ORIGINAL PAPER

The Modified Glasgow Outcome Score for the predictionof outcome in patients after cardiac arrest: a prospective clinicalproof of concept study

Obaida R. Rana • Jorg W. Schroder • Julia S. Kuhnen • Esra Saygili •

Christopher Gemein • Matthias D. H. Zink • Patrick Schauerte • Johannes Schiefer •

Robert H. G. Schwinger • Joachim Weis • Nikolaus Marx • Malte Kelm •

Christian Meyer • Erol Saygili

Received: 17 August 2011 / Accepted: 26 January 2012 / Published online: 10 February 2012

� Springer-Verlag 2012

Abstract The Glasgow–Pittsburgh cerebral performance

categories (GP-CPC) and the Glasgow Outcome Score

(GOS) have been used to categorize patients according to

their neurological outcome for prognostic predictors in

patients after cardiac arrest (CA). We postulated that

inclusion of deaths without knowing the cerebral status into

the group of patients with poor outcome after CA using the

GP-CPC and GOS will lead to dilution of the prognostic

power of the investigated biochemical marker. The present

study was conducted to verify this issue by employing a

modified outcome score, which we termed as Modified

Glasgow Outcome Score (MGOS). In the present study, 97

patients were enrolled in a prospective manner. Serum NSE

and S100B levels were measured daily for 7 days after

admission to the intensive care unit. Neurological outcome

was assessed by employing the GOS and MGOS after

6 months. By employing the GOS, 46 patients were cate-

gorized into the group of patients with poor outcome and

51 patients survived with good neurological outcome.

Patients who died without certified brain damage or with

unknown cerebral status after CA (n = 20) were separated

from patients with poor outcome in the MGOS. The

magnitude of NSE (S100B) elevation in patients with poor

outcome categorized by the MGOS was approximately

1.7-fold (1.5) higher as compared with patients divided by

the GOS. The mean calculated sensitivities and area under

the curve values of NSE and S100B predicting poor out-

come classified by the MGOS were significantly higher as

compared with the GOS. Conclusively, inclusion of deaths

without certified brain damage or with unknown cerebral

status into the group of patients with poor outcome will

lead to underestimation of the prognostic power of inves-

tigated biochemical markers such as NSE and S100B. The

MGOS will help to avoid this bias.

Keywords Cardiac arrest � Cardiopulmonary

resuscitation � Glasgow–Pittsburgh cerebral performance

categories (GP-CPC) � Glasgow Outcome Score (GOS) �Modified Glasgow Outcome Score (MGOS) � NSE � S100B

Introduction

Cardiac arrest (CA) is the leading cause of death in Europe

and the USA [1]. The consecution of an improved rate of

restoration of spontaneous circulation (ROSC) after car-

diopulmonary resuscitation (CPR) may be an increasing

number of patients suffering from various degrees of brain

damage after successful resuscitation [2]. Several clinical

O. R. Rana and J. W. Schroder contributed equally to this work.

O. R. Rana (&) � M. Kelm � C. Meyer � E. Saygili

Division of Cardiology, Pulmonology and Vascular Medicine,

University Hospital Dusseldorf, Moorenstrasse 5,

40225 Dusseldorf, Germany

e-mail: obaida.rana@med.uni-duesseldorf.de

J. W. Schroder � J. S. Kuhnen � E. Saygili � C. Gemein �M. D. H. Zink � P. Schauerte � N. Marx

Department of Cardiology, RWTH Aachen University,

Aachen, Germany

J. Schiefer

Department of Neurology, RWTH Aachen University,

Aachen, Germany

R. H. G. Schwinger

Medical Clinic II, Klinikum Weiden, Weiden, Germany

J. Weis

Institute of Neuropathology, RWTH Aachen University,

Aachen, Germany

123

Clin Res Cardiol (2012) 101:533–543

DOI 10.1007/s00392-012-0423-7

outcome scores, electrophysiological techniques, and

imaging methods have attempted to predict the presence,

degree, and course of neurological outcome after CA [3].

This has recently been complemented by neural tissue-

derived biomarkers such as neuron-specific enolase (NSE)

and S100B [4].

The Utstein Consensus Conference participants rec-

ommend the use of the Glasgow–Pittsburgh cerebral

performance categories (GP-CPC) to document the out-

come of patients after CA and ROSC [5]. However, we

recently have demonstrated that several studies evaluating

biochemical markers and electrophysiological techniques

for the prediction of outcome after CA are based on the

Glasgow Outcome Score (GOS) [6]. Although both scores

provide comparative categories, a significant number of

outcome studies using the GOS do not include patients

with severe cerebral disability (GOS 3) to the category of

poor outcome [6]. In contrast, studies applying the GP-

CPC included patients with severe cerebral disability

(GP-CPC 3) to the group of patients with poor outcome.

Furthermore, while the GP-CPC defines death, certified

brain death, or death by traditional criteria as worst cat-

egory (GP-CPC 5), the GOS is less specific and defines

the worst outcome level (GOS 1) simply as death irre-

spective of whether this was due to brain damage or other

causes or a combination of both [5, 7]. Unfortunately,

many clinical studies using the GP-CPC did not restrict

GP-CPC category five only to certified brain death, which

may have diluted sensitivity of this score and its param-

eters to predict the individual neurologic outcome [6]. We

have recently postulated that inclusion of deaths without

certified brain damage to the group of patients with poor

outcome will lead to underestimation of sensitivity and

cut-off values of NSE for the prediction of poor outcome

[6]. We speculated that a modified GOS or GP-CPC score

might help to avoid this bias. Therefore, a prospective

observational study was conducted to evaluate a modified

outcome score, which we termed as Modified Glasgow

Outcome Score (MGOS), by the use of which the well-

established biochemical markers NSE and S100B were

investigated for the prediction of neurological outcome in

patients after CA.

Methods

Patients

The study was performed between March 2009 and July

2010 at the intensive care unit (ICU) of the RWTH Aachen

University, Germany. A total of 97 consecutive adult

patients with documented out-of-hospital CA with primary

successful ROSC were prospectively enrolled. The

following patients were eligible: witnessed out-of-hospital

CA of presumed cardiac origin; absence of previous neu-

rological disease, and pregnancy. Patients were included as

soon as possible, but revascularizing procedures and sta-

bilization took precedence.

