By Alan Z. Segal, MD
Associate Professor of Clinical Neurology, Weill Cornell Medical College
Used in conjunction with the clinical exam and brain imaging, serum biomarkers, such as neurofilament light, can help refine the prognostication for patients who have experienced severe anoxic/ischemic brain injury after cardiac arrest.
Hoiland RL, Rikhraj KJK, Thiara S, et al. Neurologic prognostication after cardiac arrest using brain biomarkers: A systematic review and meta-analysis. JAMA Neurol 2022;79:390-398.
The clinical algorithm for the prediction of outcome in hypoxic-ischemic coma as a result of cardiac arrest (CA) was established with the landmark paper by David Levy, John Caronna, and Fred Plum.1 This study, a prospective series of 210 patients, established that certain examination findings, such as a lack of pupillary responses at initial evaluation, are uniformly associated with a failure to regain meaningful neurological function. On the other hand, evaluation at day 3 showing spontaneous eye movements (roving conjugate or better) or findings of motor responses (such as withdrawal to noxious stimulus) was associated with an approximately 40% chance of return to independence. Mortality in Levy’s cohort was high, with 64% dead at one week. Among 33 patients who remained comatose at one week, only a single patient recovered. A significant portion of the others remained in a persistent vegetative state, a term coined by Plum in 1972.
In the approximately 40 years since the Levy paper, major advances have occurred in the understanding of hypoxic ischemic brain injury (HIBI). Advanced neuroimaging using magnetic resonance imaging (MRI) with diffusion-weighted imaging (DWI) has allowed the rapid detection of ischemia (cytotoxic edema) suggesting foci of irreversible injury. Lesions on fluid attenuated inversion recovery sequences involving the cortical ribbon also may be associated with ischemic injury. However, despite the clarity and specificity of these findings, MRI has not become the mainstay of prognostication in CA. Although DWI abnormalities certainly suggest a less favorable prognosis, the sensitivity of MRI for poor outcomes is suboptimal, since many patients who never wake up have minimal changes on MRI scans.
Another major advance in the care of CA patients has been therapeutic hypothermia (TH). Following the positive findings of two major New England Journal of Medicine papers published in 2002, moderate TH (32°C to 34°C) was established as standard of care. However, these studies had limitations since they focused primarily on out-of-hospital cardiac arrest (OHCA) in patients with ventricular fibrillation that could rapidly achieve return of spontaneous circulation (ROSC). When generalized to other patients having long resuscitation times, such as inpatients with multiple comorbidities experiencing pulseless electrical activity (e.g., pulmonary embolism), TH has shown less profound benefits. In fact, the most recent available data suggest that strict maintenance of normothermia may be equally as effective as lowering body temperature.
The “hypothermia era,” which has evolved over 20 years, created a paradigm shift in the process of clinical prognostication, especially because patients are sedated during the initial 24-hour cooling period as well as during passive rewarming. Surrogates of the clinical examination, such as video electroencephalogram (EEG) monitoring, which did not exist in 1972, when pen-and-paper EEGs were run manually, has served as a surrogate for clinical examination. Video EEG also has allowed the increasing recognition of non-convulsive seizure activity, which itself has been found to be a poor prognostic indicator.
Prognostic serum biomarkers, neuron-specific enolase (NSE) in particular, have been in use for many years, but blood test results often are not rapidly available for real-time clinical decision making and often are without clear threshold values. An NSE level of > 33 µg/L has been deemed indicative of a poor prognosis, but NSE levels rarely are so high. NSE results often fall in the 27 µg/L to 32 µg/L range (highly suggestive of severe HIBI) or are non-diagnostic (< 27 µg/L).
In the present study, the investigators reviewed the existing literature, exploring biomarkers that generally are less well recognized than NSE. These include biomarkers reflecting astrocyte injury — glial fibrillary acidic protein (GFAP) and serum 100 calcium-binding protein β (S100β), as well as biomarkers of axonal injury — neurofilament light (Nf-L) and tau. Also included were studies of ubiquitin carboxyl hydrolase L1 (UCH-L1), which, like NSE, is a marker of neuronal cell damage. In the setting of ischemia and reperfusion, any of these substances can be released into the bloodstream through a damaged blood-brain barrier.
