Resuscitation 81 (2010) 796–803

Contents lists available at ScienceDirect

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

Clinical paper

Comparing the survival between extracorporeal rescue and conventional

resuscitation in adult in-hospital cardiac arrests: Propensity analysis of
three-year data夽
Jou-Wei Lin a , Ming-Jiuh Wang b , Hsi-Yu Yu b , Chih-Hsien Wang b , Wei-Tien Chang c ,
Jih-Shuin Jerng d , Shu-Chien Huang c , Nai-Kuan Chou b , Nai-Hsin Chi b , Wen-Je Ko b ,
Ya-Chen Wang b , Shoei-Shen Wang b , Juey-Jen Hwang a,d , Fang-Yue Lin b , Yih-Sharng Chen b,∗
a
Cardiovascular Center, National Taiwan University Hospital Yun-Lin Branch, Dou-Liou City, Yun-Lin, Taiwan
b
Department of Surgery, National Taiwan University Hospital, Taipei, Taiwan
c
Department of Emergency Care, National Taiwan University Hospital, Taipei, Taiwan
d
Department of Medicine, National Taiwan University Hospital, Taipei, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: Aim: Extracorporeal cardiopulmonary resuscitation (ECPR) has been shown to have survival benefit over
Received 29 July 2009 conventional CPR (CCPR) in patients with in-hospital cardiac arrest of cardiac origin. We compared the
Received in revised form 19 February 2010 survival of patients who had return of spontaneous beating (ROSB) after ECPR with the survival of those
Accepted 10 March 2010
who had return of spontaneous circulation (ROSC) after conventional CPR.
Methods: Propensity score-matched cohort of adults with in-hospital prolonged CPR (>10 min) of car-
Keywords:
diac origin in a university-affiliated tertiary extracorporeal resuscitation center were included in this
Cardiopulmonary resuscitation
study. Fifty-nine patients with ROSB after ECPR and 63 patients with sustained ROSC by CCPR were ana-
Extracorporeal membrane oxygenation
Return of spontaneous circulation
lyzed. Main outcome measures were survival at hospital discharge, 30 days, 6 months, and one year, and
Propensity score neurological outcome.
Cardiac arrest Results: There was no statistical difference in survival to discharge (29.1% of ECPR responders vs. 22.2% of
CCPR responders, p = 0.394) and neurological outcome at discharge and one year later. In the propensity
score-matched groups, 9 out of 27 ECPR patients survived to one month (33.3%) and 7 out of 27 CCPR
patients survived (25.9%). Survival analysis showed no survival difference (HR: 0.856, p = 0.634, 95% CI:
0.453–1.620) between the groups, either at 30 days or at the end of one year (HR: 0.602, p = 0.093, 95%
CI: 0.333–1.088).
Conclusions: This study failed to demonstrate a survival difference between patients who had ROSB after
institution of ECMO and those who had ROSC after conventional CPR. Further studies evaluating the role
of ECMO in conventional CPR rescued patients are warranted.
© 2010 Elsevier Ireland Ltd. All rights reserved.

1. Introduction resuscitation outcome.1–9 These parameters should be carefully

In-hospital cardiopulmonary resuscitation (IHCPR) remains a
critical situation with high immediate mortality despite improve-
ments in resuscitation procedures and responses.1–4 Several
factors, including initial rhythm, resuscitation duration, underly-
ing causes, and initial resuscitative effort, may be related to the
夽 A Spanish translated version of the abstract of this article appears as Appendix
in the final online version at doi:10.1016/j.resuscitation.2010.03.002.
∗ Corresponding author at: Department of Surgery, National Taiwan University
Hospital, 7 Chung-Shan South Road, Taipei 100, Taiwan. Fax: +886 2 23956934.
E-mail addresses: yschen1234@gmail.com, yschen11@mail2000.com.tw
(Y.-S. Chen).
considered when comparison or interpretation of different studies
on resuscitation is made.
Extracorporeal circulation has been gradually incorporated into
cardiopulmonary resuscitation (CPR) procedures and other scenar-
ios not only in pediatric patients, but also in adults.10–17 Studies
have demonstrated that extracorporeal membrane oxygenation
(ECMO) provides additional benefits in IHCPR with cardiac ori-
gin when compared to conventional cardiopulmonary resuscitation
(CCPR).11,12,14 It has also been shown that extracorporeal cardiopul-
monary resuscitation (ECPR) provides a benefit to victims with
prolonged CPR under CCPR.
However, the role of extracorporeal support in patients who
regain sustained return of spontaneous circulation (ROSC) after
CCPR remains unknown because previous studies have not

