Neuroprognostic accuracy of blood biomarkers for post-cardiac arrest patients: A systematic review and meta-analysis

Neuroprognostic accuracy of blood biomarkers for post-cardiac arrest patients: A systematic review and meta-analysis

Journal Pre-proof Neuroprognostic accuracy of blood biomarkers for post-cardiac arrest patients: A systematic review and meta-analysis Chih-Hung Wang,...

4MB Sizes 0 Downloads 8 Views

Journal Pre-proof Neuroprognostic accuracy of blood biomarkers for post-cardiac arrest patients: A systematic review and meta-analysis Chih-Hung Wang, Wei-Tien Chang, Ke-Ing Su, Chien-Hua Huang, Min-Shan Tsai, Eric Chou, Tsung-Chien Lu, Wen-Jone Chen, Chien-Chang Lee, Shyr-Chyr Chen

PII:

S0300-9572(20)30026-5

DOI:

https://doi.org/10.1016/j.resuscitation.2020.01.006

Reference:

RESUS 8368

To appear in:

Resuscitation

Received Date:

16 September 2019

Revised Date:

26 December 2019

Accepted Date:

9 January 2020

Please cite this article as: { doi: https://doi.org/ This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.

Neuroprognostic Accuracy of Blood Biomarkers for Post-cardiac Arrest Patients: A Systematic Review and Meta-analysis

Chih-Hung Wang, MD, PhD1,2 ; Wei-Tien Chang, MD, PhD1,2; Ke-Ing Su, MS1; ChienHua Huang, MD, PhD1,2 ; Min-Shan Tsai, MD, PhD1,2; Eric Chou, MD3; Tsung-Chien

ro of

Lu, MD, PhD1,2; Wen-Jone Chen, MD, PhD1,2,4; Chien-Chang Lee, MD, ScD1,2,*; Shyr-

1

-p

Chyr Chen, MD1,2,*

Department of Emergency Medicine, National Taiwan University Hospital, Taipei,

Department of Emergency Medicine, College of Medicine, National Taiwan

3

na

University, Taipei, Taiwan

lP

2

re

Taiwan

Department of Emergency Medicine, Baylor Scott&White All Saints Medical Center,

Division of Cardiology, Department of Internal Medicine, National Taiwan University

Jo

4

ur

Fort Worth, Texas, USA

Hospital and National Taiwan University College of Medicine, Taipei, Taiwan

Address for correspondence and reprints: *Dr. Chien-Chang Lee 1

Address: Health Economic Outcomes Research Group, Department of Emergency Medicine, National Taiwan University Hospital, No. 7, Chung-Shan South Road, Taipei 100, Taiwan. E-mail: [email protected]/ [email protected] TEL: +886-2-23123456 ext 62831

ro of

FAX: +886-2-23223150

-p

*Dr. Shyr-Chyr Chen

Address: Department of Emergency Medicine, National Taiwan University Hospital,

lP

E-mail: [email protected]

re

No. 7, Chung-Shan South Road, Taipei 100, Taiwan.

na

TEL: +886-2-23123456 ext 62831

ur

FAX: +886-2-23223150

Jo

The two corresponding authors contributed equally to this work. Both corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

2

Abstract Aim: To summarise and compare the prognostic accuracy of the blood biomarkers of brain injury, including NSE and S-100B, for neurological outcomes in adult postcardiac arrest patients.

ro of

Methods: We systematically searched PubMed and Embase databases from their inception to March 2019. We selected studies providing sufficient data of prognostic

-p

values of NSE or S-100B to predict neurological outcomes in adult post-cardiac arrest

patients. We adopted QUADAS-2 to assess risk of bias and a Bayesian bivariate

re

random-effects meta-analysis model to synthesise the prognostic data. The study

na

lP

protocol was registered with PROSPERO (CRD42018084933).

Results: We included 42 studies involving 4806 patients in the meta-analysis. The NSE

ur

was associated with a pooled sensitivity of 0.56 (95% credible interval [CrI], 0.47–0.65)

Jo

and pooled specificity of 0.99 (95% CrI, 0.98–1.00). The S-100B was associated with a pooled sensitivity of 0.63 (95% CrI, 0.46–0.78) and pooled specificity of 0.97 (95% CrI, 0.92–1.00). The heterogeneity for NSE (I2, 22.4%) and S-100B (I2, 16.1%) was low and publication bias was not significant. In subgroup analyses, both biomarkers were associated with high specificity across all subgroups with regard to different 3

populations (i.e. whether patients were out-of-hospital cardiac arrest or whether patients received targeted temperature management), different timings of measurement, and different timings of outcome assessment.

Conclusions: The prognostic performance was comparable between NSE and S-100B.

ro of

Both biomarkers may be integrated into a multimodal neuroprognostication algorithm for post-cardiac arrest patients and institution-specific cut-off points for both

-p

biomarkers should be established.

lP

100B; Neurological outcome

re

Keywords: Cardiac arrest; Cardiopulmonary resuscitation; Neuron-specific enolase; S-

na

Background

ur

Cardiac arrests, either out-of-hospital cardiac arrest (OHCA) [1] or in-hospital cardiac arrest (IHCA) [2], are critical events with poor neuroprognosis. It was reported

Jo

that approximately 65% of OHCA patients died from neurological injuries and most died following withdrawal of life sustaining therapy due to poor neuroprognosis [3]. Therefore, accurate neuroprognostication, especially testing with high specificity (i.e. low false-positive rate), is important for these patients.

4

Recent guidelines [4] recommend a multimodal strategy of neuroprognostication for post-cardiac arrest patients. Among those recommended predictors of neurological outcome, Neuron-specific enolase (NSE) and S-100B are commonly examined blood biomarkers, both of which are proteins released following injury to neurons and glial cells, respectively. Blood concentrations of NSE and S-100B are thus likely to correlate

ro of

with the extent of hypoxic-ischaemic brain injury following cardiac arrest and the potential of neurological recovery.

-p

The advantages of using biomarkers in neuroprognostication included objective

results and likely independence from the influence of targeted temperature or sedatives.

re

To maximise the value of existing evidence in the literature, we performed this

lP

systematic review and meta-analysis to assess and compare the prognostic accuracy of

na

NSE and S-100B for prediction of neurological outcomes in post-cardiac arrest patients.

ur

Materials and Methods

This systematic review and meta-analysis were performed in accordance with the

Jo

PRISMA-DTA (Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies ) statement [5], the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy [6], and recommendations for reviews of diagnostic accuracy [7]. The study protocol was registered with PROSPERO

5

(CRD42018084933).