Standard care

All patients were admitted to the ICU, monitored and

treated according to international standards. All patients

were intubated and mechanically ventilated [8]. Patients

were sedated using midazolam and/or propofol and

received adequate analgesia with fentanyl. The procedure

of mild hypothermia was performed according to interna-

tional standards with the Cool Gard 3000 (Zoll�, USA) and

Hilotherm clinic (Hilotherm� GmbH, Germany) cooling

systems [9–12]. Vasoactive or inotropic support, usually

norepinephrine or dobutamine was administered if

necessary.

Study design

Approval for the study was obtained from the relevant

ethics committee at the RWTH Aachen University, Ger-

many. This study has been performed in accordance with

the ethical standards laid down in the 1964 Declaration of

Helsinki. Because all procedures described below were

within the normal daily intensive care routine with no risk

to the patients, the institutional review board therefore

waived the need for informed patient consent. After arrival

and stabilization of patients in the ICU and approval of

inclusion criteria, serum samples for the determination of

NSE and S100B levels were taken on admission and then

every 24 h over a time period of 7 days from the super-

natant of blood collected for routine laboratory analyses.

Blood samples during CPR and after ROSC were taken by

an independent physician who was not responsible for the

patient’s care. Arterial blood gas analysis of partial pres-

sure of oxygen (PO2), partial pressure of carbon dioxide

(PCO2), lactate, pH, and blood glucose were analyzed on

admission. Furthermore, serum creatine kinase was asses-

sed on admission. Neurological outcome was assessed

6 months after CPR by use of the outcome scores as

described below.

Neurological outcome

Cerebral function was evaluated 6 months after CA by

employing the GOS as described by Janett and Bond [7]

(Table 1) and a modified six-point outcome score, which

we termed as MGOS (Table 2).

For statistical purposes, outcome categories of both

scores were split into 2–3 groups:

534 Clin Res Cardiol (2012) 101:533–543

123

GOS

Poor: consisted of patients who died of any cause or

remained in a persistent vegetative state (GOS 1–2).

Good: consisted of patients who regained consciousness

(GOS 3–5).

MGOS

Unclear: consisted of patients who died without certified

brain damage (MGOS 0).

Poor: consisted of patients who died with certified brain

damage or remained in a persistent vegetative state

(MGOS 1–2).

Good: consisted of patients who regained consciousness

(MGOS 3–5).

Assessment of brain damage

The diagnosis of global hypoxic brain damage was certified

by a study independent senior neurologist from the

Department of Neurology, RWTH Aachen University,

Germany, after extensive clinical examination and on

behalf of the results of cranial CT and/or transcranial

doppler (TCD) ultrasonography and/or somatosensory

evoked potentials (SSEP) and/or EEG. Only patients who

died with documented brain damage were included in

MGOS 1. Patients who died without certified brain dam-

age, either because of a lack of diagnostic evaluation, e.g.,

because of hemodynamic instability of the patient, or

without signs of global hypoxic brain damage after

extensive clinical and diagnostic evaluations as described

above were classified as MGOS 0. During application of

the GOS classification, both categories (MGOS 0 and

MGOS 1) were summarized as GOS 1. Patients with or

without documented brain damage who remained in the

persistent vegetative state were classified as MGOS 2 or

GOS 2.

NSE and S100B measurements

Blood samples were taken from patients with an arterial line

and collected in a standard serum tube. Samples were

immediately centrifuged for 10 min at 1,500g. Subsequently,

NSE and S100B concentrations were both measured by an

electrochemiluminescence immunoassay (ECLIA). A bio-

tinylated monoclonal NSE/S100B-specific antibody and a

monoclonal NSE/S100B-specific antibody labeled with a

ruthenium complex (Tris{2,2-bipyridyl}ruthenium(II)) form

a sandwich complex. Voltage-induced chemiluminescent

emissions were measured by a photomultiplier. Results are

determined via a calibration curve which is instrument-spe-

cifically generated by two-point calibration and a master

curve provided via the reagent barcode (COBAS�, Roche

Diagnostics GmbH, Germany).

Statistical analysis

All data are expressed as mean ± 1 SEM unless otherwise

stated. A one-way analysis of variance (ANOVA) with post-

hoc bonferroni’s test for multiple comparisons of means was

used for repeated measures. The Student’s t test was applied

for quantitative variables. A P value\0.05 was considered as

significant. Receiver operating characteristic (ROC) curve

analysis and the derived area under the curve (AUC) statistic

provide a global and standardized appreciation of the accu-

racy of a biochemical marker for predicting an event. ROC

curves were generated by plotting the sensitivity against one-

specificity. Statistical analysis was performed with the IBM

SPSS Statistics 19 software for windows.

Results

Baseline characteristics

Of the 97 patients evaluated, the neurological outcome

assessed by employing the GOS after 6 months was as

follows: 39 died of any reason (GOS 1, 40.2%), seven

Table 1 The Glasgow Outcome Score (GOS)

GOS Neurological outcome

GOS 1 Death

GOS 2 Persistent vegetative state (unable to interact with

environment)

GOS 3 Severe disability (unable to live independently but able

to follow commands)

GOS 4 Moderate disability (able to live independently but

unable to return to work)

GOS 5 Mild or no disability (able to return to work)

Table 2 The Modified Glasgow Outcome Score (MGOS): a modi-

fied outcome score for patients after CA

MGOS Neurological outcome

MGOS 0 Patient died with unknown cerebral status (e.g., death

after cardiogenic/septic shock, pulmonary embolism

or aortic dissection without assessment of

neurological status by cranial CT/MRI, SSEP, etc.)

