Anda di halaman 1dari 15

Resuscitation 84 (2013) 13241338

Contents lists available at ScienceDirect

Resuscitation
journal homepage: www.elsevier.com/locate/resuscitation

Review article

Predictors of poor neurological outcome in adult comatose survivors of cardiac arrest: A systematic review and meta-analysis. Part 2: Patients treated with therapeutic hypothermia
Claudio Sandroni a, , Fabio Cavallaro a , Clifton W. Callaway b , Sonia DArrigo a , Tommaso Sanna c , Michael A. Kuiper d , Matteo Biancone a , Giacomo Della Marca e , Alessio Farcomeni f , Jerry P. Nolan g
a

Department of Anaesthesiology and Intensive Care, Catholic University School of Medicine, Rome, Italy Department of Emergency Medicine, University of Pittsburgh, United States c Department of Cardiovascular Sciences, Catholic University School of Medicine, Rome, Italy d Department of Intensive Care, Medical Center Leeuwarden, Leeuwarden, The Netherlands e Department of Neurology, Catholic University School of Medicine, Rome, Italy f Department of Public Health and Infectious Diseases, Statistics Section, Sapienza University of Rome, Italy g Department of Anaesthesia and Intensive Care Medicine, Royal United Hospital, Bath, UK
b

a r t i c l e

i n f o

a b s t r a c t
Aims and methods: To systematically review the accuracy of early (7 days) predictors of poor outcome, dened as death or vegetative state (Cerebral Performance Categories [CPC] 45) or death, vegetative state or severe disability (CPC 35), in comatose adult survivors from cardiac arrest (CA) treated using therapeutic hypothermia (TH). Electronic databases were searched for eligible studies. Sensitivity, specicity, and false positive rates (FPR) for each predictor were calculated. Quality of evidence (QOE) was evaluated according to the GRADE guidelines. Results: 37 studies (2403 patients) were included. A bilaterally absent N20 SSEP wave during TH (4 studies; QOE: Moderate) or after rewarming (5 studies; QOE: Low), a nonreactive EEG background (3 studies; QOE: Low) after rewarming, a combination of absent pupillary light and corneal reexes plus a motor response no better than extension (M 2) (1 study; QOE: Very low) after rewarming predicted CPC 35 with 0% FPR and narrow (<10%) 95% condence intervals. No consistent threshold for 0% FPR could be identied for blood levels of biomarkers. In 6/8 studies on SSEP, in 1/3 studies on EEG reactivity and in the single study on clinical examination the investigated predictor was used for decisions to withdraw treatment, causing the risk of a self-fullling prophecy. Conclusions: in the rst 7 days after CA, a bilaterally absent N20 SSEP wave anytime, a nonreactive EEG after rewarming or a combination of absent ocular reexes and M 2 after rewarming predicted CPC 35 with 0% FPR and narrow 95% CIs, but with a high risk of bias. 2013 Elsevier Ireland Ltd. All rights reserved.

Article history: Received 16 March 2013 Received in revised form 13 June 2013 Accepted 23 June 2013

Keywords: Heart arrest Coma Prognostication Therapeutic hypothermia

Contents 1. 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1325 Materials and methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.1. Eligibility criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.1.1. Patient population . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.1.2. Interventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326

A Spanish translated version of the abstract of this article appears as Appendix in the nal online version at http://dx.doi.org/10.1016/j.resuscitation.2013.06.020. Corresponding author at: Department of Anaesthesiology and Intensive Care, Catholic University School of Medicine, Largo Gemelli 8, 00168 Rome, Italy. E-mail address: sandroni@rm.unicatt.it (C. Sandroni). 0300-9572/$ see front matter 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.resuscitation.2013.06.020

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

1325

3.

4.

5.

2.1.3. Outcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.1.4. Study type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.2. Search strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.3. Study selection and data extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.4. Statistical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 2.5. Evidence appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327 3.1. Study selection (Fig. 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327 3.2. Study characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327 3.3. Clinical examination (Table 2a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327 3.3.1. Brainstem reexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327 3.3.2. Motor response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1328 3.3.3. Myoclonus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1328 3.4. Electrophysiology (Table 2b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1328 3.4.1. Burst-suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1328 3.4.2. Seizures and status epilepticus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1329 3.4.3. Flat or low-amplitude EEG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1329 3.4.4. Nonreactive EEG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1329 3.4.5. EEG grading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1329 3.4.6. SSEP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1331 3.5. Biomarkers (Table 2c) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332 3.6. Imaging (Table 2d) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332 3.7. Predictors with highest specicity and narrow CIs (Table 3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1332 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.1. Clinical examination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.1.1. Brainstem reexes and motor response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.1.2. Myoclonus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.2. Electrophysiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.2.1. Burst suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.2.2. Low-amplitude or at EEG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333 4.2.3. Epileptiform activity and status epilepticus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1334 4.2.4. EEG reactivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1334 4.2.5. EEG grading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1334 4.2.6. N20 SSEP wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1334 4.3. Biomarkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1334 4.4. Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335 4.5. Self-fullling prophecy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335 4.6. Study limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336 Author contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336 Conict of interest statement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336 Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336

1. Introduction Mortality after resuscitation from cardiac arrest is high. Two thirds of initially resuscitated patients die before hospital discharge.1 Many of these deaths are due to post cardiac arrest brain dysfunction2 and in more than one-fourth of those who survive to hospital discharge brain hypoxia-ischaemia results in severe neurological impairment.3 Prediction of neurological outcome is an important component of the management of comatose resuscitated patients. In 2006, the Quality Standards Subcommittee of the American Academy of Neurology (AAN) published a review on available evidence. The document concluded that the presence of myoclonus status epilepticus on day 1, the bilateral absence of the N20 wave of somatosensory evoked potentials (SSEPs) or a blood concentration of neuron specic enolase (NSE) above 33 mcg/L at days 13, and absent pupillary and corneal reexes or a motor response no better than extension (M1 -2) at day 3 accurately predicted poor outcome, dened as death or unconsciousness after 1 month or unconsciousness or severe disability after 6 months. However, that review was based on evidence derived from patients not treated with therapeutic hypothermia (TH), which currently represents

the standard for the treatment of comatose patients resuscitated from out-of-hospital cardiac arrest.4 Evidence showing that AAN recommendations may not apply to TH-treated patients has been accumulating in the last years. Moreover, this and previous reviews did not comply with the currently recommended guidelines for data reporting in systematic reviews and meta-analysis, such as PRISMA,5 and did not adequately address some important limitations of the included studies, such as the risk of self-fullling prophecy, which is a bias present in most studies on prognostication after cardiac arrest wherein the treating physicians are not blinded to the results of the outcome predictor and use it to make a decision to withdraw treatment.6,7 The aim of the current work is to perform a new systematic review that, in comparison with previous reviews: (a) incorporates any missed studies or studies published more recently; (b) implements an improved approach for the evaluation of main sources of bias and statistical heterogeneity; (c) complies with the most recognised standards for evidence evaluation and data reporting; and, nally (d) addresses prognostication both in patients who have not been treated with TH and in TH-treated patients. In the rst part of this review we addressed prognostication in patients who have not been treated with TH.8 The present study deals with

1326

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

the predictors of neurological outcome in resuscitated comatose patient treated with TH. We reviewed the evidence using Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidelines,9 in order to provide grounds for future recommendations, as part of a staged approach. 2. Materials and methods This is a systematic review and aggregate data meta-analysis of prognostic accuracy studies. Data reporting is consistent with the recommendations included in the PRISMA statement.10 According to the PICOS template, the review question was formulated as follows: In adult patients who are comatose following resuscitation from cardiac arrest and who have been treated with TH (P), does the use of predictors based on clinical examination, electrophysiology, serum biomarkers or neuroimaging (I) allow accurate prediction of poor outcome (O)? Given the review question, the only eligible study design (S) for this review was an observational prognostic accuracy study in which a comparison (C) is made between the respective proportions of poor outcome among the patients having a positive test result and those having a negative test result. 2.1. Eligibility criteria 2.1.1. Patient population All studies on adult (16 years) patients who were comatose following resuscitation from cardiac arrest and were treated with TH were considered for inclusion. Patients dened as unconscious, unresponsive, or having a Glasgow Coma Scale score (GCS)11 8 were considered as comatose. Studies including non-comatose patients or patients in hypoxic coma from causes other than cardiac arrest (e.g., respiratory arrest, carbon monoxide intoxication, drowning, hanging) were excluded, except when a subpopulation of cardiac arrest patients could be evaluated separately. 2.1.2. Interventions Four types of outcome predictors were assessed: clinical examination, electrophysiology, biomarkers, and imaging (see the Part 1 of this review for details). Since we are interested only in the prediction of outcome in comatose patients in the acute phase after resuscitation, we included only studies where the outcome predictor was evaluated within seven days from cardiac arrest. 2.1.3. Outcome We included only prognostic accuracy studies in which the neurological outcome was described using the ve Cerebral Performance Categories (CPC)12 (see ESM Table E1) or in such a manner that an equivalent CPC could be determined. In those studies, the outcome is usually dichotomized as poor or good according to a predened CPC threshold, either CPC 45 vs. 13 or CPC 35 vs. CPC 12. We accepted both of those denitions for our review, but reported and pooled relevant results separately. When no threshold had been dened, we dichotomized outcomes as CPC 35 vs. 12. In order to calculate both the outcome variables and their condence intervals, we included only studies where the complete contingency table (i.e., the number of true/false negatives and positives for prediction of poor outcome) was reported, or where those variables could be calculated from reported data. 2.1.4. Study type Only clinical prognostic accuracy studies published in English, French, German, Italian or Spanish as full-text articles on indexed

journals were included. Reviews, case reports and studies published in abstract form were excluded. No publication date or publication status restrictions were imposed. 2.2. Search strategy We searched MEDLINE via PubMed, Scopus and the Cochrane Database of Systematic Reviews using the search strings included into the ESM Table E2. The automatic alert system of PubMed was activated to identify further studies published during the process of data extraction and analysis. The reference lists of relevant studies were scanned in order to identify other studies of interest. The last search was launched on May 15, 2013. 2.3. Study selection and data extraction For each of the four types of outcome predictors, two authors performed a blinded and independent eligibility assessment. Disagreements between reviewers were resolved by consensus. For each study included in the nal analysis, the following data were extracted: number of included comatose patients; age and gender; location of cardiac arrest; initial cardiac rhythm; type and timing of the assessed prognostic index; denition of poor outcome; number of patients with poor outcome; number of patients with true/false positive (TP/FP) and true/false negative (TN/FN) test result. In order to retrieve missing information, the authors of the papers were contacted whenever possible. 2.4. Statistical methods Statistical methods are described in detail in the extended version of the manuscript (ESM Appendix 1). Briey, for each included predictor we calculated sensitivity, specicity, false positive rate (FPR) expressed as 1-specicity, positive predictive value (PPV), expressed as TP/(TP + FP). and positive likelihood ratio (LR+), expressed as sensitivity/sensitivity/1-specicity. Pooling of the indices was made by summing up each member of the contingency table13 when there were two or more studies having similar time points and outcome denitions (i.e., CPC 35 or CPC 45). Condence intervals for pooled values of proportions were calculated using the F distribution method, according to Blyth.14 2.5. Evidence appraisal Given the absence of specic GRADE recommendations on prognostic accuracy studies, we adapted the GRADE recommendations for diagnostic accuracy studies9 to rate the quality of evidence (QOE). Evidence evaluation was performed independently by two authors. According to the GRADE methodology, the QOE started as high and was graded down based on the following factors: (1) limitations; (2) indirectness; (3) inconsistency; and (4) imprecision. Given the importance of the risk of self-fullling prophecy, limitations were graded as serious when the treating team was not blinded to the results of the predictor of poor outcome that was being studied, and very serious when the index was used for decisions of suspension of life sustaining therapies. Methodological details are provided in the extended version of this review, available as ESM, Appendix 1.

