NEJM CARDIOVASCULAR DISEASE ARTICLES - Part 4 potx

n engl j med 356;12 www.nejm.org march 22, 2007
1207
The new england
journal of medicine
established in 1812 march 22, 2007 vol. 356 no. 12
Emergency Duties and Deaths from Heart Disease
among Firefighters in the United States
Stefanos N. Kales, M.D., M.P.H., Elpidoforos S. Soteriades, M.D., Sc.D., Costas A. Christophi, Ph.D.,
and David C. Christiani, M.D., M.P.H.
A BS TR AC T
From the Cambridge Health Alliance,
Harvard Medical School, Cambridge, MA
(S.N.K.); the Department of Environmen-
tal Health, Harvard School of Public
Health, Boston (S.N.K., E.S.S., D.C.C.);
the Pulmonary and Critical Care Unit,
Massachusetts General Hospital, Boston
(D.C.C.); the Center for Occupational and
Environmental Medicine, Kindred Hos-
pital Northeast, Braintree, MA (D.C.C.);
and the Cyprus International Institute for
the Environment and Public Health in
association with the Harvard School of
Public Health, Nicosia, Cyprus (C.A.C.).
Address reprint requests to Dr. Kales at
the Cambridge Health Alliance, Employee
Health and Industrial Medicine, Lee B.
Macht Bldg., Rm. 427, 1493 Cambridge
St., Cambridge, MA 02139, or at skales@
challiance.org.
N Engl J Med 2007;356:1207-15.

Copyright © 2007 Massachusetts Medical Society.
Background
Heart disease causes 45% of the deaths that occur among U.S. firefighters while
they are on duty. We examined duty-specific risks of death from coronary heart
disease among on-duty U.S. firefighters from 1994 to 2004.
Methods
We reviewed summaries provided by the Federal Emergency Management Agency
of the deaths of all on-duty firefighters between 1994 and 2004, except for deaths
associated with the September 11, 2001, terrorist attacks. Estimates of the propor-
tions of time spent by firefighters each year performing various duties were obtained
from a municipal fire department, from 17 large metropolitan fire departments, and
from a national database. Odds ratios and 95% confidence intervals for death from
coronary heart disease during specific duties were calculated from the ratios of the
observed odds to the expected odds, with nonemergency duties as the reference cat-
egory.
Results
Deaths from coronary heart disease were associated with suppressing a fire (32.1%
of all such deaths), responding to an alarm (13.4%), returning from an alarm
(17.4%), engaging in physical training (12.5%), responding to nonfire emergencies
(9.4%), and performing nonemergency duties (15.4%). As compared with the odds
of death from coronary heart disease during nonemergency duties, the odds were
12.1 to 136 times as high during fire suppression, 2.8 to 14.1 times as high during
alarm response, 2.2 to 10.5 times as high during alarm return, and 2.9 to 6.6 times
as high during physical training. These odds were based on three estimates of the
time that firefighters spend on their duties.
Conclusions
Certain emergency firefighting duties were associated with a risk of death from
coronary heart disease that was markedly higher than the risk associated with
nonemergency duties. Fire suppression was associated with the highest risk, which
was approximately 10 to 100 times as high as that for nonemergency duties.

Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
T h e ne w e ngl a n d jo u r na l o f me d i c i n e
n engl j med 356;12 www.nejm.org march 22, 2007
1208
F
irefighting is known to be a dan-
gerous occupation. What is less appreciated
is that the most frequent cause of death
among firefighters is heart disease rather than
burns or smoke inhalation. Cardiovascular events,
largely due to coronary heart disease, account for
45% of deaths among firefighters on duty.
1,2
In
contrast, such events account for 22% of deaths
among police officers on duty, 11% of deaths
among on-duty emergency medical services work-
ers, and 15% of all deaths that occur on the job.
2,3

The high rate of death from cardiovascular causes
among firefighters raises questions about contrib-
uting factors. Possible factors, such as physical ex-
ertion, emergency responses, and dangerous du-
ties, are not unique to firefighting; they are also
characteristic of the work performed by police of-
ficers, military personnel, and persons in various
other occupations.
4,5

Various biologically plausible explanations for
the high mortality from cardiovascular events
among firefighters have been proposed. These
explanations include smoke and chemical expo-
sure, irregular physical exertion, the handling of
heavy equipment and materials, heat stress, shift
work, a high prevalence of cardiovascular risk fac-
tors, and psychological stressors.
6-13
Given these
occupational risks, 37 U.S. states and 2 Canadian
provinces provide benefits to firefighters in whom
certain cardiovascular diseases have developed.
14

Nevertheless, the evidence linking firefighting
to cardiovascular disease continues to be debat-
ed.
15-17
Therefore, whether deaths from coronary
heart disease among firefighters are truly precipi-
tated by their work and, if so, by which duties,
remain important questions.
The findings in our previous case–control
study of 52 deaths from coronary heart disease
among on-duty firefighters provided preliminary
evidence that coronary events may be triggered by
specific firefighting duties.
18
First, the circadian

pattern of deaths from coronary heart disease par-
alleled the pattern of emergency-response dis-
patches. Second, elevated risks of death were as-
sociated with fire suppression, alarm response,
and physical training. To confirm these findings
and further explore duty-specific risk factors for
death from coronary heart disease, we conducted
a study of all deaths that occurred among on-duty
firefighters in the United States between 1994
and 2004.
Me thod s
Deaths among Firefighters
The U.S. Fire Administration, a branch of the
Federal Emergency Management Agency, collects
narrative summaries for all reported deaths as-
sociated with firefighting in the United States.
From these publicly available summaries, we ex-
amined data on all deaths that occurred between
January 1, 1994, and December 31, 2004.
2,19
The
data included all firefighters who died while on
duty, who became ill while on duty and later died,
and who died within 24 hours after an emergency
response or training. We excluded deaths that oc-
curred during the first 48 hours after the Septem-
ber 11, 2001, terrorist attacks.
To extract study data, two reviewers indepen-
dently examined the summary of each reported
death that occurred while the firefighter was on

duty. A third reviewer resolved any classifications
that were not concordant between the first two
reviewers. On the basis of the narrative reports,
each death was classified as due to cardiovascular
causes or to noncardiovascular causes. We then
excluded those cases in which death occurred
more than 24 hours after the on-duty incident or
in which death resulted from a cardiovascular
problem other than coronary heart disease (e.g.,
certain arrhythmias, stroke, aneurysm, or genetic
cardiomyopathy).
All records of deaths that were classified by
this process as being due to coronary heart dis-
ease were selected for further study. Data extract-
ed from these records included the firefighter’s
age, sex, and job status (professional or volun-
teer); the date, cause, and mechanism of death;
and the city and state of the fire department.
Duties at the Time of Death
On the basis of the summary report of each death,
the deaths were classified according to the spe-
cific duty performed during the onset of symp-
toms or immediately preceding sudden death.
These categories were fire suppression; alarm re-
sponse; alarm return; physical training; emergen-
cy medical services, rescues, and other nonfire
emergencies; and nonemergency duties. A death
was classified as being associated with fire sup-
pression if it occurred while the person was fight-
ing a fire or at the scene of a fire after its sup-

pression. Alarm response involved responses to
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
De aths from He art Disease a mong Firefighter s
n engl j med 356;12 www.nejm.org march 22, 2007
1209
emergency incidents, including false alarms. Alarm
return included all events that occurred during
the return from incidents and those that occurred
within several hours after an emergency call.
Physical training included all job-related physical-
fitness activities, physical-abilities testing, and
simulated or live fire, rescue, emergency, and
search drills. We grouped together emergency
medical services, rescues, and other nonfire emer-
gencies in a separate category. Finally, we classi-
fied all of the following activities as nonemergen-
cy duties: administrative and fire-station tasks,
fire prevention, inspection, maintenance, meet-
ings, parades, and classroom activities.
Time Spent on Specific Duties
We used data from several sources to estimate
the average annual proportion of time that fire-
fighters spend in each category. First, we direct-
ly derived point estimates from a municipal fire
department (Cambridge Fire Department, Cam-
bridge, MA), using fiscal year 2002 data, as in our
previous study.
18
For Cambridge firefighters, the

following information was available: the number
of firefighters, the total number of alarms and
emergency responses, the distribution of emer-
gency calls and dispatches by hour of the day, a
breakdown of the types of incidents involved in
fire and nonfire emergency responses, the average
time spent per incident and the average response
time, and the estimated number of hours spent
each week in training and fire-prevention activities.
We refer to these data as the municipal estimate.
Second, to conduct a sensitivity analysis, we
obtained two additional sets of estimates, one
representing a level of emergency activity that was
higher than that of the Cambridge Fire Depart-
ment and the other representing a lower level of
emergency activity. These estimates were derived
with the use of data for the population served,
the numbers of uniformed officers, and the num-
ber of emergency incidents and the types of inci-
dents classified as fire and nonfire emergencies.
To characterize the largest and busiest fire de-
partments, an estimate was developed from 2005
survey data provided by the International Associa-
tion of Fire Fighters (Moore-Merrell L: personal
communication) for 17 large urban and suburban
fire departments (the large metropolitan esti-
mate). To represent firefighters in smaller com-
munities with lower levels of emergency activity,
an estimate was developed from nationwide Na-
tional Fire Protection Association surveys conduct-