MGOS 1 Patient died with documented hypoxic brain damage

MGOS 2 Persistent vegetative state (unable to interact with

environment)

MGOS 3 Severe disability (unable to live independently

but able to follow commands)

MGOS 4 Moderate disability (able to live independently

but unable to return to work)

MGOS 5 Mild or no disability (able to return to work)

Clin Res Cardiol (2012) 101:533–543 535

123

developed a persistent vegetative state (GOS 2, 7.2%), 12

survived with severe disability (GOS 3, 12.4%), 22 sur-

vived with moderate disability (GOS 4, 22.7%), and 17

survived with mild or no disability (GOS 5, 17.5%). For

statistical purposes, outcome categories were split into the

group of patients with poor (GOS 1 ? 2, n = 46) and good

outcome (GOS 3 ? 4 ? 5, n = 51). The clinical and

demographic characteristics of patients categorized by the

GOS are shown in Table 3. Briefly, patients with poor

outcome were significantly older compared with patients

with good neurological outcome. Furthermore, patients

with poor outcome demonstrated significantly higher arte-

rial CO2 partial pressure levels, serum lactate and blood

glucose levels, APACHE II scores, time of hypoxia, mean

duration of CPR, and required epinephrine doses before

ROSC compared with patients with good outcome, while

serum pH, numbers of defibrillation before ROSC, and

days of mechanical ventilation were significantly lower in

the poor outcome group. No significant differences were

found between patients of poor and good outcome

receiving mild hypothermia. In the poor outcome group 34

of 46 patients were treated by mild hypothermia. Of note,

from the 12 patients precluded from mild hypothermia, 9

patients died shortly after CPR and 3 patients were omitted

from mild hypothermia because of short time of hypoxia

and CPR. Consistently, due to short time of hypoxia and

CPR, 11 of 51 patients from the good outcome group were

precluded from mild hypothermia. Regarding the past

medical history, patients in the poor outcome group

showed a significantly higher prevalence of diabetes and

peripheral artery disease compared with patients with good

outcome. The number of asystoles as initial rhythm in first

ECG was significantly higher in the poor outcome group

compared with patients with good outcome, whereas in the

good outcome group ventricular fibrillation as initial

rhythm was recorded in a significantly higher quantity.

In the same 97 patients mentioned above, the neuro-

logical outcome was assessed by employing the MGOS

6 months after CPR. Because the categorization of patients

with good outcome in the MGOS is similar to the GOS, no

Table 3 Baseline

characteristics of patients

divided by the GOS

APACHE acute physiology and

chronic health evaluation, CPRcardiopulmonary resuscitation,

ROSC restoration of

spontaneous circulation

Parameters (unit) ± SD Poor

GOS 1 (n = 39)

GOS 2 (n = 7)

Good

GOS 3 (n = 12)

GOS 4 (n = 22)

GOS 5 (n = 17)

P value

Gender male no./total no. (%) 32/46 (70) 40/51 (78) 0.221

Age (years) 70.6 ± 13.2 60.9 ± 12.5 \0.001

On admission

O2 partial pressure (mmHg) 216.6 ± 175.9 233.3 ± 157.5 0.624

CO2 partial pressure (mmHg) 51.6 ± 24.4 42.3 ± 12.8 0.021

Lactate (mmol/l) 7.4 ± 5.2 4.5 ± 3.3 0.002

pH 7.17 ± 0.25 7.29 ± 0.12 0.003

Blood glucose (mmol/l) 14 ± 6.1 11.5 ± 5.4 0.040

Serum creatine kinase (mg/dl) 1.5 ± 1.4 1.2 ± 0.7 0.220

APACHE II 33.4 ± 4.4 30.6 ± 6.1 0.012

Time of hypoxia (min) 10.7 ± 3.3 6.7 ± 2.6 \0.001

Mean duration of CPR (min) 34.8 ± 33.8 18.8 ± 14.3 0.004

Bystander-initiated CPR (min) 1.2 ± 3 1.5 ± 3.1 0.686

Defibrillations before ROSC 1.5 ± 1.8 3.3 ± 3.9 0.007

Epinephrine dose before ROSC (mg) 5 ± 3.3 2.2 ± 2.2 \0.001

Hypothermia no./total no. (%) 34/46 (74) 40/51 (78) 0.607

Days of mechanical ventilation 7.3 ± 9.2 12 ± 9.9 0.037

Past medical history no./total no. (%)

Diabetes 16/46 (35) 6/51 (12) 0.014

Coronary heart disease 19/46 (41) 18/51 (35) 0.358

Cerebrovascular disease 7/46 (15) 3/51 (6) 0.056

Peripheral artery disease 10/46 (22) 1/51 (2) 0.002

First ECG

Ventricular fibrillation 16/46 (35) 41/51 (80) \0.001

Ventricular tachycardia 2/46 (4) 1/51 (2) 0.488

Asystole 21/46 (46) 6/51 (12) \0.001

Other rhythm 7/46 (15) 3/51 (6) 0.134

536 Clin Res Cardiol (2012) 101:533–543

123

differences were found in the baseline characteristics of

these patients. However, all patients who died with

uncertified brain damage (MGOS 0, n = 26) were exclu-

ded from the group of poor outcome. Of these 26 patients

(26.8%) categorized as MGOS 0, 17 patients (17.5%) died

as a consequence of a cardiogenic shock, 7 patients (7.2%)

died by a combination of cardiogenic and septic shock and

2 patients (2.1%) died after septic shock. The poor outcome

group consisted of 13 patients who died with certified

hypoxic brain damage (MGOS 1, 13.4%) and of 7 patients

who developed and remained in a persistent vegetative

state (MGOS 2, 7.2%). According to the GOS classifica-

tion, outcome categories were split into the group of

patients with unclear (MGOS 0, n = 26), poor (MGOS

1 ? 2, n = 20) and good neurological outcome (MGOS

3 ? 4 ? 5, n = 51). The clinical and demographic char-

acteristics of patients categorized by the MGOS are shown

in Table 4. Patients with good outcome were significantly

younger compared with patients with unclear outcome,

while no significance concerning age could be observed

among the other outcome groups. Likewise, patients with

good outcome had significantly lower arterial CO2 partial

pressure levels, serum lactate levels, APACHE II scores,

time of hypoxia, mean durations of CPR and, required

epinephrine doses before ROSC, but significantly higher

days of mechanical ventilation compared with patients in

the unclear outcome group. Furthermore, patients with

good outcome had significantly higher pH levels and

numbers of defibrillation before ROSC, but a significantly

lower mean duration of CPR and less required epinephrine

doses before ROSC compared with patients with poor and

unclear outcome. Comparison of patients with unclear and

poor outcome resulted in significantly higher time of

hypoxia and significantly lower days of mechanical

Table 4 Baseline characteristics of patients divided by the MGOS

Parameters (unit) ± SD Unclear (1)