Data Sources and Searches Two investigators (CHW and WTC) independently searched the PubMed and Embase databases from their inception through to March 2019. The search strings were

ro of

as follows: (1) PubMed: ((neuron specific enolase) OR s100 OR s100b OR s100-b) AND ((cardiac arrest) OR cardiopulmonary resuscitation); (2) Embase: ('neuron

-p

specific enolase'/exp OR 'neuron specific enolase' OR 's100' OR 's100b' OR 's100-b') AND ('cardiac arrest'/exp OR 'cardiac arrest' OR 'cardiopulmonary resuscitation'/exp

re

OR 'cardiopulmonary resuscitation'). No restrictions were set on publication year or

lP

language. To ensure completeness, we screened the bibliographies of the selected

ur

strategy.

na

publications and relevant review articles for references not captured by our search

Jo

Study Selection

Two investigators (CHH and MST) independently scanned the titles and abstracts

of all retrieved articles to determine whether the articles were pertinent to this review. We used the following prespecified inclusion criteria: (a) population included adult post-cardiac arrest patients, (b) evaluation of NSE or S-100B, (c) results included 6

neurological outcome or death (death was classified as the worst neurological status), and (d) sufficient data to construct a 2×2 contingency table of true-positive, falsepositive, true-negative, and false-negative counts [8]. During study selection, investigators were not blinded to authors and institutions of the publications. We excluded studies that included trauma patients. We also excluded studies having

ro of

significant overlap (more than 50%) of study patients with the selected studies. We attempted to contact authors if studies provided insufficient information to construct a

-p

2×2 table. Full-text articles were retrieved if either of the reviewers considered the abstract potentially suitable. After retrieving the full reports of potentially relevant

re

studies, two reviewers (CHH and MST) independently assessed each study’s eligibility

lP

on the basis of the inclusion criteria. Differences of opinion regarding study eligibility

na

were settled by consultation with another investigator (WJC).

ur

Data Extraction and Quality Assessment

Jo

Two investigators (CHW and WTC) independently extracted qualitative and quantitative data, and a third investigator (SCC) adjudicated discordant assessments. Data were extracted for author information, publication year, study design, study setting, patient number, patient characteristics, quantitative data required for building a 2×2 contingency table [8], information regarding the details for NSE and S-100B tests, and 7

outcomes. For each biomarker, we selected a single data set of prognostic accuracy values from each study for meta-analysis. If more than one data set was reported in a single study (e.g. different sensitivity and specificity at different time points), we selected the representative data set according to the following hierarchy of priority: (a) data of highest specificity, (b) data of highest sensitivity, (c) data with the longest

ro of

interval between cardiac arrest and timing of measurement, (d) data with the longest follow-up time for neurological outcomes, and (e) data with the largest patient number.

-p

The rationale for this hierarchy was to reflect the use of these biomarkers in clinical use

for post-cardiac arrest patients.

re

pursuing a test with maximum specificity and accuracy in predicting long-term outcome

lP

Populations were categorised according to site of cardiac arrest (OHCA vs IHCA)

na

and implementation of targeted temperature management (TTM) (TTM vs No TTM). In order to maximize the number of studies included in the subgroup analysis, we would

ur

attempt to categorize the studies according to the quartiles of the proportions of OHCA

Jo

or TTM. The patients receiving target temperature between 32 and 36℃ [9] were categorized into TTM group. Also, because the TTM was first recommended in the 2005 resuscitation guidelines [9], publications before 2006 were categorised as No TTM subgroup if the proportion of TTM was not specifically reported. For timing of test performance, early and late tests were defined as measurement time within and 8

beyond 24 hours of cardiac arrest, respectively. If the timing of test performance was a period rather than a single time point, the end of the time interval was used for categorisation. For timing of outcome assessment, early and late outcomes were defined as neurological outcomes assessed before and after hospital discharge (or one month following cardiac arrest), respectively.

ro of

Two investigators (CHH and MST) independently assessed risk of bias of the included studies using the QUADAS-2 (Quality Assessment of Diagnostic Accuracy

-p

Studies 2) tool [10]. Disagreements were resolved by consensus or consultation with a

lP

Data Synthesis and Analysis

re

third investigator (CCL).

na

We calculated the pooled sensitivity, specificity, likelihood ratios (LRs), and areas under the receiver-operating characteristic curve (AUCs) across studies with 95%

ur

credible intervals (CrIs) using a Bayesian bivariate random-effects meta-analysis model

Jo

[11]. We used the Bayesian bivariate model instead of frequentist methods because the latter may have unstable estimates when the study number is small. The Bayesian metaanalyses were implemented using non-informative priors. Results are presented separately for NSE and S-100B. The bivariate approach models the logit-transformed sensitivity and specificity, and adjusts for the negative correlation between the 9

sensitivity and specificity of the index test that may arise from different thresholds used in different studies [7,12]. We also used the model to construct a hierarchical summary receiver-operating characteristic (HSROC) curve with credible and prediction regions as a way to summarise the true- and false-positive rates from different diagnostic studies [13]. We quantified the extent of between-study heterogeneity by calculating I2

ro of

statistics. We used Deek’s test to assess possible publication bias [7]. Subgroup analyses were planned a priori to investigate potential sources of heterogeneity.

-p

Subgroups that included at least five studies per stratum were eligible for quantitative synthesis.

with

the

meta4diag

package

(https://CRAN.R-

lP

www.r-project.org)

re

Bayesian statistical analyses were conducted using R 3.2.1 software (R Foundation;

na

project.org/package=meta4diag/), while I2 and test of publication bias were done using STATA version 14 (StataCorp, College Station, TX, USA) and OpenMeta [Analyst]

ur

(http://www. cebm.brown.edu/openmeta/). All statistical tests were two-sided, with

Jo

significance defined as p < 0.05.

Results

Study characteristics and quality assessment We included 42 studies [14-55] involving 4806 patients (Supplemental Figure 1).

10

As shown in Table 1 and Supplemental Table 1, the median cohort size across included studies was 68 patients. The first quartile of the OHCA proportions was 69% and the second quartile of the TTM proportions was 71%. Therefore, we determined that if the proportion of OHCA or TTM exceeded 70% in a given cohort, the feature of OHCA or TTM was attributed to the cohort. Therefore, 22 studies (52%) were categorized as

ro of

OHCA cohort and 19 studies (45%) as TTM cohort (Table 1 Supplemental Table 1). Most of the cut-off values for NSE and S-100B were calculated statistically to identify

-p

the threshold having the highest specificity rather than being prespecified.