ed from 1994 to 2003 (the national estimate).
20
Statistical Analysis
We made the initial assumption that if specific
firefighting duties do not have a significant effect
on the risk of death from coronary heart disease,
then the number of such deaths that occur dur-
ing any given firefighting duty should be directly
proportional to the amount of time spent per-
forming that duty. For example, if 10% of a fire-
fighter’s time is spent in responding to alarms,
10% of deaths from coronary heart disease should
occur during alarm response. We then sought to
determine whether this expected pattern is or is
not supported by the actual data.
Using the chi-square goodness-of-fit test, we
assessed whether the distribution of actual deaths
associated with each duty was the same as that
of expected deaths, based on the estimates of the
average time dedicated to each firefighting duty.
We used the three different time estimates (from
the municipal, large metropolitan, and national
data) to calculate the ratios of actual to expected
deaths for each firefighting duty. The 95% confi-
dence intervals (CIs) for these ratios were calcu-
lated on the basis of the multinomial distribu-
tion. Odds ratios for death from coronary heart
disease during specific duties were calculated
from the ratios of the observed to expected odds,
with nonemergency duties used as the reference

category. The 95% CIs for the estimated odds
ratios were calculated with the use of the bino-
mial distribution.
Using data from the 2000 firefighters census,
21

which stratifies firefighters according to their age
(in decades) and job status (professionals or vol-
unteers), we calculated the rates of death from
coronary heart disease for specific duties accord-
ing to age and job status. Our calculations were
based on death counts in each category per 1 mil-
lion person-years of risk, derived from the average
number of firefighters at risk in each subgroup
over the 11-year period of observation.
Analyses were performed with the use of SAS
software for Windows (version 8.02, SAS Insti-
tute), and StatXact (version 6.0). A P value of less
than 0.05 was considered to indicate statistical
significance, and all statistical tests for differ-
ences were two-sided.
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
T h e ne w e ngl a n d jo u r na l o f me d i c i n e
n engl j med 356;12 www.nejm.org march 22, 2007
1210
R es u l t s
Between January 1, 1994, and December 31, 2004,
1144 firefighter deaths were reported to the U.S.
Fire Administration. We classified 449 deaths as

due to coronary heart disease (39%). Of these
deaths from coronary heart disease, 144 (32%)
occurred during fire suppression, 138 (31%) oc-
curred during alarm response or return, and the
remaining 167 (37%) occurred during other duties
(
Table 1
).
Table 2
shows the estimated proportion of
time that firefighters spent each year in specific
duties according to the three sources of fire-
department activity data that we used. Among
firefighters in Cambridge (our municipal data
set), approximately 2% of duty time was spent in
fire suppression. Among firefighters in our large
metropolitan data set, approximately 5% of duty
time was spent in fire suppression. Finally, among
all firefighters in the United States (as represent-
ed in our national data set), approximately 1% of
duty time was spent in fire suppression.
Table 3
shows the frequency of observed deaths
from coronary heart disease according to duty as
compared with the expected frequency. The ob-
served distribution of deaths was significantly dif-
ferent from the expected distribution based on the
estimates from each of the three data sources (P<
0.001 for the three comparisons). The ratios of ob-
served to expected deaths associated with the vari-

ous duties of firefighters were consistently higher
than 1, with the exception of nonfire emergencies
and nonemergency duties. Although 32% of deaths
occurred during fire suppression, this activity was
estimated to account for as little as 1 to 5% of the
average firefighter’s professional time per year, so
this duty was associated with the most significant-
ly elevated ratios of observed to expected deaths.
Table 1. Deaths from Coronary Heart Disease among Firefighters, Classified
According to Duty at the Time of Death.*
Duty
Deaths
(N = 449)
no. (%)
Fire suppression 144 (32.1)
Alarm response 60 (13.4)
Alarm return 78 (17.4)
Physical training 56 (12.5)
Emergency medical services and other nonfire emergencies 42 (9.4)
Fire-station and other nonemergency duties
69 (15.4)
* Data are based on narrative summaries from the records of the U.S. Fire Ad-
ministration, Federal Emergency Management Agency, for the period from
January 1, 1994, to December 31, 2004.
19
Table 2. Fire Service Activity and the Estimated Proportion of Time Spent in Specific Firefighting Duties.*
Variable
Municipal Fire
Department
Large Metropolitan Fire

Departments National Data
Fire service activity
Population served (no.) 101,355 760,935±888,916 280,000,000
Uniformed firefighters (no.) 274 1063±785 1,082,855±14,446
Population served per firefighter (no.) 370 655±218 259±3
Emergency incidents (no./firefighter/yr) 44 92±24 18±2
Fire incidents (no./firefighter/yr) 2.0 7.0±6.3 1.7±0.1
Duties (% of annual time)
Fire suppression 2 5 1
Alarm response 6 9 4
Alarm return 10 15 7
Physical training 8 8 8
Emergency medical services and other nonfire emergencies 23 34 15
Fire-station and other nonemergency duties
51 29 65
* Plus–minus values are means ±SD. Municipal data are from the Cambridge Fire Department, Cambridge, Massachusetts (2002).
18
Data
for large metropolitan fire departments are from surveys of 17 large metropolitan fire departments conducted by the International Associ-
ation of Fire Fighters (2005) (Moore-Merrell L: personal communication). National data are from annual national surveys conducted by the
National Fire Protection Association (1994 through 2003).
20
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
De aths from He art Disease a mong Firefighter s
n engl j med 356;12 www.nejm.org march 22, 2007
1211
Table 4
includes the odds ratios and 95% CIs
for the risk of death from coronary heart disease

among firefighters engaged in each emergency
duty and physical training as compared with the
reference category of nonemergency tasks. On the
basis of the three estimates of the time that fire-
fighters spent on particular duties, death from
coronary heart disease was 12 to 136 times as
likely to occur during fire suppression as during
nonemergency duties. An increased risk was also
consistently observed for other emergency duties,
as compared with nonemergency duties; the risk
was increased by a factor of 2.8 to 14.1 during
alarm response, 2.2 to 10.5 during alarm return,
and 2.9 to 6.6 during physical training.
Figure 1A shows the risk of death from coro-
nary heart disease per 1 million firefighters per
year (deaths per 1 million person-years) for each
duty according to age group, and
Figure 1B
shows
the risk of death according to job status (volun-
teer or professional). As might be expected, the
risk of coronary heart disease generally increased
with age for each type of duty, whereas the results
for job status were mixed.
Dis c u s sion
In this study, we used data from a nationwide reg-
istry of deaths among firefighters over an 11-year
period and estimates from three different sources
of time spent in various firefighting duties to
estimate the duty-specific risks of death from

coronary heart disease among firefighters. As com-
pared with nonemergency duties, certain emer-
gency duties and physical training were associat-
ed with an increased risk of death from coronary
heart disease among firefighters. These findings
are consistent with those of our previous, smaller
study
18
and with an analysis of cardiac events
that led to retirement from firefighting.
22
Fire suppression, which represents only about
1 to 5% of firefighters’ professional time each
year, accounted for 32% of deaths from coronary
heart disease and was associated with a risk of
death from coronary heart disease that was ap-
proximately 10 to 100 times as high as the risk
associated with nonemergency duties. We think
that the most likely explanation for these find-
ings is the increased cardiovascular demand of
fire suppression.
8,11
The risk of coronary heart disease events dur-
ing fire suppression may be increased because
Table 3. Observed and Expected Distributions of Deaths from Coronary Heart Disease among On-Duty Firefighters, According to Duties.*
Duty
Observed Deaths
(N = 449) Expected Deaths
Municipal Fire Department Large Metropolitan Fire Departments National Data
Expected

Deaths
(N = 449)
Observed:Expected
Deaths
Expected
Deaths
(N = 449)
Observed:Expected
Deaths
Expected
Deaths
(N = 449)
Observed:Expected
Deaths
no. (%) no. (%) ratio (95% CI) no. (%) ratio (95% CI) no. (%) ratio (95% CI)
Fire suppression 144 (32.1) 9.0 (2) 16.0 (13.2–19.1) 22.4 (5) 6.4 (5.3–7.6) 4.5 (1) 32.1 (26.4–38.1)
Alarm response 60 (13.4) 26.9 (6) 2.2 (1.6–3.0) 40.4 (9) 1.5 (1.1–2.0) 18.0 (4) 3.3 (2.4–4.5)
Alarm return 78 (17.4) 44.9 (10) 1.7 (1.3–2.2) 67.4 (15) 1.2 (0.9–1.5) 31.4 (7) 2.5 (1.8–3.2)
Physical training 56 (12.5) 35.9 (8) 1.6 (1.1–2.1) 35.9 (8) 1.6 (1.1–2.1) 35.9 (8) 1.6 (1.1–2.1)
Emergency medical services and other nonfire
emergencies
42 (9.4) 103.3 (23) 0.4 (0.3–0.6) 152.7 (34) 0.3 (0.2–0.4) 67.4 (15) 0.6 (0.4–0.9)
Fire-station and other nonemergency duties
69 (15.4) 229.0 (51) 0.3 (0.2–0.4) 130.2 (29) 0.5 (0.4–0.7) 291.8 (65) 0.2 (0.2–0.3)
* Municipal data are from the Cambridge Fire Department, Cambridge, Massachusetts (2002).
18
Data for large metropolitan fire departments are from surveys of 17 large metropolitan
fire departments conducted by the International Association of Fire Fighters (2005) (Moore-Merrell L.: personal communication). National data are from annual national surveys con-
ducted by the National Fire Protection Association (1994 through 2003).
20

Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
T h e ne w e ngl a n d jo u r na l o f me d i c i n e
n engl j med 356;12 www.nejm.org march 22, 2007
1212
many firefighters lack adequate physical fitness,
have underlying cardiovascular risk factors, and
have subclinical or clinical coronary heart disease.
Even new firefighter recruits may be overweight
and have low-to-normal aerobic capacities.
23
Such
problems are compounded during career tenure
because more than 70% of fire departments lack
programs to promote fitness and health.
1
Most
fire departments do not require firefighters to ex-
ercise regularly, undergo periodic medical exami-
nations, or have mandatory return-to-work eval-
uations after a major illness. In addition, several
studies have shown the high prevalence of risk
factors for cardiovascular disease among fire-
fighters
24-29
as well as lower-than-expected exer-
cise tolerance.
30,31
Moreover, two studies have
shown that among firefighters who had fatal

events
18
or nonfatal events
22
related to coronary
heart disease while on duty, 26% and 18%, respec-
tively, had previously received a diagnosis of coro-
nary heart disease, peripheral vascular disease,
or cerebrovascular disease, and among the remain-
der, smoking, hypertension, and diabetes melli-
tus were significantly more prevalent than among
active firefighters in the control group. Likewise,
in our study, the risk of death from coronary
heart disease increased with age for all types of
duty. Unexpectedly, professional and volunteer
firefighters had different risks of death from
coronary heart disease, depending on the type of
duty performed, although for both groups, the
risk was highest during fire suppression.
In parallel with our finding of a significantly
increased risk of death from coronary heart dis-
ease during fire suppression, as compared with
nonemergency duties, the risk was significantly
elevated during physical training. This finding is
consistent with investigations implicating intense
physical activity as a strong triggering factor, es-
pecially among physically inactive persons.
32-35

Also consistent with the triggering hypothesis

and with research documenting increased heart
rates among firefighters responding to alarms
8,9

was our finding that the risk of death from coro-
nary heart disease associated with alarm response
and alarm return was approximately five to seven
times as high as that associated with nonemer-
gency duties. Emergency medical services and
other nonfire emergency responses were not as-
sociated with a significant increase in risk. These
findings are consistent with the much lower pro-
portion of deaths from coronary heart disease
among emergency medical services workers who
are not firefighters
3
than among firefighters, and
may reflect a lower level of exposure to physically
demanding emergencies.
One limitation of our study is that the esti-
mates of odds ratios for specific job duties are
based on fairly wide approximations of time spent
on different duties. The average work year of a
professional firefighter in a major urban center
is probably much different from that of a rural
volunteer firefighter. In addition, there have been
few if any comprehensive studies of how fire-
Table 4. Risk of Death from Coronary Heart Disease among Firefighters Engaged in Emergency Duties and Physical
Training as Compared with Firefighters Engaged in Nonemergency Duties.*
Duty Municipal Fire Department

Large Metropolitan Fire
Departments National Data
Odds Ratio
(95% CI) P Value
Odds Ratio
(95% CI) P Value
Odds Ratio
(95% CI) P Value
Fire suppression 53 (40–72) <0.001 12.1 (9.0–16.4) <0.001 136 (101–183) <0.001
Alarm response 7.4 (5.1–11) <0.001 2.8 (1.9–4.0) <0.001 14.1 (9.8–20.3) <0.001
Alarm return 5.8 (4.1–8.1) <0.001 2.2 (1.6–3.1) <0.001 10.5 (7.5–14.7) <0.001
Emergency medical services and
other nonfire emergencies
1.3 (0.9–2.0) 0.16 0.5 (0.3–0.8) <0.001 2.6 (1.8–3.9) <0.001
Physical training 5.2 (3.6–7.5) <0.001 2.9 (2.0–4.2) <0.001 6.6 (4.6–9.5) <0.001
Nonemergency duties (fire sta-
tion and other)
1.0 1.0 1.0
* Municipal data are from the Cambridge Fire Department, Cambridge, Massachusetts (2002).
18
Data for large metropol-
itan fire departments are from surveys of 17 large metropolitan fire departments conducted by the International Associ-
ation of Fire Fighters (2005) (Moore-Merrell L.: personal communication). National data are from annual national sur-
veys conducted by the National Fire Protection Association (1994 through 2003).
20
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
De aths from He art Disease a mong Firefighter s
n engl j med 356;12 www.nejm.org march 22, 2007
1213

fighters spend their time. Our estimate of the
increase in risk is therefore subject to considera-
ble uncertainty. However, even in the most conser-
vative scenario (with the use of the time estimates
from the large metropolitan fire departments), the
risks associated with fire suppression remained
remarkably high and were also significantly in-
creased for alarm response, alarm return, and
physical training.
Also, our three sets of risk estimates are not
based on three completely distinct calculations.
In each case, one set of national figures for “ob-
served” deaths was used, and the resulting odds
ratios represent risk relative to nonemergency
duties, not absolute risks for one group of fire-
fighters as compared with another. Our results
should therefore not be used to suggest that the
risk of death from coronary heart disease during
fire suppression is higher in a small community
fire department than in a large metropolitan fire
department. Instead, the three calculations pro-
vide a range of estimates of the average risk for
firefighters nationwide. Because only 14% of fire-
fighters in the United States serve populations
larger than 100,000 residents,
21
we think that the
average risk for most firefighters probably falls
between the risk based on estimates of time
spent in particular duties that were derived from

a single municipal fire department and the risk
based on the nationwide time estimates. Our es-
timate that fire suppression accounts for 1 to 2%
of annual work time (for the nationwide and mu-
nicipal scenarios, respectively) is consistent with
a study of a large fire department in Montreal,
36

where fire suppression accounted for 0.7 to 2.5%
of annual work time.
A second limitation of our study was the need
to base our evaluation on brief narratives, which
lacked autopsy information for some of the deaths.
However, the misclassification of deaths due to
inadequate information would have contributed
to a random error, most likely diluting the results
of our study toward the null hypothesis. Although
26 deaths from cardiovascular but not coronary
heart disease were excluded, this small number
was unlikely to bias the overall results in a spe-
cific direction.
A third limitation of our analysis was the
starting assumption that the number of deaths
from coronary heart disease that occur during
any given firefighting duty should be directly pro-
portional to the amount of time spent perform-
ing that duty. It is well established, for example,
that the risk of coronary heart disease events var-
ies according to the time of day,
37

as well as the
season of the year.
38
In this study, we could not
examine the circadian pattern of deaths. How-
ever, in our previous, smaller study
18
and in an-
other, 10-year analysis,
2
67 to 77% of deaths from
cardiac causes among on-duty firefighters oc-
curred between noon and midnight, as did more
than 60% of emergency responses. This pattern
is in stark contrast to the peak period for cardio-
vascular events in the general population, which
is 6 a.m. to noon. With respect to season, deaths
from cardiac causes among firefighters are most
frequent in the winter, as they are in the general
population. When we analyzed duty-specific risks
22p3
20–39 Yr 40–49 Yr 50–59 Yr ≥60 Yr
Volunteer Professional
Annual No. of Deaths per 1 Million
Firefighters
40
30
10
50
20

0
Fire
Suppression
Alarm
Response
Alarm
Return
Physical
Training
Emergency
Medical
Services
Fire-
Station
Duty
60
Annual No. of Deaths per 1 Million
Firefighters
12
10
2
6
4
14
8
0
Fire
Suppression
Alarm
Response

Alarm
Return
Physical
Training
Emergency
Medical
Services
Fire-
Station
Duty
16
AUTHOR:
FIGURE:
JOB: ISSUE:
4-C
H/T
RETAKE
SIZE
ICM
CASE
EMail
Line
H/T
Combo
Revised
AUTHOR, PLEASE NOTE:
Figure has been redrawn and type has been reset.
Please check carefully.
REG F
Enon

1st
2nd
3rd
Kales
1 of 1
03-22-07
ARTIST: ts
35612
A
B
Figure 1. Duty-Specific Annual Risk of Death from Coronary Heart Disease
among Firefighters, According to Age (Panel A) and Job Status (Panel B).
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
T h e ne w e ngl a n d jo u r na l o f me d i c i n e
n engl j med 356;12 www.nejm.org march 22, 2007
1214
separately for each of the four seasons, however,
the resulting point estimates for each duty re-
mained similar in magnitude and close to the
range of our original confidence intervals. Final-
ly, although we cannot completely account for
the effects of the time of day and season, the high-
est estimates of these effects on event rates are
at least an order of magnitude smaller than the
relative risks we observed for specific duties.
In conclusion, we analyzed nationwide data
on deaths among firefighters, as well as three
separate estimates of time spent in various fire-
fighting duties, to determine the duty-specific

risks of death from coronary heart disease among
firefighters. Our analysis showed that specific
duties, especially fire suppression but also alarm
response, alarm return, and physical training, are
associated with significant increases in risk.
Supported in part by grants from the National Institute for
Occupational Safety and Health (T42/CCT122961-02, to Dr.
Kales) and the Massachusetts Public Employees Retirement
Administration Commission (to Dr. Kales). The funders had no
involvement in the study design, data collection and analysis,
writing of the paper, or decision to submit the paper for publi-
cation.
Dr. Kales and Dr. Christiani report serving as paid expert wit-
nesses, independent medical examiners, or both in workers’ com-
pensation and disability cases, including cases involving fire-
fighters. No other potential conflict of interest relevant to this
article was reported.
We thank Ken Pitts, John Gelinas, and Lori Moore-Merrell for
providing fire-department incident, response, activity, and sur-
vey data.
References
Fahy RF. U.S. firefighter fatalities
due to sudden cardiac death, 1995–2004.
Quincy, MA: National Fire Protection As-
sociation, June 2005. (Accessed February
21, 2007, at />files/PDF/OSCardiacDeath.pdf.)
Firefighter fatality retrospective study,
April 2002. (Prepared for the Federal Emer-
gency Management Agency, United States
Fire Service, National Fire Data Center.)