MGOS 0 (n = 26)

Poor (2)

MGOS 1 (n = 13)

MGOS 2 (n = 7)

Good (3)

MGOS 3 (n = 12)

MGOS 4 (n = 22)

MGOS 5 (n = 17)

P value

1 vs. 2

P value

2 vs. 3

P value

1 vs. 3

Gender male no./total no. (%) 19/26 (73) 13/20 (65) 40/51 (78) 0.570 0.176 0.489

Age (years) 72.8 ± 11 67.6 ± 15.4 60.9 ± 12.5 0.212 0.094 \0.001

On admission

O2 partial pressure (mmHg) 173.9 ± 162 272.1 ± 181.6 233.3 ± 157.5 0.064 0.408 0.130

CO2 partial pressure (mmHg) 51.9 ± 23.9 51.2 ± 25.7 42.3 ± 12.8 0.924 0.059 0.026

Lactate (mmol/l) 7.9 ± 5.7 6.9 ± 4.7 4.5 ± 3.3 0.535 0.050 0.002

pH 7.15 ± 0.28 7.2 ± 0.21 7.29 ± 0.12 0.534 0.020 0.003

Blood glucose (mmol/l) 14 ± 6.8 14 ± 5.2 11.5 ± 5.4 0.984 0.080 0.133

Serum creatine kinase (mg/dl) 1.3 ± 0.5 1.7 ± 2 1.2 ± 0.7 0.420 0.296 0.423

APACHE II 33.7 ± 3.8 33.1 ± 5.2 30.6 ± 6.1 0.651 0.093 0.010

Time of hypoxia (min) 12.1 ± 3.2 8.4 ± 1.8 6.7 ± 2.6 0.003 0.056 \0.001

Mean duration of CPR (min) 38.8 ± 41.2 29.6 ± 20.2 18.8 ± 14.3 0.350 0.048 0.004

Bystander-initiated CPR (min) 0.7 ± 2.4 1.7 ± 3.6 1.5 ± 3.1 0.337 0.810 0.300

Defibrillations before ROSC 1.7 ± 1.9 1.2 ± 1.6 3.3 ± 3.9 0.432 0.005 0.026

Epinephrine dose before ROSC (mg) 4.4 ± 2.8 6.1 ± 4 2.2 ± 2.2 0.248 \0.001 0.006

Hypothermia no./total no. (%) 14/26 (54) 20/20 (100) 40/51 (78) \0.001 0.024 0.026

Days of mechanical ventilation 3 ± 3.2 13.6 ± 11.4 12 ± 9.9 \0.001 0.626 \0.001

Past medical history no./total no. (%)

Diabetes 8/26 (31) 8/20 (40) 6/51 (12) 0.769 0.025 0.046

Coronary heart disease 11/26 (42) 8/20 (40) 18/51 (35) 0.626 0.323 0.602

Cerebrovascular disease 6/26 (23) 1/20 (5) 3/51 (6) 0.084 0.863 0.008

Peripheral artery disease 4/26 (15) 6/20 (30) 1/51 (2) 0.205 \0.001 0.026

First ECG

Ventricular fibrillation 10/26 (38) 6/20 (30) 41/51 (80) 0.764 \0.001 \0.001

Ventricular tachycardia 1/26 (4) 1/20 (5) 1/51 (2) 0.793 0.544 0.665

Asystole 11/26 (42) 10/20 (50) 6/51 (12) 0.359 \0.001 0.002

Other rhythm 4/26 (15) 3/20 (10) 3/51 (6) 0.972 0.220 0.175

APACHE acute physiology and chronic health evaluation, CPR cardiopulmonary resuscitation, ROSC restoration of spontaneous circulation

Clin Res Cardiol (2012) 101:533–543 537

123

ventilation in patients with unclear outcome. All patients in

the poor outcome group, while 14 of 26 patients in the

unclear outcome group received mild hypothermia. In the

unclear outcome group, nine patients died shortly after

CPR and three patients did not receive mild hypothermia

because of short time of hypoxia and CPR. The past

medical history and first recorded ECG data were similar to

the GOS group, when comparing patients with poor and

good outcome. In addition, the same observation was found

when comparing patients with unclear and good outcome

with the sole distinction that patients with unclear outcome

showed a significantly greater number of cerebrovascular

diseases compared to patients with good outcome.

NSE levels in patients divided by GOS and MGOS

NSE levels were investigated in patients categorized by the

GOS and MGOS (Fig. 1). Because patients who died with

unknown cerebral outcome were excluded from the group

of patients with poor outcome in the MGOS, the magnitude

of NSE elevation in patients with poor outcome in the

MGOS was higher throughout all days, especially on days

2–6, compared with patients categorized by the GOS. The

maximum of mean NSE levels of patients with poor

outcome occurred on day 3 in both outcome scores. Fur-

thermore, the mean peak NSE values of patients with poor

outcome were approximately 1.7-fold higher in the MGOS

(200 ± 45 lg/l) compared with the GOS classification

(118 ± 25 lg/l).