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

1327

Identification

PubMed: 977 records Scopus: 392 records Cochrane Database of systematic reviews: 11 records 1380 records 37 additional records identified through forward search

1417 records screened

Screening

1217 discarded (duplicates or excluded after title and abstract evaluation)

163 full-text excluded due to 136 3 5 5 10 1 1 2 Patient characteristics Intervention characteristics Comparison characteristics Outcome characteristics Study characteristics Data could not be extracted Language Unavailable

Eligibility

200 full-text articles assessed for eligibility

Included

37 studies included in qualitative and quantitative synthesis (meta-analysis)

Fig. 1. Flow-chart of study selection.

3. Results 3.1. Study selection (Fig. 1) The initial search produced 977 records from PubMed, 392 records from Scopus and 11 records from the Cochrane Database of Systematic Reviews. Thirty-seven additional records were identied through forward search. After duplicate removal and abstract screening, 200 studies were considered for full-text analysis. Among them, 163 were excluded because they did not full inclusion criteria. The remaining 37 studies were included in our review. Excluded studies with reasons for their exclusion are listed in the ESM Table E3.

electrophysiology, ve3337 on biomarkers, two38,39 on neuroimaging, and twelve4051 on multiple predictors. Sensitivity, FPR (1-specicity), and the quality of evidence for indices based on clinical examination, electrophysiology, biomarkers and neuroimaging are reported in Tables 2a2d. Evidence proles are reported in ESM Tables E4ad, while results of individual studies are reported in Tables E5ad. Predictors having a 0% FPR (100% specicity) and narrow 95% CIs (upper limit of 95% CI < 10%) are reported in Table 3.

3.3. Clinical examination (Table 2a) 3.3.1. Brainstem reexes Absence of pupillary light reex (PLR) on hospital admission17,42 was an inaccurate predictor of poor outcome (FPR 32%; 95% CI 1948). Conversely, at 72 or later absence of PLR was associated with an almost invariably poor outcome (see Table 2a). On a total of 41 patients with absent PLR after rewarming described in ve studies15,41,42,44,49 (QOE: from Moderate to Very low) only one recovered.41 Absent corneal reex (CR) at 72 h from cardiac arrest or later was described in three studies (QOE: Very low) and it was not

3.2. Study characteristics The characteristics of the 37 included studies (total 2403 patients) are summarised in Table 1. In 11/37 studies (30%) poor outcome was dened as vegetative state or death (CPC 45), while in the remaining 26/37 studies (70%) poor outcome was dened as severe disability, vegetative state or death (CPC 35). Four studies1518 were based on clinical examination, fourteen1932 on

1328 Table 1 Characteristics of the included studies. Author, year reference IHCA or OHCA No. of patients

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

Males, %

VF/VT %

Mean age, years [SD] or median (range) N/Aa 67 [13] 59 58 (4769) 65 (5077) 68 [ 11.5] 60 [17.5] 55 [16] 62 58 (4565) 50 [17] 57 [15] 65 (5074) (2268) 62 (4872) 56 [17] 60 (2375) 60 [13] 63 (2284) 63 (5371) 60 (5270) 63 (5171) 60 (2375) 57 [17] 59 [17] 63 (53.876) 64 [13.4] 48 [13.3] 61 (3381) 71 (5176) 57 [17.6] 65 (5475) 59 (1785) 63.7 [12.2] 58 (1984) 65 (2983) 57 (2477)

Predictor(s)

Denition of poor outcome (CPC) 45 vs. 13 35 vs. 12 35 vs. 12 35 vs. 12 45 vs. 13 35 vs. 12 45 vs. 13 35 vs. 12 45 vs. 13 35 vs. 12 35 vs. 12 35 vs. 12 45 vs. 13 35 vs. 12 35 vs. 12 35 vs. 12 45 vs. 13 35 vs. 12 35 vs. 12 35 vs. 12 35 vs. 12 35 vs. 12 35 vs. 12 45 vs. 13 45 vs. 13 35 vs. 12 35 vs. 12 35 vs. 12 35 vs. 12 45 vs. 13 35 vs. 12 35 vs. 12 35 vs. 12 35 vs. 12 45 vs. 13 35 vs. 12 35 vs. 12

Timing of outcome assessment 3 mo 6 mo Hospital discharge ICU discharge 1 mo 6 mo 6 mo Hospital discharge Hospital discharge Hospital discharge Hospital discharge Hospital discharge 6 mo 2 mo Hospital discharge 6 mo 6 mo 3 mo 6 mo 6 mo ICU discharge ICU discharge 6 mo 6 mo 6 mo 3 mo 6 mo Hospital discharge Hospital discharge 6 mo 6 mo 12 mo 6 mo 3 mo 3 mo 6 mo 6 mo

Clinical examination Al Thenayan, 200815 Bouwes, 2012a16, b Okada, 201217 Schefold, 200918 Electrophysiology Bouwes, 200919 Cloostermans, 201220 Kawai, 201121 Leary, 201022 Leithner, 201023 Mani, 201224 Oh, 201225 Rittenberger, 201226 Rundgren, 201027 Sakurai, 200628 Seder, 201029 Stammet, 200930 Tiainen, 200531 Zanatta, 201232 Biomarkers Mortberg, 201133 Oksanen, 200934 Steffen, 201035 Storm, 201236 Tiainen, 200337 Neuroimaging Mlynash, 201038 Wijman, 200939

N/A Mixed OHCA Mixed N/A Mixed Mixed Mixed N/A N/A N/A Mixed Mixed OHCA N/A Mixed OHCA N/A Mixed OHCA Mixed Mixed OHCA Mixed Mixed

37 79 66 72 77 56 26 62 112 38 55 101 95 26 82 45 30 11 31 90 97 35 36 21 22 103 391 19 54 34 43 106 111 61 53 75 30

N/A 73 80 75 71.4 68 53.8 58.1 67.9 52.6 60 54.5 71.6 76.9 65.9 66.7 86.7 58.6 68 79 78.4 66 89 N/A N/A 73.8 73.1 36.8 63 65 67.4 69.8 80.2 70.5 73.5 76 80

N/A 66 79 68 66.2 73 38.5 40.3 64.3 44 36.4 71.3 75.8 42.3 87.8 49 100 88.2 51.6 100 67 42.8 100 N/A N/A 70 76 0 N/A 53 32.6 51 59.5 65.6 62 65.3 100

PLR, CR, MR Myoclonus PLR, MR GCS N20 EEG, N20 EEG BIS N20 cEEG aEEG EEG aEEG BAEPs BIS, SR BIS N20 EEG, N20 S-100, NSE NSE NSE NSE NSE, S-100B MRI MRI PLR, MR, myoclonus, SSEP, EEG PLR, CR, MR, NSE, SSEP MRI, SSEP Myoclonus, cEEG PLR, CR, NSE, SSEP, EEG, MRI NSE, MRI Myoclonus, SE BR, MR, myoclonus, EEG, SSEP BR, MR, myoclonus, EEG, N20, NSE PLR, CR, MR, myoclonus, NSE, SSEP BIS, S-100B NSE, EEG

Multimodal prognostication Bisschops, 201140 N/A Bouwes, 2012b41 Mixed OHCA Choi, 201242 Crepeau, 201343 OHCA 44 Cronberg, 2011 Mixed 45 Kim, 2012 OHCA 46 Legriel, 2013 Mixed Rossetti, 201047 OHCA N/A Rossetti, 201248 Samaniego, 201149 Mixed Stammet, 201350 Mixed 51 Wennervirta, 2009 OHCA

Abbreviations: BAEPs: brain stem auditory evoked potentials; BIS: bispectral index; BR: brain reexes; CPC: Cerebral Performance Category; CR: corneal reex; EEG: electroencephalogram; GCS: Glasgow Coma Scale; IHCA: in-hospital cardiac arrest; MR: motor response; MRI: magnetic resonance imaging; N/A: not available; NSE: neuron-specic enolase; OHCA: out-of-hospital cardiac arrest; PLR: pupillary light response; SR: suppression ratio; SSEPs: short-latency somatosensory evoked potentials; VF/VT: ventricular brillation/ventricular tachycardia. a All patients were adults (>17 years). b Same population of the Bouwes 2012b study, but including patients with myoclonus or status myoclonus.

consistently associated with poor outcome. FPR ranged from 0% to 5%.41,44,49 3.3.2. Motor response In six studies,40,41,44,4749 presence of extensor or absent motor response to pain (M 2) after rewarming was still compatible with good outcome (FPR 1112%). In one study40 the coexistence of M 2 plus absent PLR and CR at 72 h from cardiac arrest predicted poor outcome with 0% [08] FPR. However, this nding was also used as a criterion for treatment withdrawal (QOE: Very low). 3.3.3. Myoclonus In one cohort study49 and in a case series,15 all patients with status myoclonus (dened as spontaneous, repetitive, unrelenting, generalised multifocal myoclonus) died or remained vegetative. Timing of clinical examination in those studies ranged from 1 to 144 h after cardiac arrest. However, in six studies16,26,43,4648

presence of myoclonus or status myoclonus within 72 h from cardiac arrest did not exclude a good (CPC 12) neurological outcome (FPR 5% [39]). Distribution and EEG patterns of myoclonus varied between studies (see ESM Table E6 for details).

