Perioperative electroencephalography in cardiac surgery with hypothermic circulatory arrest: a narrative review

Abstract OBJECTIVES Cardiac surgery with hypothermic circulatory arrest (HCA) is associated with neurological morbidity of variable severity and electroencephalography (EEG) is a sensitive proxy measure of brain injury. We conducted a narrative review of the literature to evaluate the role of perioperative EEG monitoring in cardiac surgery involving HCA. METHODS Medline, Embase, Central and LILACS databases were searched to identify studies utilizing perioperative EEG during surgery with HCA in all age groups, published since 1985 in any language. We aimed to compare EEG use with no use but due to the lack of comparative studies, we performed a narrative review of its utility. Two or more reviewers independently screened studies for eligibility and extracted data. RESULTS Fourty single-centre studies with a total of 3287 patients undergoing surgery were identified. Most were observational cohort studies (34, 85%) with only 1 directly comparing EEG use with no use. EEG continuity (18, 45%), seizures (15, 38%) and electrocerebral inactivity prior to circulatory arrest (15, 38%) were used to detect, monitor, prevent and prognose neurological injury. Neurological dysfunction was reported in almost all studies and occurred in 0–21% of patients. However, the heterogeneity of reported clinical and EEG outcome measures prevented meta-analysis. CONCLUSIONS EEG is used to detect cortical ischaemia and seizures and predict neurological abnormalities and may guide intraoperative cerebral protection. However, there is a lack of comparative data demonstrating the benefit of perioperative EEG monitoring. Use of a standardized methodology for performing EEG and reporting outcome metrics would facilitate the conduct of high-quality clinical trials.


INTRODUCTION
Hypothermia remains an essential technique to protect the brain during cardiac surgical procedures which require circulatory arrest [1]. As core temperature decreases, cerebral metabolism is reduced, thereby offering a window of neuroprotection, with lower core temperatures targeted for more complex and extensive repairs that require prolonged arrest. Despite the use of hypothermia in conjunction with cerebral perfusion techniques, neuromonitoring and neuroprotective drug regimes, new postoperative neurological deficits can still occur in both adults and children undergoing cardiac surgery, and the reason for this is not completely understood [2,3].
Electroencephalography (EEG) records the summated postsynaptic potentials of neural tissue from electrodes placed on the scalp [4]. This activity is classified as normal or abnormal based on its location, morphology and amplitude relative to the age of the patient. EEG can be used to detect seizures and signs of ischaemia and guide the depth of anaesthesia and hypothermia [5,6]. In spite of its wide availability in clinical centres, perioperative EEG monitoring is not routinely used during cardiac surgery with hypothermic circulatory arrest (HCA). A recent survey on its use during aortic arch surgery suggested that it is only performed routinely in 17% of European centres [7]. Guidelines from the American Society of Neurophysiological Monitoring consider EEG in cardiac surgery with cardiopulmonary bypass (CPB) to be a practice option rather than a standard of care, as there are no standards for patient management or established role in improving outcomes [8]. We therefore conducted a narrative review of studies reporting EEG technique and/or outcomes of perioperative EEG monitoring in children and adults undergoing cardiac surgery with HCA to evaluate whether there was consistency in methodology, synthesize the current evidence base and identify any impact on postoperative outcomes.

MATERIALS AND METHODS
This review was conducted with reference to the Cochrane handbook for reviews of interventions [9] and reported in accordance with the PRISMA statement [10]. All eligibility criteria, search terms and data items were prespecified, and the review was prospectively registered on PROSPERO (CRD42021247700) (https:// www.crd.york.ac.uk/prospero).