0300-9572/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.resuscitation.2010.03.002
 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803
addressed this issue.11–16 Conducting a randomized trial in patients
who have just regained ROSC after CPR is difficult. Therefore,
instead, we tried to match the pre-arrest conditions of ECPR and
CCPR patients. The objective of the present study was to compare
survival and neurological outcomes between patients with sus-
tained ROSC after CCPR and those regaining return of spontaneous
(ventricular) beating (ROSB) after ECPR.
2. Methods
2.1. Clinical setting
The study was conducted in a university-affiliated medical cen-
ter with 227 intensive care beds. The CPR Team, consisting of a
senior resident, several other residents, a respiratory therapist, a
head nurse, and several registered nurses, was in charge of resus-
citation for in-hospital cardiac arrest (IHCA). From 2003 onward,
each cardiac arrest event was recorded in a web-based database
system.18,19 The accompanying CPR process was also entered into
the database and reviewed by the IHCA Task Force. The underlying
etiology causing the cardiac arrest and resulting in the resuscitation
was determined by the Task Force Committee.14
2.2. Data collection
Adult IHCA victims receiving IHCPR from 2004 to 2006 were
enrolled in the study. The inclusion criteria were the following:
patients aged 18–75 years with circulatory arrest of cardiac origin,
as judged by two independent members in the IHCA Task Force
committee.
Patients who underwent CCPR for >10 min for an arrest of
proven cardiac origin were recruited as controls. Patients who
were considered for the use of extracorporeal CPR had prolonged
CPR (>10 min) without sustained ROSC. The definition of sus-
tained ROSC was continuous maintenance of the spontaneous
circulation for ≥20 min. The decision to call the ECMO team was
made mainly by the attending doctors. Strict contraindications to
the use of ECMO included severe trauma, uncontrollable hem-
orrhage, terminal malignancy, age >75 years, irreversible brain
damage, and signed consent for “do not attempt resuscitation.”
The equipment and management of ECMO has been previously
reported.14 The main component of the ECMO circuit was a
heparin-bonded surface embedded in a centrifugal pump and
hollow-fiber oxygenation (Medtronic Incorporated, Ahaheim, CA,
USA).
Only patients with arrests of proven cardiac etiology were
recruited as cases in the CCPR group and the ECPR group. In the
former, cases with sustained ROSC at the end of resuscitation were
recruited for further analysis. In the ECPR group, we defined ROSB
after extracorporeal assistance, and then included cases with ROSB
only as the comparison group. ROSB in the ECPR group, in contrast
to ROSC in the CCPR group, was maintained by artificial circulation.
The presence of atrial electrical activity without ventricular electric
activity was not considered ROSB.
2.3. Study design
We prospectively examined the survival difference between all
patients in the ECPR group with ROSB (E-rosb) and all patients
the CCPR group with sustained ROSC (C-rosc). A propensity score-
based matching process was then used to match ECPR and CCPR
patients to obtain two groups with equalized pre-arrest charac-
teristics in order to simulate the characteristics of a randomized
trial.
797
2.4. Propensity score methods
Propensity score matching is a method used to balance observed
covariates in two treatment groups, e.g., ECPR and CCPR. In
the present study, the propensity score was the conditional
probability of receiving ECPR, as a binary-dependent variable,
under a set of measurements in patients with sustained ROSC
or ROSB after CPR.20 All baseline pre-arrest characteristics and
parameters potentially related to survival were incorporated
into a logistic regression model as covariates. This matching
process made the observational studies simulate a random-
ized trial and reduced selection bias between the two study
groups.21
In ROSB and ROSC responders, a number of parameters
were investigated: age; sex; initial cardiac rhythm such as
ventricular tachycardia/ventricular fibrillation (VT/VF); pulseless
electric activity (PEA) or asystole; CPR duration; the tim-
ing and location in which the CPR occurred; presence of
comorbidities (diabetes mellitus, hypertension, dyslipidemia, and
malignancy), chronic obstructive pulmonary disease (COPD),
pre-existing coronary artery disease or cerebrovascular disease
(CVAs); abnormal liver function; and chronic need for dialy-
sis.
These parameters were combined into a non-parsimonious mul-
tivariable logistic regression model to predict the probability of the
use of ECPR. The predicted probability derived from the logistic
equation was then used as the propensity score for each individual
patient.
2.5. Study cohort assembly
ECPR and CCPR responders were pooled together and classified
according to their propensity scores in ascending order. They were
matched by their propensity score in 1:1, 2:2, 3:3, and 4:4 blocks.
Subjects that failed to find an appropriate match within the accept-
able rank range were excluded from further analysis. Remaining
patients constituted well-matched 1:1 extracorporeal-assisted CPR
responders and CCPR responders with equalized baseline prognos-
tic factors.
2.6. Endpoints
The primary endpoint was survival to hospital discharge. The
secondary endpoint was survival at 30 days, 6 months, and 365
days. Neurological status was also assessed.
2.7. Statistical analysis
The comparison between the two selected groups was made
by hazard ratio, 95% confidence interval, and p values. Patients
in the CCPR group who received extracorporeal support at a
later time were regarded as no ECMO support (on an inten-
tion to treat basis). Survival duration was determined from the
time of entry into the study to death or censorship at discharge,
one month, and one year. The proportions of each group with
better neurological outcomes, defined as Cerebral Performance
Category score 1 or 2, were compared by 2 and odds ratio
(OR).
A logistic regression model was used to carry out the propen-
sity match under SPSS 13.0 (SPSS Incorporated, Chicago, IL, USA).
OR and 95% confidence intervals (95% CI) were used to repre-
sent the predilection for a subject to be placed on extracorporeal
assistance. Propensity scores in both groups were shown as box
plots.
Patients with lower and higher propensity scores were excluded
from further analysis if an appropriate match could not be found.
798 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803

