Identifying neurological prognostic factors after cardiopulmonary resuscitation (CPR) in patients with cardiac arrest (CA) as early and accurately as possible is urgently needed to determine therapeutic strategies after successful CPR and avoid medical futility. Many investigators have previously attempted to establish them 12.

Epidemiological data on CA are generally accumulated according to the Utstein Templates 345, and retrospective analysis of these data allows, to a certain extent, prognostic prediction of the patients following CA 67. In fact, it is difficult to decide on therapeutic measures for a particular patient based solely on information obtained using the Utstein Templates, due to limitations including: the scope of these templates for registry of cases of CA are limited to individuals with disease categorized as 'cardiac etiology' and 'witnessed arrest; and the specificity of prognostic parameters examined is not satisfactory 1.

Although most of the data specified in the Utstein Templates are collected prior to 'return of spontaneous circulation' (ROSC), data obtained after ROSC also provide valuable information on the prognosis of individual patients. Therefore, the latest version of the Utstein Templates highlighting post-resuscitative care recommends development of a special template for collection and recoding of data in the post-resuscitation phase 8. Neuroimaging data (e.g., brain computed tomography and magnetic resonance imaging), electrophysiological data (e.g., electroencephalography, auditory brainstem response recording, and somatosensory evoked potential recording), and blood or cerebrospinal fluid (CSF) examination data (i.e., levels of proteins specific to the central nervous system) 12 have been tested for clinical usefulness as neurological prognostic predictors by many investigators, although they are all classified as 'supplementary data' in the latest version of the Utstein Templates.

Neuroimaging requires transfer of the patient to a dedicated examination site equipped with special devices, which makes it difficult to apply it to individuals who are receiving high doses of inotropes or circulatory support with an extracorporeal device. Acquisition and analysis of electrophysiological data requires qualified specialists in particular medical fields, making it difficult to perform such examinations in individuals presenting after weekday clinic hours. In contrast, laboratory data are easily obtained with high reproducibility as a part of normal intensive care routine, a feature favorable for clinical application. Considering the extent of insult associated with sample collection, blood may be preferable to CSF.

Serum levels of protein neuron-specific enolase (NSE) and S-100B are considered promising candidates for neurological prognostic predictors in patients with ROSC after CPR, and many investigations on the clinical usefulness of these biochemical markers in predicting neurological outcomes after CPR have been published 1910. In the present study, we performed an extensive literature review to examine the clinical usefulness of NSE and S-100B as post-resuscitation neurological prognostic predictors.

To improve applicability of study results in clinical practice, we considered the following three points when designing the present study: a consistent definition of poor (good) outcome should be used in assessing data from multiple studies; the cut-off values for biochemical markers should be set so that specificity in prediction of poor outcome is 100%; and the time points of blood sampling should be fixed in assessing the time course of change in blood levels of biochemical markers. Although few reviews of application of biological markers to prediction of neurological outcome in CA after CPR published meet the above requirements 110, the present study is the first extensive literature review that does meet them.

A total of 54 papers were retrieved by the initial text search, and 29 of them met the selection criteria. After these 29 papers were reviewed in full text and searched for cross-references, 31 papers were finally selected for the present review. The full-text contents of all 31 papers, which included 26 original articles 1415161718192021222324252627282930313233343536373839 and five review articles 1291040, were reviewed and compared. Twenty-one original articles 141821222324252627282930313233343536373839 investigated NSE, while 14 1415161718192022252628293139 investigated S-100B. Articles by Mussack and colleagues 19 and Hachimi-Idrissi and colleagues 15 were excluded from further review because they reported serum levels of S-100B in patients with CA after CPR but without comparison between different outcome groups. Therefore, we systematically reviewed a total of 24 original articles.

Generally, systematic review articles seemed not to contain any more data or results than original reports. However, inclusion of all previously published papers is one of the main purposes of this study, and therefore all the review articles were subjected to the cross-referencing and those articles were included in this study.

Biochemical markers in blood samples can be expected to serve as prognostic predictors of CA patient outcome after CPR and be more easily applicable to clinical practice than neuroimaging or electrophysiological findings. In the present study, we performed a systematic literature review to examine the clinical usefulness of NSE and S-100B (proteins specific to the central nervous system and potential biochemical markers of brain damage) as post-resuscitative predictors of neurological prognosis.

