1. health and fitness
2. disease
3. heart disease

ORIGINAL CONTRIBUTION
System Delay and Mortality Among Patients
With STEMI Treated With Primary
Percutaneous Coronary Intervention
Christian Juhl Terkelsen, MD, PhD
Jacob Thorsted Sørensen, MD
Michael Maeng, MD, PhD
Lisette Okkels Jensen, MD, DmSc
Hans-Henrik Tilsted, MD
Sven Trautner, MD
Werner Vach, PhD
Søren Paaske Johnsen, MD, PhD
Leif Thuesen, MD, DmSc
Jens Flensted Lassen, MD, PhD
T
IMELY REPERFUSION THERAPY
with either fibrinolysis or primary percutaneous coronary intervention (PCI) is recommended for patients with ST-segment
elevation myocardial infarction
(STEMI).1 However, agreeing on the definition of “timely” is difficult, because the
benefit achieved by earlier initiation of
reperfusion therapy is controversial. The
only unbiased studies that have evaluated the effects of earlier reperfusion
therapy on outcome are randomized controlled studies that compared prehospital and in-hospital fibrinolysis. In those
studies, prehospital fibrinolysis was associated with earlier initiation (1 hour)
of reperfusion therapy, resulting in an extra 15 to 21 lives saved per 1000 treated
patients.2-4 The time-dependent benefit
of primary PCI has been evaluated from
observational data only, and the nearly
neutral relationship observed between
treatment delay and mortality may lead
Context Timely reperfusion therapy is recommended for patients with
ST-segment elevation myocardial infarction (STEMI), and door-to-balloon delay
has been proposed as a performance measure in triaging patients for primary percutaneous coronary intervention (PCI). However, focusing on the time from first
contact with the health care system to the initiation of reperfusion therapy (system
delay) may be more relevant, because it constitutes the total time to reperfusion
modifiable by the health care system. No previous studies have focused on the
association between system delay and outcome in patients with STEMI treated with
primary PCI.
Objective To evaluate the associations between system, treatment, patient, and doorto-balloon delays and mortality in patients with STEMI.
Design, Setting, and Patients Historical follow-up study based on populationbased Danish medical registries of patients with STEMI transported by the emergency medical service and treated with primary PCI from January 1, 2002, to
December 31, 2008, at 3 high-volume PCI centers in Western Denmark. Patients
(N=6209) underwent primary PCI within 12 hours of symptom onset. The median
follow-up time was 3.4 (interquartile range, 1.8-5.2) years.
Main Outcome Measures Crude and adjusted hazard ratios of mortality obtained by Cox proportional regression analysis.
Results A system delay of 0 through 60 minutes (n = 347) corresponded to a
long-term mortality rate of 15.4% (n = 43); a delay of 61 through 120 minutes
(n = 2643) to a rate of 23.3% (n = 380); a delay of 121 through 180 minutes
(n = 2092) to a rate of 28.1% (n = 378); and a delay of 181 through 360 minutes
(n=1127) to a rate of 30.8% (n=275) (P⬍.001). In multivariable analysis adjusted
for other predictors of mortality, system delay was independently associated with
mortality (adjusted hazard ratio, 1.10 [95% confidence interval, 1.04-1.16] per
1-hour delay), as was its components, prehospital system delay and door-toballoon delay.
Conclusion System delay was associated with mortality in patients with STEMI treated
with primary PCI.
www.jama.com
JAMA. 2010;304(7):763-771
Author Affiliations: Departments of Cardiology
(Drs Terkelsen, Sørensen, Maeng, Thuesen, and
Lassen) and Clinical Epidemiology (Dr Johnsen),
Aarhus University Hospital, Aarhus, Denmark;
Department of Cardiology, Odense University Hospital, Odense, Denmark (Dr Jensen); Department of
Cardiology, Aalborg University Hospital, Aalborg,
Denmark (Dr Tilsted); Falck Emergency Medical
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
Service, The Falck House, Copenhagen, Denmark
(Dr Trautner); and Department of Clinical Epidemiology, University Medical Centre, Freiburg, Germany (Dr Vach).
Corresponding Author: Christian Juhl Terkelsen,
MD, PhD, Department of Cardiology, Aarhus University Hospital, DK-8200 Aarhus N, Denmark
(christian_juhl_terkelsen@hotmail.com).
(Reprinted) JAMA, August 18, 2010—Vol 304, No. 7
763
SYSTEM DELAY AND STEMI MORTALITY
to the perception that the effect of primary PCI is less time-dependent than the
effect of fibrinolysis.5-7
However, patients presenting early
have a high mortality without reperfusion therapy and receive a large benefit from primary PCI. Conversely,
those presenting late are typically lowrisk patients who have already survived the prehospital phase and benefit less from reperfusion therapy.8,9 This
difference in patient characteristics may
explain the lack of a clear association
between treatment or patient delay and
mortality, because early presenting
high-risk patients who receive optimal reperfusion therapy ultimately have
nearly the same mortality as those presenting late.10,11 Moreover, determination of treatment and patient delay is
based on information regarding symptom onset, which may be uncertain because of recall bias and because the onset of acute myocardial infarction (AMI)
may have been preceded by hours of
unstable angina. Thus, it is impossible
to establish the exact time of onset of
the AMI. To study the effect of delayed primary PCI therapy on mortality, it may be more relevant to focus on
a parameter not hampered by recall bias
and less prone to selection bias, information bias, and confounding. Several studies have focused on the association between door-to-balloon delay
and outcome, whereas the total health
care system delay, defined as the time
from first contact with the health care
system to initiation of reperfusion
therapy, has received only limited attention (FIGURE 1).12
The present study assessed the associations between treatment, patient,
system, and door-to-balloon delays and
mortality in a large population-based
cohort of patients with STEMI treated
with primary PCI.
METHODS
Setting and Design
This study was based on public medical databases that cover the entire population of Western Denmark (approximately 3 million, corresponding to 55%
of the Danish population). The Danish National Health Service provides
tax-supported health care for all inhabitants, guaranteeing access to treatment at general practitioners and hospitals, along with emergency medical
service (EMS) transportation. The EMS
system in Denmark includes 5 different EMS agencies. In Western Denmark, one agency (Falck a/s) covers approximately 95% of the study region,
and other EMS agencies working in
Western Denmark were covered by the
dispatch center operated by Falck a/s,
guaranteeing access to prehospital data
from these agencies also.
