Clinical heterogeneity and phenotyping of post cardiac arrest brain injury: one size may not fit all

Abstract

Post-cardiac arrest brain injury (PCABI) emanates from the injurious pathophysiologic sequelae that take place during and after resuscitation from cardiac arrest. Regrettably, identification of efficacious management strategies to mitigate PCABI has been disappointing with numerous well-conducted randomized control trials yielding neutral results. The reasons for this observation are likely multifactorial, however, increasingly patient and disease-specific heterogeneity is recognized as a crucial factor in clinical decision-making. Traditionally, PCABI has been stratified based upon simple historical characteristics (e.g. location of cardiac arrest, initial rhythm, witnessed vs. unwitnessed) that inadequately reflect in vivo PCABI severity or responses to clinical interventions within individual patients. It is therefore increasingly clear that this approach to PCABI is insufficient. In other syndromes, such as sepsis or acute respiratory distress syndrome, attempts to identify early “phenotypes” of patients reflect growing recognition of considerable between-patient heterogeneity in the disease mechanisms and response to therapeutic interventions. A similar approach should be taken with PCABI. In this review, we described the clinical heterogeneity and phenotypes of PCABI as related to the underlying pathophysiology, selective anatomical vulnerability and electrographic patterns. The overarching aim of the review is the propose a shift to expeditious phenotyping of PCABI severity that focuses on assessing in vivo severity and patterns of injury that could be used for future targeted therapies. We will also discuss potential causes of heterogeneous clinical responses to interventions and highlight future research areas for PCABI that focus on phenotyping and incorporating these considerations into clinical trials.

Take home message

Post-cardiac arrest brain injury is a heterogeneous disease entity that has distinct clinical phenotypes. Efforts should be made to focus on phenotyping post-cardiac arrest brain injury patients to inform clinical trial design and improve outcomes

Introduction

Post-cardiac arrest brain injury (PCABI) arises during and following resuscitation from cardiac arrest [1, 2]. Amongst survivors, it is the most common determinant of clinical outcome [3, 4]. PCABI is associated with high mortality [3, 5, 6] and significant long-term neuropsychiatric sequelae, ranging from states of impaired consciousness to psychiatric illness encompassing depression, post-traumatic stress disorder and anxiety [7–9]. These long-term consequences are associated with a significant increase in healthcare costs and a burden for patients and their families [10–12]. Recently, the incidence of PCABI has been increasing with aging populations worldwide and widespread increases in overdose-related cardiac arrest [6, 13].

Small improvements in PCABI outcomes are attributed to on-scene interventions, such as immediate cardiopulmonary resuscitation and defibrillation [14]. Regrettably, outcomes related to interventions in the intensive care setting have not appreciably improved [15, 16], despite improvements in other major critical illnesses, such as sepsis and acute respiratory distress syndrome [17, 18]. This finding is likely attributed to a paucity of efficacious therapies and management strategies for PCABI. Recently, PCABI clinical trials have yielded numerous neutral trial results [19–22], which could be attributed to underpowered studies and overly optimistic effect sizes. However, it is possible that between-patient heterogeneity of PCABI has also been under appreciated.

A central theme in the provision of healthcare in critical care is providing consistent medical interventions to all patients; however, it is increasingly appreciated that personalized approaches should be considered [23]. Consequently, authors have advocated for approaching critical illnesses with an emphasis on the underlying pathophysiology and phenotypes instead of approaching patients with “syndromic” classifications [24]. Traditionally, PCABI has been stratified based upon simple historical variables such as location of cardiac arrest (e.g. in-hospital vs. out-of-hospital), witnessed vs. unwitnessed, or shockable vs. non-shockable. These stratification schemes do not account for the complexity of PCABI and the nuances of between-patient heterogeneity. Moreover, they make assumptions of in vivo PCABI severity, only identify “probability” or “risk” of PCABI severity, disease mechanisms at play and response to clinical interventions. Doing so either selects the wrong or omits the right PCABI patient who may or may not respond to interventions. An alternative approach is needed that focusses on real-time phenotyping of PCABI patients in the early post-resuscitation period.

We provide a narrative review highlighting the heterogeneous nature of PCABI with implications for clinical trials and outcome assessments. Specifically, the aims are: (1) review current approaches to PCABI patient stratification; (2) review the heterogeneity in the physiology of circulatory arrest, (3) review the neuroanatomic, physiologic and electrographic PCABI phenotypes; (4) review the heterogeneity of PCABI patient responses to interventions; and (5) suggest an approach to phenotyping PCABI in the early post-resuscitation period.

Current approaches to stratification

Current approaches to PCABI stratification focus on historical variables of the initial cardiac arrest. Perhaps most common, the stratification of PCABI patients based upon first documented cardiac rhythm is ubiquitous in population-based observational outcome studies [5], translational research [25, 26] and randomized control trials [21, 27]. The delineation of whether the initial cardiac arrest was “shockable” vs. “non-shockable” is intended to differentiate a primary cardiac arrythmia versus a non-cardiac cause that results in a loss of cardiovascular function (Fig. 1). Causes of non-shockable cardiac arrest include hypoxemia, hypoventilation, haemorrhage, electrolyte disturbances, toxic substance ingestion, trauma, obstructive cardiovascular pathophysiology or prolonged shockable rhythms that have degenerated into pulseless electrical activity or asystole [28]. Non-shockable rhythms are associated with worse clinical outcomes and coupled with the heterogeneity in aetiology [29], many clinical trials have excluded these patients. Recently, the incidence of non-shockable PCABI patients has increased, owing to widespread increases in opioid related cardiac arrest [13, 30–32].