Quality assessments using QUADAS-2 criteria are summarised in Supplemental

re

Figure 2. Only 9 studies [15, 29, 31, 36, 38, 39, 46, 47, 49] used pre-determined cut-off

lP

values of NSE or S-100B to calculate the prognostic accuracy (Supplemental Table 1).

na

Despite that the measurements of these biomarkers were objective tests, absence of prespecified cut-off points may still introduce risk of bias according to the QUADAS-

ur

2 criteria. Therefore, in the domain of index test, the risk of bias was judged as high

Jo

risk for the studies without prespecifying cut-off points (Supplemental Figure 2). Twenty-three studies (54.8%) had unclear risk of bias in use of the reference standard (neurological outcomes) because whether neurological outcomes were assessed without knowledge of the results of the index tests was not explicitly stated.

11

Quantitative synthesis Figures 1 and 2 demonstrate forest plots of the sensitivity and specificity of NSE and S-100B reported in the 42 included studies, respectively. Summary estimates of all prognostic accuracy measures are shown in Table 2. The overall pooled sensitivity of NSE was 0.56 (95% CrI, 0.47–0.65) and specificity was 0.99 (95% CrI, 0.98–1.00).

ro of

The overall pooled sensitivity of S-100B was 0.63 (95% CrI, 0.46–0.78) and specificity was 0.97 (95% CrI, 0.92–1.00). The HSROC curves, together with bivariate summary

-p

points of sensitivity and specificity with their 95% credible regions for both NSE and

S-100B, are shown in Figure 3. The heterogeneity for the overall summary estimates of

re

NSE and S-100B was low and publication bias was not significant.

lP

The results of subgroup analyses examining the prognostic accuracy of NSE and

na

S-100B in selected populations are presented in Table 2. Basically, there were no obvious differences between NSE and S-100B in specificity across different subgroups.

ur

Because there were less than five studies in some subgroups, the data were not pooled.

Jo

Discussion

Main findings

This systematic review identified 42 studies with 4806 patients examining the prognostic accuracy of NSE and S-100B for neurological outcomes among post-cardiac

12

arrest patients. The meta-analysis indicated that the overall prognostic performance for NSE and S-100B was similarly excellent (AUC: 0.91 for both biomarkers) with high specificity (NSE: 0.99; S-100B: 0.97) for predicting poor neurological outcomes. Furthermore, the subgroup analyses revealed that for both biomarkers, the specificity was consistently high for different populations, different timings of test performance,

ro of

and different timings of outcome assessment.

-p

Comparisons with previous studies

A previous meta-analysis [57] revealed that the pooled sensitivity and specificity

re

of NSE level >33 µg/L were 0.51 and 0.88, respectively, for post-cardiac arrest patients

lP

receiving TTM. The cut-off point of 33 µg/L for NSE was mainly identified before the

na

era of TTM [58]. Because TTM had been shown to improve neurological outcomes [59,60], the prognostic performance of a specific cut-off point for NSE or S-100B may

ur

change over time. As shown in Supplemental Table 1, the used cut-off points for NSE

Jo

and S-100B varied considerably because these cut-off points were calculated post-hoc through statistical analysis with an aim to maximise specificity, which accordingly contributed to the substantial variability in sensitivity across included studies (Figures 1 and 2) and might artificially inflate the specificity. The nature of the current metaanalysis is based on reported aggregated results, so we cannot derive an optimal cut-off 13

point. This work can only be done in an individual patient data meta-analysis. Before the availability of such data, each hospital should establish its own cut-off values of these biomarkers according to the test kits used, as recommended by guidelines [4]. Furthermore, because S-100B was less documented than NSE in previous studies, European

guidelines

[4]

suggest

NSE

as

the

preferred

biomarkers

in

ro of

neuroprognostication. Nevertheless, our meta-analysis indicated that the prognostic performance of S-100B was comparable to NSE. Because of the relatively small

-p

number of included studies assessing S-100B, we chose a Bayesian bivariate model to synthesise the evidence. Traditionally, diagnostic test meta-analysis is performed using

re

frequentist methods; however, these methods are generally considered less robust than

lP

Bayesian techniques at accounting for study heterogeneity. Bayesian random-effects

na

models are superior to frequentist random-effects models with respect to estimating between-study variance, because they do not ignore the imprecision of variance

ur

estimates. Therefore, S-100B can be considered as an alternative to NSE if NSE is not

Jo

available. However, this should be verified in future studies.

Interpretation of subgroup analyses Most studies investigated the prognostic performance of NSE or S-100B among OHCA patients (Table 2); for IHCA patients, more studies should be conducted to 14

validate their use. In comparison with clinical examination and electroencephalography, tests of biomarkers are less likely to be affected by targeted temperatures or sedatives. Therefore, prognostic performance did not differ considerably between patients receiving and not receiving TTM. The half-life for NSE and S-100B is approximately 24 and 2 hours, respectively.

ro of

Therefore, it is expected that S-100B peaks earlier in the circulation after hypoxicischaemic brain injury than NSE [61-64] and most studies investigated the prognostic

-p

performance of early S-100B or late NSE measurement for post-cardiac arrest patients.

Regardless of the timing of measurement for NSE or S-100B, the specificity was

re

consistently high (Table 2), except that the number of studies assessing late S-100B

lP

measurement was not large enough for quantitative synthesis. Studies [65,66] have

na

indicated that increased S-100B level was implicated in the process of neuronal apoptosis at the early post-cardiac arrest phase while NSE may serve as a marker of

ur

neuronal injury or death as a consequence of hypoxic-ischaemic damage [67]. Early S-

Jo

100B measurement and late NSE measurement may be complementary in neuroprognosis, and should be further examined. The pooled specificity of NSE and S-100B for prediction of late neurological

outcomes was quite high with low heterogeneity (Table 2). However, heterogeneity was moderately high for prediction of early outcomes. It was reported that approximately 15

50% of OHCA patients who were discharged with unfavourable neurological outcomes after TTM did recover favourable neurological status at 6–12 months after discharge [68]. Optimal timing for evaluating neurological outcome after cardiac arrest has yet to be established. It may be suitable to assess neurological outcome at a longer follow-up

ro of

timing [56], which allows opportunity for observing improvement.

Clinical applications

-p

There are no tests with 100% specificity (false-positive rate of 0) in clinical practice. False-positive rates lower than 5% with narrow confidence intervals are

re

generally accepted as reliable [69]. As shown by the Fagan nomogram (Supplemental

lP

Figure 3), when pre-test probability could be objectively estimated, both NSE and S-

na

100B tests could efficiently help identify patients with high probability of poor neurological recovery.

ur

It should also be emphasised that although the specificity of both NSE and S-100B

Jo

tests is high, these tests should not be used alone for prognostication, as recommended by American [70] and European [4] guidelines. However, the optimal combination of prognostic modalities has not been verified. Various risk-stratification scores have been developed and validated for post-cardiac arrest patients, such as cardiac arrest hospital prognosis (CAHP) score for OHCA patients [71,72] and cardiac arrest survival 16

postresuscitation in-hospital (CASPRI) score for IHCA patients [73,74]. These riskstratification scores may also help define the pre-test probability and be used along with NSE and S-100B tests for neuroprognostication.