3.4. Electrophysiology (Table 2b) 3.4.1. Burst-suppression Presence of burst-suppression on EEG recorded during TH was not associated with an invariably poor outcome (FPR 5% [114]; QOE very low). Conversely, in a single study27 no patient with a burst-suppression pattern on continuous amplitude-integrated EEG (aEEG) recorded immediately after rewarming recovered consciousness. In another study,20 presence of burst-suppression after rewarming at 48 from cardiac arrest was 100% specic of poor outcome. Denition of burst-suppression was very inconsistent among studies (see ESM Table E7A).

C. Sandroni et al. / Resuscitation 84 (2013) 13241338 Table 2a GRADE summary of ndings for predictors based on clinical examination. Timing CPC 45 vs. 13 Within 72 h At 72 h Index Status myoclonus M2 Absent PLR Absent CR Status myoclonus M2 Absent PLR Absent CR Reference Samaniego, 201149 Al Thenayan, 200815 Al Thenayan, 200815 Al Thenayan, 200815 Al Thenayan, 200815 Samaniego, 201149 Cronberg, 201144 Samaniego, 201149 Cronberg, 201144 Samaniego, 201149 Cronberg, 201144 92 [8298] 17 [830] 27 [1541] Sensitivity % [95% CI] 8 [126] FPR % [95% CI] 0 [015] PPV 86 [5798] PPV 100 [61100] PPV 100 [61100] PPV 100 [69100] 12 [331] 0 [013] 5 [025] No. of patients 44 14 6 6 8 78 74 69

1329

Quality of evidence Very low Very low Very low Very low Very low Very low Very low Very low

At 72108 h

CPC 35 vs. 12 On admission

Absent PLR M2

Okada, 201217 Choi, 201242 Okada, 201217 Rossetti, 201048 Rossetti, 201247 Rossetti, 201048 Rossetti, 201247 Bouwes, 201216 Crepeau, 201343 Legriel, 201346 Rittenberger, 201226 Rossetti, 201048 Rossetti, 201247 Bouwes, 201241 Choi, 201242 Bouwes, 201241 Bouwes, 201241 Bisschops, 201140 Bisschops, 201140 Bisschops, 201140 Schefold, 200918 Bisschops, 201140

86 [7195] 92 [7599] 56 [4766] 75 [6683] 37 [3341]

32 [1948] 53 [3668] 9 [321] 11 [423] 5 [39]

86 66 170 170 812

Very low Very low Very low Very low Very low

At 3672 h

1 BR absenta M2

72 h

Myoclonus or status myoclonus

At 72 h

Absent PLR Absent CR M2 M 2 or no PLR or no CR M 2 and no PLR and no CR

20 [1328] 26 [1737] 60 [5367] 81 [6989] 15 [726] 46 [2866] 54 [4166]

1 [07] 4 [114] 11 [716] 11 [326] 0 [08] 5 [115] 11 [326]

207 130 387 103 103 72 103

Very low Very low Very low Very low Very low Very low Very low

At 96 h At <7 days

GCS 4 Myoclonus

Abbreviations: BR: brainstem reexes; CI: condence intervals; CPC: Cerebral Performance Categories; CR: corneal reex; FPR: false positive rates; GCS: Glasgow Coma Score; M: motor response of GCS; No: number; PLR: pupillary light response; PPV: positive predictive value. a Pupillary, corneal, oculocephalic.

3.4.2. Seizures and status epilepticus In three studies24,43,47 presence of electrographic seizures during TH or after rewarming predicted poor outcome (CPC 35) with 0% FPR. However, in a recent study,46 presence of prolonged seizures, i.e. status epilepticus (SE), on continuous or intermittent EEG monitored for a median of 48 h after cardiac arrest was followed by neurological recovery (CPC 12) in two patients. In both of those patients EEG reactivity was maintained. In another study,27 an electrographic status epilepticus (ESE) evolving from burst-suppression (SB-ESE) was associated with poor outcome in 100% of cases, but an ESE evolving from a continuous aEEG background (C-ESE) was not (FPR 4% [012]). Denitions of status epilepticus were inconsistent among studies (see ESM Table E7C).

corresponding to a at or low-amplitude EEG, predicted a CPC 35 with 0% [06] FPR (QOE: Low). However, another recent study50 did not conrm this result.

3.4.4. Nonreactive EEG In three studies,43,47,48 two of which from the same group, a nonreactive EEG background after rewarming accurately predicted CPC 35 (FPR 0% [03]; QOE: Low). However, during TH this pattern was still compatible with neurological recovery (FPR 3% [011]). Furthermore, in a large cohort study16 three patients with posthypoxic myoclonus and no EEG reactivity within 72 h from cardiac arrest had a good outcome.

3.4.3. Flat or low-amplitude EEG In one study,20 a at or low-amplitude (<20 V) EEG during TH at 24 h from cardiac arrest predicted poor outcome (CPC 35) with 0% [011] FPR. In another study, however,27 a at (<10 V) aEEG recorded during TH at a median of 8 h from cardiac arrest or immediately after rewarming was often followed by recovery of awareness (FPR 46% [3259] and 5% [115] respectively). In two studies29,30 a BIS (bispectral index) value of 6 or less during TH,

3.4.5. EEG grading Unlike the ndings in the rst part of our review8 only one study43 described the application of an EEG grading in patients treated with TH. In this study, a Grade 3 EEG, dened as the presence of low voltage, burst suppression, status epilepticus, seizures, nonreactive, or alpha/theta coma, accurately predicted a poor outcome (CPC 35; FPR 0% [09]) when recorded after rewarming, while it showed a 6% [120] FPR when recorded during TH.

1330

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

Table 2b GRADE summary of ndings for predictors based on electrophysiology. Timing CPC 45 vs. 13 During TH After RW Index Reference Sensitivity % [95% CI] 37 [2254] 55 [3871] 16 [631] 18 [834] 63 [5372] FPR % [95% CI] No. of patients Quality of evidence Low Very low Very low Low Very low

EEG EEG SSEP

Burst-suppression Flat <10 V Flat <10 Va Burst-suppression Bilaterally absent N20

Rundgren, 201027 Rundgren, 201027 Rundgren, 201027 Rundgren, 201027 Choi, 201242 Cronberg, 201144 Leithner, 201023 Samaniego, 201149 Rundgren, 201027 Rundgren, 201027

0 [05] 46 [3259] 5 [115] 0 [05] 2 [08]

93 95 95 95 180

Any time

aEEG

C-ESE SB-ESE

21 [1037] 42 [2659]

4 [012] 0 [05]

95 95

Very low Low

CPC 35 vs. 12 During TH induction During TH

BAEP EEG

Absence of wave V Burst-suppression

Sakurai, 200628 Cloostermans, 201220 Kawai, 201121 Rossetti, 201247 Wennervirta, 200951 Cloostermans, 201220 Mani, 201224 Mani, 201224 Rossetti, 201247 Legriel, 201346 Rossetti, 201247 Crepeau, 201343 Crepeau, 201343 Leary, 201022 Leary, 201022 Seder, 201029 Seder, 201029 Stammet, 200930 Seder, 201029 Bouwes, 200919 Bouwes, 201241 Cloostermans, 201220 Tiainen, 200531 Rossetti, 201048 Wennervirta, 200951 Crepeau, 201343 Rossetti, 201048 Rossetti, 201247 Crepeau, 201343 Bisschops, 201140 Bouwes, 201241 Rossetti, 201048 Rossetti, 201247 Zanatta, 201232 Zanatta, 201232 Oh, 201225 Crepeau, 201343 Rittenberger, 201226 Oh, 201225 Crepeau, 201343 Stammet, 201350

56 [3178] 55 [4367]

0 [031] 5 [114]

26 133

Very low Very low

BSR 21.61 Flat or low-voltageb Epileptiform discharges Electrographic seizures Status epilepticus Nonreactive background Grade 3 EEGc BIS < 45 BIS < 30 BIS 22 BIS 6 SR 48 Bilaterally absent N20

11 [048] 40 [1964] 62 [4180] 35 [1756] 30 [1649] 41 [3053] 63 [4976] 76 [5392] 86 [7195] 40 [2657] 86 [7394] 49 [3760] 84 [7193] 28 [2234]

0 [013] 0 [011] 8 [039] 0 [022] 0 [010] 6 [121] 3 [011] 6 [120] 35 [1559] 5 [025] 6 [120] 0 [06] 6 [120] 0 [02]

30 46 38 38 61 106 115 54 62 62 83 128 83 421

Very low Low Very low Very low Very low Very low Very low Very low Very low Very low Very low Low Very low Moderate

BIS

SSEP

After RW

EEG

Epileptiform activityd Status epilepticus Nonreactive background Grade 3 EEGc Bilaterally absent N20

43 [3355] 44 [1479] 62 [5370]

12 [331] 0 [013] 0 [03]

108 30 223

Very low Low Low

SSEP

89 [6599] 42 [3648]

0 [09] 0 [04]

51 339

Low Low

MLCEPs Any time EEG

MLCEPs absent Burst-suppressione Electrographic seizures Status epilepticus No continuous normal voltage Nonreactive backgroundf BIS <5.5g

88 [47100] 63 [4281] 24 [847] 12 [620] 78 [5891] 52 [3074] 85 [6995]

0 [063] 4 [018] 0 [09] 0 [078] 0 [010] 6 [120] 17 [732]

11 55 54 101 55 54 75

Very low Very low Very low Very low Very low Very low Very low

BIS

Abbreviations: BAEP: brainstem auditory evoked potentials; BIS: bispectral index; BSR: burst suppression ratio; C-ESE: electrographic status epilepticus from continuous pattern; CI: condence intervals; CPC: Cerebral Performance Categories; EEG: electroencephalogram; FPR: false positive rate; MLCEP: middle-latency cortical somatosensory evoked potentials; RW: rewarming; SB-ESE: electrographic status epilepticus from burst-suppression; SR: suppression ratio; SSEP: somatosensory evoked potentials; TH: therapeutic hypothermia. a Initial EEG during TH, recorded at a median of 8 h from cardiac arrest. b Recorded at 24 h from cardiac arrest. Low voltage was dened as EEG activity below 20 V. c It includes low voltage, burst-suppression, seizures, status epilepticus, nonreactive, or alpha/theta coma. d It includes electrographic seizures or periodic epileptiform discharges (PEDs) or burst-suppression with sharp bursts. e The only patient with burst suppression and good outcome had this sign recorded in the induction phase of TH. f The two patients that had a good outcome had non-reactive background during TH. g The lowest peak of BIS was detected at a median time of 5 [414.5] h from cardiac arrest.