Fig. 1.

Physiology of shockable versus non-shockable rhythms of circulatory arrest. A demonstrates the degeneration of a normal sinus rhythm to ventricular fibrillation with subsequent abrupt collapse of innate hemodynamics (B) and perfusion (C). Conversely, D demonstrates progressive slowing of the innate hemodynamics and underlying electrical cardiac rhythm. In this instance, perfusion decreases gradually over a protracted period of time, thereby exposing the brain to prolonged warm ischemia during systemic hypotension

Similarly, unwitnessed vs. witnessed cardiac arrest has been used as a stratification scheme. In general, patients suffering unwitnessed cardiac arrests with non-shockable rhythms (e.g., asystole), have been excluded from clinical trials, given an assumption that such cases would be associated with futile clinical outcomes [20, 22]. The presence of unwitnessed arrest, regardless of the initial rhythm, is an independent predictor of poor functional outcome [33].

Finally, the location of cardiac arrest, stratified by whether the arrest occurred in-hospital (IHCA) or out-of-hospital (OHCA), has been championed [34]. This viewpoint espouses that patients experiencing IHCA are likely to have the clinical outcome affected by the initial medical indication necessitating admission and thereby worse outcomes [35]; however, a recent nationwide registry analysis in Denmark demonstrated similar outcomes between IHCA and OHCA patients [34]. This unexpected result may be partly explained by the cardiac arrest being rapidly recognized in IHCA patients and thereby reducing a prolonged cerebral insult. Further, IHCA patients are typically resuscitated by advanced critical care teams in a timely manner, compared to the OHCA setting where lay responders may be required to initiate resuscitation.

Although the aforementioned PCABI stratification strategies may appear clinically justifiable, these approaches have an inherent weakness. These stratification schemes link historical variables to draw assumptions about the in vivo severity of PCABI pathophysiology within patients. Collectively, these historical stratification schemes may be associated with PCABI severity but individually, each variable explains a small proportion of the variance seen with PCABI heterogeneity. For example, non-shockable or unwitnessed arrests portend a worse prognosis but PCABI mechanisms and severity following a 60-min witnessed VF arrest patient who did not receive layperson CPR is very likely to be more severe than a brief PEA arrest from a rapidly reversible etiology that was unwitnessed by rapidly corrected. It is not that these single stratifying variables have no information, but they individually explain a very small proportion of the total variability in PCABI. A multidimensional approach is needed to capture the necessary information.

Additionally, stratification of PCABI predicated upon on-scene cardiac arrest demographics assumes that the interpretation and recording of these variables is accurately collected in an otherwise intense clinical situation where multiple human factors are simultaneously occurring at once in a time-sensitive fashion. As such, additional strategies are desired to circumvent these weaknesses and provide rapid PCABI patients phenotyping in real time.

Heterogeneity of pathophysiology

Circulatory arrest physiology

PCABI can be approached by compartmentalizing the injurious phases of pathophysiology encompassing global cerebral ischemia during circulatory arrest and followed by reperfusion after return of spontaneous circulation (ROSC) [2]. With regards to circulatory arrest, there is heterogeneity in how the brain is exposed to cerebral ischemia and the associated downstream cellular dysfunction [36]. First, sudden cardiac arrest stemming from a primary cardiac arrythmia is characterized by near instantaneous loss of innate cardiac output and brain perfusion (Fig. 1). The abrupt cessation of cerebral oxygen delivery and neuronal aerobic metabolism leads to cessation of energy ion-dependent channel function and absence of action potentials [3]. Clinically, this manifests as loss of consciousness and isoelectric rhythms on electroencephalography [36, 37]. Interestingly, clinical reports have shown a delay of 4–30 s or longer between the onset of ventricular fibrillation and clinical loss of consciousness, thereby suggesting that the human brain has limited, yet variable ability to sustain brief functioning [36]. Importantly, this clinical finding could be attributable to reduced but present cerebral blood flow immediately following VF.

Conversely, non-shockable causes of cardiac arrest may be characterized by circulatory arrest physiology that degenerates over a protracted period of time (Fig. 1) [36]. For example, prolonged antecedent hypoxemia and/or hypotension leads to a progressive loss of innate cardiovascular function and exposes the vulnerable brain to protracted cerebral ischemia. Intuitively, this may partly explain the observed worse outcomes in PCABI patients, who achieve ROSC after non-shockable rhythms [26, 29]. Clinical reports have demonstrated that brain electrical activity may be lost 4–10 min prior to complete loss of innate cardiac output during circulatory arrest stemming from progressive hypotension/hypoxemia [36], thereby suggesting the brain experiences injury prior to complete loss of innate cardiovascular function [38]. In fact, clinical outcomes are worse in both IHCA and OHCA patients who exhibit progressive deterioration of vital signs prior to the cardiac arrest as opposed to sudden cardiovascular collapse [39, 40].

The clinical outcomes of patients presenting after shockable vs. non-shockable rhythms undoubtedly differ, with PCABI “shockable” patients demonstrating a higher proportion of patients experiencing favourable neurologic outcomes compared with PCABI “non-shockable” patients [28]. Intuitively, shockable rhythms may be associated with lower no-flow durations but also are afforded the clinical efficacy of defibrillation, in contrast to non-shockable rhythms. Given these different outcomes which likely reflect the underlying circulatory arrest physiology, patients with non-shockable versus shockable rhythms are routinely approached as distinct phenotypes [41], however, opportunities exist to identify outliers in both groups based upon within-patient pathophysiology and responses to clinical interventions.