Grubb and colleagues 14 performed the multiple logistic regression analysis on mortality in these biomarkers and clinical scores (i.e., arrest rhythm, bystander CPR and GCS score). They showed that NSE was an independently significant predictor among them. Pfeifer and colleagues 25 compared the neurological predictive value between these biomarkers and the clinical predictors, such as time of anoxia, GCS score, presence of bystander CPR, and so on. Prohl and colleagues 39 also compared this value using clinical examination score reflected some brain stem reflexes. Both studies showed the odds ratio of these clinical predictors was lower than those of NSE and S-100B. Those results indicate that predictive value of biomarkers was superior to that of clinical predictors. Clinical predictors were often affected by the clinical situations (e.g., using of sedative agents, relying on information of emergency medical service personnel who collected information at a chaotic emergency scene). Therefore, the specificity of biomarkers is higher than that of clinical predictors, even though the clinical predictors can be more easily assessed.

NSE is the neuronal form of the cytoplasmic glycolytic enzyme enolase. It is a dimeric enzyme composed of two γ subunits (γγ isomer) with a total molecular weight of 78 kDa and a biological half-life of 24 hours. It is mainly located in neurons and neuroendocrine cells 4142. The S-100 protein is a calcium-binding protein with a total molecular weight of 21 kDa and a biological half-life of two hours. S-100B, a homodimer composed of two β subunits (ββ form), is secreted from glial cells and Schwann cells 43. Recent studies have suggested that elevated levels of S-100B might cause neuronal apoptosis, suggesting that S-100B may play a role as a cytokine in brain inflammatory responses 4445. Accordingly, in patients with high serum levels of S-100B after CPR, S-100B when present at high levels in the brain is suspected to induce brain cell apoptosis leading to aggravation of post-CA brain injury. S-100B is also considered a putative 'Alarmin' released during an early stage of the inflammatory response 46.

The findings of the present study suggest that, when assayed 'on admission' (i.e., within eight hours after CA), serum levels of S-100B might be more clinically useful than those of NSE in predicting neurological outcomes such as regaining consciousness and returning to independent daily life. Assay of serum S-100B level focuses on the process of aggravation of brain injury, while brain imaging, physical examination, and electrophysiology all focus on the consequences of brain injury. NSE is a protein located in nerve cells and detectable in body fluids as a marker enzyme indicative of nerve cell injury 47. Monitoring for increases in the serum NSE level thus focuses on cell death as a result of post-CA brain injury. Consequently, S-100B serves as a prognostic predictor within 24 hours after CA, and thus at an earlier stage than other factors (including NSE), which focus on the consequences of brain injury and are therefore meaningful as prognostic predictors one to three days after CA (i.e., only after manifestation of brain injury is completed) 122.

Many preceding studies recognized an increase in serum NES level over time in patients with poor outcome after CPR, and also demonstrated a decrease in serum S-100B level over time in those with favorable outcome 14212526282930. In those studies, these changes were ascribed to the difference in biological half-life between these two proteins. However, it should be noted that these changes can also be explained by considering S-100B a cause of post-CA brain injury and NSE, an enzyme released from nerve cells, as a result of brain injury: NSE as a marker of brain cell injury increases over time and accumulates as a result of aggravation of brain injury, while persistence of S-100B as a cause of brain injury at high levels leads to aggravation of disease, a conclusion consistent with the previously reported findings.

Therefore, we emphasize that the time-course change in early phase of serum S-100B and NSE levels would offer a more reliable indication of what is happening in the brain of the CA patient, which might be useful for optimization of therapeutic intervention in future cases of CA. However, there are few investigations on the time-course of these biomarkers. Usui and colleagues 48 reported the detailed time-course of these biomarkers in serum and cerebrospinal fluid every one to two hours within 18 hours after CA in mongrel dog models. Bottiger and colleagues 18 reported the time-course of human serum S-100B every 15 minutes within 1 hour and at 2, 8, and 24 hours after CA. However, this study by Bottiger and colleagues is limited by the number of subjects. Therefore, there is no investigation on the time-course of these biomarkers involving a large number of human subjects. The future investigators should put more weight on the time-course change of the biomarkers, especially the changes in early phase (i.e., within 24 hours following CA).

The present study identified 24 original articles involving investigation of the clinical usefulness of NSE and S-100B as prognostic predictors of CA patients after CPR. The design adopted varied from study to study, making a systematic literature review extremely difficult 1040. The major problems encountered during the review process involved variation or heterogeneity among studies in the following three respects: definitions of outcome to be assessed (e.g., some studies adopted grouping criteria for outcome different from others); the value of specificity corresponding to each cut-off value reported for a particular biochemical marker predictive of a poor prognosis (i.e., not always fixed at 100%); and specification of blood sampling time points (i.e., not always uniform application of a blood sampling schedule to all subjects with sampling intervals specified).

Consistency in the three respects noted above had not been considered in the five review articles previously published on the same subject 1291040, and the present review is the first attempt to consider it.