C. Sandroni et al. / Resuscitation 84 (2013) 13241338 Table 2c GRADE summary of ndings for predictors based on biomarkers. Timing Index Cutoff (g l1 ) Reference Sensitivity % [95% CI] 61 [4179] 32 [1553] 32 [2539] 20 [356] 18 [834] 80 [5993] 10 [045] 65 [4483] 50 [4357] 50 [2575] 22 [360] 53 [4660] 86 [6497] 30 [2437] FPR% [95% CI] No. of patients

1331

Quality of evidence Very low Very low Very low Very low Very low Very low Very low Very low Very low Very low Very low Very low Very low Very low

CPC 45 vs. 13 NSE At 48 h CPC 35 vs. 12 NSE At ROSC At 12 h At 24 h NSE NSE

33.0 41.5 33.0 31.2 41.0 49.6 52.4 0.180.21 33.0 4.97 25.0 33.0 44.3 54.559.3

Cronberg, 201144 Kim, 201245 Bouwes, 201241 Tiainen, 200337 Oksanen, 200934 Kim, 201245 Wennervirta, 200951 Mortberg, 201133 Tiainen, 200337 Bouwes, 201241 Rossetti, 201247 Mortberg, 201133 Tiainen, 200337 Bouwes, 201241 Oksanen, 200934 Kim, 201245 Bouwes, 201241 Storm, 201236 Wennervirta, 200951 Bouwes, 201241 Tiainen, 200337 Stammet, 201350 Samaniego, 201149 Storm, 201236 Steffen, 201035

0 [039] 0 [029] 10 [615] 4 [020] 4 [014] 0 [029] 0 [014] 0 [07] 9 [514] 7 [032] 0 [012] 6 [310] 0 [034] 6 [311]

34 35 358 35 89 34 30 66 395 31 33 397 29 373

S-100B At 2448 h At 48 h NSE NSE

S-100B At 72 h NSE

81.8 0.23 0.30 33.0 57.2 78.9

18 [1325] 22 [360] 21 [938] 75 [5390] 46 [2667] 48 [3263]

0 [02] 4 [020] 0 [07] 22 [648] 0 [028] 0 [06]

310 34 75 42 33 97

Moderate Very low Very low Very low Very low Very low

Abbreviations: CI: condence intervals; CPC: Cerebral Performance Categories; FPR: false positive rates; NSE: neuron-specic enolase; ROSC: recovery of spontaneous circulation.

3.4.6. SSEP In four studies,19,20,31,41 two of which from the same group, a bilateral absence of N20 SSEP wave during TH accurately predicted a CPC 35 (FPR 0% [02] QOE: Moderate). Absence of N20 SSEP

after rewarming was also invariably associated with poor outcome, either dened as CPC 3519,32,40,47,48 or CPC 45,31,42,44,49 except in one study,23 where one case of full neurological recovery with reappearance of the N20 SSEP wave was observed.

Table 2d GRADE summary of ndings for predictors based on imaging. Timing CPC 45 vs. 13 Median 80 h (IQR 55117) Index Reference Sensitivity % [95% CI] 90 [55100] 80 [4497] 30 [765] 77 [4695] FPR % [95% CI] No. of patients 21 21 21 22 Quality of evidence Very low Very low Very low Very low

MRI DWI or FLAIR

Extensive cortical lesion pattern Abnormalities in basal ganglia Abnormalities in brainstem ADC < 650 106 mm2 /s in >10% of brain volume

Mlynash, 201038 Mlynash, 201038 Mlynash, 201038 Wijman, 200939

9 [041] 9 [041] 0 [024] 0 [028]

At 49108 h

MRI DWI (ADC)

CPC 35 vs. 12 On admission Median 46 h (IQR 3752) Median 74 h (IQR 6186)

CT MRI MRI DWI

Lost grey/white matter interface (CT) ADC occipital cortex < 616 106 mm2 /s Abnormalities in both cortex and basal ganglia Abnormalities in both cortex and basal ganglia

Choi, 201242 Kim, 201245 Cronberg, 201144

100 [55100] 91 [7598] 58 [3380]

0 [063] 0 [024] 0 [063]

8 43 22

Very low Very low Very low

At <5 days

MRI DWI

Choi, 201242

100 [55100]

0 [063]

Very low

Abbreviations: ADC: apparent diffusion coefcient; CI: condence intervals; CPC: Cerebral Performance Categories; CT: computed tomography; DWI: diffusion weighted imaging; FPR: false positive rates; IQR: interquartile range; MRI: magnetic resonance imaging.

1332

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

Table 3 Predictors of poor outcome with 0% FPR and upper 95% CI limit <10% (for comments see text). Timing Index Sensitivity % [95% CI] FPR % [95% CI] LR+ [95% CI] No. of patients with positive test 14 7 16 No. of studies Used for WLST Quality of evidence Reference

CPC 45 vs. 13 During TH After RW Any time CPC 35 vs. 12 During TH During TH (at 24 h) After RW After RW After RW (at 48 h) After RW (at 72 h)

Burst-suppression Burst-suppression SB-ESE


a

37 [2254] 18 [834] 42 [2659]

0 [05] 0 [05] 0 [05]

42 [3678] 22 [1379] 49 [3794]

1 1 1

No No No

Low Low Low

27 27 27

Bilaterally absent N20 S-100B 0.180.21 mcg/L Bilaterally absent N20b Nonreactive background NSE 81.8 g l1 S-100B 0.3 g l1 NSE 78.9 g l1 M 2 and no PLR and no CR

28 [2234] 65 [4483] 42 [3648] 62 [5370] 18 [1325] 21 [938] 48 [3263] 15 [726]

0 [02] 0 [07] 0 [04] 0 [03] 0 [02] 0 [07] 0 [06] 0 [08]

13 [532] 22 [3156] 15 [544] 33 [7163] 56 [3909] 18 [1304] 52 [3828] 11 [1190]

63 17 109 76 29 7 21 10

4 2 5 3 1 1 1 1

Yes (2/4) N/A Yes (4/5)c Yes (1/3) No N/A Yes Yes

Moderate Very low Low Low Moderate Very low Very low Very low

19,20,31,41

33,37

32,40,41,47,48

43,47,48

41 50 35 40

Abbreviations: CI: condence intervals; CPC: Cerebral Performance Categories; CR: corneal reex; FPR: false positive rates; LR+: positive likelihood ratio; M: motor response of GCS; No: number; NSE: neuron-specic enolase; PLR: pupillary light response; RW: rewarming; SB-ESE: electrographic status epilepticus evolving from burst-suppression; TH: therapeutic hypothermia; WLST: withdrawal of life sustaining treatment. a The pattern was present during TH and immediately after regaining normothermia. b In studies where poor outcome was dened as CPC 45, FPR of bilaterally absent N20 SSEP after RW was 1% [07] because one patient had a good outcome. c The total number of independent studies documenting N20 SSEP was 8 (1 of 9 studies is reported twice because it included both SSEPs during TH and after rewarming).

3.5. Biomarkers (Table 2c) Only one small study44 investigated on the predictive value of biomarkers for poor outcome dened as CPC 45. In this study, a NSE above 33 mcg/L at 48 h predicted poor outcome with 0% FPR (95% CI 039). In papers where poor outcome was dened as CPC 35, the threshold for 0% FPR varied between 49.6 mcg/L and 52.4 mcg/L at 24 h from cardiac arrest,45,51 between 25 mcg/L and 81.8 mcg/L at 48 h,36,37,41,45,51 and between 57.2 mcg/L and 78.9 mcg/L at 72 h.35,36 In two papers33,37 a S-100B value of 0.18 and 0.21 mcg/L at 24 h predicted poor outcome (CPC 35) in 100% of cases, but not at 48 h, when the threshold for 0% FPR was 0.3 mcg/L.50 3.6. Imaging (Table 2d) In a single study42 including 20 patients, eight of whom underwent a brain CT scan at 1 h after cardiac arrest, presence of diffuse brain swelling with loss of grey/white matter interface predicted poor outcome (CPC 35) with 0% [063] FPR. This study, however, has a limited generalizability, since it included only patients with cardiac arrest caused by drowning. All studies on MRI were carried out after rewarming. Presence of diffuse ischaemic lesions involving both cortex and deep grey matter nuclei,42,44 detected using diffusion weighted imaging (DWI) MRI within ve days from CA predicted poor outcome (CPC 35) with 0% FPR, while the presence of ischaemic lesions in either of these structures at a median of 80 h from CA did not.38 The presence of isolated lesions in the brainstem also predicted CPC 35 with 0% FPR, but only three patients with this sign were described38 (95% CI 024). Quantitative evaluation of diffusion with MRI using ADC (absolute diffusion coefcient) was described in two studies.39,45 In one study,39 the presence of more than 10% of brain volume with an

ADC <650 106 mm2 /s at 49108 h from cardiac arrest predicted death or persistent vegetative state with 0% [028] FPR. In the other study45 the area with highest discriminatory power was identied in the occipital cortex, where an ADC < 616 106 mm2 /s at a median of 45.8 h from cardiac arrest predicted death, vegetative state or severe neurological disability with 0% [024] FPR (see Table 2d).

3.7. Predictors with highest specicity and narrow CIs (Table 3) Table 3 includes the list of predictors with 0% FPR and upper 95% CI limit below 10%. For prediction of poor outcome dened as CPC 45, the predictors fullling these characteristics were the presence of burst-suppression after rewarming (QOE: Low) and an SB-ESE during TH and after rewarming (QOE: Low). Both these results came from a single study on aEEG and the number of patients showing this sign was very small. For prediction of poor outcome dened as CPC 35, the list of predictors included (1) a bilaterally absent N20 SSEP wave during TH (4 studies; QOE: moderate) or after rewarming (5 studies; QOE: Low); (2) a nonreactive background on EEG after rewarming (3 studies; QOE: Low); (3) a combination of absent PLR and CR plus a motor score 12 after rewarming (1 study; QOE: Very low); and (4) a S-100B > 0.180.21 mcg/L at 24 h, a NSE > 81.8 mcg/L or a S-100B 0.3 at 48 h, and an NSE >78.9 mcg/L at 72 h from cardiac arrest (5 studies; QOE: from moderate to very low). The presence of electrographic seizures during TH was also associated to a CPC 35 with 0% FPR24,47 but it was not included in this list, since the presence of status epilepticus during TH did not predict an invariably poor outcome. In 6/14 studies listed in Table 3 the investigated predictor was used as a criterion for decisions regarding withdrawal of lifesustaining treatment (WLST).