At the cellular level, reperfusion of the neurovascular unit following ROSC is associated with numerous pathophysiologic sequelae that likely occur to varying degrees in a broad PCABI population. Immediately upon reperfusion, calcium infiltration ensues into neuron and glial cells with subsequent inhibition of mitochondrial function (Fig. 2) [3]. Additionally, the formation of reactive oxygen species in the early reperfusion phase can further elicit mitochondrial dysfunction and provide insufficient adenosine triphosphate production to sustain homeostatic cellular functions (Fig. 2) [3]. In the setting of mitochondrial dysfunction, activation of intracellular caspases also results in delayed programmed cell death of neurons and glial cells. Finally, the concept of immune system dysregulation in the early ROSC period has emerged recently as a potentially modifiable target in PCABI patients. The complexities of this topic are centered upon whether the innate immune system’s response is adaptive of maladaptive and clinical trials thus far have been conflicting [42]. For example, the IMICA trial [43] randomized patients placebo versus tocilizumab (interleukin-6 blockade) and did not demonstrate differences in outcomes, whereas the STEROHCA trial [44] found a signal towards improved outcomes in patients receiving intravenous methylprednisolone. It is likely that heterogeneous responses exist between patients in regards to the innate immune system’s response to reperfusion of the neurovascular unit and elucidating this nuance is a clear next step in the research for this field.

Fig. 2.

Cellular Pathophysiology of post-cardiac arrest brain injury. A demonstrates reperfusion of a previously ischemic neurovascular unit. Microglial (yellow) activation with release of pro-inflammatory cytokines predominates with potential consequent pyroptosis of vulnerable neurons. Astrocyte (purple) activation also is seen with ischemic-reperfusion injury and may contribute to porous blood–brain barrier mechanisms. Axonal and neuron degeneration ensues from a culmination of pathophysiologic sequelae. B demonstrates intracellular mechanisms encompassing calcium infiltration and reactive oxygen species-induced mitochondrial dysfunction. Apoptosis can ensue from mitochondrial pathways

Anatomic PCABI phenotypes

Anatomically, the human brain is comprised distinct regions that can be divided into the cerebrum, brainstem and cerebellum. Important differences exist between these structures in regard to cerebrovascular physiology and cellular composition. Within the cerebrum, grey matter is densely populated with neurons and cell bodies of key nuclei. Additionally, the cerebrum also houses the subcortical white matter which is composed of myelinated axons that are responsible for signal transduction. Given the anatomical differences, it unsurprising that the cerebral metabolic rate of oxygen utilization and CBF characteristics vary greatly between grey and white matter. Specifically, the cerebral blood flow requirements of grey matter tissue are much higher (60–80 mL/100 g min) than sub-cortical white matter (20–30 mL/100 g min) to sustain normal metabolism and neuronal homeostasis [45]. As such, grey matter tissue locations are especially susceptible to PCABI following global ischemia—reperfusion injury [45]. Additionally, the deep nuclei of the brain (thalami and basal ganglia) are located at distal watershed locations within the cerebrum and in states of global cerebral ischemia, would be especially sensitive to critical reductions in cerebral blood flow [3].

Tissue regions exhibiting selective vulnerability include the hippocampi, cerebral cortex, thalami and basal ganglia (Fig. 3) [46]. Injury to each of these locations in isolation can expose PCABI survivors to differential long-term outcome phenotypes. Importantly, the brainstem is relatively resistant to ischemic injury in comparison to supra-tentorial structures within the cerebrum, however, the precise mechanisms underpinning this observation in humans are not well delineated [47]. Given the heterogeneity of anatomical injury patterns observed in PCABI, the nuances of such anatomical injury patterns may not be accounted for by dichotomization of neurologic outcome scales into ‘favourable’ versus ‘unfavourable’, and considerable research is needed to clarify this area.

Fig. 3.

Anatomical injury patterns of post-cardiac arrest brain injury. On the left panels, illustrations of the brain in coronal (top) and axial (bottom) planes are exhibited. Anatomical foci of selective vulnerability are shown with the following designations: (A) cerebral cortex—grey matter; (B) sub-cortical white matter; (C) hippocampi; (D) basal ganglia (comprised of the caudate, putamen and globus pallidus); (E) thalamus. On the right side of the figure, axial MRI head images are shown revealing selective injury patterns of PCABI with the corresponding alphabetical letter designations corresponding to their anatomic foci labelled on the illustrations (left)

Physiologic PCABI phenotypes

The majority of PCABI critical care research has targeted the reperfusion phase after ROSC. Undoubtedly, PCABI has a dynamic pathophysiology with phases of injury being characterized by various mechanisms. Historically, the mitigation of early reperfusion injury has been approached by the rapid implementation of therapeutic hypothermia [48]. In recent years, large well-designed clinical trials have called into question the efficacy of therapeutic hypothermia [20, 21]. An alternative approach of augmentation and optimization of cerebral oxygen delivery in the post-ROSC setting has taken focus [49]. To this extent, strategies aimed at increasing cerebral blood flow via increased mean arterial pressure or mild hypercapnia have been explored. Regrettably, the results from clinical trials have not consistently demonstrated improved outcomes [19, 22].