Prediction power of serum NSE for poor outcome

in patients divided by GOS and MGOS

Next, we measured cut-off values, sensitivity, and the area

under the curve (AUC) by use of ROC analysis. The

specificity was set 100% for the prediction of poor outcome

for NSE in each group, respectively. The cut-off values,

sensitivity, and the AUC of NSE for MGOS and GOS are

depicted in Table 5. Best prediction of NSE for poor out-

come was found on days 3 and 7 in both outcome scores

(Fig. 2). Mean cut-off values calculated from days 1 to 7 in

the MGOS group were not statistically significant as com-

pared with the GOS classification (MGOS: 60.3 ± 33.2 lg/l,

GOS: 59.6 ± 42.4 lg/l, P = 0.9). However, the mean

calculated sensitivity from days 1 to 7 of NSE predicting

poor outcome in the MGOS was significantly higher com-

pared with the GOS (MGOS: 60 ± 29.3%, GOS:

48.4 ± 22.8%, P \ 0.01). The same observation was found

0

30

60

90

120

150

180

210

240

270

peak

NS

E in

µg/

l (G

OS

)

p<0.001

goodpoor-

118

33

n=46 n=511 2 3 4 5 6 7

30

60

90

120

150

180

210

240

NS

E in

µg/

l (G

OS

)

*

***

*

days

poor (GOS 1+2)

good (GOS 3+4+5)

*

*p<0.05 vs. good

0

30

60

90

120

150

180

210

240

270

peak

NS

E in

µg/

l (M

GO

S)

p<0.001

goodpoorunclear

p<0.001

45

200

33

n=26n=20 n=511 2 3 4 5 6 7

30

60

90

120

150

180

210

240

*

*

**

*

NS

E in

µg/

l (M

GO

S)

days

unclear (MGOS 0)

poor (MGOS 1+2)

good(MGOS 3+4+5)

*

*p<0.05 vs. unclear and vs. good

A B

C D

Fig. 1 The magnitude of NSE

elevation in patients with poor

outcome was higher with the

MGOS compared with the GOS.

a NSE serum levels at different

time points after CPR and

ROSC in patients with unclear

(n = 20), poor (n = 26) and

good (n = 51) outcome

assessed by the MGOS. b NSE

peak serum levels in patients

with unclear, poor, and good

outcome assessed by the

MGOS. c NSE serum levels at

different time points in patients

with poor (n = 46) and good

(n = 51) outcome assessed by

the GOS. d NSE peak serum

levels in patients with poor and

good outcome assessed by the

GOS

538 Clin Res Cardiol (2012) 101:533–543

123

for the mean calculated AUC value (MGOS: 0.89 ± 0.1,

GOS: 0.8 ± 0.1, P \ 0.001).

S100B levels in patients divided by GOS and MGOS

In analogy to NSE measurements, S100B levels were

investigated in patients categorized by the GOS and the

MGOS (Fig. 3). We found higher serum S100B levels in

patients with poor outcome on days 2–6 and especially on

days 2 and 3 while using the classification by the MGOS

compared with the GOS. In addition, the mean peak S100B

values in patients with poor outcome were 1.5-fold higher

in the MGOS compared with the GOS (MGOS: 5.2 ±

1.2 lg/l, GOS: 3.5 ± 1.2 lg/l). The maximum of mean

S100B levels of patients with poor outcome was achieved

on day 2 in both outcome scores.

Prediction power of serum S100B for poor outcome

in patients divided by GOS and MGOS

The cut-off values, sensitivities, and the AUC values of

S100B for MGOS and GOS are depicted in Table 6. Best

Table 5 Cut-off points,

sensitivities, specificities and

AUC values of NSE predicting

poor outcome in patients

divided by the GOS and MGOS

AUC area under the curve,

CI confidence interval

Cut-off value Specificity Sensitivity AUC 95% CI of AUC P value

GOS

NSE day 1 151.5 100 2 0.599 0.481–0.718 0.11

NSE day 2 65 100 35 0.743 0.636–0.851 \0.001

NSE day 3 57.5 100 61.5 0.807 0.704–0.909 \0.001

NSE day 4 34.5 100 59 0.834 0.740–0.927 \0.001

NSE day 5 39 100 58 0.878 0.788–0.967 \0.001

NSE day 6 39 100 56 0.855 0.763–0.947 \0.001

NSE day 7 31 100 67 0.899 0.812–0.985 \0.001

NSE peak value 137 100 36 0.766 0.668–0.864 \0.001

MGOS

NSE day 1 129 100 0 0.703 0.530–0.875 0.171

NSE day 2 70 100 43 0.825 0.696–0.953 \0.001

NSE day 3 57.5 100 80 0.922 0.828–1.000 \0.001

NSE day 4 56.5 100 73 0.938 0.873–1.000 \0.001

NSE day 5 39 100 75 0.952 0.888–1.000 \0.001

NSE day 6 39 100 69 0.936 0.863–1.000 \0.001

NSE day 7 31 100 80 0.932 0.816–1.000 \0.001

NSE peak value 137 100 53 0.878 0.759–0.997 \0.001

0,0 0,2 0,4 0,6 0,8 1,0

0,0

0,2

0,4

0,6

0,8

1,0

31

MGOS: AUC 0.932 GOS: AUC 0.899

NSE day 7

Sen

sitiv

ity (

%)

1-Specificity (%)

31

0,0 0,2 0,4 0,6 0,8 1,0

0,0

0,2

0,4

0,6

0,8

1,0

57.5

MGOS: AUC 0.922 GOS: AUC 0.807

NSE day 3

Sen

sitiv

ity (

%)

1-Specificity (%)

57.5

Fig. 2 ROC analysis of serum NSE predicting unfavorable outcome in patients divided by the MGOS and GOS. Solid circles mark NSE cut-off

points with 100% specificity. NSE cut-off values are given in lg/l. AUC denotes area under the curve

Clin Res Cardiol (2012) 101:533–543 539

123

prediction of S100B for poor outcome was found on days 2

and 3 in both outcome groups (Fig. 4). The mean cut-off

value calculated from days 1 to 7 in the MGOS was similar

without significant difference compared with the GOS

(MGOS: 1.12 ± 1.15 lg/l, GOS: 1.09 ± 1.15 lg/l, P =

0.356). However, the mean calculated sensitivity from days

1 to 7 of S100B for the prediction of poor outcome in the

MGOS was significantly higher compared with the GOS

(MGOS: 33.3 ± 12%, GOS: 26.6 ± 6.7%, P \ 0.02).

Moreover, the mean calculated AUC value for S100B in

the MGOS was also significantly higher as compared with

the GOS (MGOS: 0.79 ± 0.1, GOS: 0.7 ± 0.1, P \ 0.01).