The EMS system is organized as a
partially 1-tier and partially 2-tier system with initial dispatcher triage. All
emergencies deemed in need of an ambulance result in the dispatch of a primary unit manned with 2 EMS personnel trained in basic life support and the
use of a defibrillator in automated external defibrillator mode. Subject to
availability and determined by either
dispatcher triage or evaluation by the
Figure 1. Delays From Symptom Onset to Primary Percutaneous Coronary Intervention in Patients With ST-Segment Elevation Myocardial
Infarction Transported by the Emergency Medical Service
Field-triaged to a PCI center
Symptom
onset
Arrival at
PCI center
EMS call
Patient delay
Transportation delay
Primary PCI
Door-to-balloon delay
Prehospital system delay
System delay
Treatment delay
Transferred from local hospitals
Symptom
onset
Patient delay
EMS call
Arrival at
local hospital
Transportation delay
Departure from
local hospital
Local hospital delay
Arrival at
PCI center
Interhospital delay
Primary PCI
Door-to-balloon delay
Prehospital system delay (before arrival at PCI center)
System delay
Treatment delay
Stratified according to prehospital triage. EMS indicates emergency medical service; PCI, percutaneous coronary intervention.
764
JAMA, August 18, 2010—Vol 304, No. 7 (Reprinted)
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
©2010 American Medical Association. All rights reserved.
SYSTEM DELAY AND STEMI MORTALITY
primary unit, a physician-manned ambulance or a unit manned with a paramedic or nurse anesthetist also attend
the scene. Equipment for telecommunication was gradually implemented beginning in 1999, and in 2006 all ambulances had equipment for acquisition
and transmission of electrocardiographic data. The use of field triage for
primary PCI gradually increased during the study period.
Throughout the study period, the
catheterization laboratory was notified when the diagnosis of STEMI was
established, whether in the prehospital phase or at the local hospital, and
patients were admitted directly to the
catheterization laboratory. Unambiguous individual-level linkage between the
databases used in this study was possible using the civil registration number, a unique 10-digit personal identification number assigned to every
Danish citizen at birth.13,14 The study
was approved by the Danish Data Protection Agency ( J.No. 2008-412299).
Patients and Procedures
The study population consisted of 6209
patients with STEMI or bundlebranch block myocardial infarction admitted for primary PCI between January 1, 2002, and December 31, 2008,
at the 3 Western Denmark highvolume PCI centers: Aarhus University Hospital, Skejby; Odense University Hospital, Odense; and Aalborg
University Hospital, Aalborg. Patients
were identified in the Western Denmark Heart Registry (WDHR), which
collects baseline characteristics and patient- and procedure-specific information on all angiographies and coronary interventions performed in
Western Denmark. Self-presenters and
patients without EMS data were
excluded.
Primary PCI has been the recommended treatment for STEMI in Denmark since the publication of the Danish Trial of Acute Myocardial Infarction
2 (DANAMI-2) in 2003.15 Patients must
meet the following criteria to be eligible for primary PCI: symptom dura-
tion of 12 hours or less and STsegment elevation of 0.1 mV or greater
in at least 2 contiguous leads (ⱖ0.2 mV
in V1-V3) or presumed new-onset left
bundle-branch block. Pretreatment with
fibrinolysis was used in 118 patients
(2.0%).
Treatment, Patient, System,
and Door-to-Balloon Delay
The estimates of various delays to the
initiation of reperfusion therapy were
based on prehospital data registered by
the EMS provider (Falck a/s, Copenhagen, Denmark) and data registered
Table 1. Characteristics of Patients With ST-Segment Elevation Myocardial Infarction
Transported by the Emergency Medical Service and Treated With Primary Percutaneous
Coronary Intervention (N = 6209)
Field-Triaged Directly
to PCI Center
(n = 2183)
Characteristic
Demographics
Age, median (IQR), y
Women
Comorbid conditions
Treated hypertension
Diabetes
Previous myocardial infarction
Previous PCI
Previous congestive heart failure
Active or previous smoker
Delays and transportation
Delay, median (IQR), min a
Treatment
Patient
System
Prehospital system
Door-to-balloon
Transportation, median (IQR), km
Clinical characteristics
Body mass index, median (IQR) b
Blood pressure, median (IQR),
mm Hg
Systolic
Diastolic
Killip class
I
II
III
IV
Anterior STEMI or BBBMI
Culprit vessel
Left main artery
Left anterior descending artery
Circumflex artery
Right coronary artery
Multivessel disease
Valid
Cases
P
Value
2183
65 (56-74)
2183 1106 (28)
4026
4026
.37
.04
1521 924 (30)
1564 332 (11)
1502 337 (11)
1500 185 (6.2)
2014 134 (3.5)
1462 2316 (79)
3053
3161
3008
3006
3779
2916
.36
.39
.03
.90
.22
.97
172 (117-261)
74 (34-160)
97 (78-124)
50 (37-67)
39 (24-70)
25 (7-46)
2183
1945
2183
1638
1618
2183
240 (175-340)
106 (49-210)
139 (103-180)
89 (49-134)
29 (21-72)
71 (42-104)
4026
3548
4026
3014
3008
4026
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
26 (24-29)
1157
26 (24-29)
1903
.62
130 (115-150)
78 (65-85)
1286
1279
2181
126 (110-143)
75 (65-80)
2252
2242
4023
⬍.001
⬍.001
No. (%)
64 (55-74)
548 (25)
481 (32)
151 (9.7)
136 (9.1)
94 (6.3)
59 (2.9)
1160 (79)
Valid
Cases
Admitted at Local
Hospital and Transferred
to PCI Center
(n = 4026)
No. (%)
1976 (91)
108 (5.0)
3653 (91)
189 (4.7)
46 (2.1)
51 (2.3)
909 (45)
94 (2.3)
87 (2.2)
2009 1658 (46)
2107
39 (1.9)
911 (43)
299 (14)
60 (1.5)
1684 (43)
550 (14)
858 (41)
917 (46)
1608 (41)
1998 1780 (48)
.87
3624
3902
.72
.82
3717
.16
Abbreviations: BBBMI, bundle-branch block myocardial infarction; IQR, interquartile range; PCI, percutaneous coronary
intervention; STEMI, ST-segment elevation myocardial infarction.
a Door-to-balloon delay indicates time from arrival at PCI center to primary PCI; patient delay, time from symptom onset to contact with the health care system; prehospital system delay, time from contact with the health care system to
arrival at the PCI center; system delay, time from contact with the health care system to primary PCI.
b Calculated as weight in kilograms divided by height in meters squared.