A critical physiologic assumption in critical care that is that increased blood flow to a target organ results in increased diffusion of oxygen into tissue and with intact mitochondrial function, aerobic metabolism ensues with maintenance of cellular homeostasis. This coupling between convective oxygen delivery and normal diffusion into the end organ tissues underpins the physiologic functioning of the oxygen cascade in humans [45]. Importantly, in critical illness, decoupling may occur between convective oxygen delivery and diffusion [45, 50, 51]. The presence of diffusion limitation as a mechanism of brain tissue hypoxia in humans with PCABI has been demonstrated using multimodal neuromonitoring. [51] In this instance, a proportion of PCABI patients exhibited decoupling between convective oxygen delivery and diffusion, whereas other PCABI patients demonstrated intact diffusion [51]. PCABI patients exhibiting intact diffusion would likely benefit from interventions that aim to augment cerebral blood flow, whereas those demonstrating diffusion limitation would not. These physiologic PCABI phenotypes likely represent heterogeneous underlying pathophysiology and clinical implications of these phenotypes may partly explain the neutrality of recent large clinical trials [19, 22].

An additional consideration of PCABI physiologic phenotypes pertains to the extra-cranial severity of illness and organ dysfunction. Global ischemia–reperfusion injury of extra-cranial organs can elicit downstream deleterious sequelae such as myocardial dysfunction with cardiogenic shock, severe hepatic dysfunction, disseminated intra-vascular coagulation, acute lung and kidney injury [52–54]. In these instances, clinical outcome may be determined by the severity of multi-organ failure as opposed to PCABI, specifically. Identification of this particular severe phenotype is in the immediate post-ROSC setting may be challenging and result in inclusion of patients into trials that are testing ‘neuroprotective’ interventions. Although a randomized design will likely balance these phenotypes between treatment groups, their inclusion may still result in underpowered trials.

Future work will require methods to detect their presence in PCABI patients using generalizable assessment techniques to inform clinical trial enrolment and results. To do so, timely identification of these phenotypes is crucial to enable clinicians to alter management strategies. As of yet, it is unknown whether these pathophysiologic phenotypic changes are present early after ROSC, thereby making it a requisite line of inquiry for cardiac arrest research. A clear next step would be to link the downstream anatomic and electrophysiologic phenotypes to differences in the physiology of cardiac arrest (i.e. shockable versus non-shockable). This would enable clinicians and trialists to make informed decisions on the expected natural history of PCABI within a patient in the early hours following ROSC.

Electrographic PCABI phenotypes

There has been increasing interest in using electroencephalography (EEG) to assess PCABI. The emergence of increased access to EEG services and automated software to provide detection of abnormal rhythms and seizures has spurred increased research into the electrographic patterns of PCABI. During the circulatory arrest phase, EEG patterns are characterized by a progressive slowing of EEG rhythms that culminate in isoelectric patterns once the cortical infarction threshold is surpassed by reductions of cerebral blood flow (< 10 mL/100 g min) [55]. Upon the reperfusion phase following successful resuscitation, release of excitotoxic neurotransmitters (e.g. glutamate) and internalization of inhibitory neurotransmitter receptors (e.g. GABA) can predispose the vulnerable tissue to exhibit abnormal EEG rhythms that are consistent with seizures or non-convulsive status epilepticus (NCSE) [56].

During this excitotoxicity phase of PCABI pathophysiology, the detection and management of NCSE has garnered significant interest [57]. With regards to diagnosis, clinical manifestations of seizures may be obscured by concomitant sedative administration, thereby necessitating EEG monitoring in PCABI patients. The precise duration of EEG monitoring to identify NCSE is unclear, however, studies suggest increased sensitivity of diagnosing NCSE with prolonged EEG recording > 8 h in duration [57]. The specific EEG montage to apply is also a point of debate with full montage monitoring exhibiting improved sensitivity of seizure detection but also often requiring neurophysiologic expertise in its interpretation [58]. Automated software devices have been evaluated and appear to be promising in serving as surrogates for NCSE evaluation [59].

Post-anoxic myoclonus is a particular electrographic phenotype that has drawn attention in regard to PCABI prognostication, but also management considerations. Historically, myoclonic status epilepticus was regarded as a definitive sign of poor neurological prognosis however, in recent years it has been established that the false positive rate of ~ 5% in predicting adverse clinical outcome [60–62]. As such, multiple guidelines now suggest a trial of pharmacologic management with the aim of terminating myoclonic statis epilepticus prior to prognostication [60].

Similarly, generalized periodic discharges (GPDs) have been theorized to be a therapeutic target in PCABI patients. In the TELSTAR trial, PCABI patients exhibiting GPDs were randomized to pharmacologic therapy aimed at suppressing GPDs versus standard of care [63]. Interestingly, there were no differences in the clinical outcomes of PCABI patients in either trial group, thereby raising the possibility that GPDs are a sign of injury severity and interventions aimed at treating this EEG phenotype do not modify that outcome in PCABI patients [63]. Although associations between malignant EEG patterns and clinical outcome have been shown, a key research gap pertains to whether these malignant patterns are, in fact, injurious or a clinical sign of underlying PCABI severity. Such a fundamental research question must be addressed prior to undertaking clinical trials aimed evaluating the efficacy of anti-epileptic therapies in PCABI.

Recently, an alternate use of EEG to identify PCABI patients who demonstrated reactive backgrounds and exhibit favourable neurological recovery has emerged [64]. In this instance, use of EEG within the first 24–48 h following ROSC could identify patients whom have not suffered severe PCABI and stratify patient groups in clinical trials.