Fig. 1. Flow chart for IHCA selection over a 36-month period. Panel A: Inclusion criteria for the conventional cardiopulmonary resuscitation group. Abbreviations: IHCA:
in-hospital cardiac arrest; CPR: cardiopulmonary resuscitation; CCPR: conventional CPR; ROSC: return of spontaneous circulation. Panel B: Inclusion criteria for the
extracorporeal-assisted CPR group. Abbreviations: ECPR: extracorporeal-assisted CPR; ROSB: return of spontaneous beating.

Previously, we have reported that between the propensity-score 3. Results

matched groups, there was about a 50% risk reduction in sur-
vival to discharge (hazard ratio [HR] 0.51, 95% CI 0.35–0.74,
p < 0.0001), 30-day survival (HR 0.47, 95% CI 0.28–0.77, p = 0.003),
and one-year survival (HR 0.53, 95% CI 0.33–0.83, p = 0.006) favor-
ing extracorporeal CPR over conventional CPR.14 For example, if
80% of ICHA patients died and if ECPR would result in a risk
reduction of 50%, it was estimated that 27 patients in either
arm would be needed to achieve 80% statistical power (given
˛ = 0.05).
After the 1:1 matched groups were assembled, another box
plot was drawn to show the equally distributed propensity scores
between matched groups. Underlying patient characteristics were
compared with the 2 test and Fisher’s exact test (when case num-
bers were small) for categorical data or with Student’s t-test for
continuous variables. Kaplan–Meier curves were plotted to demon-
strate the survival trend and survival difference. A Cox regression
model was used to compare the hazard [hazard ratio (HR), 95% con-
fidence interval, and p value] between the two matched groups,
with adjustment for potential prognostic factors. Neurological
outcomes were compared by a multi-scale 2 test (Cerebral Per-
formance Category scores 1 and 2: good neurological outcomes;
3 and 4: poor neurological outcomes; and 5: death). p < 0.05 was
considered significant.
3.1. Patient characteristics
A total of 1039 adult IHCA events occurred in the three years
of the study, with 975 cases receiving CCPR and 64 cases receiv-
ing ECPR (Fig. 1A for CCPR, and 1B for ECPR). Among the CCPR
group, 545 patients were considered to have arrest of a cardiopul-
monary origin, 187 patients were aged 18–75 years, and 113 had
events that were witnessed and received prolonged CPR (CPR dura-
tion > 10 min). We restricted our analysis to the 63 cases (Table 1,
the C-rosc group; 41 men and 22 women) with ROSC after pro-
longed CCPR. In addition, 59 patients received prolonged ECPR. Four
of these patients did not regain ROSB after extracorporeal support,
and were excluded from the analysis. Finally, 55 patients (Table 1,
E-rosb; 47 men and 8 women) were recruited into the ECPR group
(E-rosb group; Fig. 1B).
Baseline characteristics between these two selected groups
were compared. Among the 55 patients who received ECPR, there
were more male subjects, a higher prevalence of dyslipidemia, and
possibly more previously known cardiovascular diseases, uneven
causes of cardiac arrest, and a higher need of inotropic agents before
cardiac arrest (Table 1). Analysis of pre- and post-CPR variables
revealed that more CCPR took place at midnight. CPR duration was
 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803 799

Table 1
Baseline Characteristics of the ECPR group with ROSB (group E-rosb) and the con-
ventional CPR group with ROSC (group C-rosc).