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
(Reprinted) JAMA, August 18, 2010—Vol 304, No. 7
765
SYSTEM DELAY AND STEMI MORTALITY
Figure 2. Flow of Patient Inclusion Into
Study
13 439 Patients admitted for
primary PCI
562 Excluded (previous index primary PCI)
12 877 Had first index STEMI
during study period
6668 Excluded
3291 Primary PCI not performed
1552 Treatment delay >12 h
44 Missing treatment delay data
1433 EMS data not available
223 System delay >6 h
125 Mortality data not available
120 Foreign citizen
5 Emigrated
6209 Included in analysis
System delay indicates time from contact with the
health care system to time of primary percutaneous
coronary intervention (PCI); treatment delay, time from
symptom onset to time of primary PCI. EMS indicates emergency medical service; STEMI, STsegment elevation myocardial infarction.
in the WDHR. Time of ambulance call
was registered by a time stamp at the
dispatch center, whereas time of arrival on scene, departure from scene,
arrival at the local hospital, departure
from the local hospital, and arrival at
the PCI center were registered electronically in the ambulance by the EMS
personnel by pressing a radio button.
Symptom onset and time of first guiding-catheter insertion were registered
in the WDHR.
Treatment delay was defined as the
time from symptom onset to guidingcatheter insertion during primary PCI;
patient delay as the time from symptom onset to contact with the EMS;
system delay as the time from contact
with the EMS to guiding-catheter
insertion during primary PCI; prehospital system delay as the time from
contact with the EMS to arrival at the
PCI center; and door-to-balloon delay
as the time from arrival at the PCI center to guiding-catheter insertion during primary PCI (Figure 1). The use of
guiding-catheter insertion as a surrogate for time of intervention was chosen because time of balloon inflation
was only available in a minority of
patients, only a few minutes elapse
766
from guiding-catheter insertion to first
intervention, and the majority of
patients achieve reperfusion before
balloon inflation.16
Mortality and Covariates
Data on mortality were obtained from
The Danish Civil Registration System,
which has recorded changes in vital status of the entire Danish population
since 1968.17 Vital status is updated
daily.
Baseline characteristics and other covariates (TABLE 1) were derived from
the Danish Civil Registration System,
the WDHR, and the National Registry
of Patients.
Statistical Analysis
Dichotomous data are presented as percentages. Continuous variables are presented as medians (interquartile
ranges). The Fisher exact test, ␹2 test,
Mann-Whitney test, and KruskalWallis test were used for comparisons
of categorical and continuous variables, as appropriate. Follow-up began on the date of primary PCI and
ended on the date of death, emigration, or September 24, 2009, whichever came first.
We computed Kaplan-Meier cumulative mortality curves, stratified according to intervals of system delay, and
made comparisons between groups
with log-rank statistics. Cox proportional hazards regression analysis was
used to examine the association between the covariates and the intervals
of delay to reperfusion described above
and mortality. The proportional hazards assumption was checked for each
categorical variable by visual inspection and by the method described by
Grambsch and Therneau,18 using the
scaled Schoenfeld residuals.
For continuous variables, the linearity assumption was checked graphically using the Martingale residuals.
Cox-Snell residuals were used to assess the overall model fit. Systolic and
diastolic blood pressure levels were converted to categorical values, because
they did not fulfill the linearity
assumption.
JAMA, August 18, 2010—Vol 304, No. 7 (Reprinted)
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
Crude and mutually adjusted hazard ratios (HRs) with 95% confidence
intervals (CIs) were computed. Variables associated with time to death in
the univariable Cox regression analyses (Wald test P⬍.05) were included
in multivariable Cox regression models. Missing values among covariates
were replaced with their conditional
means, obtained as predictions from a
regression model using all nonmissing covariates for each patient.19,20 This
method for assigning missing values
was also used for categorical variables
without rounding the binary outcome, as previously proposed by Allison21 when proportions are not close
to 0 or 1.21,22 Because of colinearity between infarct location and culprit vessel, only infarct location was entered in
the multivariable models; because of colinearity between systolic and diastolic blood pressure, only systolic
blood pressure was entered in the models. Separate models were performed
considering only nonoverlapping intervals of time to reperfusion: model 1
considered treatment delay, model 2
considered patient and system delay,
and model 3 considered patient, prehospital system, and door-to-balloon
delay.
All statistical analyses were performed using Stata 10.0 (StataCorp,
College Station, Texas).
RESULTS
A total of 13 439 patients with suspected STEMI or bundle-branch block
myocardial infarction were transferred to or admitted directly to 1 of the
3 PCI centers. The first index STEMI
during the study period (n=12 877 patients) was included for further analyses. Patients were excluded if primary
PCI was not performed (n = 3291 patients) or if they had a treatment delay
greater than 12 hours (n=1552), missing treatment delay data (n=44), or a
system delay greater than 6 hours
(n=223). Mortality data were not available for patients who were foreign citizens (n=120) or had emigrated (n=5).
In 1433 patients considered selfpresenters, EMS data were not
©2010 American Medical Association. All rights reserved.
SYSTEM DELAY AND STEMI MORTALITY
available. Thus, the study cohort consisted of 6209 patients (FIGURE 2), of
whom 2183 (35%) were field-triaged directly to a PCI center, bypassing the local hospital. The proportion of patients field-triaged directly to the PCI
centers increased from 386 of 1414
(27%) in the first 2 years of the study
period to 873 of 1864 (47%) in the last
2 years (P ⬍.001).
When stratifying according to
whether patients were field-triaged directly to the PCI center or transferred
from other hospitals, there were significant differences in several baseline
characteristics and in the door-toballoon and system delays (Table 1).