PCABI heterogeneity with clinical interventions

In recent years, numerous clinical trials have demonstrated the lack of efficacy for clinical interventions in a broad population of patients with PCABI despite substantial preclinical data demonstrating benefit. A central question when interpreting the results of a clinical trial is “whether or not the study population in the trial is representative of the patient in front of us”. This dogma is not truer than in the field of resuscitation sciences where well-designed trials have question or refuted widespread use of clinical interventions that had historically been used to mitigate PCABI [20, 21].

Specifically, the use of therapeutic hypothermia gained significant traction following the publication of the HACA trial and Bernard et al. trial [48, 65]. Clinical efficacy was established for therapeutic hypothermia and became widespread in resuscitations sciences. However, in 2013, the TTM trial demonstrated equivocal results in regards to mortality for PCABI patients undergoing therapeutic hypothermia versus normothermia (36 °C) [20]. These results were further corroborated in 2021 by the TTM-2 trial, which demonstrated similar results between therapeutic hypothermia and normothermia with early fever management (> 37.7 °C) [21]. Notwithstanding the rigorous methodology and outcome assessments on both trials, important considerations are to be noted. First, the mortality rate of trial participants approached ~ 50%, which is lower than the typical non-trial PCABI population [21]. Second, there were tremendous rates of bystander cardiopulmonary resuscitation [20, 21] compared with population-based analyses [66, 67] but this discrepancy is not unique to the TTM trials [22]. Clinically, it is important to note that the severity of PCABI seen in clinical practices is likely worse than that reflected in current large-scale trials [20, 21]. Finally, both trials included patients with a “likely cardiac” cause and patients presenting after unwitnessed out-of-hospital-cardiac-arrest (e.g. asystole as the initial rhythm) were excluded [20, 21]. Such important demographics may explain the surprisingly low mortality rates. Collectively, these trial demographic data suggest a relatively less severe PCABI cohort and one that is not entirely generalizable to all jurisdictions. Conversely, the HYPERION trial, which enrolled PCABI patients following non-shockable rhythms demonstrated efficacy of therapeutic hypothermia in this PCABI population, however, the fragility index of 1 suggests confirmatory studies are required prior to widespread implementation of the trial results [27]. Interestingly, a greater effect was seen in the IHCA patients (absolute risk reduction ~ 10%) compared to the OHCA patients (absolute risk reduction ~ 2%) enrolled in the trial [27]. Upcoming large clinical trials such as STEPCARE, ICECAP and PRINCESS-II plan to include broad and heterogeneous PCABI populations which will undoubtedly illuminate the importance of these considerations for the use of therapeutic hypothermia or fever treatment.

Alternatively, strategies aimed at augmenting and optimizing cerebral oxygen delivery have been championed for PCABI. Chiefly among these interventions is arterial pressure augmentation. Three clinical trials (NEUROPROTECT, COMACARE, BOX) cumulatively randomized out-of-hospital-cardiac-arrest patients presenting following shockable rhythms to ‘normal’ (> 65 mmHg) versus ‘augmented’ (85–100 mmHg in NEUROPROTECT & COMACARE; > 77 mmhg in BOX) mean arterial pressure [19, 68–70]. To date, there is no convincing evidence of consistent clinical efficacy for such strategies; however, it is important to note that these current trials represent a relatively small proportion of overall PCABI patients and did not include those with demographics suggesting increased PCABI severity. Further, sub-analyses suggest that there may be extracranial physiologic benefits to augmented strategies including decreased serum troponin levels and increased urine output with augmented mean arterial pressure [71]. Future studies are required to clarify whether augmented mean arterial pressure strategies are efficacious in a broad PCABI population and whether sub-groups of ‘responders’ may exist.

Considerations for clinical trials

Traditional clinical trials aim to estimate the average causal effect of one or more randomly assigned interventions on outcomes. Using randomization as an instrument, randomized controlled trials (RCTs) seek to balance across treatment arms both measurable and unmeasurable confounders that are present at the time of randomization [72]. Between-patient heterogeneity threatens the usefulness of traditional RCTs in several ways that are particularly relevant to patients resuscitated from cardiac arrest [73].

By design, RCTs are powered to detect a minimal “clinically relevant” treatment effect. Consider an intervention for which a 5% reduction in mortality and functional dependence measured 180 days after randomization is considered relevant. Imagine also that some patients who are enrolled in the trial are, at the time of enrollment, either too well or too ill to benefit from the trial intervention. In other words, baseline clinical characteristics are deterministic of these patients’ outcomes. In an RCT, these non-responders are evenly allocated across treatment arms, diluting the observed population-level average treatment effect. Thus, even if most enrolled subjects are more likely to benefit than the effect size for which the trial is powered, the trial may still be neutral. Equally as important, undertaking blanket interventions in all PCABI patients risks exposing individual patients to potentially harmful therapeutic maneuvers without biologically plausible benefit. For example, mean arterial pressure augmentation with vasoactive agents in a PCABI patient exhibiting diffusion limitation would be unlikely to improve brain tissue oxygenation and expose the patient to harmful effects of vasopressors (e.g. mesenteric or digital ischemia) or inotropes (e.g. arrythmias). This example of ‘personalized resuscitation’ by using real time phenotyping approach could mitigate risks to patients by avoiding therapies that may not have a biological rationale within the individual.