Discussion

The present study demonstrates that (1) by the use of the

GOS the calculated prediction power of the biochemical

markers NSE and S100B with respect to sensitivities and

AUC values is underestimated as compared with the

modified outcome score, which we termed as MGOS; (2)

this bias is mainly generated, because patients who died

with an unclear neurological status were not excluded from

the group of patients with poor outcome in the GOS; (3) the

best prediction of NSE was achieved on days 3 and 7; (4)

the best prediction of S100B was achieved on day 3, and

(5) the prediction power of NSE was higher as compared

with S100B in our clinical setup.

In 1991 the GP-CPC were proposed to complement the

GOS [5]. The GOS was first introduced by Janett and Bond

in 1975 [7]. The main difference between both scores lies

in the fact that GOS does not further differentiate the cause

of death in GOS 1, whereas GP-CPC 5 acknowledges death

as certified brain death or death by traditional criteria.

Because of the unspecific definition of GP-CPC 5, conse-

quently, most studies investigating biochemical markers or

somatosensory evoked potentials (SSEP) for the prediction

of neurological outcome included patients who died either

from primary brain damage or death from any cause into

this category [6]. Thus, no profound difference is found

between both outcome groups, and both outcome groups

are yet used in clinical studies evaluating the prediction

power of biochemical markers or SSEP [13–16]. However,

in the present prospective clinical study we chose to use the

DC

BA

0

1

2

3

4

5

6

7

8

9

S10

0B in

µg/

l (G

OS

)

goodpoor-

3,5

0,5

p<0.01

n=46 n=511 2 3 4 5 6 7

0

1

2

3

4

5

6

7

8

****

*

S10

0B in

µg/

l (G

OS

)

poor (GOS 1+2)

good (GOS 3+4+5)

days*p<0.05 vs. good

*

1 2 3 4 5 6 7

0

1

2

3

4

5

6

7

8

*S

100B

in µ

g/l (

MG

OS

)

unclear (MGOS 0)

poor (MGOS 1+2)

good (MGOS 3+4+5)

*

#

#p<0.05 vs. unclear and good days*p<0.05 vs. good

*

#

0

1

2

3

4

5

6

7

8

9

p=0.65

S10

0B in

µg/

l (M

GO

S)

p<0.001

goodpoorunclear

p=0.07

2,0

5,2

0,5

n=26n=20 n=51

Fig. 3 The magnitude of

S100B elevation in patients with

poor outcome was higher with

the MGOS compared with the

GOS. a S100B serum levels at

different time points after CPR

and ROSC in patients with

unclear (n = 20), poor

(n = 26), and good (n = 51)

outcome assessed by the

MGOS. b S100B peak serum

levels in patients with unclear,

poor, and good outcome

assessed by the MGOS.

c S100B serum levels at

different time points in patients

with poor (n = 46) and good

(n = 51) outcome assessed by

the GOS. d S100B peak serum

levels in patients with poor and

good outcome assessed by the

GOS

540 Clin Res Cardiol (2012) 101:533–543

123

GOS to evaluate the calculated prediction power in com-

parison with the MGOS.

Daily clinical experience shows that many patients,

studies vary between 16 and 21% [17–19], die after pri-

mary successful CPR in consequence of a cardiogenic

shock. This is in line with our study with 17.5% deaths

after cardiogenic shock. In these patients further neuro-

logical diagnostic investigations are often not possible. In

addition, further patients may die during the intensive care

interval due to a septic shock (1–2%) or other complica-

tions [17, 18], before an evaluation of their neurological

state was feasible. Because the neurological state of the

patients in both the aforementioned groups is unclear and

the number of patients may not rather be small in the

unclear outcome group (MGOS 0), the calculated course of

the here investigated biomarkers in the GOS 1 group would

be decreased if patients with unknown cerebral outcome

are not excluded from the group of poor outcome. This

would result in a dilution of the calculated AUC and sen-

sitivity for both NSE and S100B. Additionally, the MGOS

0 group with unclear neurological state may have a very

heterogeneous constellation with only few differences in

baseline characteristics compared with the group with poor

outcome (MGOS 1 ? 2). Because of its inhomogeneous

Table 6 Cut-off points,

sensitivities, specificities and

AUC values of S100B

predicting poor outcome in

patients divided by the GOS and

MGOS

AUC area under the curve,

CI confidence interval

Cut-off value Specificity Sensitivity AUC 95% CI of AUC P value

GOS

S100B day 1 3.59 100 16 0.647 0.561–0.786 \0.01

S100B day 2 0.9 100 31 0.705 0.591–0.819 \0.01

S100B day 3 1.2 100 36 0.792 0.675–0.909 \0.001

S100B day 4 0.72 100 27 0.722 0.606–0.837 \0.01

S100B day 5 0.63 100 30 0.725 0.605–0.844 \0.01

S100B day 6 0.3 100 25 0.635 0.502–0.768 0.05

S100B day 7 0.26 100 21 0.673 0.530–0.817 \0.03

S100B peak value 3.59 100 31 0.727 0.626–0.829 \0.001

MGOS

S100B day 1 3.59 100 12.5 0.734 0.583–0.886 \0.01

S100B day 2 1.17 100 40 0.819 0.676–0.961 \0.001

S100B day 3 1.2 100 50 0.953 0.900–1.000 \0.001

S100B day 4 0.72 100 36 0.841 0.736–0.947 \0.001

S100B day 5 0.63 100 38.5 0.771 0.619–0.923 \0.01

S100B day 6 0.3 100 31 0.717 0.545–0.890 \0.02

S100B day 7 0.26 100 25 0.683 0.503–0.864 0.07

S100B peak value 3.59 100 35 0.773 0.627–0.920 \0.01

0,0 0,2 0,4 0,6 0,8 1,0

0,0

0,2

0,4

0,6

0,8

1,0

1.17

MGOS: AUC 0.819 GOS: AUC 0.705

S100B day 2

Sen

sitiv

ity (

%)

1-Specificity (%)

0.9

0,0 0,2 0,4 0,6 0,8 1,0

0,0

0,2

0,4

0,6

0,8

1,0

1.2

MGOS: AUC 0.953 GOS: AUC 0.792

S100B day 3

Sen

sitiv

ity (

%)

1-Specificity (%)

1.2

Fig. 4 ROC analysis of serum S100B predicting unfavorable outcome in patients divided by the MGOS and GOS. Solid circles mark S100B

cut-off points with 100% specificity. S100B cut-off values are given in lg/l. AUC denotes area under the curve

Clin Res Cardiol (2012) 101:533–543 541

123

constellation and the unclear neurological state of its

patients, the time course of the biomarkers NSE and S100B

in the MGOS 0 group, which resembles the one in the good

outcome group (MGOS 3–5), may not be reliable for fur-

ther interpretations and conclusions.