Study eligibility
We included patients of all ages undergoing cardiac surgery with HCA. We identified studies assessing perioperative EEG monitoring (processed or unprocessed) and whether the study compared the use of EEG monitoring with no EEG monitoring. Outcomes of interest were neurological status, which encompassed any observer-reported outcome indicating neurological dysfunction and perioperative EEG technique. We included all randomized controlled trials (RCTs), non-randomized trials, prospective and retrospective observational cohorts, case series and cross-sectional studies published in any language since 1985 so that retrieved articles better reflected current perioperative management.
Cardiac surgery was defined as any therapeutic clinical procedure performed on the heart or great vessels and HCA as cessation of the circulation following systemic cooling via CPB. EEG was defined as the subjective or quantitative interpretation of cortical activity recorded from at least 2 electrodes placed on the scalp. Secondary publications, sub-studies or long-term outcomes of previously reported studies were excluded unless the results were specifically related to the utility of EEG monitoring or reported additional neurological outcome measures. Studies published only as a conference abstract, or for which all options to obtain the full text were exhausted, were excluded due to insufficient data.

Search strategy
We searched international primary research databases (Medline, Embase, Central, LILACS) from 1 January 1985 to 13 May 2022 and reference lists of relevant articles, systematic reviews and meta-analyses to identify all eligible studies. The search terms used were comprehensive and adapted for each database, with database-specific filters to identify the population and intervention of interest (see Supplementary material).

Statistical analysis
Continuous data were expressed as median with interquartile range, mean with standard deviation or range. Categorical data were expressed as counts and percentages where relevant. We did not plan to perform a meta-analysis or analyse sensitivity and homogeneity because we did not anticipate retrieving homogenous studies relatively resistant to bias. In the event of no or limited studies comparing EEG monitoring with no EEG monitoring, we planned to undertake a narrative review of studies describing perioperative EEG monitoring.

RESULTS
From 341 unique records, we identified 40 studies with a description of perioperative EEG monitoring that included 3287 patients undergoing cardiac surgery with HCA. Only 1 study directly compared detailed outcomes between groups with perioperative EEG monitoring versus no monitoring. We therefore performed a narrative review of the non-comparative studies, and 1 comparative study, on the utility of perioperative EEG monitoring (Fig. 1).

Outcome measures
The most commonly reported clinical outcome measure was postoperative neurological dysfunction, which occurred in 0-21% of patients. This broad term encompassed neurological signs and symptoms ranging from confusion to paralysis. When measuring clinical outcome, a variety of methods were used, including recognized clinical examinations (Glasgow Coma Scale, neurological examination) and imaging (ultrasound, computed tomography, magnetic resonance imaging). Outcome measures were typically assessed before hospital discharge. Only 6 (15%) studies used scales to assess long-term neuro-developmental outcome, exclusively in children [39,41,43,44,46,47] (as shown in the Supplementary material). The majority of these studies identified that both the presence of seizures and the increasing duration of electrocerebral inactivity (ECI) on EEG were associated with poor neuro-developmental outcomes.
One retrospective study directly compared clinical outcome in patients undergoing perioperative EEG monitoring against a control group [29]. They identified early detection of stroke (75% sensitivity) and accurate prediction of no stroke (97% negative predictive value) using a neuromonitoring protocol that included EEG. A single-centre study reported outcomes of aortic arch surgery using EEG-guided DHCA [16] and compared outcomes to their previously reported cohort [50]. They identified a lower rate of mortality, stroke and reoperation for bleeding in the more recent study and report the only difference between the cohorts was the use of ECI-guided DHCA.
Other infrequently reported EEG parameters included changes in the amplitude or frequency of cortical activity to detect intraoperative ischaemia in 4 (10%) studies [13,19,20,29]. When detected, this resulted in a change in cannulation strategy; the depth of hypothermia prior to circulatory arrest or the CPB/cerebral perfusion pump flow/blood pressure augmentation; and postoperative imaging and catheter-based interventions. A retrospective study of aortic arch procedures identified asymmetries in EEG activity between left and right hemispheres following innominate artery cannulation, reflecting uneven active cooling of the brain [34].
The interpretation of EEG activity in relation to ECI was often defined as cortical activity <2 mV for 2-3 min. Whilst national and international guidelines for EEG interpretation are widely available [8,[51][52][53][54][55][56][57], few explicitly cited these standards. The number of electrodes used to record the EEG varied between 2 and 21, with 15 (38%) utilizing the full international 10-20 scalp positions [51] followed by a limited montage consisting of between 2 and 12 electrodes (13, 33%). These studies frequently cited the 10:20 system for electrode application, but few reported recording parameters such as filter settings (6, 15%) and EEG sampling rate (4, 10%). Seventeen (43%) studies provided limited or no information on the technical standards of EEG recording.