Study limitations

ro of

First, many of the studies excluded during the literature search were due to insufficient information to build a 2×2 contingency table. Most studies examining the

-p

prognostic performance of serial NSE or S-100B tests provided only serial sensitivity and specificity, without case number and total participant number provided at

re

corresponding time points. Studies reporting diagnostic test accuracy should still follow

lP

the STARD guidelines [75] to explicitly report contingency tables of the index test

na

results by the results of the reference standard at each time point. Second, the metaanalysis could only examine the overall prognostic performance of NSE and S-100B

ur

tests and their performance across different subgroups. A specific cut-off point for NSE

Jo

or S-100B could not be identified through meta-analysis. Rather, as guideline recommendations [4,70], each hospital should establish its own normal values and cutoff points for these biomarkers. Third, like most classification systems for neurological function or brain injury, such as cerebral performance category score [76], Glasgow outcome score [77], modified Glasgow outcome score [78], modified Rankin scale[79], 17

death was graded as the worst outcome on the scale in our study. Also, many of the included studies did not explicitly indicate blinding of outcome assessors to the results of index tests when evaluating the reference standard of poor neurologic outcome. Both factors could have biased the assessment, introducing the bias of ‘self-fulfilling prophecies.’ The prognostic performance of these biomarkers may thus been falsely

ro of

exaggerated. Fourth, the proportion of 70% used to categorize a study as OHCA vs IHCA or TTM vs No TTM was determined by distribution of the proportions of a

-p

certain character. This classification was used to maximize the number of studies

included in the subgroup analysis, which may introduce selection bias. Therefore, the

re

results of subgroup analysis may best be viewed as being exploratory rather than

lP

definite. Fifth, despite that there was no significant statistical heterogeneity for the

na

overall synthesized results, there was obvious methodological and clinical heterogeneity across the included studies, such as different timings of measurement or

ur

definitions of neurological outcomes. Although subgroup analyses had been conducted

Jo

to reduce the confounding effect of each variable, a large-scale prospective study may still be necessary to clarify the prognostic accuracy of these biomarkers in a specific group. Finally, in the process of calculating the pooled prognostic estimates, the data sets with the highest specificity were prioritised to select. We selected to present the pooled accuracy estimate in favour of specificity with an aim to reflect the use of these 18

biomarkers in clinical practice [4, 70]. A marker with high specificity and low false negative can help clinician to make a difficult clinical decision.

Conclusions Prognostic performance was comparable between NSE and S-100B, both of which

ro of

demonstrated high specificity in predicting poor neurological outcome for post-cardiac arrest patients. The prognostic performance may not be influenced by different

populations, different timings of test performance, and different timings of outcome

-p

assessment. Both NSE and S-100B may be integrated into a multimodal

Conflicts of Interests

lP

re

neuroprognostication algorithm for post-cardiac arrest patients.

ur

na

The authors declare that they have no conflict of interest.

Acknowledgments

Jo

We thank the staff of the 3rd Core Lab, Department of Medical Research, National Taiwan University Hospital for technical support. Author Chih-Hung Wang recieved a grant (108-S4091) from the National Taiwan University Hospital. National Taiwan University Hospital had no involvement in designing the study, collecting, analysing or

19

interpreting the data, writing the manuscript, or deciding whether to submit the manuscript for publication. The remaining authors have disclosed that they do not have any conflicts of interest. References 1.

Ong ME, Shin SD, De Souza NN, Tanaka H, Nishiuchi T, Song KJ, et al.

ro of

Outcomes for out-of-hospital cardiac arrests across 7 countries in Asia: The Pan Asian Resuscitation Outcomes Study (PAROS). Resuscitation 2015;96:100-8.

Benjamin EJ, Blaha MJ, Chiuve SE, Cushman M, Das SR, Deo R, et al. Heart

-p

2.

Disease and Stroke Statistics-2017 Update: A Report From the American Heart

Lemiale V, Dumas F, Mongardon N, Giovanetti O, Charpentier J, Chiche JD, et

lP

3.

re

Association. Circulation 2017;135:e146-e603

na

al. Intensive care unit mortality after cardiac arrest: the relative contribution of shock and brain injury in a large cohort. Intensive Care Med 2013;39:1972-80. Nolan JP, Soar J, Cariou A, Cronberg T, Moulaert VR, Deakin CD, et al. European

ur

4.

Jo

Resuscitation Council and European Society of Intensive Care Medicine Guidelines for Post-resuscitation Care 2015: Section 5 of the European Resuscitation Council Guidelines for Resuscitation 2015. Resuscitation 2015;95:202-22. 5.

McInnes MDF, Moher D, Thombs BD, McGrath TA, Bossuyt PM, the PRISMA-

DTA Group, et al. Preferred Reporting Items for a Systematic Review and Meta20

analysis of Diagnostic Test Accuracy Studies: The PRISMA-DTA Statement. JAMA 2018;319:388-96. 6.

Deeks JJ, Bossuyt PM, Gatsonis C. Cochrane Handbook for Systematic Reviews

of Diagnostic Test Accuracy. Version 1.0.0. 2013. Available from srdta.cochrane.org Accessed 1st June 2019. Leeflang MM, Deeks JJ, Gatsonis C, Bossuyt PM; Cochrane Diagnostic Test

ro of

7.

Accuracy Working Group. Systematic reviews of diagnostic test accuracy. Ann Inter

8.

-p

Med 2008;149:889-97.

Kim KW, Lee J, Choi SH, Huh J, Park SH. Systematic Review and Meta-Analysis

re

of Studies Evaluating Diagnostic Test Accuracy: A Practical Review for Clinical

Part 7.5: Postresuscitation Support. Circulation. 2005;112:IV-84-IV-8.

na

9.

lP

Researchers-Part I. General Guidance and Tips. Korean J Radiol 2015;16:1175-87.

10. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al.

ur

QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies.