We recently demonstrated that on average, mean NSE

cut-off values and sensitivities were higher (56 ± 35 ng/

ml, 56 ± 18%) in studies excluding patients with unknown

cerebral outcome compared with studies which included all

patients who died into the category of poor outcome

(41 ± 17 ng/ml, 44 ± 25%) [6]. This is partly in line with

the results of the present study. While the cut-off values of

both classifications were not significantly different, mean

sensitivities calculated from days 1 to 7 by the MGOS

(60 ± 29.3%) were significantly higher compared with the

GOS (48.4 ± 22.8%). In addition, the AUC was also found

to be significantly higher in the MGOS (0.89 ± 0.1)

compared with the GOS (0.8 ± 0.1). In analogy, these

findings could be complemented for S100B in a similar

manner.

Based on the aforementioned findings, we propose the

use of the MGOS to allow a more stringent distinction: in

studies evaluating neurological outcome by biochemical

or clinical markers it is of utmost importance to differ-

entiate between death with certified brain damage and

death without certified brain damage (e.g., cardiogenic

and/or septic shock, pulmonary embolism, aortic dis-

section). Hypoxic brain damage should be documented

by neurological investigations, cranial CT/MRI, elec-

trophysiological evaluations, and/or TCD ultrasonogra-

phy. Only those patients who die with certified brain

damage should be included in MGOS 1. Likewise, GOS

1 may be specified into subgroup 1a (death from any

cause) and 1b (death with certified brain damage), and

the GP-CPC score category 5 should be divided into 5a

(death from any cause) and 5b (death with certified brain

death).

While some studies focusing on the GP-CPC allocated

GP-CPC 3 to GP-CPC 4–5 into the group of poor outcome,

all studies using the GOS classification included patients

with GOS 3 (*GP-CPC 3) to those with GOS 4–5 in the

group with good outcome [6]. In future studies, a consensus

may be established to align the complementary categories

of the two scores to make comparison between studies

easier.

Several other studies aimed to investigate the prediction

power for long-term neurologic outcome either of NSE or

S100B in patients with CA after primary successful CPR.

But as different as the particular study design, which varies

from applied neurological outcome tests to the defined

times for blood samples, so dissimilar are the results:

Shinozaki et al. measured NSE and S100B on admission to

hospital and at 6 and 24 h after CA [20]. They report

S100B as more reliable as an early predictor of poor neu-

rological outcome measured by GP-CPC within 24 h after

CA than NSE and recommend already a clinical applica-

tion. Tiainen et al. found in a comparison of patients with

CA followed by therapeutic hypothermia versus those

followed by normothermia a decrease in levels of serum

NSE but not S100B, which were assessed 24, 36, and 48 h

after CA [21]. After neurologic assessment by using the

GP-CPC, they concluded that the decreased NSE levels

may indicate selective attenuation of delayed neuronal

death by therapeutic hypothermia in those patients. Grubb

et al. finally reported that both NSE and S100B concen-

trations measured 24–48 h after CA provided useful

information to identify patients at risk of significant cog-

nitive impairment [22]. In our clinical study, we also found

that both NSE and S100B are capable prediction parame-

ters for long-term neurologic outcome of patients with CA

and initially successful CPR. But similar to Tiainen et al.

the prediction power of NSE was higher as compared with

S100B in our setup. The reason for the obvious differences

in all these studies with regard to our results may be

attributed not only to the different classification of patient’s

neurological state by MGOS, but also to the various time

points of blood sample assessment. Whether there is such a

strong dependency of NSE and S100B prediction power

from the point of their assessment, e.g., a better prediction

power of S100B in the first 24 h after CA or a higher

prediction power of NSE in the following period as sup-

posed by the studies mentioned above may be topic of

further investigations.

Study limitations

The assessment of neurological status of patients who died

as a consequence of cardiogenic shock (17.5%), a combi-

nation of cardiogenic and septic shock (7.2%) and septic

shock (2.1%), was not possible. These patients were clas-

sified as patients with unclear neurological outcome

(MGOS 0). Only if the diagnosis of hypoxic brain damage

could be documented prior to death, e.g., by cranial CT or

MRI, these patients could be included in the MGOS 1

group. However, because stabilization procedures took

precedence, cranial imaging was not eligible in every

patient.

Conclusions

Inclusion of deaths without certified brain damage into the

group of patients with poor outcome in neurological out-

come studies will lead to underestimation of the prognostic

power of the investigated biochemical marker. The MGOS

will help to avoid this bias.

542 Clin Res Cardiol (2012) 101:533–543

123

Acknowledgments We would like to thank the medical and nursing

staff of the ICU of the Medical Clinic I, Department of Cardiology,

RWTH Aachen University, Germany, for supporting this study. This

work contains data from the doctoral thesis of Julia S. Kuhnen,

RWTH Aachen University, Germany.

Conflict of interest The authors declare that they have no conflict

of interest.