DISCUSSION
Perioperative EEG during cardiac surgery with HCA is used to detect neurological abnormalities, seizures and ischaemia; to guide the depth of anaesthesia and hypothermia before arrest; and to predict clinical outcome. However, we found no prospective studies comparing perioperative EEG monitoring versus no EEG monitoring in adults and children undergoing cardiac surgery.
In the 40 articles included in this review, there were only 3 RCTs, none of which used EEG monitoring as the primary intervention. The remaining prospective and retrospective cohort studies demonstrated heterogeneity in the interventions used and outcome measures reported. Almost all lacked a control group of patients who were not monitored with perioperative EEG, limiting our evaluation of the role of EEG in improving postoperative outcomes. As a result, this review provides limited evidence to support EEG monitoring during HCA surgery, a technique that is used in some centres to guide surgical decision-making.

Seizures
Perioperative seizures (which occurred in up to 21% of patients) and the cumulative burden of seizures are associated with unfavourable neurological outcomes in children [58] and adults [59] following admission to intensive care. In our review, a high proportion of seizures in the perioperative period were only detected by EEG. It would therefore seem appropriate to utilize EEG to detect these events, as recommended by guidelines for continuous EEG monitoring in neonates [52].

Electroencephalography continuity
Monitoring EEG continuity during cooling, HCA and rewarming could indicate whether a relationship exists between the depth/ rate of hypothermia, rate of rewarming and postoperative morbidity. As cooling progresses, periodic complexes intermix with background EEG activity, and the amplitude of cortical activity decreases, becoming separated by ever-increasing periods of relative electrical suppression (i.e. a burst-suppression pattern) until ECI is achieved; the reverse occurs with rewarming [21,22,28,37]. The time and temperature at which these milestones occur are highly variable between patients and, thus, cannot be accurately predicted by other clinical metrics [17]. If the timing and duration of ECI are associated with neurological outcome, this would support the utility of EEG as an intraoperative neuroprotection tool to ensure that ECI has been obtained prior to DHCA.

Electroencephalography utility
In this review, postoperative neurological dysfunction was associated with prolonged time to the return of continuous EEG activity [11,22] and longer periods of ECI in select studies [39]. One study reviewed stroke rates in 364 adults who required aortic arch repair with HCA [29]. Of these, 223 were monitored with evoked potentials (EPs) and EEG. Surgeons were alerted when specific EP/EEG monitoring criteria were breached. Twelve developed early stroke, which was detected using EPs (9/12) and EEG (1/12). Although there were no significant differences in stroke rate between monitored and unmonitored groups, intraoperative detection of stroke in the monitored group led to earlier intervention, which may have limited brain injury. Authors provide a summary of stroke detection and intervention criteria; however, EEGcriteria breaches were infrequent, limiting the EEG monitoring evidence base.
No study reported an association between outcome and the rate and depth of cooling before DHCA. This has been shown to affect synaptic activity in animal studies, with higher rates of cooling causing a progressive decrease in activity and lower rates promoting the preservation of activity and tissue plasticity [60]. Fast rewarming can cause brain injury in both animal [61] and human studies [62]. This may explain some of the variation seen in the time to achieve ECI and return of continuous EEG activity, although many other factors including anaesthetic regime may contribute to this finding.
As moderate hypothermia is being slowly introduced in surgical practice, the role of EEG could move from ensuring ECI prior to circulatory arrest to preserving some degree of continuity in the EEG. In a recent series, EEG monitoring was used during hemiarch replacement with moderate HCA and selective antegrade cerebral perfusion (SACP) [13]. Immediately after circulatory arrest, ECI occurred in 45% of patients, which was indicative of cerebral ischaemia. EEG activity was re-established following SACP in all but 2 cases; in one, asymmetric activity was restored following bilateral ACP, and in the other, CPB was re-established and the depth of hypothermia increased before circulatory arrest. They concluded that intraoperative EEG may have specific value in identifying patients with persistent cerebral ischaemia, even after SACP.