The meta-analysis initially screened 2,953 studies. Among these, 1,652 were found to be duplicated and another 1,254 were excluded after title and abstract review. There remained 398 studies, which were thoroughly evaluated, yielding 86 studies of 10,567 patients (73% men) that were entered into the analysis. There were 44 studies that included only patients with OHCA, one with only inpatients, and 31 that included both. There were 10 studies that did not specify.
Biomarkers were analyzed at variable time intervals after CA, with the primary outcome determined by values 48 hours after ROSC. Even though NSE has been the most widely accepted biomarker, analysis showed that Nf-L had the highest predictive value. The authors noted that, because of heterogeneity in the data, with studies using disparate thresholds for abnormal biomarker values, a “two-stage random effects model” rather than typical bivariate analysis was used. A receiver operating curve (ROC) was generated for each marker, and the area under the ROC curve (AUC) was calculated. An AUC value close to 1 indicates a very high level of specificity, with a minimum false-positive rate.
For Nf-L, the AUC was 0.92. Using the highest threshold for an abnormality, the specificity for Nf-L in predicting a poor neurological outcome was 98%. Subgroup analysis of patients treated with TH as well as specific diagnoses (such as OHCA) showed the same magnitude of results — 0.92 for TH and 0.93 for OHCA. This consistency was important because OHCA is believed to represent a unique subset of CA patients and TH has been a source of significant uncertainty in HIBI prognostication.
Interestingly, the second highest AUC after Nf-L was associated with another white matter marker (tau) with an AUC of 0.89. The AUC for NSE was lower at 0.84, with a sensitivity as low as 46%. The lowest AUC was found for GFAP at 0.77, which was not surprising, since astrocytes are thought to be most resistant to ischemia.
Although the primary result of the study was outcome at 48 hours, similar results were found when analyzing three-month prognosis. As the authors noted, Nf-L is unique among biomarkers, since it has a long half-life. By contrast, S100β, which degrades within two hours, is only of use within the first 24 hours after CA. Also, S100β, as well as NSE, can have extra-neuronal sources, a limitation that does not apply to Nf-L.
COMMENTARY
A better understanding of predictive biomarkers of prognosis in CA would be a crucial step in advancing care beyond its current state, which remains largely one of clinical judgment. As the authors noted, biomarker thresholds must be set to achieve a high specificity for poor neurological outcomes, since false-positive results (suggesting futility in a patient who could recover) are unacceptable. However, this directly limits their utility, since markers would only be positive in the most extreme HIBI. In these cases, biomarkers are not necessary, since a poor outcome could otherwise be predicted by imaging or, more importantly, by the clinical exam.
A further weakness of biomarkers is that they are most elevated when injury to the brain is generalized. However, HIBI is not a uniform process. The cerebral cortical pyramidal (Betz) cells (layers 3 and 5 of the six-layered neocortex) have high energy demands and are preferentially affected by anoxia. Also, the middle (pyramidal) layer of the three-layered hippocampal archi-cortex, the basal ganglia, and the dentate nuclei of the cerebellum may be uniquely susceptible. Focal damage to these localized areas would be capable of producing poor outcomes, without being widespread enough to raise biomarkers.
In contrast to susceptible cortical areas, regions with lesser oxygen demands (subcortical white matter tracts) are affected when HIBI is most catastrophic. It would be in these cases that white-matter indicators (NF-L and tau) shown in this study would be positive. As the authors noted, involvement of the subcortical white matter increasingly has been recognized as a major anatomical site of HIBI, and when white-matter injury is observed on MRI, it is a strong predictor of an unfavorable neurological outcome.
As was noted by Levy in his seminal 1985 Journal of the American Medical Association paper, outcomes in cardiac arrest are driven by physiology, but also by the “self-
fulfilling prophecy” generated when care is withdrawn from patients deemed to have a poor prognosis. This continues to be a reality. The study authors noted that the withdrawal of life-sustaining therapies remains a significant cause of death in patients with HIBI and may not directly correlate with the severity of the neurological injury when mortality is driven by major comorbidities.
REFERENCE
- Levy DE, Caronna JJ, Singer BH, et al. Predicting outcome from hypoxic-ischemia coma. JAMA 1985;253:1420-1426.