Variable Group E-rosb Group C-rosc p-Value

N 55 63
Sex
Male, n (%) 47 (85.5) 41 (65.1) 0.011

Age (year)
Mean (SD) 59.0 (11.7) 60.6 (12.7) 0.463
Median (range) 62.3 (21–73) 65.0 (22–75) 0.229
Age >60 years, n (%) 34 (61.8) 40 (63.3) 0.851

Pre-existing comorbidity
Diabetes mellitus 21 (38.2) 19 (30.2) 0.358
Hypertension 29 (52.7) 30 (47.6) 0.580
Dyslipidemia 10 (18.2) 2 (3.2) 0.012
Malignancy 5 (9.1) 15 (13 3) 0.769
Lung insufficiency 2 (3.6) 5 (7.9) 0.447
Stroke 10 (18.2) 6 (9.5) 0.171
Chronic renal disease 5 (9.1) 10 (15.9) 0.270
Cardiovascular disease 35 (63.6) 30 (47.6) 0.081
Chronic hepatitis 2 (3.6) 6 (9.5) 0.281

Causes of arrest
ACS 36 (65.5) 46 (73.0) 0.022
Congestive heart failure 4 (7.3) 9 (14.3)
Myocarditis 3 (5.5) 1 (1.6)
Post-cardiotomy 7 (12.7) 0
Pulmonary embolism 1 (1.8) 0
Unspecified cardiac causes 3 (5.5) 7 (11.1)

Initial rhythm
VT/VF 28 (50.9%) 26 (41.3%) 0.360
PEA 15 (27.3%) 25 (39.7%)
Asystole 12 (21.8%) 12 (19.0%)

Department type
Internal medicine 35 (63.6%) 45 (71.4%) 0.205
Surgery 20 (36.4%) 16 (25.4%)
Other 0 2 (3.2%)

IE before CPR
n (%) 51 (92.7) 21 (39.6) <0.001
Mean (SD) 54 (66.6) 25.0 (20.3) 0.366
Median (range) 40 (0–324) 19.1 (4.6–73)

ACS: acute coronary syndrome, CPR: cardiopulmonary resuscitation, C-rosc:

conventional CPR with sustained ROSC, IE: intropic equivalent (␮g/kg/min,
dopamine + dobutamine + 100 × epinephrine + 100 × norepinephrine + 100 × Fig. 2. Distribution of propensity scores in the CCPR and ECPR groups before match-
isoproterenol + 15 × milrinone), E-rosc: extracorporeal CPR with ROSB, PEA: ing (upper panel A) and after matching (lower panel B).
pulseless electrical activity, VT/VF: ventricular tachycardia/ventricular fibrillation.

selection for extracorporeal assistance (OR: 7.816, p = 0.032, 95%

shorter, and the maximal concentration of lactic acid was lower.
Patients who regained ROSC after CCPR had less opportunity to
receive further intervention (Table 2).
The outcome for the two groups is shown in Table 3. A sur-
vival benefit of extracorporeal support only was seen in the first
three days. However survival to discharge did not differ signif-
icantly between the E-rosb group (29.1%, n = 16) and the C-rosc
group (22.2%, n = 14) (p = 0.394). Neurological outcome, defined as
the Glasgow–Pittsburgh cerebral performance categories (CPC), at
discharge and at one year, was similar in the extracorporeal support
and conventional resuscitation groups.
3.2. Assignment of propensity score
Propensity analysis with a logistic regression model demon-
strated the odds with which a patient would have undergone
ECPR if certain clinical characteristics were present. All of the
patients included in the propensity analysis were ROSB/ROSC
responders from ECPR or CCPR. The logistic regression analy-
sis showed that each minute increase in CPR time would result
in a 1.046-fold increase in the odds (OR: 1.046, p = 0.002, 95%
CI: 1.020–1.073) of being selected for extracorporeal assistance.
The presence of dyslipidemia was also positively associated with
CI: 1.19–51.12). The presence of end-stage renal disease (OR:
0.105, p = 0.027, 95% CI: 0.014–0.769) was less likely to be asso-
ciated with ECPR. CPR occurring during the midnight shift was
less likely to result in the use of ECPR (OR: 0.177, p = 0.018,
95% CI: 0.042–0.747). Parameters such as initial rhythm (VT/VF,
PEA, or asystole), the presence of diabetes, hypertension, can-
cer, COPD, CVA, and liver function, did not affect the decision
to choose ECPR or CCPR (p > 0.05). All the predictors remained
in the model and were used to calculate the propensity score
for ECPR. Propensity score was 0.6657 ± 0.2498 in the 55 cases
of ECPR and 0.2918 ± 0.2403 in the 63 CCPR patients. The dif-
ference between the two groups in propensity predisposing
to the selection of ECPR was statistically significant (p < 0.001)
(Fig. 2A).
3.3. Characteristics of the matched groups
The propensity score-matching process selected 27 ECPR
responders (21 men and 6 women; mean age, 59 ± 11 years) and
27 CCPR responders (23 men and 4 women; mean age, 60 ± 13
years). None of the 27 CCPR responders received extracorporeal
support later in their hospital course. Baseline characteristics were
comparable between the two groups (Table 4). Propensity score
800 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803

Table 2
CPR and post-CPR variables in the E-rosb group (group E) and C-rosc group.