For field-triaged, transferred, and all
EMS-transported patients, the proportion treated with a system delay of 120
minutes or less was 72% (n = 1566),
35% (n=1424), and 48% (n=2990), respectively, and among patients with
available door-to-balloon time
(n=4626) the proportion treated with
a door-to-balloon delay of 90 minutes
or less was 86% (n = 1399), 80%
(n=2407), and 82% (n=3806). The median time from guiding-catheter insertion to balloon inflation was 4 (interquartile range, 1-8) minutes in patients
in whom time of balloon inflation was
registered (n=1836). The median follow-up time was 3.4 (interquartile
range, 1.8-5.2) years, with a cumulative 1-year mortality of 9.3% (n=579).
The majority of covariates were associated with mortality at follow-up in
the univariable analysis (TABLE 2). According to Wald statistics, system delay had the strongest association with
mortality among the covariates modifiable in the acute phase, with an HR
of 1.22 (95% CI, 1.15-1.29; P ⬍.001)
per 1-hour increase in system delay
(Table 2). When stratifying according
to intervals of system delay, long-term
cumulative mortality was 15.4% (n=43)
in patients with system delays of 0
through 60 minutes (n = 347), 23.3%
(n = 380) in those with delays of 61
through 120 minutes (n=2643), 28.1%
(n = 378) in those with delays of 121
through 180 minutes (n = 2092), and
30.8% (n=275) in those with delays of
181 through 360 minutes (n = 1127)
(P ⬍ .001). Kaplan-Meier mortality
curves are presented in FIGURE 3.
For the different intervals of system
delay, no differences were observed in
infarct location, culprit vessel, or Killip
class, whereas significant differences
were observed in the majority of remaining covariates (TABLE 3). In the
multivariable analyses, after adjusting
for other covariates, treatment delay and
patient delay were not associated with
mortality (TABLE 4), whereas system delay remained independently associ-
Table 2. Crude Hazard Ratios of Covariates Associated With Long-term Mortality in
Univariable Cox Regression Analysis (N = 6209).
Characteristics
Demographics
Age, per 1-year increase
Women
Comorbid conditions
Treated hypertension
Diabetes
Previous myocardial infarction
Previous PCI
Previous congestive heart failure
Valid
Cases
Deaths a
6209
6209
364
4574
4725
4510
4506
5793
280
146
122
53
102
1.45 (1.25-1.69)
2.29 (1.91-2.73)
1.55 (1.28-1.89)
1.19 (0.90-1.58)
4.22 (3.43-5.17)
25
82
20
1.5
189
⬍.001
⬍.001
⬍.001
.22
⬍.001
4378
505
0.78 (0.66-0.93)
7.6
.006
6209
5493
6209
4652
4626
6209
1.054 (1.029-1.080)
1.042 (1.014-1.071)
1.22 (1.15-1.29)
1.19 (1.11-1.27)
1.13 (1.048-1.22)
1.00 (0.999-1.002)
19
8.8
51
26
10
0.45
⬍.001
.003
⬍.001
⬍.001
.002
.50
3060
0.94 (0.91-0.96)
25
⬍.001
1 [Reference]
0.49 (0.40-0.61)
78
42
⬍.001
⬍.001
0.40 (0.31-0.51)
0.45 (0.35-0.56)
54
46
⬍.001
⬍.001
1 [Reference]
0.55 (0.44-0.69)
0.53 (0.43-0.65)
0.37 (0.29-0.48)
69
26
35
55
⬍.001
⬍.001
⬍.001
⬍.001
1 [Reference]
2.63 (2.13-3.24)
5.09 (4.02-6.44)
8.65 (6.94-10.8)
1.35 (1.19-1.54)
1.28 (1.14-1.45)
2.19 (1.92-2.49)
545
81
184
370
21
16
138
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
Active or previous smoker
Delays and transportation
Delay, per 1-h increase c
Treatment
Patient
System
Prehospital system
Door-to-balloon
Transportation distance,
per 1-km increase
Clinical characteristics
Body mass index, per 1-unit
increase
Systolic blood pressure, mm Hg
⬍110
110-129
3537
641
1111
550
170
161
130-144
ⱖ145
Diastolic blood pressure, mm Hg
⬍65
65-74
75-84
ⱖ85
Killip class
I
II
III
IV
Anterior infarct location
Culprit vessel LM/LAD
Multivessel disease
873
912
3521
812
819
1092
798
6204
5629
297
140
138
5633
6009
5715
101
118
546
191
119
157
79
1073
811
97
76
89
467
520
640
HR (95% CI)
Wald
Test
1.074 (1.069-1.080) 666
1.46 (1.29-1.66)
35
P
Value b
⬍.001
⬍.001
Abbreviations: CI, confidence interval; HR, hazard ratio; LAD, left anterior descending artery; LM, left main artery, PCI, percutaneous coronary intervention.
a Total deaths=1076.
b By univariable Cox regression.
c Door-to-balloon delay indicates time from arrival at the PCI center to primary PCI; patient delay, time from symptom onset
to contact with the health care system; prehospital system delay, time from contact with the health care system to arrival
at the PCI center; system delay, time from contact with the health care system to primary PCI; treatment delay, time from
symptom onset to primary PCI.
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
(Reprinted) JAMA, August 18, 2010—Vol 304, No. 7
767
SYSTEM DELAY AND STEMI MORTALITY
Figure 3. Kaplan-Meier Cumulative Mortality Estimates for Patients With ST-Segment
Elevation Myocardial Infarction Treated With Primary Percutaneous Coronary Intervention
(N=6209)
System delay, min
0-60
61-120
121-180
181-360
Mortality, %
30
20
10
Log-rank P<.001
0
0
1
2
3
4
5
6
192
1006
842
491
138
667
533
332
87
375
278
172
7
Follow-up, y
No. at risk
System delay, min
347
0-60
2643
61-120
2092
121-180
1127
181-360
311
2339
1836
923
278
1906
1503
765
230
1420
1183
647
Stratified according to intervals of system delay (time from contact with the health care system to the time
of primary PCI). PCI indicates percutaneous coronary intervention; STEMI, ST-segment elevation myocardial
infarction.
ated with mortality, with an adjusted
HR of 1.10 (95% CI, 1.04-1.16; P=.002)
per 1-hour delay (Table 4). The main
components of system delay were also
associated with mortality: prehospital
system delay had an adjusted HR of 1.10
(95% CI, 1.02-1.18; P=.02), and doorto-balloon delay had an adjusted HR of
1.14 (95% CI, 1.05-1.24; P = .001) per
1-hour delay (Table 4).