RCTs after cardiac arrest have generally aimed to be pragmatic, excluding patients unlikely to respond to treatment based on simple combinations of historical features (e.g., unwitnessed asystole), exam findings (e.g., able to follow verbal commands) or high likelihood of death within 24–48 h (e.g. severe refractory multi-organ failure following ROSC) [20–22]. The extent to which these simple measures adequately identify treatment non-responders is uncertain. For example, in a large cohort of patients who underwent brain CT-scan after resuscitation from cardiac arrest, one in five had severe early cerebral edema interpreted to reflect irrecoverable primary hypoxic-ischemic brain injury, and ≥ 50% of these had a witnessed collapse and/or presenting rhythm other than asystole [20].

Less deterministic manifestations of between-patient heterogeneity may also undermine the usefulness of population-level treatment estimates. Growing data suggest that different subgroups of treatment-responsive post-arrest patients may benefit from different therapeutic strategies. For example, those with intermediate to severe illness severity (quantified by multimodal assessment including neuroimaging, clinical exam and electroencephalography) may derive benefit from hypothermic temperature control to 33 °C while those with mild illness severity may not require this treatment or benefit from a different target [74–76]. When portions of the trial population each benefit from a different treatment arm, the overall trial may be neutral. An intriguing trial (STEPCARE, NCT05564754) will examine the effect of sedation minimization after ROSC, thereby enabling unconfounded clinical examination as a clinical indicator of determining PCABI severity. Insights garnered from this trial could help provide insights into stratifying PCABI severity in the early ROSC setting for clinicians and researchers, alike.

To account for this, RCTs have tested for interactions between treatment and patient subgroups defined by simple historical (e.g., age, sex) or arrest characteristics (e.g., presenting rhythm, arrest duration). Again, whether these simple measures are sufficient to identify individual patients’ optimal treatment targets is uncertain. Indeed, in contrast to the fundamental premise of an RCT, more important to the clinician delivering clinical care is to estimate the potential benefit of an intervention to the individual patient being treated. This is a quantity that traditional RCTs cannot capture.

Several strategies can overcome limitations of historical trial designs in the face of between-patient heterogeneity. A variety of tools based on initial post-arrest exam, blood-based biomarkers of illness severity, neuro-diagnostics, or imaging may capture more nuanced clinical information that can inform predictive or prognostic enrichment strategies to target the right intervention to the right patients. These rich clinical data sources can also be combined to create more complex data-driven clinical phenotypes [77]. Moreover, continuous or repeated clinical monitoring of physiologic processes proximate to the therapeutic target could be used to titrate treatment based on an individual patients’ needs over time. For example, rather than comparing two predefined blood pressure targets after cardiac arrest, a trial might serially assess an individual patient’s cerebrovascular autoregulation and compare an individualized hemodynamic target to usual care. Notably, there are pragmatic challenges with such an approach including which monitoring devices to use, how quickly to implement them at the bedside and an approach to conducting standardized stepwise modulation of MAP to ascertain the ‘optimal’ patient-specific threshold.

Contemporary PCABI phenotyping approaches

Incorporating patient-specific heterogeneity considerations underscores the importance of research directed at phenotyping PCABI patients in an expeditious manner. Focus on this area also holds promise in moving towards personalized approaches to post-ROSC management strategies. Using a similar multimodal strategy that is employed for PCABI prognostication, phenotyping PCABI patients could also be conducted with analogous approaches. Such an approach to PCABI phenotyping may be compartmentalized into the following assessment domains: (a) clinical examination; (b) neuromonitoring; (c) point-of-care blood-based neurological biomarkers; and (d) neuroimaging (Table 1).

Table 1.

Clinical methods of PCABI phenotyping

Method	Variable	Resolution	Advantages	Disadvantages	Risks
Physiologic
TCD	MCAFV	Focal	Non-invasive	Expertise and window acquisition	None
SjvO2	SjvO2	Global	Generalizable	Unilateral	Infection and vascular injury
Invasive MMM	ICP, PbtO2, microdialysis,CBF	Focal	Direct parenchymal monitoring	Generalizability, Expertise Cost	Hemorrhage and infection
NIRS	rSO2	Focal	Generalizability	Accuracy, scalp contamination Affected by melanin	None
Pupillometry	NPi	Focal	Generalizable and reproducible	None	None
Neuroimaging
CT-scan	Cerebrum	Global	Accessible	Limited sensitivity	Radiation
MRI	Cerebrum and Brainstem	Global	High specificity and sensitivity	Access	Lengthy time and patient transfer
CTA/CTP	Vasculature	Global	Global cerebral hemodynamics	Accessibility, Limited PCABI data	Contrast
Electrophysiology
EEG	Neural activity	Global	Global	Expertise	None
SSEP	Neural Function	Focal	Cerebrum and brainstem	Expertise	None
Biomarkers
NSE	Cell body	Global	Highly studied and clinical assays	Extracranial source (red cells)	None
Nf-L	Axonal	Global	Brain-specific	Clinical assay availability	None
GFAP	Astrocyte	Global	Point of care and Brain-specific	None	None
Tau	Axonal	Global	Axonal marker	Extracranial source (muscle)	None
UCHL-1	Cell body	Global	Point of care and brain-specific	Short half-life	None

TCD transcranial Doppler, SjvO₂ jugular venous bulb oximetry, MMM multimodal monitoring, NIRS near infrared spectroscopy, MCA_FV middle cerebral artery blood flow velocity, ICP intracranial pressure, PbtO₂ brain tissue oxygen tension, CBF cerebral blood flow, rSO₂ regional saturation of oxygen, NPi pupillary index, CT computed tomography, MRI magnetic resonance imaging, EEG electroencephalography, SSEP somatosensory evoked potentials, NSE neuron specific enolase, Nf-L neurofilament-light, GFAP glial fibrillary acidic protein, UCHL-1 ubiquitin carboxyl hydrolase L1, CTA computed tomography angiography, CTP computed tomography perfusion

Historically, prompt clinical examination of PCABI patients has been confounded by the administration of sedative medications to facilitate timely therapeutic hypothermia [78]. Therefore, studies of the comprehensive assessment of immediate post-ROSC clinical examination have not been widely conducted. As an indicator of cerebral function, the clinical examination holds tremendous promise to help select patients who truly have experienced a significant PCABI burden versus those who have not. With the publication of TTM1 [20] and TTM2 [21], immediate sedative administration for hypothermia appears to be waning [79]. As such, the concept of a ‘sedation hold’ to assess the neurological examination could be possible in the immediate post-ROSC setting.