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

1333

4. Discussion 4.1. Clinical examination 4.1.1. Brainstem reexes and motor response In almost all studies included in our review, predictors based on clinical examination were recorded either on admission or at 36 h or more from cardiac arrest, after the end of TH. Use of clinical examination in patients undergoing therapeutic hypothermia is hampered by the interference from hypothermia itself and from sedatives or muscle relaxants used to maintain it. However, even when performed after TH, clinical examination could have been still affected by sedation, rstly because analgesia/sedation is often continued after rewarming for various reasons (e.g., to facilitate mechanical ventilation or to treat seizures or myoclonus),26,49 and secondly because even after stopping sedative drugs their residual effect may persist, because hypothermia reduces drug clearance.5257 In our review, absence of PLR after rewarming was the most accurate predictor among those based on clinical examination. Among 41 patients showing this sign, only one recovered. Moreover, in two studies not included in our review58,59 absent PLR was invariably associated with death or poor neurological outcome. CR and motor response were less reliable predictors than PLR in our review. This may partly be because both of those signs are likely to be affected by the residual effects of neuromuscular blocking drugs unlike the PLR, which involves the non-striated ciliary muscle as the effector. In one study49 included in our review, an absent CR and a M 2 at 72 h or later showed a 5% and 11% FPR, respectively, in patients rewarmed from TH who had received sedative drugs 12 h before neurological examination, while the FPR was 0% in patients who did not receive sedation. This was not observed for an absent PLR, which in that study predicted poor outcome with 0% FPR in both sedated and non-sedated patients. However, even in patients not treated using TH, the predictive value of CR and motor response was lower than that of PLR.8 4.1.2. Myoclonus Clinical and electrophysiological characteristics of myoclonus varied widely in studies included in our review (see Table E6 for details). Some studies15,49 described status myoclonus or myoclonus status epilepticus (MSE) dened as a spontaneous repetitive unrelenting and generalised myoclonus.60 Another study26 dened MSE as the presence of more than 30 min of myoclonic jerks time locked with either bursts in a burst suppression pattern or generalised periodic epileptiform discharges on EEG. Others47,48 reported multifocal prolonged myoclonus with no specic denition but similar to the previously described status myoclonus or MSE. The associated EEG activity, when reported, was mainly epileptiform. According to recent recommendations,61 generalised myoclonus appearing during the rst 24 h after cardiac arrest is usually, but not always,62 a sign of unfavourable prognosis in patients treated with TH. Indeed, in studies performed in patients not treated with TH presence of myoclonus6365 (mostly generalised and multifocal) in the rst 2448 h was associated almost invariably with a poor outcome,8 except in a few case reports.6669 In TH-treated patients detection of myoclonus during the rst 24 h is difcult, due to the combined presence of neuromuscular blocking and sedation.49,70 In the present review only one study26 focused on myoclonic status epilepticus occurring within the rst 24 h. In that study MSE was not consistently generalised and all patients showing that sign had a poor outcome. No single specic feature of myoclonus was consistently predictive of poor outcome in this review (see Table E6). In particular, multifocal distribution, prolonged duration or

absence of EEG reactivity were still compatible with neurological recovery,16,47,48 although the outcome was generally worse with status myoclonus.15,16,26,49 4.2. Electrophysiology Using EEG as a predictor in resuscitated comatose patients may have some limitations. Like clinical examination, the EEG is prone to interference from both sedation and hypothermia itself in patients treated with TH after cardiac arrest. Moreover, interpreting the EEG after resuscitation requires the analysis of a large and continuous ow of data during several hours. The techniques of EEG data acquisition used by the authors of studies in our review varied (see also Table E7A). Some20 have automatically sampled 5 min of EEG every hour for the rst 24 h, others46,47 analysed a mix of continuous and discontinuous (30-min) EEG, others51 analysed the median values of EEG-derived variables over an entire 24-h period. Inevitably, sampling and use of signal processing (e.g. BIS) implies a certain loss of information. Finally, the EEG pattern immediately after resuscitation is not stable, but it shows a complex combination of various patterns whose evolution is not entirely known. For this reason, the predictive value of EEG can be inuenced by timing of recording. 4.2.1. Burst suppression Similarly to that observed in patients not treated with TH,8 the presence of a burst suppression pattern on EEG was not invariably predictive of a poor outcome. In patients with favourable outcome, burst suppression may occur during TH as a transient pattern, which usually disappears shortly after rewarming.20,27 This evolution from burst suppression towards a continuous pattern during neurological recovery early after the arrest was already described in the pre-hypothermia era.71 The denition of burst-suppression was inconsistent among studies included in our review (see ESM Table E7A for details) and it may have inuenced the relevant observed predictive value of this index. In fact, the study which used the most restrictive denition27 had 37% sensitivity and 0% FPR, another with a less restrictive denition20 had 55% sensitivity and 4% FPR, and the one with the least restrictive denition47 had a 79% sensitivity and 7% FPR. A more denitive evaluation of the predictive value of burst suppression will need a consistent denition of discontinuous EEG backgrounds. A recent consensus document from the American Clinical Neurophysiology Society72 could help to standardise the EEG terminology in critical care patients. 4.2.2. Low-amplitude or at EEG During TH, presence of a at or low-amplitude EEG (corresponding either to a at amplitude-integrated EEG27 or to a BIS 629,30,50 ) recorded during TH was not 100% accurate in predicting poor outcome. Similarly to burst suppression pattern, low EEG amplitude can be observed in the rst hours after resuscitation in patients with subsequent EEG and clinical recovery.20,27,71,73 In TH-treated patients, the combined effect of hypothermia and profound sedation can contribute to this reversible EEG depression. At normothermia, a at aEEG was associated with 0% FPR in one study,44 while in another study27 three patients having this sign subsequently regained consciousness (FPR 5% [115]). However, all these three false positives had a at EEG recorded immediately after rewarming and stopping of sedative drugs, whose effects may have still been present at the time of recording. In summary, the presence of a at or low-amplitude EEG during TH or after rewarming is not consistently associated to a poor outcome. Its predictive value may be affected by factors like timing of recording and interference from sedatives and body temperature.

1334

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

4.2.3. Epileptiform activity and status epilepticus Epileptiform activity is common following cardiac arrest.74 The simplest manifestation is presence of epileptiform discharges (EDs) which consist of spikes, polyspikes or sharp waves followed or not by a slow wave.72 EDs may occur independently and randomly or periodically (PEDs24 or PDs72 ). Electrographic seizures (ESz) consist of EDs occurring repetitively and continuously for at least 10 s.24,26 The ESM Table E7B provides a comparison of ESz denitions used in papers included in this review. A prolonged (>30 min) continuous or recurrent series of electrographic seizures is generally named electrographic status epilepticus (ESE), although denitions vary among studies26,27,70,75 in terms of minimal duration and type of electrical activity (see ESM Table E7C for details). Despite those differences, presence of status epilepticus on EEG recorded either during TH or after rewarming was almost invariably associated to poor outcome in our review. An important exception was an ESE evolving from a continuous EEG background (C-ESE). In a single study,27 two of ten patients with C-ESE recovered consciousness (CPC 23) as opposed to none of sixteen patients with ESE evolving from burst-suppression (SBESE). In a case series from Rossetti et al. not included in this review,75 six comatose patients treated with TH and showing status epilepticus from 2 to 9 days after resuscitation from cardiac arrest regained consciousness, and four had a CPC 12 at six months. In that paper, status epilepticus was dened as rhythmic focal or generalised EDs, or periodic or rhythmic evolving waves lasting >5 min. Unlike other patients within their original cohorts, all of these patients had intact brain stem reexes and showed EEG reactivity. All these results suggest that status epilepticus in postanoxic patients is a heterogeneous condition which may include different variants with possibly different prognosis, and that other clinical and EEG signs should be evaluated to improve its interpretation. Although there is a clear association between seizures or status epilepticus and poor neurological outcome in resuscitated comatose patients, it is not clear whether epileptiform activity in those patients is just a marker of irreversible postanoxic brain injury or it contributes to poor outcome by causing direct or indirect neuronal damage.76,77 In this last case, seizure detection and treatment may have a potential additional clinical benet. 4.2.4. EEG reactivity EEG reactivity is tested by assessing reproducible changes in amplitude or frequency of EEG background following external stimuli, such as tactile or nociceptive stimulation, auditory stimuli (clapping, voice sounds) or eye opening. Absence of EEG reactivity both during TH and after rewarming predicted poor outcome with 100% specicity in two studies from the same group.47,48 In another study40 an unreactive EEG recorded within 14 days after cardiac arrest in 13 resuscitated patients was associated with an invariably poor outcome. However, a recent cohort study did not conrm these results for patients undergoing TH.43 Moreover, in a case series16 three resuscitated comatose patients with absent EEG reactivity and clinical myoclonus had a good outcome. Evidence about EEG reactivity in postanoxic coma is still limited and warrants further investigation. Limitations of this index include being both operator dependent and non-quantitative, and lacking standardisation. The criteria for eliciting and recording EEG reactivity have been outlined in a recent consensus document.72 4.2.5. EEG grading Only one study43 reported EEG grading for prognostication. The classication that study adopted differed from the general scheme of the classical EEG grading systems (see ESM Table E8Comparison of EEG Grading systems in Part 1 of this review). The EEG patterns classied as Grade 3 in that study showed a 0% FPR after rewarming.

Those patterns were described separately in other studies included in this review where, in at least one case, they did not conrm 0% FPR.27 Moreover, one of those EEG patterns, alpha-theta coma, did not demonstrate 100% specicity in patients not treated using normothermia (see Table 2a in the rst part of this review8 ).

4.2.6. N20 SSEP wave The bilateral absence of N20 wave of short-latency somatosensory evoked potentials was among the most accurate predictors of poor outcome in patients treated with TH. In studies included in our review, no patient whose N20 SSEP wave was bilaterally absent during TH recovered, and among 538 patients studied with SSEPs after rewarming there was only one case of false positive result.23 SSEP were early predictors of poor outcome, a feature demonstrated in patients not treated with TH as well.8 In comatose patients treated with TH, SSEPs compare favourably with clinical examination and EEG, being resistant to the effects of both sedative drugs and mild hypothermia.31,78,79 However, in the ICU environment evoked potentials are prone to electrical interference coming from muscular artefacts and electrical equipment, which is the most important cause of interobserver variation in the interpretation of the median nerve SSEP in post-anoxic comatose patients.80 For this reason, utmost care should be taken to maximise signal/noise ratio when recording SSEPs in these patients. When noise makes unequivocal detection of N20 impossible, the most correct classication for this result should be indeterminate rather than absent N20.41 Like other predictors, SSEPs are also prone to the risk of selffullling prophecy. In our review, among all predictors, absence of N20 SSEP wave was the one most commonly used for treatment decisions (see ESM Table E8). One observational study81 found that results of SSEPs are more likely to inuence physicians and families decision to withdraw life-sustaining therapies than those of clinical examination or EEG.