COMMENT
To our knowledge, this study is the
first to evaluate the association
between system delay and outcome in
an unselected cohort of patients with
STEMI transported by an EMS and
treated with primary PCI. In contrast
to treatment and patient delay, system
delay was independently associated
with mortality. Moreover, it was the
highest ranking among the covariates
studied that can be modified in the
acute phase, and it comprises the total
delay that is modifiable by the health
care system.
The mortality benefit obtained by
earlier initiation of reperfusion
therapy is difficult to assess in observational studies. Previous studies have
plotted mortality according to treatment delay and reported a nearly horizontal association between time to
768
reperfusion and mortality.6,23 However, confounding and selection bias
may hamper such analyses. High-risk
patients tend to present early, whereas
those presenting late have already survived the early hours, ie, the period in
which they are at highest risk of death.
This so-called survivor-cohort effect is
supported by Löwel et al.8 In the prefibrinolytic era, they reported that 88%
of patients with AMI who contacted
the health care system within 1 hour
of symptom onset died during the prehospital phase or in the hospital. In
comparison, among patients who contacted the health care system from 1 to
24 hours after symptom onset, 43%
died during the prehospital phase or
in the hospital.8,24
This finding is consistent with those
reported by Aquaro et al and Nallamothu et al, who demonstrated that
early presenters had the highest risk
scores11 and the largest ST-segment elevations.25 Without reperfusion therapy,
patients presenting early have the highest mortality, but with optimal reperfusion therapy, they may attain nearly
the same mortality as those presenting late.10 This phenomenon may explain the neutral association previously observed between treatment delay
and mortality.6,23,26 Paradoxically, the
JAMA, August 18, 2010—Vol 304, No. 7 (Reprinted)
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
phenomenon also supports a timedependent benefit of primary PCI, because it implies that the benefit of primary PCI is more pronounced in the
early hours after symptom onset and
confirms that the reduction in mortality achieved by earlier reperfusion
therapy is underestimated when evaluated from observational data.10
We were able to adjust for major risk
factors in the analysis of the effect of
treatment delay and patient delay and
hence able to reduce the effect of confounding. Nevertheless, we found no association with mortality. This may be
explained by the selection bias, which
is still present. It also may be related to
the fact that patient delay and treatment delay depend on the time from onset of symptoms and are affected by substantial measurement error, because
patients have to recall this onset. Moreover, the biologically relevant point is
the time of onset of infarction, which
may not to be identical to the time of
first symptoms.
Door-to-balloon delay is suggested by
the D2B Alliance as “A key indicator of
quality of care in STEMI patients treated
with [primary] PCI.”27 Undoubtedly,
this parameter is associated with mortality,7,28,29 as is also documented in the
present study. It is useful for monitoring primary PCI performance at PCI
centers, and various initiatives have successfully reduced door-to-balloon delay.30 However, door-to-balloon delay
comprises only a minor part of the
health care system delay. A strategy of
prehospital diagnosis and rerouting patients directly to a catheterization laboratory may shorten system delay by as
much as 1 hour,31 although such a strategy may be associated with longer doorto-balloon delay, because shorter notice gives the catheterization laboratory
less time to prepare for the arrival of
field-triaged patients.31 Therefore, doorto-balloon delay should be used as a
performance measure at the PCI centers to ensure a focus on optimal center performance, but it may not be an
ideal general health care system performance measure in patients with
STEMI.
©2010 American Medical Association. All rights reserved.
SYSTEM DELAY AND STEMI MORTALITY
The effects of different primary PCI
delays on mortality cannot be subject
to randomized assessments. Given that
confounding, selection bias, and recall bias may hamper patient and treatment delays, it seems reasonable that
the optimal way to evaluate the association between delayed initiation of
reperfusion therapy and mortality in a
nonrandomized study is to focus on system delay. Even though patient and
treatment delays are theoretically applicable to all patients, data on these delays are available only in the selected
cohort of patients surviving until making contact with the health care system and only if the patient is able to recall the exact time of symptom onset.
Moreover, it is questionable if the exact onset of AMI can be determined, because it is based on subjective information and the AMI may have been
preceded by hours of unstable angina.