Neuromonitoring devices have long been championed as point-of-care tools to assess the severity of brain injury in vivo in humans. Invasive neuromonitoring remains the gold standard in monitoring and evaluating the pathophysiology of acute brain injuries. In the setting of PCABI, monitoring techniques such as brain tissue oxygen tension and intracranial pressure monitoring have recently been used in descriptive analyses to shed light on the heterogeneity within patients in regards to intracranial compliance [80], cerebrovascular reactivity [51] and oxygen diffusion from the microcirculation to the parenchyma. Further efforts utilizing invasive neuromonitoring should incorporate metabolic assessment with microdialysis but also functional monitoring using depth electroencephalography. It is imperative to note that invasive neuromonitoring has risks and limited generalizability. Thus, invasive monitoring studies should seek to simultaneously draw agreements between gold standard invasive monitors and generalizable non-invasive methods to enhance widespread clinical applicability of study findings. While invasive neuromonitoring is the gold standard for achieving these purposes [81], numerous drawbacks prohibit its widespread use as a generalizable tool to facilitate phenotyping (Table 1). In particular, procedural risks to the patient, neurosurgical expertise, delay to placement and cost are obstacles to using invasive neuromonitoring [45]. Furthermore, a significant portion of PCABI patients receive concomitant anticoagulants or anti-platelet medications for acute coronary syndrome management, thereby presenting contraindications to invasive neuromonitoring placement. Thus, non-invasive monitoring devices are desirable. Transcranial Doppler has emerged, especially with automated devices that no longer require personnel with technical expertise [82]. Additional options include electroencephalography for the detection of electrographic phenotypes. Future work should focus on drawing electrographic correlates that link to other domains of PCABI pathophysiology, including cerebrovascular physiology. Near infrared spectroscopy has demonstrated utility for identifying the immediate period of ROSC and there may be utility in identifying patients at risk of downstream physiologic sequelae [83]. Monitoring studies in the delayed ROSC period demonstrate that near-infrared spectroscopy is likely not an accurate reflection of the cerebrovascular physiology in PCABI patients [84, 85].

An additional assessment tool that has emerged in the PCABI field is blood-based brain biomarkers (Fig. 4). Traditionally, markers reflecting cell body injury (neuron-specific enolase, ubiquitin carboxyl hydrolase L1), astroglial (glial fibrillary acidic protein) and axonal injury (neurofilament-light, tau) have been used for PCABI prognostication [86]. However, a shift towards the use of biomarkers as measures of the severity of PCABI injury has emerged. Hoiland et al. demonstrated de novo release of the aforementioned biomarkers from the brain in PCABI patients who exhibited concomitant brain tissue hypoxia [87]. Preliminary data suggested these biomarkers have increased sensitivity and specificity in detecting in vivo brain tissue hypoxia as compared to traditional blood-based investigations conducted in PCABI patients such as serum lactate, liver enzymes and creatinine [87]. The advent of point-of-care biomarker analytical platforms raises the possibility that biomarkers could be used in the future to detect in vivo insights into patient-specific pathophysiology. [88]

Fig. 4.

The cellular origin of blood-based biomarkers of injury to the neurovascular unit. On the left panel, the cellular structure and components of the neurovascular unit are shown comprising of astrocytes (purple) encasing cerebral arterioles, degenerating neuron cell body and axon (pink) and a reactive microglial cell (yellow). The origins of biomarkers are shown in proximity to the cells from which they are released: NSE neuron specific enolase (neuron cell body), Nf-L neurofilament-light (axonal injury), GFAP glial fibrillary acidic protein (astrocytes), UCHL-1 ubiquitin carboxyl hydrolase L1 (neuron cell body), Tau axonal injury. On the right panel, a magnified depiction of the cerebral microvasculature is shown with a porous blood–brain barrier with disruption of astrocyte end foot processes. Leakage of the aforementioned biomarkers into the bloodstream and circulation is demonstrated

Neuroimaging is a core evaluation technique for neurologically injured patients. An advantage of a spatial technique is that it can provide detailed evaluation of the whole brain with insights into anatomical locations that may be selectively injured. Standard head computed tomography is widely conducted and relatively ubiquitous in healthcare settings. Advances using artificial intelligence models have provided preliminary data that could identify PCABI patients with subtle or early grey-white differentiation loss that indicates the severity of underlying PCABI [89]. Additional CT-scan techniques can provide cerebrovascular physiologic data pertaining to semi-quantitative cerebral blood flow, regional cerebral hemodynamic differences and identification of penumbral tissue [90]. With historical concerns of contrast-induced nephropathy waning, CT perfusion scanning could provide invaluable insights into the cerebral hemodynamic profile in PCABI in the early ROSC period. Head magnetic resonance imaging (MRI) has long been considered the gold standard of neuroimaging. Portable MRI techniques have emerged that do not necessitate PCABI patients to be removed from controlled settings [91]. Feasibility data of image acquisition has been shown in humans with PCABI for prognostication purposes [91]. A clear next research step would be conducted these examinations early after ROSC with serial evaluations to describe the natural history of PCABI MRI findings.