References

1. (2005) American Heart Association guidelines for cardiopulmo-

nary resuscitation and emergency cardiovascular care part 3:

overview of CPR. Circulation 112:1–203

2. Puttgen HA, Pantle H, Geocadin RG (2009) Management of

cardiac arrest patients to maximize neurologic outcome. Curr

Opin Crit Care 15:118–124

3. Wijdicks EF, Hijdra A, Young GB, Bassetti CL, Wiebe S (2006)

Quality Standards Subcommittee of the American Academy of

Neurology. Practice parameter: prediction of outcome in coma-

tose survivors after cardiopulmonary resuscitation (an evidence-

based review): report of the Quality Standards Subcommittee of

the American Academy of Neurology. Neurology 67:203–210

4. Shinozaki K, Oda S, Sadahiro T, Nakamura M, Hirayama Y, Abe

R, Tateishi Y, Hattori N, Shimada T, Hirasawa H (2009) S-100B

and neuron-specific enolase as predictors of neurological out-

come in patients after cardiac arrest and return of spontaneous

circulation: a systematic review. Crit Care 13:121

5. Cummins RO, Chamberlain DA, Abramson NS, Allen M, Baskett

PJ, Becker L, Bossaert L, Delooz HH, Dick WF, Eisenberg MS

(1991) Recommended guidelines for uniform reporting of data

from out-of-hospital cardiac arrest: the utstein style. A statement

for health professionals from a task force of the American Heart

Association, the European Resuscitation Council, the Heart and

Stroke Foundation of Canada, and the Australian Resuscitation

Council. Circulation 84:960–975

6. Rana OR, Saygili E, Schiefer J, Marx N, Schauerte P (2011)

Biochemical markers and somatosensory evoked potentials in

patients after cardiac arrest: the role of neurological outcome

scores. J Neurol Sci 305:80–84

7. Jennett B, Bond M (1975) Assessment of outcome after severe

brain damage. Lancet 1:480–484

8. Abanador-Kamper N, Kamper L, Wolfertz J, Dinh W, Thurmann

P, Seyfarth M (2011) Influence of algorithm-based analgesia and

sedation in patients after sudden cardiac arrest. Clin Res Cardiol.

doi:10.1007/s00392-011-0378-0

9. Rana M, W Schroder J, Saygili E, Hameed U, Benke D, Hoff-

mann R, Schauerte P, Marx N, Rana OR (2011) Comparative

evaluation of the usability of 2 different methods to perform mild

hypothermia in patients with out-of-hospital cardiac arrest. Int J

Cardiol 152:321–326

10. Jacobshagen C, Pelster T, Pax A, Horn W, Schmidt-Schweda S,

Unsold BW, Seidler T, Wagner S, Hasenfuss G, Maier LS (2010)

Effects of mild hypothermia on hemodynamics in cardiac arrest

survivors and isolated failing human myocardium. Clin Res

Cardiol 99:267–276

11. Pfeifer R, Hopfe J, Ehrhardt C, Goernig M, Figulla HR, Voss A

(2011) Autonomic regulation during mild therapeutic hypother-

mia in cardiopulmonary resuscitated patients. Clin Res Cardiol

100:797–805

12. Koester R, Kaehler J, Barmeyer A, Mullerleile K, Priefler M,

Soeffker G, Braune S, Nierhaus A, Meinertz T, Kluge S (2011)

Coronary angiography and intervention during hypothermia can

be performed safely without cardiac arrhythmia or vasospasm.

Clin Res Cardiol 100:1013–1019

13. Damian MS, Ellenberg D, Gildemeister R, Lauermann J, Simonis

G, Sauter W, Georgi C (2004) Coenzyme Q10 combined with

mild hypothermia after cardiac arrest: a preliminary study. Cir-

culation 110:3011–3016

14. Turedi S, Gunduz A, Mentese A, Dasdibi B, Karahan SC, Sahin

A, Tuten G, Kopuz M, Alver A (2009) Investigation of the

possibility of using ischemia-modified albumin as a novel and

early prognostic marker in cardiac arrest patients after cardio-

pulmonary resuscitation. Resuscitation 80:994–999

15. Zandbergen EG, Hijdra A, Koelman JH, Hart AA, Vos PE,

PROPAC Study Group et al (2006) Prediction of poor outcome

within the first 3 days of postanoxic coma. Neurology 66:62–68

16. Reisinger J, Hollinger K, Lang W, Steiner C, Winter T, Ze-

indlhofer E, Mori M, Schiller A, Lindorfer A, Wiesinger K,

Siostrzonek P (2007) Prediction of neurological outcome after

cardiopulmonary resuscitation by serial determination of serum

neuron-specific enolase. Eur Heart J 28:52–58

17. Lettieri C, Savonitto S, De Servi S, Guagliumi G, LombardIMA

Study Group et al (2009) Emergency percutaneous coronary

intervention in patients with ST-elevation myocardial infarction

complicated by out-of-hospital cardiac arrest: early and medium-

term outcome. Am Heart J 157:569–575

18. Garot P, Lefevre T, Eltchaninoff H, Morice MC, Tamion F, Abry

B, Lesault PF, Le Tarnec JY, Pouges C, Margenet A, Monchi M,

Laurent I, Dumas P, Garot J, Louvard Y (2007) Six-month out-

come of emergency percutaneous coronary intervention in

resuscitated patients after cardiac arrest complicating ST-eleva-

tion myocardial infarction. Circulation 115:1354–1362

19. Kouraki K, Schneider S, Uebis R, Tebbe U, Klein HH, Janssens

U, Zahn R, Senges J, Zeymer U (2011) Characteristics and

clinical outcome of 458 patients with acute myocardial infarction

requiring mechanical ventilation. Results of the BEAT registry of

the ALKK-study group. Clin Res Cardiol 100:235–239

20. Shinozaki K, Oda S, Sadahiro T, Nakamura M, Abe R, Nakada

TA, Nomura F, Nakanishi K, Kitamura N, Hirasawa H (2009)

Serum S-100B is superior to neuron-specific enolase as an early

prognostic biomarker for neurological outcome following car-

diopulmonary resuscitation. Resuscitation 80:870–875

21. Tiainen M, Roine RO, Pettila V, Takkunen O (2003) Serum

neuron-specific enolase and S-100B protein in cardiac arrest

patients treated with hypothermia. Stroke 34:2881–2886

22. Grubb NR, Simpson C, Sherwood RA, Abraha HD, Cobbe SM,

O’Carroll RE, Deary I, Fox KA (2007) Prediction of cognitive

dysfunction after resuscitation from out-of-hospital cardiac arrest

using serum neuron-specific enolase and protein S-100. Heart

93:1268–1273

Clin Res Cardiol (2012) 101:533–543 543

123