Accordingly, the lack of an association between patient delay and mortality may be explained by confounding,
selection bias, recall bias, and measure-
Table 3. Characteristics of Patients With ST-Segment Elevation Myocardial Infarction Transported by the Emergency Medical Service and
Treated With Primary Percutaneous Coronary Intervention, Stratified According to System Delay (N = 6209)
System Delay, min
0-60
(n = 347)
Variables
Demographics
Age, median (IQR), y
Women
Comorbid conditions
Treated hypertension
Diabetes
Previous myocardial infarction
Previous PCI
Previous congestive heart failure
Active or previous smoker
Delays, transportation, and triage
Delay, median (IQR), min b
Treatment
Patient
Prehospital system
Door-to-balloon
Transportation, median (IQR), km
Field-triaged to PCI center
Clinical characteristics
Body mass index, median (IQR) c
Blood pressure, median (IQR),
mm Hg
Systolic
Diastolic
Killip class
I
II
III
IV
Anterior STEMI or BBBMI
Culprit vessel
Left main artery
Left anterior descending artery
Circumflex artery
Right coronary artery
Multivessel disease
61-120
(n = 2643)
121-180
(n = 2092)
181-360
(n = 1127)
No. (%)
Valid
Cases
No. (%)
Valid
Cases
No. (%)
Valid
Cases
No. (%)
Valid
Cases
P
Value a
62 (54-72)
91 (26)
347
347
63 (55-73)
645 (24)
2643
2643
65 (56-75)
584 (28)
2092
2092
67 (58-77)
334 (30)
1127
1127
⬍.001
.003
81 (27)
33 (11)
16 (5.7)
18 (6.4)
7 (2.1)
231 (81)
297
300
283
280
327
287
615 (31)
194 (9.4)
197 (9.9)
121 (6.1)
74 (3.0)
1585 (82)
2008
2074
1982
1979
2453
1944
449 (30)
158 (10)
133 (9.1)
77 (5.3)
59 (3.0)
1105 (78)
1494
1537
1458
1456
1959
1414
260 (34)
98 (12)
127 (16)
63 (8.0)
53 (5.0)
555 (76)
775
814
787
791
1054
733
.18
.18
⬍.001
.10
.007
.005
170 (99-265)
131 (60-226)
29 (23-35)
19 (15-23)
7 (4-37)
153 (44)
347
337
201
199
347
347
180 (125-277)
99 (45-199)
55 (41-71)
28 (20-46)
38 (13-59)
1413 (54)
2643
2435
1920
1901
2643
2643
221 (173-315)
92 (41-188)
101 (50-123)
37 (24-92)
74 (33-106)
495 (24)
2092
1827
1598
1593
2092
2092
299 (243-378)
86 (39-173)
159 (55-193)
55 (26-154)
90 (47-133)
122 (11)
1127
894
933
933
1127
1127
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
26 (24-29)
165
26 (24-29)
1446
26 (24-29)
963
26 (24-29)
486
.97
130 (115-149)
78 (67-85)
189
189
347
130 (115-147)
79 (68-85)
1598
1592
2641
125 (110-140)
75 (65-80)
1141
1135
2090
125 (110-140)
72 (60-80)
609
605
1126
⬍.001
⬍.001
316 (91)
2418 (92)
1899 (91)
996 (89)
16 (4.6)
7 (2.0)
8 (2.3)
148 (47)
117 (4.4)
60 (2.3)
46 (1.7)
1085 (44)
100 (4.8)
45 (2.2)
6 (2.2)
868 (46)
64 (5.7)
28 (2.5)
38 (3.4)
466 (47)
3 (0.9)
158 (48)
37 (11)
129 (39)
140 (44)
312
327
319
39 (1.5)
1091 (43)
359 (14)
1068 (42)
1084 (45)
2441
2557
2419
31 (1.5)
888 (44)
274 (13)
846 (42)
929 (48)
1895
2039
1936
26 (2.4)
458 (42)
179 (17)
423 (39)
544 (52)
.15
985
1086
.40
.06
1041
⬍.001
Abbreviations: BBBMI, bundle-branch block myocardial infarction; IQR, interquartile range; PCI, percutaneous coronary intervention; STEMI, ST-segment elevation myocardial infarction.
a By Kruskal-Wallis test.
b Door-to-balloon delay indicates time from arrival at PCI center to primary PCI; patient delay, time from symptom onset to contact with the health care system; prehospital system
delay, time from contact with the health care system to arrival at the PCI center; system delay, time from contact with the health care system to primary PCI.
c Calculated as weight in kilograms divided by height in meters squared.
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
(Reprinted) JAMA, August 18, 2010—Vol 304, No. 7
769
SYSTEM DELAY AND STEMI MORTALITY
ment bias, but neither this observation nor the lack of any long-lasting
effect of media campaigns on patient or
treatment delays should deter encouraging patients to seek medical help as
soon as possible after the onset of symptoms.32,33
In comparison, system delay is by
definition only defined in patients
surviving until contact with the health
care system; hence, studying the effect
of system delay is not affected by
selection bias from survival. It can be
studied in all patients contacting the
health care system, and it is an objective parameter not prone to recall
bias. Most importantly, however, system delay and its components appear
to be the only risk factors that can be
modified in the acute phase, by opti-
mizing prehospital and in-hospital triage.31,34
This study has a number of limitations. System delay may have been
underestimated, because data were
not available on contacts with general
practitioners in the acute phase. In
reviewing a sample of 130 hospital
records, we found that 6% of the
EMS-transported patients were not in
the EMS registry, resulting in a minor
underestimation of the number of
patients transported by the EMS. Usually, the time of the first balloon inflation is used as the time of reperfusion,
but reperfusion often takes place
before balloon inflation (eg, during
wiring or thrombectomy). Therefore,
the time of first wiring of the vessel
might be a better parameter for repre-
senting the time of reperfusion. 16
However, data on first wiring were
not available, and because insertion of
the guiding catheter is followed
within a few minutes by the first coronary intervention, we decided to use
the time of guiding-catheter insertion
as the time of first intervention.
Acknowledging the widespread acceptance of the door-to-balloon delay as a
performance measure, we decided to
use the door-to-balloon delay synonymously with the time from arrival at
the PCI center to the first insertion of
the guiding catheter.
We conclude that health care system delay is valuable as a performance
measure when patients with STEMI are
treated with primary PCI, because it is
associated with mortality, it consti-
Table 4. Multivariable Cox Regression Analysis of Covariates Associated With Long-term Mortality in Patients With ST-Segment Elevation
Myocardial Infarction Treated With Primary Percutaneous Coronary Intervention (N = 6209) a
Model 1
Covariates Remaining Significant
in Models b
Demographics
Age, per 1-y increase
Comorbid conditions
Diabetes
Previous congestive heart failure
Previous or current smoking
Delays, per 1-h increase d
Treatment
Patient
System
Prehospital system
Door-to-balloon
Clinical characteristics
Systolic blood pressure, mm Hg
⬍110
110-129
130-144
ⱖ145
Killip class
I
II
III
IV
Anterior STEMI or BBBMI
Multivessel disease
Model 2
Model 3
HR (95% CI)
P
Value c
HR (95% CI)
P
Value c
HR (95% CI)
P
Value c
1.067 (1.060-1.074)
⬍.001
1.066 (1.060-1.073)
⬍.001
1.066 (1.060-1.073)
⬍.001
1.97 (1.64-2.37)
1.73 (1.39-2.16)
1.24 (1.05-1.48)
⬍.001
⬍.001
.01
1.94 (1.61-2.33)
1.78 (1.43-2.22)
1.25 (1.05-1.48)
⬍.001
⬍.001
.01
1.95 (1.63-2.35)
1.78 (1.43-2.22)
1.26 (1.06-1.50)
⬍.001
⬍.001
.008
1.00 (0.98-1.03)
.54
1.00 (0.98-1.03)
.63
1.00 (0.98-1.03)
.64
1.10 (1.04-1.16)
.002
1.10 (1.02-1.18)
1.14 (1.05-1.24)
.02
.001
1 [Reference]
0.76 (0.63-0.92)
0.66 (0.55-0.81)
0.52 (0.40-0.66)
.004
⬍.001
⬍.001
1 [Reference]
0.74 (0.61-0.90)
0.68 (0.56-0.82)
0.53 (0.41-0.67)
.004
⬍.001
⬍.001
1 [Reference]
0.74 (0.61-0.89)
0.66 (0.55-0.80)
0.52 (0.41-0.67)
.001
⬍.001
⬍.001
1 [Reference]
1.61 (1.29-2.01)
2.47 (1.92-3.18)
4.73 (3.69-6.06)
1.30 (1.14-1.48)
1.53 (1.34-1.76)
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
1 [Reference]
1.65 (1.33-2.06)
2.55 (1.98-3.28)
4.71 (3.67-6.05)
1.29 (1.13-1.47)
1.51 (1.32-1.73)
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
1 [Reference]
1.63 (1.31-2.03)
2.51 (1.95-3.23)
4.57 (3.56-5.87)
1.29 (1.13-1.47)
1.51 (1.31-1.73)
⬍.001
⬍.001
⬍.001
⬍.001
⬍.001
Abbreviations: BBBMI, bundle-branch block myocardial infarction; CI, confidence interval; HR, hazard ratio; STEMI, ST-segment elevation myocardial infarction.