Phenotyping PCABI

A key consideration in regards to PCABI phenotyping is time. PCABI represents a dynamic clinical entity in which the adage “time is brain” is imperative for the interventions that can be applied. The further from ROSC that interventions are implemented, the diminishing clinical efficacy is apparent [20, 21]. Examples of neutral clinical trials in which delayed interventions have not shown improved outcomes include mean arterial pressure augmentation [19], mild hypercapnia [22], and therapeutic hypothermia [20, 21]. In each case, the target intervention was applied > 4–6 h following ROSC; a timeline during a significant proportion of the reperfusion injury of PCABI has occurred [92].

Traditionally, the use of multimodal neurological assessment for PCABI patients has been applied for neurological prognostication at the 72-h mark or later following ROSC [60]. However, implementation of such a strategy of comprehensive anatomic and physiologic neurological evaluation has not been standard in the immediate post-ROSC period for phenotyping purposes. For decades, the implementation of expeditious therapeutic hypothermia had taken precedence in the immediate post-ROSC period, thereby negating the possibility of phenotyping the severity and in vivo within-patient heterogeneity PCABI. With the publication of the landmark TTM1 [20] and TTM2 [21] trials, the importance of expeditious therapeutic hypothermia in all PCABI patients may not be necessary, thereby providing a window of opportunity to conduct strategies to phenotype PCABI in the immediate post-ROSC setting. A multimodal approach to phenotyping could be achieved by integration of generalizable assessment tools including bedside neuromonitoring devices, neuroimaging, electrophysiology and blood-based brain biomarker evaluation (Fig. 5). Integration of clinical scoring systems that stratify PCABI patients severity should be incorporated into phenotyping assessments. Scoring systems such as the Pittsburgh Cardiac Arrest Category [93] and Cardiac Arrest Survival Post-Resuscitation In-hospital [94] score present clinically relevant risk-stratification schemes for clinicians and trialists to incorporate into clinical practice and trial design, respectively.

Fig. 5.

An approach phenotyping post-cardiac arrest patients to facilitate individualized management decisions. A suggested approach of multimodal assessment of PCABI patients is shown as a model of evaluating PCABI disease pathophysiology to facilitate rapid phenotyping of patients. This approach would utilize expeditious implementation of neuroimaging (CT ± MRI), incorporating electroencephalography, blood-based biomarker measurements and bedside neuromonitoring. Doing so may lead to important clinical phenotyping that could be utilized for future clinical trial enrollment, design and outcome assessments

Lessons can be learned from ischemic stroke studies in which the timeliness of neuroimaging and phenotyping take precedence [95]. Recently, doing so has resulted in improved outcomes by enabling nuanced approaches to reperfusion strategies in individual patients. A similar phenotyping approach, which prioritizes timely comprehensive assessment of PCABI severity and within-patient pathophysiologic evaluation should be sought in the early post-ROSC period (Fig. 5). The precise phenotyping methods, order of operationalization and personalized interventions stemming from PCABI phenotyping strategies remain to be determined, however, such a dogmatic shift holds promise. For example, emergency room prioritization upon expeditious neuroimaging (CT head with possible perfusion imaging) coupled with a sedation hold facilitate a baseline neurological exam could provide clinicians with immediate datapoints to risk-stratify PCABI severity (Fig. 6). Subsequent analysis of blood-based neurologic biomarkers and implementation of non-invasive and generalizable neuromonitoring (e.g. EEG or TCD) could provide further injury severity, cerebral hemodynamic and functional assessments. The challenge with this approach is undertaking a comprehensive evaluation of PCABI severity with emphasis upon generalizable methods (e.g. non-contrast CT, sedation hold for neurological examination and point-of-care biomarker assessment) to facilitate widespread implementation strategies in all healthcare settings (Fig. 6). Doing so will enable similar approaches across healthcare settings and enhance ethically driven principles of justice and equality to a broad PCABI population. Research focus should be placed on studies that employ expeditious PCABI phenotyping using devices that can be conducted at the bedside with high generalizability (Fig. 6).

Fig. 6.

Phenotyping pathways for post-cardiac arrest brain injury. A suggested pathway for rapid phenotyping could consist of enabling an immediate unconfounded clinical examination with a focus on brainstem evaluation. Thereafter, expeditious neuroimaging with non-contrast CT ± perfusion sequencing followed by point-of-care biomarker assessment and application of generalizable and non-invasive neuromonitoring to evaluate both brain hemodynamics and function

Conclusions

PCABI represents a complex entity that is associated with adverse outcomes and without widescale efficacious therapies in the intensive care setting. There are significant heterogeneity considerations for PCABI with regard to its pathophysiology and possibly, responses to interventions. Future research should focus on establishing PCABI phenotypes and incorporating them into clinical trials to examine the effects of interventions on ‘responders’ versus ‘non-responders’. Such strategies may help personalize the post-resuscitation care.

Data availability

There are no data presented because it is a narrative review.

Declaration

Conflicts of interest

CR is Deputy Editor for Intensive Care Medicine. She has not taken part in the review or selection process of this article. All other authors have no conflicts of interest.