Arlington, VA: TriData Corp., 2002.
Maguire BJ, Hunting KL, Smith GS,
Levick NR. Occupational fatalities in emer-
gency medical services: a hidden crisis.
Ann Emerg Med 2002;40:625-32.
Franke WD, Anderson DF. Relation-
ship between physical activity and risk
factors for cardiovascular disease among
law enforcement officers. J Occup Med
1994;36:1127-32.
Fisher NG, Nicol ED. Cardiological
disease in the Armed Forces: a clear and
present danger. J R Nav Med Serv 2005;
91:112-7.
Melius J. Occupational health for fire-
fighters. Occup Med 2001;16:101-8.
Guidotti TL. Human factors in fire-
fighting: ergonomic-, cardiopulmonary-,
and psychogenic stress-related issues. Int
Arch Occup Environ Health 1992;64:
1-12.
Barnard RJ, Duncan HW. Heart rate
and ECG responses of fire fighters. J Oc-
cup Med 1975;17:247-50.
Kuorinka I, Korhonen O. Firefighters’
reaction to alarm, an ECG and heart rate
study. J Occup Med 1981;23:762-6.
Burgess JL, Nanson CJ, Bolstad-John-
son DM, et al. Adverse respiratory effects
following overhaul in firefighters. J Occup

Environ Med 2001;43:467-73.
Smith DL, Manning TS, Petruzzello SJ.
Effect of strenuous live-fire drills on car-
diovascular and psychological responses
of recruit firefighters. Ergonomics 2001;
44:244-54.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Kawachi I, Colditz GA, Stampfer MJ,
et al. Prospective study of shift work and
risk of coronary heart disease in women.
Circulation 1995;92:3178-82.
Friel JK, Stones M. Firefighters and
heart disease. Am J Public Health 1992;82:
1175-6.
International Association of Fire Fight-
ers. Presumptive legislation. (Accessed Feb-
ruary 21, 2007, at />content/presumptive/infselect.asp.)
Steenland K. Epidemiology of occupa-
tion and coronary heart disease: research
agenda. Am J Ind Med 1996;30:495-9.

Guidotti TL. Occupational mortality
among firefighters: assessing the associa-
tion. J Occup Environ Med 1995;37:1348-56.
Haas NS, Gochfeld M, Robson MG,
Wartenberg D. Latent health effects in
firefighters. Int J Occup Environ Health
2003;9:95-103.
Kales SN, Soteriades ES, Christoudias
SG, Christiani DC. Firefighters and on-
duty deaths from coronary heart disease:
a case control study. Environ Health 2003;
2:14.
United States Fire Administration.
Firefighter fatalities. (Accessed February
21, 2007, at />fatalities/.)
National Fire Protection Association.
Fire statistics, U.S. fire service. (Accessed
February 21, 2007, at a.
org/categoryList.asp?categoryID=955&
URL=Research%20&%20Reports/
Fire%20statistics/Fire%20service.)
Karter MJ. U.S. fire department pro-
file through 2000. Quincy, MA: National
Fire Protection Association, 2001.
Holder JD, Stallings LA, Peeples L,
Burress JW, Kales SN. Firefighter heart
presumption retirements in Massachusetts:
1997-2004. J Occup Environ Med 2006;48:
1047-53.
Roberts MA, O’Dea J, Boyce A, Mannix

ET. Fitness levels of firefighter recruits
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
before and after a supervised exercise
training program. J Strength Cond Res
2002;16:271-7.
Kales SN, Polyhronopoulos GN, Al-
drich JM, Leitao EO, Christiani DC. Cor-
relates of body mass index in hazardous
materials firefighters. J Occup Environ
Med 1999;41:589-95.
Clark S, Rene A, Theurer WM, Mar-
shall M. Association of body mass index
and health status in firefighters. J Occup
Environ Med 2002;44:940-6.
Soteriades ES, Hauser R, Kawachi I,
Liarokapis D, Christiani DC, Kales SN.
Obesity and cardiovascular disease risk
factors in firefighters: a prospective cohort
study. Obes Res 2005;13:1756-63.

Ide CW. A longitudinal survey of the
evolution of some cardiovascular risk fac-
tors during the careers of male firefight-
ers retiring from Strathclyde Fire Brigade
from 1985–1994. Scott Med J 2000;45:79-
83.
Glueck CJ, Kelley W, Wang P, Gartside
PS, Black D, Tracy T. Risk factors for coro-
nary heart disease among firefighters in
Cincinnati. Am J Ind Med 1996;30:331-
40.
Soteriades ES, Kales SN, Liarokapis D,
Christoudias SG, Tucker SA, Christiani DC.
Lipid profile of firefighters over time:
opportunities for prevention. J Occup Envi-
ron Med 2002;44:840-6.
Lemon PW, Hermiston RT. Physiolog-
ical profile of professional fire fighters.
J Occup Med 1977;19:337-40.
Kales SN, Christiani DC. Cardiovas-
cular fitness in firefighters. J Occup Envi-
ron Med 2000;42:467-8.
Johnstone MT, Mittleman M, Tofler
G, Muller JE. The pathophysiology of the
onset of morning cardiovascular events.
Am J Hypertens 1996;9:22S-28S.
Mittleman MA, Maclure M, Tofler GH,
Sherwood JB, Goldberg RJ, Muller JE.
Triggering of acute myocardial infarction
24.

25.
26.
27.
28.
29.
30.
31.
32.
33.
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
De aths from He art Disease a mong Firefighter s
n engl j med 356;12 www.nejm.org march 22, 2007
1215
by heavy physical exertion: protection
against triggering by regular exertion.
N Engl J Med 1993;329:1677-83.
Franklin BA, Bonzheim K, Gordon S,
Timmis GC. Snow shoveling: a trigger for
acute myocardial infarction and sudden
coronary death. Am J Cardiol 1996;77:
855-8.
Willich SN, Klatt S, Arntz HR. Circa-
dian variation and triggers of acute coro-
34.
35.
nary syndromes. Eur Heart J 1998;19:
Suppl C:C12-C23.
Austin CC, Dussault G, Ecobichon DJ.
Municipal firefighter exposure groups,

time spent at fires and use of self-con-
tained-breathing-apparatus. Am J Ind Med
2001;40:683-92.
Cohen MC, Rohtla KM, Lavery CE,
Muller JE, Mittleman MA. Meta-analysis
of the morning excess of acute myocardial
36.
37.
infarction and sudden cardiac death. Am
J Cardiol 1997;79:1512-6. [Erratum, Am J
Cardiol 1998;81:260.]
Spencer FA, Goldberg RJ, Becker RC,
Gore JM. Seasonal distribution of acute
myocardial infarction in the second Na-
tional Registry of Myocardial Infarction.
J Am Coll Cardiol 1998;31:1226-33.
Copyright © 2007 Massachusetts Medical Society.
38.
Copyright © 2007 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
original article
The
new england journal of medicine
1660
Endarterectomy versus Stenting in Patients
with Symptomatic Severe Carotid Stenosis
Jean-Louis Mas, M.D., Gilles Chatellier, M.D., Bernard Beyssen, M.D.,
Alain Branchereau, M.D., Thierry Moulin, M.D., Jean-Pierre Becquemin, M.D.,
Vincent Larrue, M.D., Michel Lièvre, M.D., Didier Leys, M.D., Ph.D.,
Jean-François Bonneville, M.D., Jacques Watelet, M.D.,

Jean-Pierre Pruvo, M.D., Ph.D., Jean-François Albucher, M.D.,
Alain Viguier, M.D., Philippe Piquet, M.D., Pierre Garnier, M.D.,
Fausto Viader, M.D., Emmanuel Touzé, M.D., Maurice Giroud, M.D.,
Hassan Hosseini, M.D., Ph.D., Jean-Christophe Pillet, M.D.,
Pascal Favrole, M.D., Jean-Philippe Neau, M.D., and Xavier Ducrocq, M.D.,
for the EVA-3S Investigators*
From Hôpitaux Sainte-Anne (J L.M., B.B.,
E.T.) and Europeén Georges Pompidou
(G.C.), Université René Descartes, Paris;
Hôpitaux La Timone (A.B.) and Sainte-
Marguerite (P.P.), Université de la Médi-
terranée, Marseille; Hôpital Jean Minjoz,
Université de Franche-Comté, Besançon
(T.M., J F.B.); Hôpital Henri Mondor,
Université Paris-Val-de-Marne, Créteil
(J P.B., H.H.); Hôpitaux Rangueil (V.L.,
A.V.) and Purpan (J F.A.), Université Paul
Sabatier, Toulouse; Université Claude Ber-
nard, Lyon (M.L.); Hôpital Roger Salen-
gro, Université du Droit et de la Santé,
Lille (D.L., J P.P.); Hôpital Charles Nicolle,
Université de Rouen, Rouen (J.W.); Hôpi-
tal de Bellevue, Université Jean Monnet,
Saint-Etienne (P.G.); Hôpital Côte de Na-
cre, Université de Caen, Caen (F.V.); Hôpi-
tal Général, Université de Bourgogne,
Dijon (M.G.); Nouvelles Cliniques Nan-
taises, Nantes (J C.P.); Hôpital Lari-
boisière, Université Denis Diderot, Paris
(P.F.); Hôpital La Milétrie, Université de