Role of electroencephalography in perioperative monitoring
The EACTS/ESVS 2019 expert consensus document on the management of thoracic aortic and aortic arch disease identified widespread use of perioperative EEG monitoring but a lack of evidence for an incremental benefit [63]. Similarly, a systematic review and meta-analysis on outcomes in children following DHCA with EEG monitoring identified 19 studies published in English since 1990 [64]. They reported similar pooled event rates of clinical seizures (12.9%), EEG seizures (14.9%), neurological abnormalities (29.8%) and EEG abnormalities (17.3%) to our findings in adults and children. They concluded that despite its frequent use EEG remains poorly studied.
To facilitate the synthesis of findings from multiple studies, valid and comparable outcome measures must be reported [65]. We found variation in the utilization, acquisition and recording period of EEG, with inconsistent use of measures to evaluate the same outcome. EEG continuity, seizure and ECI monitoring were the most commonly used metrics but not all mention how EEG patterns and seizures were classified. Similarly, metrics of neurological dysfunction were the most commonly reported clinical outcome measure, but signs and symptoms used to define this were broad, ranging from confusion to paralysis, and measurements were performed at variable time points. This disparity reflects the absence of a standardized method for reporting perioperative EEG and the measurement of clinical outcomes following cardiac surgery.
Long-term assessment of neurological outcomes following surgery with HCA represents the gold standard to detect more subtle yet persistent neurological deficits, but these were infrequently performed, perhaps due to the burden it places on participants and researchers. In addition, those who require surgery with HCA typically have a multitude of pre-existing comorbidities, heterogeneous and complex heart disease and variability in the length of postoperative hospital stay; measures used to define neurological outcomes during early childhood are also dependent on age. These factors make it inherently difficult to attribute postoperative injury to any one factor and a challenge to compare the neonatal, child and adult populations. Other than for the detection of postoperative seizures, there is currently limited evidence supporting EEG metrics, such as disappearance/return of EEG continuity, or the time and duration of ECI in guiding perioperative care. In studies that monitored ECI attainment before DHCA, a consistent definition of ECI, and how long it needs to be established before circulatory arrest is required.

Limitations
The limitations of this review include the lack of RCTs directly comparing the use of EEG monitoring with a control group; our inability to perform a meta-analysis due to the lack of comparable outcome measures; and a risk of reporting bias, although minimized by performing an extensive search across multiple databases and including non-English language articles.

CONCLUSIONS
Perioperative EEG monitoring in cardiac surgery with HCA can detect seizures and neurological abnormalities, identify intraoperative ischaemia and may guide cerebral protection and predict outcome. However, the inconsistent metrics used to record, acquire and interpret EEG, and clinical outcome measures limit the evidence base to inform clinical practice. No prospective studies compared perioperative EEG versus no perioperative EEG monitoring, and thus, there is a lack of direct evidence to demonstrate whether EEG monitoring may have a role in improving clinical outcomes. An assessment of EEG continuity during HCA could provide insight into improving perioperative cerebral protection. A standardized approach to EEG monitoring during HCA and postoperative clinical outcome reporting is required to inform the design of future clinical trials.