Jo

Ann Inter Med 2011;155:529-36. 11. Guo J, Riebler A, Rue H. Bayesian bivariate meta-analysis of diagnostic test studies with interpretable priors. Stat Med 2017;36:3039-58. 12. Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary 21

measures in diagnostic reviews. J Clin Epidemiol 2005;58:982-90. 13. Harbord RM, Whiting P, Sterne JA, Egger M, Deeks JJ, Shang A, et al. An empirical comparison of methods for meta-analysis of diagnostic accuracy showed hierarchical models are necessary. J Clin Epidemiol 2008;61:1095-103. 14. Roine RO, Somer H, Kaste M, Viinikka L, Karonen SL. Neurological outcome

ro of

after out-of-hospital cardiac arrest. Prediction by cerebrospinal fluid enzyme analysis. Arch Neurol 1989;46:753-6.

-p

15. Martens P, Raabe A, Johnsson P. Serum S-100 and neuron-specific enolase for prediction of regaining consciousness after global cerebral ischemia. Stroke

re

1998;29:2363-6.

lP

16. Schoerkhuber W, Kittler H, Sterz F, Behringer W, Holzer M, Frossard M, et al.

na

Time course of serum neuron-specific enolase. A predictor of neurological outcome in patients resuscitated from cardiac arrest. Stroke 1999;30:1598-603.

ur

17. Hachimi-Idrissi S, Van der Auwera M, Schiettecatte J, Ebinger G, Michotte Y,

Jo

Huyghens L. S-100 protein as early predictor of regaining consciousness after out of hospital cardiac arrest. Resuscitation 2002;53:251-7. 18. Mussack T, Biberthaler P, Kanz KG, Wiedemann E, Gippner-Steppert C, Mutschler W, et al. Serum S-100B and interleukin-8 as predictive markers for comparative neurologic outcome analysis of patients after cardiac arrest and severe 22

traumatic brain injury. Crit Care Med 2002;30:2669-74. 19. Fries M, Kunz D, Gressner AM, Rossaint R, Kuhlen R. Procalcitonin serum levels after out-of-hospital cardiac arrest. Resuscitation 2003;59:105-9. 20. Meynaar IA, Oudemans-van Straaten HM, van der Wetering J, Verlooy P, Slaats EH, Bosman RJ, et al. Serum neuron-specific enolase predicts outcome in post-anoxic

ro of

coma: a prospective cohort study. Intensive Care Med 2003;29:189-95. 21. Tiainen M, Roine RO, Pettila V, Takkunen O. Serum neuron-specific enolase and

-p

S-100B protein in cardiac arrest patients treated with hypothermia. Stroke 2003;34:2881-6.

re

22. Zingler VC, Krumm B, Bertsch T, Fassbender K, Pohlmann-Eden B. Early

lP

prediction of neurological outcome after cardiopulmonary resuscitation: a multimodal

na

approach combining neurobiochemical and electrophysiological investigations may provide high prognostic certainty in patients after cardiac arrest. Eur Neurol

ur

2003;49:79-84.

Jo

23. Pfeifer R, Borner A, Krack A, Sigusch HH, Surber R, Figulla HR. Outcome after cardiac arrest: predictive values and limitations of the neuroproteins neuron-specific enolase and protein S-100 and the Glasgow Coma Scale. Resuscitation 2005;65:49-55. 24. Rech TH, Vieira SR, Nagel F, Brauner JS, Scalco R. Serum neuron-specific enolase as early predictor of outcome after in-hospital cardiac arrest: a cohort study. 23

Crit Care 2006;10:R133. 25. Wessels T, Harrer JU, Jacke C, Janssens U, Klötzsch C. The prognostic value of early transcranial Doppler ultrasound following cardiopulmonary resuscitation. Ultrasound Med Biol 2006;32:1845-51. 26. Reisinger J, Hollinger K, Lang W, Steiner C, Winter T, Zeindlhofer E, et al.

ro of

Prediction of neurological outcome after cardiopulmonary resuscitation by serial determination of serum neuron-specific enolase. Eur Heart J 2007;28:52-8.

-p

27. Wennervirta JE, Ermes MJ, Tiainen SM, Salmi TK, Hynninen MS, Särkelä MO, et al. Hypothermia-treated cardiac arrest patients with good neurological outcome differ

re

early in quantitative variables of EEG suppression and epileptiform activity. Crit Care

lP

Med 2009;37:2427-35.

na

28. Steffen IG, Hasper D, Ploner CJ, Schefold JC, Dietz E, Martens F, et al. Mild therapeutic hypothermia alters neuron specific enolase as an outcome predictor after

ur

resuscitation: 97 prospective hypothermia patients compared to 133 historical non-

Jo

hypothermia patients. Crit Care 2010;14:R69. 29. Cronberg T, Rundgren M, Westhall E, Madigan N, Pascual-Leone A, Meehan WP. Neuron-specific enolase correlates with other prognostic markers after cardiac arrest. Neurology 2011;77:623-30. 30. Daubin C, Quentin C, Allouche S, Etard O, Gaillard C, Seguin A, et al. Serum 24

neuron-specific enolase as predictor of outcome in comatose cardiac-arrest survivors: a prospective cohort study. BMC Cardiovasc Disord 2011;11:48. 31. Samaniego EA, Mlynash M, Caulfield AF, Eyngorn I, Wijman CA. Sedation confounds outcome prediction in cardiac arrest survivors treated with hypothermia. Neurocrit Care 2011;15:113-9.

ro of

32. Bouwes A, Binnekade JM, Kuiper MA, Bosch FH, Zandstra DF, Toornvliet AC, et al. Prognosis of coma after therapeutic hypothermia: a prospective cohort study. Ann

-p

Neurol 2012;71:206-12.

33. Helánová K1, Pařenica J, Jarkovský J, Dostálová L, Littnerová S, Klabenešová I,

re

et al. [S-100B protein elevation in patients with the acute coronary syndrome after

lP

resuscitation is a predictor of adverse neurological prognosis]. Vnitr Lek 2012;58:266-

na

72.

34. Kim J, Choi BS, Kim K, Jung C, Lee JH, Jo YH, et al. Prognostic performance of

ur

diffusion-weighted MRI combined with NSE in comatose cardiac arrest survivors

Jo

treated with mild hypothermia. Neurocrit Care 2012;17:412-20. 35. Lee BK, Jeung KW, Lee HY, Matsuyama T4, Katayama Y5, Hirose T, et al. Combining brain computed tomography and serum neuron specific enolase improves the prognostic performance compared to either alone in comatose cardiac arrest survivors treated with therapeutic hypothermia. Resuscitation 2013;84:1387-92. 25

36. Scheel M, Storm C, Gentsch A, Nee J, Luckenbach F, Ploner CJ, et al. The prognostic value of gray-white-matter ratio in cardiac arrest patients treated with hypothermia. Scand J Trauma Resusc Emerg Med 2013;21:23. 37. Stammet P, Wagner DR, Gilson G, Devaux Y. Modeling serum level of s100beta and bispectral index to predict outcome after cardiac arrest. J Am Coll Cardiol

ro of

2013;62:851-8. 38. Zellner T, Gartner R, Schopohl J, Angstwurm M. NSE and S-100B are not

-p

sufficiently predictive of neurologic outcome after therapeutic hypothermia for cardiac arrest. Resuscitation 2013;84:1382-6.

re

39. Hasslacher J, Lehner GF, Harler U, Beer R, Ulmer H, Kirchmair R, et al.

lP

Secretoneurin as a marker for hypoxic brain injury after cardiopulmonary resuscitation.

na

Intensive Care Med 2014;40:1518-27.