Variable Group E-rosb Group C-rosc P

N 55 63
ROSC (ROSB) 55 (100) 63 (100) 1
Witnessed (%) 55 (100) 63 (100) 1
Defibrillationa , * 55 (100) 43 (68.2) <0.001

Time period of CPR episode*

Period A (7:01–15:00 h) 19 (34.5) 16 (25.4) 0.007
Period B (15:01–23:00 h) 31 (55.4) 26 (41.3)
Period C (23:01–7:00 h) 5 (9.1) 21 (33.3)

Event of location
ICU/OR/Cath lab 31 (56.4) 24 (38.1) 0.104
ER/ward 24 (43.6) 38 (60.3)
Others 0 1 (1.6)

Intubated prior to arrest

39 (70.1) 45 (71.4) 0.950
*
CPR duration
Mean (SD), min N = 55 48.7 (26.9) N = 63 31.4 (17.1) <0.001
Median (range) 40 (16–150) 30 (11–80) <0.001

Available maximal lactic acid level in 24 h*

Mean (SD) N = 51 12.5 (6 3) N = 206.2 (5 5) <0.001
Median (range) 12.0 (2.7–39.7) 3.7 (1.1–20)

Subsequent intervention
Any of the following 35 (63.6) 14 (22.2) <0.001
Revascularization 26 (47.3) 6 (9.5) <0.001
VAD 2 (3.6) 0 0.215
HTx 4 (7.3) 0 0.044
ECMO – 3 (14.2)
Others 9 (16.4) 6 (9.5)
Valve replacement 3, SVR 2 Tapping 3, pacing 2, DAA graft1

Hospital stay after CPR (day)

N = 55 N = 63
Mean 19.2 (21.9) 17.5 (34.8) 0.752
Median (range)* 11 (1–93) 12.0 (1–174) 0.007
*
p < 0.05.
a
Defibrillation before or during CPR, Cath lab: catheterization laboratory, CPR: cardiopulmonary resuscitation, C-rosc: conventional CPR with ROSC, ER: emergency room,
ECMO: extracorporeal membrane oxygenation, E-rosb: extracorporeal CPR with ROSB, HTx: heart transplantation, ICU: intensive care unit, OR: operation room, ROSB: return
of spontaneous heart beating (for group E), ROSC: return of spontaneous circulation (for group C), and VAD: ventricular assist device.

Table 3
Outcome of the E-rosb group and C-rosc group.

Variables Group E-rosb Group C-rosc P

N 55 63

ECMO duration (h)

Mean (SD) 112 (130) – –
Median (range) 71 (2–77.1) –
Weaned off ECMO 27 (49.1) – –
Survival to discharge 16 (29.1) 14 (22.2) 0.394
OR (95% CI) 1.436 (0.625–3.298)

CPC status at discharge

1 or 2 13 (23.6) 12 (19.1) 0.543
OR (95% CI) 1.315 (0.543–3.186)
Survival at 24 h* 53 (96.3) 43 (68.3) 0.001
OR (95% CI) 12.33 (2.73–55.7)
Survival at 3 days* 43 (78.2) 35 (55.6) 0.011
OR (95% CI) 2.867(1.275–6.445)
Survival at 14 days 24 (43.6) 24 (38.1) 0.541
OR (95% CI) 1.258 (0.602–2.628)
Survival at 30 days 19 (34.5) 17 (27 0) 0.374
OR (95% CI) 1.428 (0.650–3.135)
Survival at 6 m 16 (29.1) 13 (20.6) 0.289
OR (95% CI) 1.578 (0679–3.667)
Survival at one year 11 (20.0) 11 (17.5) 0.724
OR (95% CI) 1.182(0.468–2.987)

CPC status at one year

1 or 2 8 (14.5) 10 (15.9) 0.841
OR (95% CI) 0.902 (0.329–2.475)
*
p < 0.05; ECMO: extracorporeal membrane oxygenation, m: month, CI: confidence interval, CPC: Cerebral Performance Category score, OR: odds ratio, and ROSC: return
of spontaneous circulation.
 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803 801

Table 4
Clinical characteristics and outcomes in propensity-score matched cases.

