strictly regulating the time for prognostication and criteria allowing for WLST, we have aimed to avoid false and premature predictions. The majority of patients with a favorable prognosis will wake up during the first three days after cardiac arrest [25] and we decided to postpone prognostication an extra 1.5 days to account for the effects of sedation during cooling and a possible delay of the recovery process by hypothermia. Recently published studies have shown that, although awakening itself may not be delayed by hypothermia [42], effects of sedation could explain why the clinical examination is less reliable [22]. Clinical examinations may still be unreliable at 4.5 days after cardiac arrest and data supporting this time-point is admittedly scarce [21]. Therefore, an extensor or absent motor response to pain at 4.5 days must be combined with 1) absent SSEPs or 2) a treatment refractory status epilepticus to allow for WLST in the TTM-trial. If any of these two conditions is not fulfilled, at least an extra day of observation is demanded.

Many intensivists would consider it unethical to continue intensive care in a patient with early generalized and persistent myoclonus after cardiac arrest, but false predictions may occur both with [35] and without [43] hypothermia. We therefore combined an early myoclonus status with another strong predictor, absent cortical responses on the SSEP, and allowed for SSEP to be performed immediately after rewarming in patients with status myoclonus. Early prognostication is also allowed for the small fraction of patients who become brain dead and these patients should be diagnosed according to national legislation. We recommend, however, that the clinical diagnosis of brain death should be avoided during the first 24 hours after ROSC and be supported by radiological evidence of herniation and loss of intracerebral blood-flow when there is any doubt about the diagnosis [36]. Finally, strong ethical reasons for an early withdrawal of care may include generalized malignant disease or a clearly stated wish not to be resuscitated. Such reasons may become evident only after the patient has arrived in the ICU.

As therapeutic hypothermia for cardiac arrest was introduced at the Skane University Hospital in Lund in 2003, a decision was made to postpone neurological prognostication for all patients who remain in coma until 72 hours after rewarming to account for a delayed recovery process. In a report of this strategy, 52% of hypothermia-treated patients awoke before the stipulated time for prognostication, 17% died early and 31% were still in coma 72 hours after rewarming (4.5 days after CA) [21]. Only 6/34 patients who were in coma at the time of prognostication awoke at some time-point and the rest remained in coma until death. In the majority of the deceased patients, clinical, neurophysiological, biochemical (neuron specific enolase (NSE)) and neuroradiologic findings supported a massive brain injury and this was confirmed by post mortem examinations. However, a sub-group of 8/34 patients with low-range NSE (≤20 μg/l) was identified, of whom 5/8 had normal MRI and 6/8 had normal SSEP. All these 8 patients had a generalized status epilepticus pattern on EEG and a lack of motor or extensor response to pain. Only one patient regained consciousness but was severely neurologically handicapped.