a Adjusted for sex, body mass index, history of previous myocardial infarction, and treatment for hypertension.
b Nonoverlapping intervals of treatment delay were considered for inclusion in the multivariable models: model 1 considered treatment delay, model 2 considered patient and system
delay, and model 3 considered patient, prehospital system, and door-to-balloon delay.
c By multivariable Cox regression.
d Door-to-balloon delay indicates time from arrival at the PCI center to primary percutaneous coronary intervention (PCI); patient delay, time from symptom onset to contact with the
health care system; prehospital system delay, time from contact with the health care system to arrival at the PCI center; system delay, time from contact with the health care system
to primary PCI.
770
JAMA, August 18, 2010—Vol 304, No. 7 (Reprinted)
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
©2010 American Medical Association. All rights reserved.
SYSTEM DELAY AND STEMI MORTALITY
tutes the part of treatment delay modifiable by the health care system in the
acute phase, and it applies to patients
field-triaged directly to the PCI center
as well as to patients transferred from
local hospitals. Increased focus on the
total health care system delay may optimize triage of patients with STEMI and
may be the key to further improving
survival of these patients.
Author Contributions: Dr Terkelsen had full access to
all of the data in the study and takes responsibility for
the integrity of the data and the accuracy of the data
analysis.
Study concept and design: Terkelsen, Maeng, Trautner,
Thuesen, Lassen.
Acquisition of data: Terkelsen, Sørensen, Jensen,
Tilsted, Trautner, Lassen.
Analysis and interpretation of data: Terkelsen,
Sørensen, Maeng, Trautner, Vach, Johnsen, Lassen.
Drafting of the manuscript: Terkelsen, Sørensen,
Jensen, Thuesen, Lassen.
Critical revision of the manuscript for important intellectual content: Terkelsen, Sørensen, Maeng, Jensen,
Tilsted, Trautner, Vach, Johnsen, Lassen.
Statistical analysis: Terkelsen.
Obtained funding: Terkelsen, Sørensen, Lassen.
Administrative, technical, or material support:
Terkelsen, Sørensen, Jensen, Trautner, Lassen.
Study supervision: Maeng, Jensen, Tilsted, Johnsen,
Thuesen, Lassen.
Financial Disclosures: Dr Sørensen reported receiving an unrestricted grant from Falck EMS, Denmark,
to perform studies unrelated to the present study. No
other authors reported disclosures.
Funding/Support: This study was supported by grants
from the Helga and Peter Kornings Foundation, Aarhus,
Denmark ( J.No. 40-134918) and the Health Research Fund of Central Denmark Region, Aarhus, Denmark ( J.No. 1-45-72-1-08).
Role of the Sponsors: The funders had no role in the
design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or
in the preparation of the manuscript.
Additional Contributions: We thank Tim Lash, DSc,
MPH, Department of Epidemiology, Aarhus University, Aarhus, Denmark, for his advice during the revision of the manuscript. Dr Lash received no compensation for his contributions.
REFERENCES
1. Antman EM, Hand M, Armstrong PW, et al; Canadian Cardiovascular Society; American Academy of Family Physicians; American College of Cardiology; American Heart Association. 2007 Focused Update of the
ACC/AHA 2004 Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction: a report of the American College of Cardiology/American
Heart Association Task Force on Practice Guidelines.
J Am Coll Cardiol. 2008;51(2):210-247.
2. European Myocardial Infarction Project Group. Prehospital thrombolytic therapy in patients with suspected acute myocardial infarction. N Engl J Med. 1993;
329(6):383-389.
3. Morrison LJ, Verbeek PR, McDonald AC, Sawadsky
BV, Cook DJ. Mortality and prehospital thrombolysis
for acute myocardial infarction: a meta-analysis. JAMA.
2000;283(20):2686-2692.
4. Boersma E, Maas ACP, Deckers JW, Simoons ML.
Early thrombolytic treatment in acute myocardial infarction: reappraisal of the golden hour. Lancet. 1996;
348(9030):771-775.
5. Brodie BR, Stone GW, Morice MC, et al; Stent Primary Angioplasty in Myocardial Infarction Study Group.
Importance of time to reperfusion on outcomes with
primary coronary angioplasty for acute myocardial infarction (results from the Stent Primary Angioplasty
in Myocardial Infarction Trial). Am J Cardiol. 2001;
88(10):1085-1090.
6. Zijlstra F, Patel A, Jones M, et al. Clinical characteristics and outcome of patients with early (⬍2 h),
intermediate (2-4 h) and late (⬎4 h) presentation
treated by primary coronary angioplasty or thrombolytic therapy for acute myocardial infarction. Eur
Heart J. 2002;23(7):550-557.
7. McNamara RL, Wang Y, Herrin J, et al; NRMI
Investigators. Effect of door-to-balloon time on mortality in patients with ST-segment elevation myocardial infarction. J Am Coll Cardiol. 2006;47(11):
2180-2186.