Poitiers, Poitiers (J P.N.); and Hôpital
Saint-Julien Université Henri Poincaré,
Nancy (X.D.) — all in France. Address
reprint requests to Dr. Mas at the Service
de Neurologie, Hôpital Sainte-Anne, 1 Rue
Cabanis, 75674 Paris CEDEX 14, France,
or at
*
Investigators and committees of the
Endarterectomy versus Angioplasty
in Patients with Symptomatic Severe
Carotid Stenosis (EVA-3S) trial are
listed in
the Appendix.
N Engl J Med 2006;355:1660-71.
Copyright © 2006 Massachusetts Medical Society.
ABSTRACT
Background
Carotid stenting is less invasive than endarterectomy, but it is unclear whether it is as
safe in patients with symptomatic carotid-artery stenosis.
Methods
We conducted a multicenter, randomized, noninferiority trial to compare stenting
with endarterectomy in patients with a symptomatic carotid stenosis of at least 60%.
The primary end point was the incidence of any stroke or death within 30 days after
treatment.
Results
The trial was stopped prematurely after the inclusion of 527 patients for reasons of
both safety and futility. The 30-day incidence of any stroke or death was 3.9% after
endarterectomy (95% confidence interval [CI], 2.0 to 7.2) and 9.6% after stenting
(95% CI, 6.4 to 14.0); the relative risk of any stroke or death after stenting as com-

pared with endarterectomy was 2.5 (95% CI, 1.2 to 5.1). The 30-day incidence of dis-
abling stroke or death was 1.5% after endarterectomy (95% CI, 0.5 to 4.2) and 3.4%
after stenting (95% CI, 1.7 to 6.7); the relative risk was 2.2 (95% CI, 0.7 to 7.2). At
6 months, the incidence of any stroke or death was 6.1% after endarterectomy and
11.7% after stenting (P = 0.02). There were more major local complications after stent-
ing and more systemic complications (mainly pulmonary) after endarterectomy, but
the differences were not significant. Cranial-nerve injury was more common after
endarterectomy than after stenting.
Conclusions
In this study of patients with symptomatic carotid stenosis of 60% or more, the rates
of death and stroke at 1 and 6 months were lower with endarterectomy than with
stenting. (ClinicalTrials.gov number, NCT00190398.)
n engl j med 355;16 www.nejm.org october 19, 2006
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
endarterectomy versus stenting in severe carotid stenosis
n engl j med 355;16 www.nejm.org october 19, 2006
1661
F
indings from two large randomized,
clinical trials
1-3
have established endarterec-
tomy as the standard treatment for severe
symptomatic carotid-artery stenosis. As compared
with endarterectomy, stenting avoids the need for
general anesthesia and an incision in the neck that
could lead to nerve injury and wound complica-
tions. The costs may be less than those of surgery,
mainly because the hospital stay is shorter. How-

ever, stenting also carries a risk of stroke and local
complications, and the long-term efficacy of this
technique is not well known. A systematic review
4

of five randomized trials comparing stenting with
endarterectomy
5-10
concluded that the current evi-
dence does not support a change from the recom-
mendation of carotid endarterectomy as the stan-
dard treatment for carotid stenosis. Several more
trials are in progress in Europe
11-13
and the United
States.
14
We conducted this trial, which started in No-
vember 2000, to evaluate whether stenting is not
inferior to endarterectomy with regard to the
risks of the procedure and its long-term efficacy
in patients with symptomatic carotid stenosis.
In September 2005, the safety committee recom-
mended that enrollment in the trial be stopped.
We report on the risks of stroke or death within
30 days and 6 months after treatment.
Methods
The Endarterectomy versus Angioplasty in Patients
with Symptomatic Severe Carotid Stenosis (EVA-3S)
trial, a publicly funded, randomized, noninferiority

trial, was conducted in 20 academic and 10 non-
academic centers in France. The study was ap-
proved by the ethics committee of Hôpital Cochin
in Paris. All patients provided written informed
consent.
Centers and Investigators
To join the trial, each center was required to as-
semble a team of physicians comprising at least one
neurologist, one vascular surgeon, and one inter-
ventional physician. The neurologist was respon-
sible for the initial evaluation and follow-up of the
patients. The vascular surgeon had to have per-
formed at least 25 endarterectomies in the year
before enrollment. The interventional physician
had to have performed at least 12 carotid-stenting
procedures or at least 35 stenting procedures in
the supraaortic trunks, of which at least 5 were in
the carotid artery. Centers fulfilling all require-
ments except those with regard to the interven-
tional physician could join the EVA-3S study and
randomly assign patients, but all stenting proce-
dures had to be performed under the supervision
of an experienced tutor (a clinician who qualified
to perform stenting in this study) until the local
interventional physician became self-sufficient (ac-
cording to the tutor) and performed a sufficient
number of procedures according to the predefined
criteria.
Patients
Patients were eligible if they were 18 years of age

or older, had had a hemispheric or retinal transient
ischemic attack or a nondisabling stroke (or reti-
nal infarct) within 120 days before enrollment, and
had a stenosis of 60 to 99% in the symptomatic
carotid artery, as determined by the North Amer-
ican Symptomatic Carotid Endarterectomy Trial
(NASCET) method.
15
The degree of stenosis war-
ranting treatment, set at 70% or more at the start of
the trial, was subsequently (in October 2003) set at
60% or more because endarterectomy was shown
to benefit patients with symptomatic stenosis of
50 to 69%.
3
The presence of an ipsilateral carotid
stenosis of 60% or more had to be confirmed by
means of catheter angiography or both duplex
scanning and magnetic resonance angiography
of the carotid artery.
Patients were excluded if one of the following
was present: a modified Rankin score
16
of 3 or
more (disabling stroke) (on a scale of 0 to 5, with
higher scores indicating more severe disability);
nonatherosclerotic carotid disease; severe tan-
dem lesions (stenosis of proximal common carotid
artery or intracranial artery that was more severe
than the cervical lesion); previous revasculariza-

tion of the symptomatic stenosis; history of bleed-
ing disorder; uncontrolled hypertension or diabe-
tes; unstable angina; contraindication to heparin,
ticlopidine, or clopidogrel; life expectancy of less
than 2 years; or percutaneous or surgical interven-
tion within 30 days before or after the study pro-
cedure. The appearance of the stenotic lesion on
angiography was not a factor in the selection of
patients.
Patients who were suitable candidates for both
techniques were randomly assigned to undergo
endarterectomy or stenting. Randomization was
carried out centrally by means of a computer-gen-
erated sequence, involving randomized blocks of
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
The new england journal of medicine
n engl j med 355;16 www.nejm.org october 19, 2006
1662
two, four, or six patients that were stratified accord-
ing to study center and degree of stenosis (steno-
sis of ≥90% or <90%).
Endarterectomy and Stenting
The goal was for endarterectomy and stenting to
be performed within 2 weeks after randomization.
Surgeons performed endarterectomy according to
customary practice. Carotid stenting had to be car-
ried out through the femoral route with the use of
stents and protection devices approved by the ac-
creditation committee. Interventional physicians

had to have performed at least two stenting pro-
cedures with any new device before its use in the
trial. In January 2003, the safety committee recom-
mended the systematic use of stents with cerebral
protection devices because of a higher risk of stroke
in patients treated without cerebral protection
17
;
centers began using them on February 1, 2003. The
daily use of aspirin (100 to 300 mg) and clopido-
grel (75 mg) or ticlopidine (500 mg) for 3 days be-
fore and 30 days after stenting was also recom-
mended.
Follow-up and End Points
The study neurologists performed follow-up eval-
uations at 48 hours, 30 days, 6 months after treat-
ment, and every 6 months thereafter. The primary
end point was a composite of any stroke or death
occurring within 30 days after treatment. Second-
ary outcomes were myocardial infarction, transient
ischemic attack, cranial-nerve injury, major local
complications, and systemic complications with-
in 30 days after treatment; and composites of any
stroke or death within 30 days after treatment plus
ipsilateral stroke, any stroke, or any stroke or death
within 31 days through the end of follow-up. Neu-
rologists assessed the degree of disability from
stroke 30 days and 6 months after the event. Func-
tional disability from cranial-nerve injury was cat-
egorized as absent, mild, moderate, or severe at the