4.3. Biomarkers Serum biomarkers NSE and S-100B have important theoretical advantages, such as ease of sampling, quantitative results, and likely independence from the effects of sedative drugs. However, similarly to our previous ndings in patients not treated using TH,8 their thresholds for prediction of poor outcome with 0% FPR in patients treated with TH varied largely. For NSE, that threshold ranged from 44.3 to 81.8 g l1 at 48 h, and from 57.2 to 78.9 g l1 at 72 h, but in a very recent cohort study,50 values close to 100 g l1 at 48 h were still compatible with good outcome. S-100B values for 0% FPR ranged from 0.18 at 24 h and 0.30 at 48 h (see Table 2c). One major reason for this variability may be the presence of extracerebral sources of biomarkers. NSE values are markedly increased in the presence of haemolysis because red blood cells contain NSE. Other less common sources of NSE include neuroendocrine tumours and small-cell lung carcinoma. S-100B is contained in muscle and adipose tissue,82 therefore its levels could be increased by a thoracic trauma caused by prolonged CPR. Another source of variability for biomarkers values is the use of heterogeneous immunoenzyme measurement techniques. A third issue in the interpretation of biomarker results could be our incomplete knowledge of the kinetics of their blood concentrations in the rst few days after cardiac arrest. It is likely that the best threshold for outcome prediction with those biomarkers actually varies over time. Finally, serum concentrations of biomarkers are per se continuous variables, which limits their applicability for predicting a dichotomous outcome as CPC, especially when a threshold for 0% FPR is required.

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

1335

4.4. Imaging The main CT nding of anoxic-ischaemic cerebral insult is brain swelling, which appears as a reduction of ventricles and sulci and an attenuation of the greywhite matter interface. In a single study42 included in our review, the presence of cerebral oedema on brain CT scan, quantitatively measured as a CT density ratio of grey to white matter below 1.22 measured at the basal ganglia level, was associated with death or vegetative state at discharge in 100% of cases. This study included only 20 patients and had a limited generalizability since it was made on victims of cardiac arrest caused by drowning. Its ndings will need conrmation from larger cohorts. MRI can be used to detect the presence of ischaemic brain lesions in resuscitated comatose patients. In a paper included in our review,38 presence of moderate to severe abnormalities on DWI or FLAIR sequences in either cortical or deep grey matter at a median of 80 h (3.5 days) from cardiac arrest correctly predicted poor outcome in all but three patients. Specicity increased to 100% when patients evaluated within 14 h were excluded. The authors also performed quantitative MRI measurements using ADC and found that poor outcome patients exhibited a nadir in ADC values at 35 days after cardiac arrest, which therefore appeared to be the optimal time window for prognostication. Indeed, another study39 from the same group showed that a global reduction of ADC values predicted poor outcome with 100% specicity within that time window. Two other studies42,44 included in our review42,44 showed that presence of extensive DWI changes in both cortical and deep grey matter nuclei within 5 days from cardiac arrest were 100% specic for poor outcome. In summary, detection of diffuse brain cytotoxic oedema using MRI DWI at 35 days after cardiac arrest can accurately predict poor neurological outcome after cardiac arrest. Being a qualitative technique, DWI is prone to interobserver variability, but it can be standardised using quantitative methods like ADC. ADC measurements, however, at present require an off-line data analysis with dedicated software and are not universally available. Brain MRI in resuscitated comatose patients has mainly been used as a research rather than a prognostication tool and it has not attained a widespread use yet. Relatively long measurement times and lack of bedside availability may limit MRI use in the most unstable resuscitated patients.

patients resuscitated from cardiac arrest and treated using TH were given a poor prognosis before the end of TH. At present, several authors agree that in patients treated with TH the time to prognostication should be delayed beyond 72 h after rewarming,44,8486 especially with respect to clinical examination. However, there is no denite evidence on how long this should be in order to avoid missing cases of late recovery. In two studies included in our review,15,26 recovery of consciousness occurred up to 6 and 25 days from cardiac arrest, respectively. In a recent case report on post-cardiac arrest myoclonus,87 it occurred at 14 days. Prevention of self-fullling prophecy bias would require blinding of test results to the treating team and providing sufciently prolonged life support in patients who do not recover consciousness after resuscitation and rewarming. Both those tasks are difcult to accomplish. Some predictors, such as results of clinical examination, cannot be concealed to the treating team. Others, such as an EEG, should not be concealed as they can reveal the presence of potentially treatable complications, like seizures. In some institutions, having a dedicated investigator not involved in patient management who will ensure blinding of collected data may not be feasible.30 On the other hand, indenite supportive care in potentially hopeless patients raises both ethical and nancial concerns. A potential source of data, however, could be represented by those familial or society contexts where withdrawal of life support is not accepted.88 In 1998 a cross-sectional interview in outpatient practices of three university hospitals in USA revealed that even in a theoretical scenario where there was no hope of waking up, 1520% of patients would still choose to have aggressive supportive care maintained.89 Data from studies on patients resuscitated from cardiac arrest in Seattle, Washington, show that about 20% of those still unconscious at 23 days have full supportive care maintained indenitely.90 A long-term assessment of patients with predicted poor prognosis in a prospective study may therefore be feasible, at a cost of a likely slow patient recruitment. Another way of limiting the risk of falsely pessimistic predictions would be using a multimodal approach. Combining predictors of poor outcome seems to be the most logical solution to reduce the risk of false positives even though this may reduce sensitivities. Unfortunately, only a few studies40,48,50 to date have evaluated this multimodal approach, mainly deriving results post hoc without a prospective validation. 4.6. Study limitations

4.5. Self-fullling prophecy Predictors of poor outcome in comatose patients resuscitated from cardiac arrest are prone to self-fullling prophecy.6,7 In our review, only 4/37 studies (11%) two of which from the same group reported blinding of the treating team from the results of the investigated predictor. In two of these studies,19,41 results of the predictor (absence of N20 SSEP wave) recorded during TH were not disclosed, but if patients remained comatose after rewarming, a second SSEP was performed and results were disclosed to the treating team, who used this information for treatment decisions. A treatment suspension policy was reported in 22/37 studies (54%). Nine of those studies the treatment suspension policy was based, at least in part, on one or more of the investigated predictors (see ESM Table E8). In two studies,39,48 the 2005 AAN guidelines60 were used for treatment decisions, although these guidelines were based on evidence from patients not treated with TH. Treatment limitations were applied at a minimum of 3 days or less from cardiac arrest in 8 studies and from 3 to 7 days in 6 studies, while in the remaining 23 studies the minimal duration of life support measures was not reported. One study43 reported suspension of support as early as day 2, still during TH, in three patients upon family request. One recent observational study83 also documented that 18/49 (37%) Our review has several limitations. Firstly, the lack of specic GRADE guidelines for evaluation of prognostic accuracy studies required us to adapt the GRADE guidelines for diagnostic studies. Some of our choices, such as assigning a serious limitation to studies that lacked blinding, may be considered arbitrary. However, lack of blinding and the consequent risk of self-fullling prophecy have long been recognised as a major limitation of prognostic accuracy studies in post-cardiac arrest patients.7,91 Secondly, studies included in our review did not have a consistent timing of outcome measurement, since we chose to include studies regardless of the length of their follow-up. In our review, prediction of poor outcome refers to a period ranging from discharge from intensive care unit to six months, even if all but one study on the two most robust predictors in this review SSEPs and EEG reactivity measured the outcome at least one month after cardiac arrest. In studies where outcome evaluation was undertaken too early the number of patients with poor outcome could have been overestimated, since patients assigned CPC 3 may further improve in the rst few months after cardiac arrest.46,65 Thirdly, most predictors were documented in only one or two studies and their reproducibility needs to be veried in further studies. A publication bias cannot therefore be excluded for these predictors. Fourthly, we included both studies

1336

C. Sandroni et al. / Resuscitation 84 (2013) 13241338

with CPC 35 and studies with CPC 45 in our review and described those outcomes separately. This may be seen as an inconsistency. However, there is no general consensus on what represents a poor neurological outcome after resuscitation. While the majority (70%) of studies included in this review dened poor outcome as CPC 35, the opposite was true for studies performed on patients not treated with TH,8 where poor outcome was dened as CPC 45 in 76% of cases. Ideally, both outcomes should have been reported for each predictor. However, this would have required access to original patient data, which was beyond the scope of this aggregated data meta-analysis. 5. Conclusions

Revision of the manuscript: Clifton Callaway, Giacomo Della Marca, Michael Kuiper. Data analysis: Fabio Cavallaro, Claudio Sandroni. Statistical revision: Alessio Farcomeni. Conict of interest statement Claudio Sandroni, Fabio Cavallaro, Clifton Callaway, Sonia DArrigo, Tommaso Sanna, Michael Kuiper, Matteo Biancone Giacomo Della Marca and Alessio Farcomeni have no conicts of interest to declare. Jerry Nolan is Editor-in-Chief of Resuscitation. Acknowledgements

Our analysis identied a series of early predictors of poor neurological outcome with 100% specicity and narrow CIs in comatose patients resuscitated from cardiac arrest and treated using TH. For poor outcome dened as CPC 45 these included presence of either burst-suppression or electrographic status epilepticus evolving from a burst-suppression anytime (QOE: Low). However, these predictors were described in a small number of patients in a single study, which severely limits the relevance of these ndings. Moreover, their use appear problematic, due to the inconsistent denitions of both burst-suppression and status epilepticus (see Table E7). For poor outcome dened as CPC 35, these included a bilaterally absent N20 SSEP wave during TH (4 studies; QOE: moderate) or after rewarming (5 studies; QOE: Low), a nonreactive EEG background after rewarming (3 studies; QOE: Low) and a combination of absent pupillary light and corneal reexes plus a motor response no better than extension (M 2) (1 study; QOE: very low) after rewarming. An S-100B > 0.180.21 mcg/L at 24 h, a NSE > 81.8 mcg/L or a S-100B 0.3 at 48 h, and an NSE >78.9 mcg/L at 72 h also predicted a CPC 35 with narrow 95% CIs (5 studies; QOE: from moderate to very low) but no consistent threshold for 0% FPR could be identied. All these predictors had important limitations, the most important being the lack of blinding in included studies and the frequent use of the investigated predictor to support decisions of WLST, with a consequent risk of self-fullling prophecy. This occurred in the single study on the combination of absent PLR and CR plus M 2, in 1/3 studies documenting a nonreactive EEG background and in 6/8 studies documenting absence of N20 SSEP wave. Despite its limitations, bilateral absence of the N20 SSEP wave appears as the most reproducible predictor with 0% FPR. Although not attaining 0% FPR, status myoclonus and status epilepticus showed a high specicity and can be considered as important predictors, but were both inconsistently described in literature and a common denition based on consensus is warranted. PLR after rewarming also showed a FPR close to 0% but its sensitivity was low. Given the paucity of robust univariate predictors with 100% specicity, predicting outcome on the basis of a single index appears unsafe. An integrated approach using a combination of predictors along with a careful evaluation of all available clinical information at present is probably the best strategy for early prognostication after cardiac arrest. Author contributions Study design: Claudio Sandroni, Fabio Cavallaro, Jerry Nolan. Data search and collection, grading: Matteo Biancone, Clifton Callaway, Fabio Cavallaro, Sonia DArrigo, Michael Kuiper, Claudio Sandroni, Tommaso Sanna. Drafting of the manuscript: Claudio Sandroni, Fabio Cavallaro, Jerry Nolan (Manuscripts body text); Matteo Biancone, Sonia DArrigo, Giacomo Della Marca (Tables).