40. Oddo M, Rossetti AO. Early multimodal outcome prediction after cardiac arrest

ur

in patients treated with hypothermia. Crit Care Med 2014;42:1340-7.

Jo

41. Leão RN, Ávila P, Cavaco R, Germano N, Bento L. Therapeutic hypothermia after cardiac arrest: outcome predictors. Rev Bras Ter Intensiva 2015;27:322-32. 42. Roger C, Palmier L, Louart B, Molinari N4, Claret PG5, de la Coussaye JE, et al. Neuron specific enolase and Glasgow motor score remain useful tools for assessing neurological prognosis after out-of-hospital cardiac arrest treated with therapeutic 26

hypothermia. Anaesth Crit Care Pain Med 2015;34:231-7. 43. Ok G, Aydın D, Erbüyün K, Gürsoy C, Taneli F, Bilge S, et al. Neurological outcome after cardiac arrest: a prospective study of the predictive ability of prognostic biomarkers neuron-specific enolase, glial fibrillary acidic protein, S-100B, and procalcitonin. Turk J Med Sci 2016;46:1459-68.

ro of

44. Helwig K, Seeger F, Holschermann H, Lischke V4, Gerriets T5,6, Niessner M, et al. Elevated Serum Glial Fibrillary Acidic Protein (GFAP) is Associated with Poor Outcome

After

Cardiopulmonary

Resuscitation.

Neurocrit

Care

-p

Functional

2017;27:68-74.

re

45. Kaneko T, Fujita M, Ogino Y, Yamamoto T, Tsuruta R, Kasaoka S. Serum

lP

neutrophil gelatinase-associated lipocalin levels predict the neurological outcomes of

na

out-of-hospital cardiac arrest victims. BMC Cardiovasc Disord 2017;17:111. 46. Moseby-Knappe M, Pellis T, Dragancea I, Friberg H, Nielsen N, Horn J, et al.

ur

Head computed tomography for prognostication of poor outcome in comatose patients

Jo

after cardiac arrest and targeted temperature management. Resuscitation 2017;119:8994.

47. Rossetti AO, Tovar Quiroga DF, Juan E, Novy J, White RD, Ben-Hamouda N, et al. Electroencephalography Predicts Poor and Good Outcomes After Cardiac Arrest: A Two-Center Study. Crit Care Med 2017;45:e674-82. 27

48. Streitberger KJ, Leithner C, Wattenberg M, Tonner PH, Hasslacher J, Joannidis M, et al. Neuron-Specific Enolase Predicts Poor Outcome After Cardiac Arrest and Targeted Temperature Management: A Multicenter Study on 1,053 Patients. Crit Care Med 2017;45:1145-51. 49. Tsetsou S, Novy J, Pfeiffer C, Oddo M, Rossetti AO. Multimodal Outcome

ro of

Prognostication After Cardiac Arrest and Targeted Temperature Management: Analysis at 36 degrees C. Neurocrit Care 2018;28:104-9.

-p

50. Vondrakova D, Kruger A, Janotka M, Malek F, Dudkova V, Neuzil P, et al.

Association of neuron-specific enolase values with outcomes in cardiac arrest survivors

re

is dependent on the time of sample collection. Crit Care 2017;21:172.

lP

51. Chung-Esaki HM, Mui G, Mlynash M, Eyngorn I, Catabay K, Hirsch KG. The

na

neuron specific enolase (NSE) ratio offers benefits over absolute value thresholds in post-cardiac arrest coma prognosis. J Clin Neurosci 2018;57:99-104.

ur

52. Jang JH, Park WB, Lim YS, Choi JY, Cho JS, Woo JH, et al. Combination of

Jo

S100B and procalcitonin improves prognostic performance compared to either alone in patients with cardiac arrest: A prospective observational study. Medicine (Baltimore) 2019;98:e14496. 53. Kim JH, Kim MJ, You JS, Lee HS, Park YS, Park I, et al. Multimodal approach for neurologic prognostication of out-of-hospital cardiac arrest patients undergoing 28

targeted temperature management. Resuscitation 2019;134:33-40. 54. Park JH, Wee JH, Choi SP, Oh JH, Cheol S. Assessment of serum biomarkers and coagulation/fibrinolysis markers for prediction of neurological outcomes of out of cardiac arrest patients treated with therapeutic hypothermia. Clin Exp Emerg Med 2019;6:9-18.

Neuron-specific-enolase

as

a

predictor

of

the

ro of

55. Schrage B, Rubsamen N, Becher PM, Roedl K, Söffker G, Schwarzl M, et al. neurologic

outcome

after

-p

cardiopulmonary resuscitation in patients on ECMO. Resuscitation 2019;136:14-20.

56. Becker LB, Aufderheide TP, Geocadin RG, Callaway CW, Lazar RM, Donnino

re

MW, et al. Primary outcomes for resuscitation science studies: a consensus statement

lP

from the American Heart Association. Circulation 2011;124:2158-77.

na

57. Golan E, Barrett K, Alali AS, Duggal A, Jichici D, Pinto R, et al. Predicting neurologic outcome after targeted temperature management for cardiac arrest:

ur

systematic review and meta-analysis. Crit Care Med 2014;42:1919-30.

Jo

58. Zandbergen EG, Hijdra A, Koelman JH, Hart AA, Vos PE, Verbeek MM, et al. Prediction of poor outcome within the first 3 days of postanoxic coma. Neurology 2006;66:62-8.