Group E-rosb-ps Group C-rosc-ps p

Matched case number 27 27 1.000

ROSB/ROSC 100% 100% 1.000
Age (mean ± SD) 59 ± 11 60 ± 13 0.850

CPR duration
(min, mean ± SD) 41.8 ± 19.8 40.0 ± 26.4 0.781
Male 77.8% 85.2% 0.484
Initial rhythm 0.817
VT/VF 55.6% 66.7%
PEA 25.9% 22.2%
Asystole 18.5% 11.1%

Baseline clinical characteristics

Diabetes 29.6% 18.5% 0.340
Hypertension 48.1 37.0% 0.407
Dyslipidemia 7.4% 3.7% 0.552
Malignancy 7.4% 11.1% 0.639
COPD 7.4% 7.4% 1.000
CVA 7.4% 14.8% 0.386
Abnormal liver function 7.4% 7.4% 1.000
Hemodialysis 11.1% 11.1% 1.000

Pre- and post-CPR characteristics

Location (intensive care unit/operation room/catheterization room) (n, %) 13 (48.1%) 13 (48.1%) 1.000
Timing (midnight) (n, %) 5 (18.5%) 5 (18.5%) 1.000
Department (n, %) 0.440
Medicine 17 (63.0%) 19 (70.4%)
Surgery 10 (37%) 7 (25.9%)
Others 0 1 (3.7%)
Subsequent intervention 13 (48.1%) 4 (14.8%) 0.018
CABG 7 (25.9%) 0 0.010
PCI 4 (14.8%) 3 (11.1%) 1.000
VAD 2 (7.4%) 0 0.491
HTx 1 (3.7%) 0 1.000

Cumulative survival
1 day 19 (70.4%) 16 (59.3%) 0.393
3 days 14 (51.9%) 14 (51.9%) 1.000
7 days 12 (44.4%) 11 (40.7%) 0.783
One month 9 (33.3%) 7 (25.9%) 0.551
6 months 8 (29.6%) 4 (14.8%) 0.327
One year 6 (22.2%) 3 (11.1%) 0.467

Neurological outcomes
CPC status at discharge (n, %) 0.460
1 or 2 7 5
3 or 4 1 0
CPC status at one year (n, %) 0.427
1 or 2 5 3
3 or 4 1 0

was 0.4797 ± 0.2219 in the ECPR group and 0.4909 ± 0.2277 in the vival analysis trimmed at 30 days did not show a survival difference
matched CCPR group (p = 0.984) (Fig. 2B). (HR: 0.856, p = 0.634, 95% CI: 0.453–1.620). There was no survival
difference at 6 months (HR: 0.588, p = 0.089, 95% CI: 0.319–1.084)
3.4. Primary endpoint and at one year (HR: 0.602, p = 0.093, 95% CI: 0.333–1.088) (Fig. 3).