We gratefully thank the following investigators for having provided original data from their studies: 1. Dr. Amy Z. Crepeau, Department of Neurology, Mayo Clinic, Rochester, MN; 2. Dr. Stephane Legriel, Intensive Care Department, CH Versailles Site Andr Mignot, Le Chesnay, France; 3. Dr. Pascal Stammet, Department of Anaesthesia and Intensive Care, Centre Hospitalier, Luxembourg; 4. Dr. Michael Mlynash, Stanford Stroke Center, Department of Neurology and Neurological Sciences, Stanford University, Palo Alto, CA, USA; 5. Dr. Kyu Nam Park, Department of Emergency Medicine, Seoul St Marys Hospital, Seoul Korea; 6. Dr. Andrea Rossetti, Department of Clinical Neurosciences, University Hospital and Faculty of Biology and Medicine, Lausanne, Switzerland; 7. Dr. G. Bryan Young, FRCPC, Department of Clinical Neurological Sciences, London Health Sciences Centre, London, Ontario, Canada. We also gratefully thank Dr. Lorenzo Compieta for providing statistical advice during data analysis. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.resuscitation.2013.06.020. References
1. Nolan JP, Laver SR, Welch CA, Harrison DA, Gupta V, Rowan K. Outcome following admission to UK intensive care units after cardiac arrest: a secondary analysis of the ICNARC Case Mix Programme Database. Anaesthesia 2007;62:120716. 2. Laver S, Farrow C, Turner D, Nolan J. Mode of death after admission to an intensive care unit following cardiac arrest. Intensive Care Med 2004;30:21268. 3. Stiell IG, Nichol G, Leroux BG, et al. Early versus later rhythm analysis in patients with out-of-hospital cardiac arrest. N Engl J Med 2011;365:78797. 4. Nolan JP, Morley PT, Hoek TL, Hickey RW. Advancement Life support Task Force of the International Liaison committee on R. Therapeutic hypothermia after cardiac arrest. An advisory statement by the Advancement Life support Task Force of the International Liaison committee on Resuscitation. Resuscitation 2003;57:2315. 5. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009;339:b2535. 6. Nolan JP, Neumar RW, Adrie C, et al. Post-cardiac arrest syndrome: epidemiology, pathophysiology, treatment, and prognostication. A Scientic Statement from the International Liaison Committee on Resuscitation; the American Heart Association Emergency Cardiovascular Care Committee; the Council on Cardiovascular Surgery and Anesthesia; the Council on Cardiopulmonary, Perioperative, and Critical Care; the Council on Clinical Cardiology; the Council on Stroke. Resuscitation 2008;79:35079. 7. Geocadin RG, Peberdy MA, Lazar RM. Poor survival after cardiac arrest resuscitation: a self-fullling prophecy or biologic destiny. Crit Care Med 2012;40:97980.

C. Sandroni et al. / Resuscitation 84 (2013) 13241338 8. Sandroni C, Cavallaro F, Callaway CW, et al. Predictors of poor neurological outcome in adult comatose survivors of cardiac arrest: a systematic review and meta-analysis. Part 1: Normothermia. Resuscitation 2013;84: 131023. 9. Schunemann HJ, Oxman AD, Brozek J, et al. Grading quality of evidence and strength of recommendations for diagnostic tests and strategies. BMJ 2008;336:110610. 10. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ 2009;339:b2700. 11. Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A practical scale. Lancet 1974;2:814. 12. Brain Resuscitation Clinical Trial I Study Group. Randomized clinical study of thiopental loading in comatose survivors of cardiac arrest. N Engl J Med 1986;314:397403. 13. Deeks JJ. Systematic reviews of evaluations of diagnostic and screening tests. In: Egger M, Smith GD, Altman DG, editors. Systematic reviews in health care: meta-analysis in context. London: BMJ Books; 2001. p. 265. 14. Blyth CR. Approximate binomial condence limits. J Am Stat Assoc 1986;81:84355. 15. Al Thenayan E, Savard M, Sharpe M, Norton L, Young B. Predictors of poor neurologic outcome after induced mild hypothermia following cardiac arrest. Neurology 2008;71:15357. 16. Bouwes A, van Poppelen D, Koelman JH, et al. Acute posthypoxic myoclonus after cardiopulmonary resuscitation. BMC Neurol 2012;12:63. 17. Okada K, Ohde S, Otani N, et al. Prediction protocol for neurological outcome for survivors of out-of-hospital cardiac arrest treated with targeted temperature management. Resuscitation 2012;83:7349. 18. Schefold JC, Storm C, Kruger A, Ploner CJ, Hasper D. The Glasgow Coma Score is a predictor of good outcome in cardiac arrest patients treated with therapeutic hypothermia. Resuscitation 2009;80:65861. 19. Bouwes A, Binnekade JM, Zandstra DF, et al. Somatosensory evoked potentials during mild hypothermia after cardiopulmonary resuscitation. Neurology 2009;73:145761. 20. Cloostermans MC, van Meulen FB, Eertman CJ, Hom HW, van Putten MJ. Continuous electroencephalography monitoring for early prediction of neurological outcome in postanoxic patients after cardiac arrest: a prospective cohort study. Crit Care Med 2012;40:286775. 21. Kawai M, Thapalia U, Verma A. Outcome from therapeutic hypothermia and EEG. J Clin Neurophysiol 2011;28:4838. 22. Leary M, Fried DA, Gaieski DF, et al. Neurologic prognostication and bispectral index monitoring after resuscitation from cardiac arrest. Resuscitation 2010;81:11337. 23. Leithner C, Ploner CJ, Hasper D, Storm C. Does hypothermia inuence the predictive value of bilateral absent N20 after cardiac arrest. Neurology 2010;74:9659. 24. Mani R, Schmitt SE, Mazer M, Putt ME, Gaieski DF. The frequency and timing of epileptiform activity on continuous electroencephalogram in comatose post-cardiac arrest syndrome patients treated with therapeutic hypothermia. Resuscitation 2012;83:8407. 25. Oh SH, Park KN, Kim YM, et al. The prognostic value of continuous amplitude-integrated electroencephalogram applied immediately after return of spontaneous circulation in therapeutic hypothermia-treated cardiac arrest patients. Resuscitation 2013;84:2005. 26. Rittenberger JC, Popescu A, Brenner RP, Guyette FX, Callaway CW. Frequency and timing of nonconvulsive status epilepticus in comatose post-cardiac arrest subjects treated with hypothermia. Neurocrit Care 2012;16:11422. 27. Rundgren M, Westhall E, Cronberg T, Rosen I, Friberg H. Continuous amplitudeintegrated electroencephalogram predicts outcome in hypothermia-treated cardiac arrest patients. Crit Care Med 2010;38:183844. 28. Sakurai A, Kinoshita K, Moriya T, et al. Reduced effectiveness of hypothermia in patients lacking the wave V in auditory brainstem responses immediately following resuscitation from cardiac arrest. Resuscitation 2006;70:528. 29. Seder DB, Fraser GL, Robbins T, Libby L, Riker RR. The bispectral index and suppression ratio are very early predictors of neurological outcome during therapeutic hypothermia after cardiac arrest. Intensive Care Med 2010;36: 2818. 30. Stammet P, Werer C, Mertens L, Lorang C, Hemmer M. Bispectral index (BIS) helps predicting bad neurological outcome in comatose survivors after cardiac arrest and induced therapeutic hypothermia. Resuscitation 2009;80:43742. 31. Tiainen M, Kovala TT, Takkunen OS, Roine RO. Somatosensory and brainstem auditory evoked potentials in cardiac arrest patients treated with hypothermia. Crit Care Med 2005;33:173640. 32. Zanatta P, Messerotti Benvenuti S, Baldanzi F, Bosco E. Pain-related middlelatency somatosensory evoked potentials in the prognosis of post anoxic coma: a preliminary report. Minerva Anestesiol 2012;78:74956. 33. Mortberg E, Zetterberg H, Nordmark J, Blennow K, Rosengren L, Rubertsson S. S-100B is superior to NSE, BDNF and GFAP in predicting outcome of resuscitation from cardiac arrest with hypothermia treatment. Resuscitation 2011;82: 2631. 34. Oksanen T, Tiainen M, Skrifvars MB, et al. Predictive power of serum NSE and OHCA score regarding 6-month neurologic outcome after out-ofhospital ventricular brillation and therapeutic hypothermia. Resuscitation 2009;80:16570. 35. Steffen IG, Hasper D, Ploner CJ, et al. Mild therapeutic hypothermia alters neuron specic enolase as an outcome predictor after resuscitation: 97 prospective

1337

36.

37.

38. 39. 40.

41. 42. 43.

44. 45.

46.

47. 48. 49.

50.

51.

52.

53. 54.

55.

56.

57.

58.

59. 60.

61.

62.

63.