59. Bernard SA, Gray TW, Buist MD, Jones BM, Silvester W, Gutteridge G, et al. Treatment of comatose survivors of out-of-hospital cardiac arrest with induced 29

hypothermia. N Engl J Med 2002;346:557-63. 60. Hypothermia after Cardiac Arrest Study Group. Mild therapeutic hypothermia to improve the neurologic outcome after cardiac arrest. N Engl J Med 2002;346:549-56. 61. Bottiger BW, Mobes S, Glatzer R, Bauer H, Gries A, Bärtsch P, et al. Astroglial protein S-100 is an early and sensitive marker of hypoxic brain damage and outcome

ro of

after cardiac arrest in humans. Circulation 2001;103:2694-8. 62. Shinozaki K, Oda S, Sadahiro T, Nakamura M, Abe R, Nakada TA, et al. Serum

-p

S-100B is superior to neuron-specific enolase as an early prognostic biomarker for neurological outcome following cardiopulmonary resuscitation. Resuscitation

re

2009;80:870-5.

lP

63. Mortberg E, Zetterberg H, Nordmark J, Blennow K, Rosengren L, Rubertsson S.

na

S-100B is superior to NSE, BDNF and GFAP in predicting outcome of resuscitation from cardiac arrest with hypothermia treatment. Resuscitation 2011;82:26-31.

ur

64. Minami T, Sainte S, De Praetere H, Rega F, Flameng W, Verbrugghe P, et al.

Jo

Hospital cost savings and other advantages of sutureless vs stented aortic valves for intermediate-risk elderly patients. Surg Today 2017;47:1268-73. 65. Bianchi R, Adami C, Giambanco I, Donato R. S100B binding to RAGE in microglia stimulates COX-2 expression. J Leukoc Biol 2007;81:108-18. 66. Van Eldik LJ, Wainwright MS. The Janus face of glial-derived S100B: beneficial 30

and detrimental functions in the brain. Restor Neurol Neurosci 2003; 21:97-108. 67. Marangos PJ, Schmechel DE. Neuron specific enolase, a clinically useful marker for neurons and neuroendocrine cells. Annu Rev Neurosci 1987;10:269-95. 68. Terman SW, Hume B, Meurer WJ, Silbergleit R. Impact of presenting rhythm on short- and long-term neurologic outcome in comatose survivors of cardiac arrest treated

ro of

with therapeutic hypothermia. Crit Care Med 2014;42:2225-34. 69. Sandroni C, Cariou A, Cavallaro F, Cronberg T, Friberg H, Hoedemaekers C, et

-p

al. Prognostication in comatose survivors of cardiac arrest: an advisory statement from the European Resuscitation Council and the European Society of Intensive Care

re

Medicine. Resuscitation 2014;85:1779-89.

lP

70. Callaway CW, Donnino MW, Fink EL, Geocadin RG, Golan E, Kern KB, et al.

na

Part 8: Post-Cardiac Arrest Care: 2015 American Heart Association Guidelines Update for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation

ur

2015;132:S465-82.

Jo

71. Maupain C, Bougouin W, Lamhaut L, Deye N, Diehl JL, Geri G, et al. The CAHP (Cardiac Arrest Hospital Prognosis) score: a tool for risk stratification after out-ofhospital cardiac arrest. Eur Heart J 2016;37:3222-8. 72. Wang CH, Huang CH, Chang WT, Tsai MS, Yu PH, Wu YW, et al. Prognostic performance of simplified out-of-hospital cardiac arrest (OHCA) and cardiac arrest 31

hospital prognosis (CAHP) scores in an East Asian population: a prospective cohort study. Resuscitation 2019;137:133-9. 73. Chan PS, Spertus JA, Krumholz HM, Berg RA, Li Y, Sasson C, et al: A validated prediction tool for initial survivors of in-hospital cardiac arrest. Arch Intern Med 2012;172:947-53.

ro of

74. Wang CH, Chang WT, Huang CH, Tsai MS, Yu PH, Wu YW, et al. Validation of the Cardiac Arrest Survival Postresuscitation In-hospital (CASPRI) score in an East

-p

Asian population. PLoS One 2018;13:e0202938.

75. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, et al.

Becker LB, Aufderheide TP, Geocadin RG, Callaway CW, Lazar RM, Donnino

na

76.

lP

studies. BMJ 2015;351:h5527.

re

STARD 2015: an updated list of essential items for reporting diagnostic accuracy

MW, et al. Primary outcomes for resuscitation science studies: a consensus statement

ur

from the American Heart Association. Circulation 2011;124: 2158-77.

Jo

77. Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet 1975; 1: 480-4. 78. Rana OR, Schröder JW, Kühnen JS, Saygili E, Gemein C, Zink MDH, et al. The Modified Glasgow Outcome Score for the prediction of outcome in patients after cardiac arrest: a prospective clinical proof of concept study. Clin Res Cardiol 32

2012;101:533-43. 79. Wilson JTL, Hareendran A, Grant M, Baird T, Schulz UGR, Muir KW, et al. Improving the assessment of outcomes in stroke: use of a structured interview to assign grades on the modified Rankin Scale. Stroke 2002;33:2243-6.

ro of

Figure 1. Forest plots of sensitivity (top) and specificity (bottom) for NSE. CrI, credible interval; FN, false-negative; FP, false-positive; TN, true-negative; TP, true-positive.

Jo

ur

na

lP

re

-p

(TIFF)

33

ro of -p re lP na ur Jo

Figure 2. Forest plots of sensitivity (top) and specificity (bottom) for S-100B. CrI, credible interval; FN, false-negative; FP, false-positive; TN, true-negative; TP, truepositive. (TIFF)

34

ro of

-p

Figure 3. Hierarchical summary receiver-operating characteristic (HSROC) curves and

Jo

ur

na

lP

NSE (top) and S-100B (bottom).

re

bivariate summary points of (specificity, sensitivity) and their 95% credible regions for

35

36

ro of

-p

re

lP

na

ur

Jo

Table 1. Summary characteristics of the included 42 studies Frequency, n (%)b

Description Continent America

3 (7)

Europe

27 (64)

Asia

7 (17)

Australia/ Oceania

3 (7)

Cross-continent

2 (5)

Before 2005

10 (24)

2006-2010

5 (12)

2011-2015

14 (33)

2016-2019

13 (31)

-p

Study design Prospective cohort

26 (57)

Retrospective cohort

12 (29)

68 (43-127)

re

Median sample size (IQR), n Population

22 (52)

IHCA (IHCA patients >70% of study cohort)

2 (5)

TTM (TTM performed in >70% of study

19 (45)

na

lP

OHCA (OHCA patients >70% of study cohort)

cohort)

No TTM (TTM performed in <30% of study

ur

cohort)

ro of

Year of publication

15 (36)

Blood biomarkders tested

29 (69)

S-100B

6 (14)

Both NSE and S-100B

7 (17)

Jo

NSE

Timing of measurement after cardiac arresta Early NSE measurement

6 (14)

Late NSE measurement

30 (71)

Early S-100B measurement

10 (24)

Late S-100B measurement

3 (7) 37

Cut-off value (IQR), µg/L NSE

45.95 (28.95-68.05)

S-100B

0.7 (0.3-1.25)

Cerebral performance category scale

28 (67)

Modified Rankin scale

1 (2)

Glasgow outcome score

8 (19)

Modified Glasgow outcome score

1 (2)

Glasgow coma scale

1 (2)

No formal scale used

3 (7)

ro of

Definition of poor neurological outcome

Timing of outcome assessment after arrest Early outcome

19 (45)

Late outcome

23 (55)

-p

IHCA: in-hospital cardiac arrest; IQR: interquartile range; NSE: neuron specific enolase; OHCA: outof-hospital cardiac arrest; TTM:targeted temperature management a

Early and late measurements were defined as measurement time within and beyond 24 hours of

Unless otherwise indicated.

lP

b

re

cardiac arrest, respectively.