30-day follow-up visit. Neurologists also recorded
whether treatment-related outcomes were associ-
ated with a delay in discharge. The occurrence of
stroke, death, and other outcomes was assessed by
the events committee, which was unaware of the
treatment assignments (except for patients who
had local complications).
Statistical Analysis
We calculated
18
that we would need to enroll 872
patients for the study to have a statistical power
of 80% to assess whether stenting was not infe-
rior to endarterectomy with regard to the 30-day
incidence of stroke or death, given an expected
30-day incidence of stroke or death of 5.6% af-
ter endarterectomy
19
and 4% after stenting,
20,21

a true absolute difference between groups in
the 30-day risk of stroke or death of no more
than 2% (noninferiority margin), and a one-
sided alpha of 0.05. A similar difference in the
30-day risk of stroke or death between endarter-
ectomy and medical treatment was observed in
NASCET.
22
Our protocol required that an inde-

pendent safety committee review safety issues
each time 10 new validated primary outcome
events occurred, with no predetermined rule for
stopping the trial, and reassess the number of
patients required to show an effect after 30 pri-
mary outcome events had occurred. In Septem-
ber 2005, the safety committee recommended
stopping enrollment for reasons of both safety
and futility. On the basis of the observed 30-day
risk of stroke or death after endarterectomy, we
would have needed to enroll more than 4000
patients to test the noninferiority of stenting
(assuming that the relative noninferiority limit
was unchanged). Given the observed 30-day
risks of stenting, the committee considered it to
be extremely unlikely that the trial, should it
continue with more patients, would reach its
objectives.
Analyses of the 30-day outcomes were based on
all patients who were randomly assigned to treat-
ment and who underwent carotid repair. The re-
sults are presented as relative risks with 95% con-
fidence intervals (CIs), calculated with the use of
superiority analysis. We also assessed noninferior-
ity, as initially planned. We assessed homogeneity
of the relative risks of stroke or death among cen-
ters using the Breslow–Day test. For this purpose,
centers were categorized into three groups, ac-
cording to the numbers of patients included in the
study (<21, 21 to 40, and >40 patients). Analyses of

the 6-month outcomes were based on all patients
who were randomly assigned to treatment. Rates
of stroke and death were estimated with the use
of the Kaplan–Meier method. All data were ana-
lyzed according to the intention-to-treat principle.
All P values are two-sided and were not adjusted
for multiple testing. We used SAS software (version
8.2) for all analyses. The authors vouch for the
completeness and veracity of the data and data
analyses.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
endarterectomy versus stenting in severe carotid stenosis
n engl j med 355;16 www.nejm.org october 19, 2006
1663
Results
Patients and Treatments
By September 2005, 527 patients had been random-
ly assigned to treatment, 7 of whom did not un-
dergo carotid repair (Fig. 1). The remaining 520
patients were included in the analysis of the 30-day
risk of stroke or death. Three strokes that occurred
between randomization and treatment were not
included in the analysis of the 30-day risk of stroke
or death but were included in the 6-month analy-
sis of outcomes. The two groups were similar with
respect to baseline characteristics, except for a
greater proportion of patients 75 years of age or
older and more patients with a history of stroke
in the endarterectomy group and a higher propor-

tion of contralateral carotid occlusion in the stent-
ing group (
Table 1
).
Characteristics of the endarterectomy and stent-
ing procedures are listed in Table 2. Two patients
underwent a repeated procedure within 48 hours
after the initial endarterectomy owing to residu-
al stenosis or dissection. In 13 patients, stenting
was converted intraoperatively to endarterectomy
owing to problems with access. Two of these pa-
tients had a stroke before endarterectomy.
End Points
Although the trial was intended to assess noninfe-
riority, we observed that stenting carried a greater
risk than did endarterectomy. When we analyzed
the data as planned, the 95% CI of the difference
in the 30-day incidence of stroke or death between
stenting and endarterectomy (2.1 to 9.3%) did not
include the 2% limit used to define noninferior-
ity. The 30-day incidence of any stroke or death was
3.9% (95% CI, 2.0 to 7.2) after endarterectomy and
9.6% (95% CI, 6.4 to 14.0) after stenting, with a
relative risk of 2.5 (95% CI, 1.2 to 5.1). The abso-
lute risk increase was 5.7%, suggesting that one
additional stroke or death resulted when 17 pa-
tients underwent stenting rather than endarter-
ectomy. The 30-day incidence of disabling stroke
or death was 1.5% (95% CI, 0.5 to 4.2) after end-
arterectomy and 3.4% (95% CI, 1.7 to 6.7) after

stenting, resulting in a relative risk of 2.2 (95%
CI, 0.7 to 7.2) (
Table 3
). A greater proportion of
strokes occurred on the day of the procedure in
the stenting group than in the endarterectomy
group (17 of 24 vs. 3 of 9, P = 0.05).
The relative risk of stroke or death did not dif-
fer significantly among the centers that enrolled
fewer than 21 patients (relative risk, 1.9; 95% CI, 0.6
to 6.2), those that enrolled 21 to 40 patients (rela-
tive risk, 3.3; 95% CI, 0.7 to 15.2), and those that
enrolled more than 40 patients (relative risk, 2.7;
95% CI, 0.9 to 8.1) (P = 0.83). The 30-day incidence
of stroke or death was similar among patients
treated by interventional physicians who were
experienced (11 of 105, or 10.5%), tutored during
training (7 of 98, or 7.1%), and tutored after train-
ing (7 of 57, or 12.3%) (P = 0.54; chi-square sta-
tistic, 1.25).
527 Patients underwent
randomization
262 Assigned to
endarterectomy
3 Did not undergo endarterectomy
1 Declined
1 Had carotid occlusion
1 Had a stroke before treatment
259 Underwent carotid repair and
were included in analysis of

the primary outcome
2 Had a TIA between randomi-
zation and endarterectomy
2 Underwent stenting
265 Assigned to stenting
4 Did not undergo stenting
1 Declined
2 Had <60% stenosis
1 Had a stroke before treatment
261 Underwent carotid repair and
were included in analysis of
the primary outcome
1 Had a nondisabling stroke
1 Had a TIA and 1 had a myo-
cardial infarction between
randomization and stenting
1 Underwent endarterectomy
Endarterectomy attempted
in 257
Stenting attempted
in 260
Endarterectomy completed
in 257
Stenting failed in 13, who then
underwent endarterectomy
Stenting completed in 247
Figure 1. Randomization and Treatment.
Carotid repair was successful in the five patients who had a transient is che-
mic attack (TIA), a nondisabling stroke, or myocardial infarction between
randomization and carotid repair. The three strokes occurred within 2 days

after randomization.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
The new england journal of medicine
n engl j med 355;16 www.nejm.org october 19, 2006
1664
The 30-day incidence of stroke or death was
lower among patients who underwent stenting
with cerebral protection (18 of 227, or 7.9%) than
among those treated with stenting alone (5 of 20,
or 25%; P = 0.03). However, the relative risk of stroke
or death for stenting over endarterectomy did not
differ significantly before systematic use of a ce-
rebral protection device was recommended (2.0;
95% CI, 0.8 to 5.0) or after (3.4; 95% CI, 1.1 to
10.0; P = 0.50).
Table 1. Baseline Characteristics of the Patients.*
Characteristic
Endarterectomy
Group
(N = 259)
Stenting Group
(N = 261) P Value
Age — yr 70.3±10.7 69.1±10.2 0.21
Age ≥75 yr — no. of patients (%) 105 (40.5) 84 (32.2) 0.06
Male sex — no. of patients (%) 202 (78.0) 189 (72.4) 0.16
Vascular risk factors
Hypertension — no. of patients (%)† 188 (72.6) 192 (73.6) 0.84
Diabetes — no. of patients (%)† 66 (25.5) 58 (22.2) 0.41
Hypercholesterolemia — no. of patients (%)† 144 (55.6) 151 (57.9) 0.66

Tobacco use — no. of patients (%)‡ 61 (23.6) 63 (24.1) 0.92
Systolic blood pressure — mm Hg 140.8±17.8 140.9±16.5 0.95
Diastolic blood pressure — mm Hg 77.1±10.8 77.8±9.9 0.42
Body-mass index§ 26.3±4.1 26.1±4.6 0.63
History of vascular disease — no. of patients (%)
Stroke 52 (20.1) 33 (12.6) 0.02
Transient ischemic attack 60 (23.2) 66 (25.3) 0.61
Myocardial infarction 34 (13.1) 28 (10.7) 0.42
Peripheral arterial disease 30 (11.6) 37 (14.2) 0.43
Cardiac failure leading to hospitalization 7 (2.7) 7 (2.7) 1.00
Prior surgery or angioplasty — no. of patients (%)
Coronary artery 34 (13.1) 35 (13.4) 1.00
Carotid artery 10 (3.9) 5 (1.9) 0.20
Other artery 23 (8.9) 16 (6.1) 0.25
Medications before qualifying event — no. of patients (%)
Antiplatelet therapy 136 (52.5) 128 (49.0) 0.43
Antihypertensive medication 177 (68.3) 179 (68.6) 1.00
Antidiabetic agents 64 (24.7) 53 (20.3) 0.25
Lipid-lowering medication 125 (48.3) 129 (49.4) 0.79
Qualifying event — no. of patients (%) 0.47
Cerebral transient ischemic attack 78 (30.1) 95 (36.4)
Ocular transient ischemic attack 37 (14.3) 33 (12.6)
Ischemic stroke 139 (53.7) 127 (48.7)
Retinal infarct 5 (1.9) 6 (2.3)
Modified Rankin score at randomization¶ 0.94
0 145 (56.0) 139 (53.3)
1 68 (26.3) 71 (27.2)
2 42 (16.2) 47 (18.0)
3 4 (1.5) 4 (1.5)
Copyright © 2006 Massachusetts Medical Society. All rights reserved.

Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
endarterectomy versus stenting in severe carotid stenosis
n engl j med 355;16 www.nejm.org october 19, 2006
1665
The relative risk of stroke or death adjusted for
age was 2.4 (95% CI, 1.2 to 4.8) and adjusted for
the presence or absence of a history of stroke was
2.6 (95% CI, 1.3 to 5.2). More patients in the stent-
ing group had contralateral carotid occlusion; none
of them had a stroke after stenting. The 30-day
incidence of stroke or death after stenting did not
differ significantly between patients who received
dual antiplatelet therapy (19 of 211, or 9.0%) and
those who received single antiplatelet therapy (4 of
36, or 11.1%; P = 0.75).
There were more systemic complications (main-
ly pulmonary) after endarterectomy and more se-
vere local complications after stenting than after
endarterectomy, but these differences were not
significant. Cranial-nerve injury was significantly
more common after endarterectomy than after
stenting (7.7% vs. 1.1%, P<0.001). The median
duration of the hospital stay was shorter after
stenting (3 days; interquartile range, 2 to 5) than
after endarterectomy (4 days; interquartile range,
3 to 5; P = 0.01).
Table 1. (Continued.)
Characteristic
Endarterectomy
Group

(N = 259)
Stenting Group
(N = 261) P Value
Brain imaging — no. of patients (%)
Computed tomography 217 (83.8) 230 (88.1) 0.17
Magnetic resonance imaging 161 (62.2) 156 (59.8) 0.59
Infarct corresponding to the qualifying event 133 (51.4) 117 (44.8) 0.16
Previous infarct 70 (27.0) 68 (26.1) 0.84
Diagnostic carotid angiography — no. of patients (%)
Catheter angiography 110 (42.5) 113 (43.3) 0.86
Magnetic resonance angiography 163 (62.9) 161 (61.7) 0.79
Ultrasonography 245 (94.6) 253 (96.9) 0.20
Degree of symptomatic carotid stenosis — no. of patients (%)∥ 0.68
60–69% 21 (8.1) 15 (5.7)
70–79% 55 (21.2) 56 (21.5)
80–89% 77 (29.7) 86 (33.0)
90–99% 106 (40.9) 104 (39.8)
Contralateral carotid occlusion — no. of patients (%) 3 (1.2) 13 (5.0) 0.02
Contralateral stenosis of 60–99% — no. of patients (%) 44 (17.0) 31 (11.9) 0.11
Time from qualifying event to treatment — no. of patients (%) 0.62
<2 wk 41 (15.8) 53 (20.3)
2–4 wk 68 (26.3) 66 (25.3)
4–12 wk 124 (47.9) 118 (45.2)
>12 wk 26 (10.0) 24 (9.2)
Time from randomization to treatment
Median — days 6.0 6.0 0.52
Interquartile range — days 2–10 3–9
<2 wk — no. of patients (%) 240 (92.7) 249 (95.4) 0.26
* Plus–minus values are means ±SD. Proportions, means, and medians were compared with the use of Fisher’s exact
test, Student’s t-test, and the Wilcoxon nonparametric test, respectively.

† This condition was diagnosed before the qualifying event.
‡ Tobacco use was defined as the smoking of one cigarette or more per day.
§ The body-mass index is the weight in kilograms divided by the square of the height in meters.
¶ The modified Rankin score ranges from 0 to 5, with higher scores indicating more severe disability.
∥ The degree of stenosis was measured with the use of digital subtraction angiography or magnetic resonance angiogra-
phy, according to the NASCET method.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
The new england journal of medicine
n engl j med 355;16 www.nejm.org october 19, 2006
1666
Table 2. Characteristics of Treatment for 257 Patients Who Completed Endarterectomy and 247 Patients Who
Completed Stenting.*
Treatment Group Value
Endarterectomy
Anesthesia — no. of patients (%)†
General 187 (73.0)
Locoregional 69 (27.0)
Cerebral monitoring — no. of patients (%)† 66 (25.8)
Surgical technique — no. of patients (%)
Endarterectomy
With the use of a patch 129 (50.2)
Without the use of a patch 53 (20.6)
Eversion 63 (24.5)
Carotid–carotid bypass 11 (4.3)
Transposition 1 (0.4)
Shunt 50 (19.5)
Duration of surgery — min
Median 80
Interquartile range 60–110

Medical treatment — no. of patients (%)
Preprocedure‡
Antiplatelet therapy 168 (69.1)
Anticoagulant therapy 30 (12.3)
Both 45 (18.5)
During procedure§
Heparin 253 (99.2)
Postprocedure¶
Antiplatelet therapy 97 (38.2)
Anticoagulant therapy∥ 16 (6.3)
Both 141 (55.5)
Stenting
Anesthesia — no. of patients (%)
General 16 (6.5)
Neuroleptanalgesia 56 (22.7)
Local 175 (70.9)
Mean length of lesion— mm 15.6±7.9
Femoral route — no. of patients (%) 238 (96.4)
Predilatation of stenosis — no. of patients (%) 42 (17.0)
Number of stents — no. of patients (%)
0 1 (0.4)
1 236 (95.5)
2 or more 10 (4.0)
Stent across external carotid artery — no. of patients (%)** 202 (82.1)
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
endarterectomy versus stenting in severe carotid stenosis
n engl j med 355;16 www.nejm.org october 19, 2006
1667
Table 2. (Continued.)

Treatment Group Value
Stenting
Duration of procedure (min)
Median 70
Interquartile range 50–90
Type of stent used — no. of patients (%)**
Carotid Wallstent Monorail (Boston Scientific) 140 (56.9)
Acculink (Abbott) 70 (28.5)
Precise RX (Cordis) 26 (10.6)
Carotid Wallstent OTW (Boston Scientific) 5 (2.0)
Zilver (Cook) 5 (2.0)
Cerebral protection — no. of patients (%) 227 (91.9)
Before systematic use of protection devices recommended
by safety committee
58 (78.4)††
After systematic use of protection devices recommended
by safety committee
169 (97.7)††
Device used
GuardWire Plus (Medtronic) 67 (29.5)
FilterWire EZ (Boston Scientific) 61 (26.9)
Spider RX (ev3) 30 (13.2)
EmboShield (Abbott) 24 (10.6)
Angioguard RX (Cordis) 21 (9.3)
Spider (ev3) 19 (8.4)
Accunet (Abbott) 5 (2.2)
Medical treatment — no. of patients (%)
Preprocedure†
Dual antiplatelet therapy 204 (82.9)
Single antiplatelet therapy 42 (17.1)

During procedure
Heparin 241 (97.6)
Atropine 192 (77.7)
Postprocedure
Dual antiplatelet therapy 211 (85.4)
Single antiplatelet therapy 36 (14.6)
* Plus–minus values are means ±SD. Percentages may not total 100 because of rounding.
† Data are missing for one patient.
‡ Data are missing for 14 patients.
§ Data are missing for two patients.
¶ Data are missing for three patients.
∥ Anticoagulant therapy consisted of low-molecular-weight heparins at prophylactic doses for a few days.
** The stent could not be implanted in one patient.
†† Among the 247 patients who completed stenting, 74 underwent stenting before the recommendation was given and
173 underwent stenting afterward.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
The new england journal of medicine
n engl j med 355;16 www.nejm.org october 19, 2006
1668
Table 4
lists the incidence of primary outcome
events at 6 months. The three composite outcomes
were significantly more common after stenting
than after endarterectomy.
Discussion
This trial was stopped early for reasons of both
safety and futility. The 30-day risk of any stroke
or death was significantly higher after stenting
(9.6%) than after endarterectomy (3.9%), resulting

in a relative risk of 2.5 (95% CI, 1.2 to 5.1). Al-
though early stopping of randomized clinical tri-
als carries a risk of the overestimation of treat-
ment effects (i.e., analyzing the data at a “random
high”),
23
the excess of primary outcome events af-
ter stenting was considered large enough (one ad-
ditional stroke or death among each 17 patients
treated by stenting) for the safety committee to rec-
ommend stopping the trial. In addition, the ob-
served rates of the primary outcome made it very
unlikely that the trial would show the noninferior-
ity of stenting.
The 30-day incidence of stroke or death after
endarterectomy was lower in our trial than in pre-
vious trials of endarterectomy in symptomatic pa-
tients.
1,2
The lower surgical risk in our study is
unlikely to be explained by the selection of sur-
geons with a very high level of expertise. Indeed,
the surgeons worked in academic and nonaca-
demic centers in various areas of France and had
only to have performed 25 endarterectomies in
the year before enrollment; there was no upper
limit for perioperative stroke and death. The base-
line characteristics of our patients were similar
to those included in other trials of endarterec-
tomy,

2,5
which makes it unlikely that our find-
ings are explained by the inclusion of patients at
low risk for perioperative stroke or death. More-
over, to prevent the underreporting of minor
strokes in patients who underwent surgery un-
der general anesthesia and then were returned to
surgical wards, all patients were examined 2 days
after the procedure. Therefore, the most likely
explanation for the low rate of complications
from endarterectomy in our trial is that the risks
of this procedure have decreased since the piv-
otal trials
1,2
were conducted.
The combination of results of previous trials
4

yielded a 30-day incidence of stroke or death after
endovascular repair of the carotid artery of 8.1%
(51 of 632 patients; range, 0.0 to 12.1%). There
was significant heterogeneity among these trials,
which may have resulted from the use of differ-
Table 3. Risk of Stroke or Death and Other Treatment-Related Outcomes within 30 Days after Endarterectomy
or Stenting.*
Outcome Event
Endarterectomy
(N = 259)
Stenting
(N = 261)