In the 27 ECPR responders, 8 survived to discharge (29.6%). Five 4. Discussion

out of the 27 CCPR responders survived to discharge (18.5%). Mean
survival was 29.72 ± 7.18 days and 25.19 ± 9.66 days, respectively.
The multi-variable Cox regression model showed that extracorpo-
real assistance did not significantly reduce the hazard of mortality
during admission (HR: 0.618, p = 0.413, 95% CI: 0.325–1.176). In
addition, CPR duration and age increased the hazard of mortality
during admission (HR: 1.015 for each min of CPR, p = 0.056, 95%
CI: 1.000–1.031; HR: 1.043 for each year of age, p = 0.031, 95% CI:
1.004–1.084). Initial rhythm with VT/VF reduced the risk of in-
hospital mortality (HR: 0.389, p = 0.012, 95% CI: 0.186–0.814).
3.5. Secondary endpoint
At the end of one month, 9 out of 27 of the ECPR patients survived
(33.3%) and 7 out of 27 of the CCPR patients survived (25.9%). Sur-
In this prospective observational study, there was no significant
difference in short-term and one-year survival when the com-
parison was made only between the ECPR responders and CCPR
responders. Neurological outcomes were also comparable. The
propensity-score method that assembled comparable subgroups
also resulted in this conclusion. These results imply that survival
after ECPR was not necessarily better than that from CCPR once
circulation was regained after prolonged CPR.
Extracorporeal support could offer the last chance to survive.
In patients with ROSC, myocardial function may have the reserve
capability to maintain the circulation even though catecholamines
may be required for a short period. ROSB in the ECPR group may
not be comparable to ROSC in the CCPR group. ROSB after extra-
corporeal support was due to unloading of the ventricular and
802 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803
Fig. 3. Kaplan–Meier plot of the survival curves in the matched extracorporeal CPR
and conventional CPR groups; survival difference reaching borderline significance
at the end of one year.
systemic circulation. The phenomenon is similar to resumption of
the heart beat after cardiac arrest in cardiac surgery. ROSC by CCPR,
i.e., maintenance of spontaneous circulation for ≥20 min, required
more preserved myocardial function than the status in ROSB by
ECPR.
Even with the neutral results in this study, extracorporeal sup-
port may have some potential benefits. Extracorporeal support
could provide a quick and sustained method for cooling and re-
warming if a hypothermic strategy were to be applied. It could
also prevent and treat possible neurological complications after
hypothermia.22 Extracorporeal support could further supplement
circulatory output if shock persisted after ROSC. In our series, three
patients required extracorporeal support later in the clinical course
despite initial stabilization after CCPR. The interval between ROSC
and late extracorporeal support was 24 ± 12 h. Persistent shock
under extremely high concentration of catecholamines is an indi-
cation for extracorporeal support.23
Extracorporeal support also plays a role in the management
of shock.12,24,25 When adequate systolic blood pressure cannot be
maintained with inotropic equivalents at >40 ␮g/kg/min, extracor-
poreal support should be considered. In our series, two out of three
cases who received late extracorporeal support in the CCPR group
survived to hospital discharge. Whether the indication for extra-
corporeal support in these three patients should be re-emphasized
in persistent shock or in CPR is debatable.
4.1. Limitations
The major limitation of this study is the comparability between
the patients who had ROSB after institution of ECMO and those
who had ROSC after conventional CPR. The appropriate method for
obtaining matched study groups should be a randomization of the
patients who regained ROSC (ECMO vs. control) and had follow-up.
However, this approach is not available yet in our institute. The rea-
son we thought that the two groups studied here were comparable
was that all patients included in the analysis were of the highest
risk because CPR was not successful in the first 10 min.
Another assumption about the comparability of the two treat-
ment groups is that the differences between them could be
minimized by propensity analysis. However, although propensity
analysis increased the comparability of the two groups, non-
randomization is the primary problem in the present study. Also
the quality of CPR may not have been well-controlled between the
two groups. Although the propensity-score approach can reduce
the selection bias and confounding factors inherent in an observa-
tional study, a key limitation is that bias can remain if there are
unmeasured or unknown confounders that are not incorporated
into the propensity score.26,27 Moreover, the median propensity
scores for the patients in both groups were low, meaning that a
significant group of patients in the propensity matched cohort had
a lower probability for receiving ECPR and that the influence of this
on the survival outcomes remained uncertain.
The strict selection criteria and matching process might have
reduced the statistical power of the study. The analysis assumed
that if there was 50% risk reduction, including 27 patients in each
group would reach 80% statistical power. However, the results
showed only 38% risk reduction. Probably a much bigger case num-
ber will be needed to identify a significant difference. Conversely,
there were more subsequent interventions (the composite of
revascularization, late extracorporeal support, ventricular-assisted
device, and heart transplantation) in the ECPR group. The barely
seen survival benefit of ECPR may be attributed to observer bias
rather than extracorporeal support. A study involving interventions
such as extracorporeal support cannot be blinded, so observer bias
may confound the interpretation of results.
5. Conclusion
As compared with our previous study,14 we restricted this anal-
ysis to those who regained ROSC from CCPR and those who obtained