8. Löwel H, Lewis M, Hörmann A. Prognostic significance of pre-hospital phase in acute myocardial infarct: results of the Augsburg Myocardial Infarct Registry, 1985-1988 [in German]. Dtsch Med Wochenschr.
1991;116(19):729-733.
9. Boersma E; Primary Coronary Angioplasty vs.
Thrombolysis Group. Does time matter? a pooled
analysis of randomized clinical trials comparing
primary percutaneous coronary intervention and
in-hospital fibrinolysis in acute myocardial infarction
patients. Eur Heart J. 2006;27(7):779-788.
10. Terkelsen CJ, Christiansen EH, Sørensen JT, et al.
Primary PCI as the preferred reperfusion therapy in
STEMI: it is a matter of time. Heart. 2009;95(5):
362-369.
11. Aquaro GD, Pingitore A, Strata E, et al. Relation of
pain-to-balloon time and myocardial infarct size in patients transferred for primary percutaneous coronary
intervention. Am J Cardiol. 2007;100(1):28-34.
12. Terkelsen CJ, Sørensen JT, Nielsen TT. Is there any
time left for primary percutaneous coronary intervention according to the 2007 updated American College of Cardiology/American Heart Association STsegment elevation myocardial infarction guidelines and
the D2B Alliance? J Am Coll Cardiol. 2008;52
(15):1211-1215.
13. Frank L. Epidemiology: when an entire country
is a cohort. Science. 2000;287(5462):2398-2399.
14. Sorensen HT. Regional administrative health
registries as a resource in clinical epidemiology: a
study of options, strengths, limitations and data quality provided with examples of use. Int J Risk Saf Med.
1997;10:1-22. http://www.iospress.nl/loadtop
/load.php?isbn=09246479. Accessed July 22,
2010.
15. Andersen HR, Nielsen TT, Rasmussen K, et al;
DANAMI-2 Investigators. A comparison of coronary
angioplasty with fibrinolytic therapy in acute myocardial infarction. N Engl J Med. 2003;349(8):733742.
16. Terkelsen CJ, Kaltoft AK, Nørgaard BL, et al. ST
changes before and during primary percutaneous coronary intervention predict final infarct size in patients
with ST elevation myocardial infarction. J Electrocardiol.
2009;42(1):64-72.
17. Pedersen CB, Gøtzsche H, Møller JO, Mortensen
PB. The Danish Civil Registration System: a cohort of
eight million persons. Dan Med Bull. 2006;53(4):
441-449.
18. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals.
Biometrika. 1994;81:515-526.
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 05/06/2013
19. Schemper M, Smith TL. Efficient evaluation of
treatment effects in the presence of missing covariate values. Stat Med. 1990;9(7):777-784.
20. Schemper M, Heinze G. Probability imputation revisited for prognostic factor studies. Stat Med. 1997;
16(1-3):73-80.
21. Allison PD. Missing Data. Thousand Oaks, CA:
Sage Publications; 2002.
22. Vach W, Schumacher M. Logistic-regression with
incompletely observed categorical covariates—a comparison of 3 approaches. Biometrika. 1993;80:
353-362.
23. Cannon CP, Gibson CM, Lambrew CT, et al. Relationship of symptom-onset-to-balloon time and doorto-balloon time with mortality in patients undergoing angioplasty for acute myocardial infarction. JAMA.
2000;283(22):2941-2947.
24. The pre-hospital management of acute heart
attacks: recommendations of a Task Force of the
European Society of Cardiology and the European Resuscitation Council. Eur Heart J. 1998;19(8):11401164.
25. Nallamothu BK, Bates ER, Herrin J, Wang Y, Bradley
EH, Krumholz HM; NRMI Investigators. Times to
treatment in transfer patients undergoing primary
percutaneous coronary intervention in the United
States: National Registry of Myocardial Infarction
(NRMI)-3/4 analysis. Circulation. 2005;111(6):
761-767.
26. Brodie BR, Stuckey TD, Wall TC, et al. Importance of time to reperfusion for 30-day and late survival and recovery of left ventricular function after primary angioplasty for acute myocardial infarction. J Am
Coll Cardiol. 1998;32(5):1312-1319.
27. Nallamothu BK, Bradley EH, Krumholz HM. Time
to treatment in primary percutaneous coronary
intervention. N Engl J Med. 2007;357(16):16311638.
28. McNamara RL, Herrin J, Bradley EH, et al; NRMI
Investigators. Hospital improvement in time to reperfusion in patients with acute myocardial infarction,
1999 to 2002. J Am Coll Cardiol. 2006;47(1):
45-51.
29. Rathore SS, Curtis JP, Chen J, et al; National
Cardiovascular Data Registry. Association of doorto-balloon time and mortality in patients admitted
to hospital with ST elevation myocardial infarction:
national cohort study. BMJ. 2009;338:b1807.
30. Bradley EH, Nallamothu BK, Herrin J, et al. National efforts to improve door-to-balloon time results
from the Door-to-Balloon Alliance. J Am Coll Cardiol.
2009;54(25):2423-2429.
31. Terkelsen CJ, Lassen JF, Nørgaard BL, et al. Reduction of treatment delay in patients with STelevation myocardial infarction: impact of prehospital diagnosis and direct referral to primary
percutanous coronary intervention. Eur Heart J. 2005;
26(8):770-777.
32. Hedges JR, Feldman HA, Bittner V, et al; REACT
Study Group. Impact of community intervention to reduce patient delay time on use of reperfusion therapy
for acute myocardial infarction: Rapid Early Action for
Coronary Treatment (REACT) trial. Acad Emerg Med.
2000;7(8):862-872.
33. Blohm MB, Hartford M, Karlson BW, Luepker RV,
Herlitz J. An evaluation of the results of media and
educational campaigns designed to shorten the time
taken by patients with acute myocardial infarction to
decide to go to hospital. Heart. 1996;76(5):430434.
34. Le May MR, So DY, Dionne R, et al. A citywide
protocol for primary PCI in ST-segment elevation myocardial infarction. N Engl J Med. 2008;358(3):231240.
(Reprinted) JAMA, August 18, 2010—Vol 304, No. 7
771