64. 65.

hypothermia patients compared to 133 historical non-hypothermia patients. Crit Care 2010;14:R69. Storm C, Nee J, Jorres A, Leithner C, Hasper D, Ploner CJ. Serial measurement of neuron specic enolase improves prognostication in cardiac arrest patients treated with hypothermia: a prospective study. Scand J Trauma Resusc Emerg Med 2012;20:6. Tiainen M, Roine RO, Pettila V, Takkunen O. Serum neuron-specic enolase and S-100B protein in cardiac arrest patients treated with hypothermia. Stroke 2003;34:28816. Mlynash M, Campbell DM, Leproust EM, et al. Temporal and spatial prole of brain diffusion-weighted MRI after cardiac arrest. Stroke 2010;41:166572. Wijman CA, Mlynash M, Cauleld AF, et al. Prognostic value of brain diffusionweighted imaging after cardiac arrest. Ann Neurol 2009;65:394402. Bisschops LL, van Alfen N, Bons S, van der Hoeven JG, Hoedemaekers CW. Predictors of poor neurologic outcome in patients after cardiac arrest treated with hypothermia: a retrospective study. Resuscitation 2011;82:696701. Bouwes A, Binnekade JM, Kuiper MA, et al. Prognosis of coma after therapeutic hypothermia: a prospective cohort study. Ann Neurol 2012;71:20612. Choi SP, Youn CS, Park KN, et al. Therapeutic hypothermia in adult cardiac arrest because of drowning. Acta Anaesthesiol Scand 2012;56:11623. Crepeau AZ, Rabinstein AA, Fugate JE, et al. Continuous EEG in therapeutic hypothermia after cardiac arrest: prognostic and clinical value. Neurology 2013;80:33944. Cronberg T, Rundgren M, Westhall E, et al. Neuron-specic enolase correlates with other prognostic markers after cardiac arrest. Neurology 2011;77:62330. Kim J, Choi BS, Kim K, et al. Prognostic performance of diffusion-weighted MRI combined with NSE in comatose cardiac arrest survivors treated with mild hypothermia. Neurocrit Care 2012;17:41220. Legriel S, Hilly-Ginoux J, Resche-Rigon M, et al. Prognostic value of electrographic postanoxic status epilepticus in comatose cardiac-arrest survivors in the therapeutic hypothermia era. Resuscitation 2013;84:34350. Rossetti AO, Carrera E, Oddo M. Early EEG correlates of neuronal injury after brain anoxia. Neurology 2012;78:796802. Rossetti AO, Oddo M, Logroscino G, Kaplan PW. Prognostication after cardiac arrest and hypothermia: a prospective study. Ann Neurol 2010;67:3017. Samaniego EA, Mlynash M, Cauleld AF, Eyngorn I, Wijman CA. Sedation confounds outcome prediction in cardiac arrest survivors treated with hypothermia. Neurocrit Care 2011;15:1139. Stammet P, Wagner DR, Gilson G, Devaux Y. Modeling serum level of S100B and bispectral index to predict outcome after cardiac arrest. J Am Coll Cardiol 2013, http://dx.doi.org/10.1016/j.jacc.2013.04.039 [Epub ahead of print]. Wennervirta JE, Ermes MJ, Tiainen SM, et al. Hypothermia-treated cardiac arrest patients with good neurological outcome differ early in quantitative variables of EEG suppression and epileptiform activity. Crit Care Med 2009;37:242735. Beaufort AM, Wierda JM, Belopavlovic M, Nederveen PJ, Kleef UW, Agoston S. The inuence of hypothermia (surface cooling) on the time-course of action and on the pharmacokinetics of rocuronium in humans. Eur J Anaesthesiol Suppl 1995;11:95106. Caldwell JE, Heier T, Wright PM, et al. Temperature-dependent pharmacokinetics and pharmacodynamics of vecuronium. Anesthesiology 2000;92:8493. Fritz HG, Holzmayr M, Walter B, Moeritz KU, Lupp A, Bauer R. The effect of mild hypothermia on plasma fentanyl concentration and biotransformation in juvenile pigs. Anesth Analg 2005;100:9961002. Fukuoka N, Aibiki M, Tsukamoto T, Seki K, Morita S. Biphasic concentration change during continuous midazolam administration in brain-injured patients undergoing therapeutic moderate hypothermia. Resuscitation 2004;60:22530. Leslie K, Sessler DI, Bjorksten AR, Moayeri A. Mild hypothermia alters propofol pharmacokinetics and increases the duration of action of atracurium. Anesth Analg 1995;80:100714. Tortorici MA, Kochanek PM, Poloyac SM. Effects of hypothermia on drug disposition, metabolism, and response: a focus of hypothermia-mediated alterations on the cytochrome P450 enzyme system. Crit Care Med 2007;35: 2196204. Daubin C, Quentin C, Allouche S, et al. Serum neuron-specic enolase as predictor of outcome in comatose cardiac-arrest survivors: a prospective cohort study. BMC Cardiovasc Disord 2011;11:48. Fugate JE, Wijdicks EF, Mandrekar J, et al. Predictors of neurologic outcome in hypothermia after cardiac arrest. Ann Neurol 2010;68:90714. Wijdicks EF, Hijdra A, Young GB, Bassetti CL, Wiebe S. Practice parameter: prediction of outcome in comatose survivors after cardiopulmonary resuscitation (an evidence-based review): report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology 2006;67:20310. Cronberg T, Brizzi M, Liedholm LJ, et al. Neurological prognostication after cardiac arrestrecommendations from the Swedish Resuscitation Council. Resuscitation 2013;84:86772. Lucas JM, Cocchi MN, Salciccioli J, et al. Neurologic recovery after therapeutic hypothermia in patients with post-cardiac arrest myoclonus. Resuscitation 2012;83:2659. Thomke F, Marx JJ, Sauer O, et al. Observations on comatose survivors of cardiopulmonary resuscitation with generalized myoclonus. BMC Neurol 2005;5:14. Wijdicks EF, Young GB. Myoclonus status in comatose patients after cardiac arrest. Lancet 1994;343:16423. Zandbergen EG, Hijdra A, Koelman JH, et al. Prediction of poor outcome within the rst 3 days of postanoxic coma. Neurology 2006;66:628.

1338

C. Sandroni et al. / Resuscitation 84 (2013) 13241338 79. Stecker MM, Cheung AT, Pochettino A, et al. Deep hypothermic circulatory arrest. I. Effects of cooling on electroencephalogram and evoked potentials. Ann Thorac Surg 2001;71:1421. 80. Zandbergen EG, Hijdra A, de Haan RJ, et al. Interobserver variation in the interpretation of SSEPs in anoxic-ischaemic coma. Clin Neurophysiol 2006;117:152935. 81. Geocadin RG, Buitrago MM, Torbey MT, Chandra-Strobos N, Williams MA, Kaplan PW. Neurologic prognosis and withdrawal of life support after resuscitation from cardiac arrest. Neurology 2006;67:1058. 82. Anderson RE, Hansson LO, Nilsson O, Dijlai-Merzoug R, Settergren G. High serum S100B levels for trauma patients without head injuries. Neurosurgery 2001;48:12558, discussion 125860. 83. Perman SM, Kirkpatrick JN, Reitsma AM, et al. Timing of neuroprognostication in postcardiac arrest therapeutic hypothermia. Crit Care Med 2012;40:71924. 84. Blondin NA, Greer DM. Neurologic prognosis in cardiac arrest patients treated with therapeutic hypothermia. Neurologist 2011;17:2418. 85. Samaniego EA, Persoon S, Wijman CA. Prognosis after cardiac arrest and hypothermia: a new paradigm. Curr Neurol Neurosci Rep 2011;11:1119. 86. Friberg H, Rundgren M, Westhall E, Nielsen N, Cronberg T. Continuous evaluation of neurological prognosis after cardiac arrest. Acta Anaesthesiol Scand 2013;57:615. 87. Greer GM. Unexpected good recovery in a comatose post-cardiac arrest patient with poor prognostic features. Resuscitation 2013;84:e812. 88. Topcuoglu MA, Oguz KK, Buyukserbetci G, Bulut E. Prognostic value of magnetic resonance imaging in post-resuscitation encephalopathy. Intern Med 2009;48:163545. 89. Sulmasy DP, Terry PB, Weisman CS, et al. The accuracy of substituted judgments in patients with terminal diagnoses. Ann Intern Med 1998;128:6219. 90. Tirschwell D. Improved prediction of awakening or nonawakening in severe anoxic coma using tree-based classication. Crit Care Med 2006;34:15735. 91. Zandbergen EG, de Haan RJ, Stoutenbeek CP, Koelman JH, Hijdra A. Systematic review of early prediction of poor outcome in anoxic-ischaemic coma. Lancet 1998;352:180812.

66. Morris HR, Howard RS, Brown P. Early myoclonic status and outcome after cardiorespiratory arrest. J Neurol Neurosurg Psychiatry 1998;64:2678. 67. Celesia GG, Grigg MM, Ross E. Generalized status myoclonicus in acute anoxic and toxic-metabolic encephalopathies. Arch Neurol 1988;45: 7814. 68. Datta S, Hart GK, Opdam H, Gutteridge G, Archer J. Post-hypoxic myoclonic status: the prognosis is not always hopeless. Crit Care Resusc 2009;11:3941. 69. English WA, Gifn NJ, Nolan JP. Myoclonus after cardiac arrest: pitfalls in diagnosis and prognosis. Anaesthesia 2009;64:90811. 70. Legriel S, Bruneel F, Sediri H, et al. Early EEG monitoring for detecting postanoxic status epilepticus during therapeutic hypothermia: a pilot study. Neurocrit Care 2009;11:33844. 71. Jorgensen EO, Holm S. The natural course of neurological recovery following cardiopulmonary resuscitation. Resuscitation 1998;36:11122. 72. Hirsch LJ, Laroche SM, Gaspard N, et al. American Clinical Neurophysiology Societys Standardized Critical Care EEG Terminology: 2012 version. J Clin Neurophysiol 2013;30:127. 73. Aibiki M. Bispectral index and suppression ratio during hypothermia after cardiac arrest. Intensive Care Med 2011;37:1399. 74. Lu-Emerson C, Khot S. Neurological sequelae of hypoxic-ischemic brain injury. NeuroRehabilitation 2010;26:3545. 75. Rossetti AO, Oddo M, Liaudet L, Kaplan PW. Predictors of awakening from postanoxic status epilepticus after therapeutic hypothermia. Neurology 2009;72:7449. 76. VanLandingham KE, Lothman EW. Self-sustaining limbic status epilepticus. I. Acute and chronic cerebral metabolic studies: limbic hypermetabolism and neocortical hypometabolism. Neurology 1991;41:19429. 77. DeGiorgio CM, Tomiyasu U, Gott PS, Treiman DM. Hippocampal pyramidal cell loss in human status epilepticus. Epilepsia 1992;33:237. 78. Kottenberg-Assenmacher E, Armbruster W, Bornfeld N, Peters J. Hypothermia does not alter somatosensory evoked potential amplitude and global cerebral oxygen extraction during marked sodium nitroprusside-induced arterial hypotension. Anesthesiology 2003;98:11128.