ROSB from ECPR. The present study failed to demonstrate a survival
difference between the patients who had ROSC after institution
of ECMO as compared to those who had ROSC after conventional
CPR. However, this study was limited by small case number and
the potential comparability problem between two groups. Fur-
ther trials randomizing post-resuscitation ROSC responders are still
warranted.
Conflict of interest
All authors did not have conflict of interest to declare.
Acknowledgments
This study was supported by grants from the National Science
Council, Taiwan: NSC 96-2314-B-002-039, 97-2314-B-002-046-
MY3, 98-2314-B-002-036-MY2, NTUH 97-S857, 97-END08, 98P17
and 98-S1061. We would like to thank Dr. Chia-Hsiun Chang for the
assistance in independent statistical review.
References
1. Peberdy MA, Kaye W, Ornato JP, et al. Cardiopulmonary resuscitation of adults
in the hospital: a report of 14720 cardiac arrests from the National Registry of
Cardiopulmonary Resuscitation. Resuscitation 2003;58:297–308.
2. Weil MH, Fries M. In-hospital cardiac arrest. Crit Care Med 2005;33:2825–30.
3. Peberdy MA, Ornato JP, Larkin GL, et al. Survival from in-hospital cardiac arrest
during nights and weekends. JAMA 2008;299:785–92.
4. Cooper S, Janghorbani M, Cooper G. A decade of in-hospital resuscitation: out-
comes and prediction of survival? Resuscitation 2006;68:231–7.
5. Ali B, Dudley Jr SC, Zafari AM. Time to defibrillation after in-hospital cardiac
arrest. N Engl J Med 2008;358:1632–3 [author reply 1633–34].
6. Manfredini R, Salmi R, Manfredini F. Survival patterns with in-hospital cardiac
arrest. JAMA 2008;299:2625 [author reply 2626–27].
7. Nadkarni VM, Larkin GL, Peberdy MA, et al. First documented rhythm and clin-
ical outcome from in-hospital cardiac arrest among children and adults. JAMA
2006;295:50–7.
8. Petursson P, Gudbjornsdottir S, Aune S, et al. Patients with a history of dia-
betes have a lower survival rate after in-hospital cardiac arrest. Resuscitation
2008;76:37–42.
 J.-W. Lin et al. / Resuscitation 81 (2010) 796–803
9. Danciu SC, Klein L, Hosseini MM, Ibrahim L, Coyle BW, Kehoe RF. A predic-
tive model for survival after in-hospital cardiopulmonary arrest. Resuscitation
2004;62:35–42.
10. de Mos N, van Litsenburg RR, McCrindle B, Bohn DJ, Parshuram CS. Pediatric in-
intensive-care-unit cardiac arrest: incidence, survival, and predictive factors.
Crit Care Med 2006;34:1209–15.
11. Thiagarajan RR, Laussen PC, Rycus PT, Bartlett RH, Bratton SL. Extracorporeal
membrane oxygenation to aid cardiopulmonary resuscitation in infants and
children. Circulation 2007;116:1693–700.
12. Chen YS, Chao A, Yu HY, et al. Analysis and results of prolonged resuscitation in
cardiac arrest patients rescued by extracorporeal membrane oxygenation. J Am
Coll Cardiol 2003;41:197–203.
13. Chen YS, Yu HY, Huang SC, et al. Extracorporeal membrane oxygenation sup-
port can extend the duration of cardiopulmonary resuscitation. Crit Care Med
2008;36:2529–35.
14. Chen YS, Lin JW, Yu HY, et al. Cardiopulmonary resuscitation with assisted extra-
corporeal life-support versus conventional cardiopulmonary resuscitation in
adults with in-hospital cardiac arrest: an observational study and propensity
analysis. Lancet 2008;372:554–61.
15. Nagao K, Hayashi N, Kanmatsuse K, et al. Cardiopulmonary cerebral resuscitation
using emergency cardiopulmonary bypass, coronary reperfusion therapy and
mild hypothermia in patients with cardiac arrest outside the hospital. J Am Coll
Cardiol 2000;36:776–83.
16. American heart association guidelines for cardiopulmonary resuscitation and
emergency cardiovascular care. Circulation 2005;112:IV1–203.
17. International consensus on cardiopulmonary resuscitation and emergency car-
diovascular care science with treatment recommendations. Part 4: advanced life
support. Resuscitation 2005;67:213–47.
18. Cummins RO, Chamberlain D, Hazinski MF, et al. Recommended guidelines
for reviewing, reporting, and conducting research on in-hospital resuscitation:
the in-hospital ‘Utstein style’. A statement for healthcare professionals from
803
the American Heart Association, the European Resuscitation Council, the Heart
and Stroke Foundation of Canada, the Australian Resuscitation Council, and the
Resuscitation Councils of Southern Africa. Resuscitation 1997;34:151–83.
19. Shih CL, Lu TC, Jerng JS, et al. A web-based Utstein style registry sys-
tem of in-hospital cardiopulmonary resuscitation in Taiwan. Resuscitation
2007;72:394–403.
20. Haviland A, Nagin DS, Rosenbaum PR. Combining propensity score matching
and group-based trajectory analysis in an observational study. Psychol Methods
2007;12:247–67.
21. Ahmed A, Zannad F, Love TE, et al. A propensity-matched study of the association
of low serum potassium levels and mortality in chronic heart failure. Eur Heart
J 2007;28:1334–43.
22. Scaife ER, Connors RC, Morris SE, et al. An established extracorporeal membrane
oxygenation protocol promotes survival in extreme hypothermia. J Pediatr Surg
2007;42:2012–6.
23. Liden H, Wiklund L, Haraldsson A, Berglin E, Hultman J, Dellgren G. Temporary
circulatory support with extra corporeal membrane oxygenation in adults with
refractory cardiogenic shock. Scand Cardiovasc J 2008;December:1–7 [Epub
ahead of print].
24. Chen JS, Ko WJ, Yu HY, et al. Analysis of the outcome for patients experiencing
myocardial infarction and cardiopulmonary resuscitation refractory to conven-
tional therapies necessitating extracorporeal life support rescue. Crit Care Med
2006;34:950–7.
25. Chen YS, Ko WJ, Chi NH, et al. Risk factor screening scale to optimize treat-
ment for potential heart transplant candidates under extracorporeal membrane
oxygenation. Am J Transpl 2004;4:1818–25.
26. Abella BS, Alvarado JP, Myklebust H, et al. Quality of cardiopulmonary resusci-
tation during in-hospital cardiac arrest. JAMA 2005;293:305–10.
27. Abella BS, Edelson DP, Kim S, et al. CPR quality improvement during in-hospital
cardiac arrest using a real-time audiovisual feedback system. Resuscitation
2007;73:54–61.