Table 2. Summary Estimates of Overall and Subgroup Analyses for NSE and S-100B Studies

Sensitivit

Specificit

AUC

Positive

Negative

I2

Publicatio

(patients

y

y

(95%

likelihoo

likelihoo

(%)

n

),

(95%

(95%

CrI)

d ratio

d ratio

CrI)

CrI)

(95%

(95%

CrI)

CrI)

na

Variable

ur

n (n)

bias*

Overall

36

0.56

0.99

0.91

102.73

0.44

22.

(4508)

(0.47-

(0.98-

(0.9-

(34.69-

(0.34-

4

0.65)

1.00)

0.91)

648.75)

0.54)

13

0.63

0.97

0.91

23.88

0.39

16.

(677)

(0.46-

(0.92-

(0.89

(7.23-

(0.21-

1

0.78)

1.00)

-

212.62)

0.56)

Jo

NSE (14-

16, 20-32,

34-36, 38-

0.06

51, 53-55) S-100B (15, 17-19, 21-23, 27,

0.69

38

33, 37, 43,

0.93)

52, 54) Population NSE for OHCA (14,

17

0.53

0.99

0.97

78.00

0.48

(2494)

(0.34-

(0.98-

(0.95

(19.98-

(0.27-

0.71)

1.00)

-

769.48)

0.69)

16, 21, 2729, 34, 39,

0

NA

0.98)

40, 42, 45, 46, 48-50,

for OHCA

8

0.72

0.97

0.95

27.65

(388)

(0.52-

(0.91-

(0.92

(7.16-

0.87)

1.00)

-

441.00)

(17-19, 21, 27, 33, 52,

0.99)

54) 0.52

1.00

0.97

(2586)

(0.32-

(0.99-

(0.97

0.71)

1.00)

29, 32, 3436, 38, 40-

51, 53-55) 5

(329)

(15, 17-19, 21-23)

(0.28-

48791.7)

0.71)

5.3

NA

NA

42.96

0.44

33.

(0.30-

(0.91-

(0.87

(5.56-

(0.18-

4

0.80)

1.00)

-

3019.48)

0.72)

0.97) 0.99

0.91

41.70

0.47

(844)

(0.44-

(0.96-

(0.89

(12.64-

(0.36-

0.64)

1.00)

-

367.52)

0.58)

26, 28, 49)

for No TTM

(48.14-

0.54

(14-16, 20-

S-100B

0.48

12

Jo

No TTM

0.53)

0.92

52, 54)

NSE for

6

0.99

ur

(27, 33, 37,

0.57

na

for TTM

(0.10-

NA

0.97)

42, 48, 50,

S-100B

-

38.

410.94

re

TTM (27,

16

lP

NSE for

0.29

-p

S-100B

ro of

53, 54)

0

NA

0

NA

0.92)

7

0.62

0.97

0.97

19.59

0.40

(318)

(0.37-

(0.91-

(0.88

(6.08-

(0.15-

0.84)

0.99)

-

125.08)

0.66)

1.00) 39

Timing of test performan ce Early NSE

6

0.56

0.97

0.64

19.53

0.46

50.

(342)

(0.45-

(0.82-

(0.59

(2.81-

(0.33-

4

0.67)

1.00)

-

500.42)

0.61)

measureme nt (14, 15,

NA

0.68)

21, 22, 43,

Early S100B

10

0.70

0.98

0.93

36.36

(403)

(0.52-

(0.92-

(0.91

(7.67-

0.85)

1.00)

-

827.55)

measureme

0.95)

19, 21, 22,

52) 0.56

1.00

(4166)

(0.44-

(0.99-

0.66)

1.00)

nt (16, 20, 23-32, 34-

3

0.53)

134.14

0.45

(0.92

(35.41-

(0.33-

-

1244.63)

0.57)

0

NA

na

44-51, 54,

ur

55)

100B

(0.13-

NA

0.93)

36, 38-42,

Late S-

16.

0.92

lP

measureme

30

re

27, 33, 43,

Late NSE

0.30

-p

nt (15, 17-

ro of

53)

3

NAe

NA

NA

NA

NA

NA

NA

0.59

0.99

0.93

53.32

0.42

44.

NA

(274)

Jo

measureme nt (23, 37, 54)

Timing of outcome assessment NSE for

16

40

early

(2434)

outcome

(0.40-

(0.96-

(0.92

(14.99-

(0.25-

0.75)

1.00)

-

539.37)

0.63)

(14, 20, 23,

5

0.94)

25, 28, 35, 36, 39, 4345, 48, 50, 53-55)

for early

6

0.80

0.88

0.91

6.44

0.23

45.

(335)

(0.53-

(0.67-

(0.85

(2.19-

(0.05-

3

0.95)

0.97)

-

29.69)

0.57)

outcome (17, 19, 23,

0.97)

33, 43, 54)

late

20

0.54

1.00

0.88

230.89

(2074)

(0.44-

(0.99-

(0.87

(34.80-

0.63)

1.00)

-

outcome (15, 16, 21,

0.88)

34, 38, 4042, 46, 47, 49, 51)

(342)

NA

0

NA

(0.36-

10514.9)

0.56)

1.00

1.00

133.40

0.52

(0.32-

(0.97-

(0.9-

(12.30-

(0.34-

0.65)

1.00)

1.00)

8534.33)

0.69)

ur

outcome

0.49

na

for late

7

lP

27, 29-32,

S-100B

0

re

22, 24, 26,

0.46

-p

NSE for

NA

ro of

S-100B

(15, 18, 21, 22, 27, 37,

Jo

52)

AUC: area under the receiver operating characteristic curve; CrI: credible interval; NA: not available; NSE: neuron specific enolase; OHCA: out-of-hospital cardiac arrest; TTM:targeted temperature management *

Deek’s test p-value

41