A recently published article discusses pitfalls in critical care meta-analysis 49, highlighting publication bias and trial-level heterogeneity as pitfalls to be carefully avoided in a systematic review of observational studies such as the present review. To avoid publication bias, we performed an extensive literature search to identify all papers published previously on the clinical usefulness of NSE and S-100B in prediction of prognosis after resuscitation from CA and closely reviewed the full-text contents of all papers thus identified. Trial-level heterogeneities encountered in papers reviewed in the present study include grouping criteria for outcome (definitions of outcome), specification of time points for blood sampling, and assay procedures for individual biochemical markers of interest. Consistency in definitions of outcome and specification of time points for blood sampling appears to be particularly important in the present literature review investigating the clinical usefulness of serum levels of NSE and S-100B as prognostic predictors.

Rapid interdisciplinary treatment and monitoring are required in the early post-resuscitative period, sometimes including percutaneous coronary intervention as well. Consequently, a study design involving uniform blood sampling within a few hours after resuscitation would be difficult to adopt in a multicenter study. Rosen and colleagues 29 studied 66 out-of-hospital CA and collected blood samples at various time points according to their ward routines. Their mean first sample times (± standard error) were 10.5 ± 0.9 hours after CA. In consideration of these results, we think that blood sampling at least once between 4 and 12 hours after resuscitation would be practicable and adoptable. Therefore, a multicenter prospective study involving blood sampling between 4 and 12 hours after resuscitation at a time point specified by interval from onset of CA would be most helpful in investigating the clinical usefulness of S-100B and NSE as early predictors of neurological outcome of CA patients after CPR.

The present review, which included all previous papers identified in a literature search, included a paper published in 1989 38 as the earliest published report. In the past 20 years, however, techniques of assay for both NSE and S-100B have been greatly improved, with concomitant increase in sensitivity of detection 5051. It is therefore difficult, and even inappropriate, to assess the cut-off values reported for serum levels of these biochemical markers during this period using a uniform scale or standard.

Finally, we emphasize that extracellular S100B at μM concentration is harmful to astrocyte and neurons but at nM concentrations is beneficial to those 4552. Thus, at least at the very beginning of brain injury the secretion and release of S100B (and hence elevation of serum S100B levels, if any) might not necessarily be indicative of aggravation of brain injury; it may be indicative of activation of astrocytes and attempt to provide neurons with a trophic factor. Thus, it may be important that the levels of serum S100B and NSE be measured at the very onset of CA and at intervals during the next few hours. An analysis of the time-course of serum S100B and NSE levels would offer a more reliable indication of what is going on in the brain of the patient, which might be useful for optimization of therapeutic intervention in future cases.

The present study shows that the measurements of serum levels of S100B within 24 hours after CA might be clinically more relevant than those of NSE in predicting neurological outcomes.

As noted above, no systematic literature review has been performed including all previously published papers on the clinical usefulness of NSE and S-100B as neurological prognostic predictors. In such circumstances, the findings of the present study should aid future investigators in examining the clinical usefulness of these biochemical markers and determination of cut-off values predictive of poor neurological outcome.

• A consistent definition of poor (good) outcome should be used in assessing data from multiple studies.

• The cut-off values for neurological predictors after CPR should be set so that specificity in prediction of poor outcome is 100%.

• The time points of blood sampling should be fixed in assessing the time course of change in blood levels of NSE and S-100B.

• S-100B assayed on admission may be more useful than NSE assayed concomitantly as an early biochemical predictor of remaining comatose and no-return to independent daily life.

• A multicenter prospective study involving blood sampling after resuscitation at a time point specified by interval from onset of CA would be most helpful in investigating the clinical usefulness of S-100B and NSE as early predictors of neurological outcome.

CA: cardiac arrest; CPC: cerebral performance categories; CPR: cardiopulmonary resuscitation; CSF: cerebrospinal fluid; GCS: Glasgow coma scale; GOS: Glasgow outcome scale; NSE: neuron-specific enolase; ROC: receiver-operating characteristics; ROSC: return of spontaneous circulation.

The authors declare that they have no competing interests.

KS conceived and designed the study. OS and HH made critical revision of the manuscript for important intellectual content. TS, MN, YH, RA, YT, NH, and TS drafted the manuscript. All authors read and approved the final manuscript.

KS: Clinical Fellow, Department of Emergency and Critical Care Medicine, Chiba University Hospital, Board Certified Member of Japanese Society of Intensive Care Medicine.

OS: Professor and Chairman, Department of Emergency and Critical Care Medicine, Chiba University Graduate School of Medicine, Board Certified Member of Japanese Society of Intensive Care Medicine.

HH: Professor Emeritus and Former Chairman, Department of Emergency and Critical Care Medicine, Chiba University Graduate School of Medicine, Immediate Past President of Japanese Society of Intensive Care Medicine