Unadjusted
Relative Risk
(95% CI) P Value
no. of patients (%)
Nonfatal stroke 7 (2.7)† 23 (8.8)‡ 3.3 (1.4–7.5) 0.004
Symptoms lasting 7 days or more 6 (2.3) 20 (7.7)
Nondisabling 6 (2.3) 16 (6.1)
Disabling§ 1 (0.4) 7 (2.7)
Death 3 (1.2) 2 (0.8) 0.7 (0.1–3.9) 0.68
Fatal stroke 2 (0.8)† 1 (0.4)‡
Other cause 1 (0.4)¶ 1 (0.4)∥
Any stroke or death 10 (3.9) 25 (9.6) 2.5 (1.2–5.1) 0.01
Any disabling stroke or death 4 (1.5) 9 (3.4) 2.2 (0.7–7.2) 0.26
Transient ischemic attack 2 (0.8) 6 (2.3) 3.0 (0.6–14.6) 0.28
Myocardial infarction** 2 (0.8) 1 (0.4) 0.5 (0.04–5.4) 0.62
Bradycardia or hypotension†† 0 11 (4.2) Not computable <0.001
Systemic complications 8 (3.1)‡‡ 5 (1.9)§§ 0.6 (0.2–1.9) 0.42
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
endarterectomy versus stenting in severe carotid stenosis
n engl j med 355;16 www.nejm.org october 19, 2006
1669
Table 3. (Continued.)
Outcome Event
Endarterectomy
(N = 259)
Stenting
(N = 261)
Unadjusted
Relative Risk

(95% CI) P Value
no. of patients (%)
Major local complications 3 (1.2) 8 (3.1)¶¶ 2.6 (0.7–9.9) 0.22
Cervical or groin hematoma∥∥ 2 (0.8) 1 (0.4)
Infection*** 1 (0.4) 1 (0.4)
Femoral pseudoaneurysm or arteriove-
nous fistula at puncture site†††
— 4 (1.5)
Lower-limb arterial occlusion
or thrombosis‡‡‡
— 4 (1.5)
Cranial-nerve injury 20 (7.7)§§§ 3 (1.1)¶¶¶ 0.15 (0.04–0.49) <0.001
* Proportions were compared with the use of Fisher’s exact test. Relative risks were calculated with endarterectomy as
the reference group.
† Among patients who underwent endarterectomy, stroke was caused by cerebral infarction in six patients (including
one who had a disabling nonfatal stroke and none who had a fatal stroke) and cerebral hemorrhage in three (includ-
ing two who had a fatal stroke). All but one of the strokes were ipsilateral to the treated artery. Of the nine strokes,
three occurred on the day of the procedure. Cerebral hemorrhage occurred 1 hour after the procedure in one patient
and the day after in the two other patients. At the time of cerebral hemorrhage, the first patient had received intrave-
nous heparin during the procedure (0.5 mg per kilogram of body weight) and the two other patients were receiving
prophylactic doses of low-molecular-weight heparin.
‡ Among patients who underwent stenting, stroke was caused by cerebral infarction in 21 patients (including 5 who
had disabling nonfatal strokes and 1 who had a fatal stroke) and cerebral hemorrhage in 3 (2 who had disabling
nonfatal strokes and none who had a fatal stroke). All but two of the strokes were ipsilateral to the treated artery. Of
the 24 strokes, 17 occurred on the day of the procedure. Cerebral hemorrhage occurred 24 hours, 7 days, or 10 days
after the procedure. At the time of cerebral hemorrhage, the three patients were receiving dual antiplatelet therapy.
§ Stroke was defined as disabling if the modified Rankin score (on a scale of 0 to 5, with higher scores indicating
more severe disability) was 3 or more for at least 30 days after the event, with an increase of 2 points or more over
the prestroke score.
¶ This patient committed suicide 17 days after endarterectomy.

∥ This patient died suddenly 30 days after stenting.
** Myocardial infarction was defined by at least two of the following criteria: typical chest pain lasting 20 minutes or
more; serum levels of creatine kinase, creatine kinase MB, or troponin at least twice the upper limit of the normal
range; and new Q wave on at least two adjacent derivations or predominant R waves in V
1
(R wave ≥1 mm >S wave
in V
1
).
†† Bradycardia or hypotension was listed if it required treatment or prolonged monitoring.
‡‡ Systemic complications in the endarterectomy group were infection (mainly pulmonary) in five patients, unstable
angina in one, gastrointestinal bleeding in one, and subdural hematoma in one. Six of the eight events were associ-
ated with a delay in discharge.
§§ Systemic complications in the stenting group were infection in two patients, pacemaker implantation in one, throm-
bocytopenia in one, and venous thrombosis in one. Four of these five events were associated with a delay in dis-
charge.
¶¶ Two patients had two major local complications each.
∥∥ Hematoma was listed if it required surgery or blood transfusion.
*** Infection was listed if it required surgery or parenteral antibiotic therapy.
††† Femoral pseudoaneurysm or arteriovenous fistula was listed if it required surgery.
‡‡‡ Occlusion or thrombosis was listed if it required percutaneous or surgical treatment.
§§§ Nerve injury in the endarterectomy group was hypoglossal-nerve palsy in 10 patients, palsy of the marginal mandib-
ular branch of the facial nerve in 7, recurrent laryngeal-nerve palsy in 2, and glossopharyngeal-nerve palsy in 1. At
the 30-day follow-up visit, two of the cranial-nerve injuries (hypoglossal-nerve palsy in one patient and recurrent la-
ryngeal-nerve palsy in one patient) were categorized as severe, one of them leading to delayed discharge.
¶¶¶ Nerve injury in the stenting group was hypoglossal-nerve palsy in two patients and Horner’s syndrome in one pa-
tient. The patient with Horner’s syndrome had carotid dissection during angioplasty. The other two patients had
aborted angioplasty with conversion to surgery. At the 30-day follow-up visit, no cranial-nerve injury was categorized
as severe.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.

Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
The new england journal of medicine
n engl j med 355;16 www.nejm.org october 19, 2006
1670
ent endovascular techniques or different criteria for
patient selection. The 30-day incidence of stroke
after stenting in our study (9.2%) was higher than
that in the Stenting and Angioplasty with Pro-
tection in Patients at High Risk for Endarterec-
tomy (SAPPHIRE) trial
10
(3.6%), despite the use
of similar endovascular techniques. However, most
patients (70%) included in the SAPPHIRE trial had
asymptomatic stenosis, which carries a lower risk
of stroke during carotid repair than does symp-
tomatic stenosis.
20,24
Patients in the Carotid and
Vertebral Artery Transluminal Angioplasty Study
(CAVATAS)
5
were similar to those in our trial, but
the majority (77%) underwent carotid angioplasty
without stenting, and procedures were not per-
formed with the use of cerebral protection devices.
A potential bias in the comparison of a rela-
tively new procedure such as stenting with an es-
tablished procedure such as endarterectomy is the
effect of the learning curve. Our trial involved cen-

ters with staff members who had various degrees
of experience in carotid stenting, including centers
in which investigators treated enrolled patients
under the supervision of a tutor. We tried to limit
the effect of the learning curve through the care-
ful training and supervision of interventional phy-
sicians. We did not find any significant differ-
ences in outcome related to the number of stenting
procedures performed in individual centers or to
the experience of the interventional physicians,
although these analyses were able to detect only
large differences. There may also be a learning
curve related to changes in technique. Centers in
our trial were not required to use a device from
a particular manufacturer for stenting or cerebral
protection, but experience with any new device was
required before its use in the trial.
Cerebral-protection devices have been devel-
oped to reduce embolization of plaque fragments
during stenting. Uncontrolled studies
11,20,21
sug-
gest that these devices may reduce the risk of
procedural stroke. However, one could argue that
protection devices may cause additional adverse
events in some patients and increase costs.
In summary, our results indicate that in pa-
tients with symptomatic carotid stenosis of 60%
or more, treatment with endarterectomy results
in lower rates of stroke or death at 30 days and

6 months than does stenting. Long-term follow-
up is ongoing to determine whether the advan-
tage of endarterectomy is sustained. A larger
number of patients are required to provide defi-
nite answers about the risk–benefit profile of
stenting, as compared with endarterectomy, and
to permit meaningful subgroup analyses.
Supported by a grant from the Programme Hospitalier de
Recherche Clinique of the French Ministry of Health (AOM
97066), Assistance Publique–Hôpitaux de Paris.
Dr. Beyssen reports having received lecture fees from ev3 and
Guidant; and Dr. Becquemin, lecture fees from Cordis, Guidant,
and Cook. No other potential conflict of interest relevant to this
article was reported.
We are indebted to Véronique Favret, Ouafia Lakat, and Chris-
tine Mandet for their outstanding efforts in data management;
to Ludovic Trinquart for statistical help; and to Joël Ménard and
Nicolas Best for their constant support.
Table 4. Incidence of Primary Outcome Events at 6 Months.
Event
Endarterectomy Group
N = 262
Stenting
Group
N = 265 P Value*
no. of patients (%)
Any stroke or death at 30 days† plus ipsilateral stroke
between 31 days and 6 mo
11 (4.2) 27 (10.2) 0.008
Any stroke or death at 30 days† plus any stroke between

31 days and 6 mo
12 (4.6) 29 (10.9) 0.007
Any stroke or death within 6 mo† 16 (6.1) 31 (11.7) 0.02
* P values were obtained with the use of the log-rank test.
† Any stroke or death included a stroke in one patient between randomization and planned endarterectomy (which was can-
celed) and strokes in two patients between randomization and planned stenting (which was canceled in one patient).
appendix
The following investigators (with the number of patients randomly assigned at each center given in parentheses) and committees par-
ticipated in the EVA-3S trial: Hôpital Purpan, Toulouse (52) — J F. Albucher, F. Chollet, H. Rousseau, C. Cognard, M. Degeilh, A.
Barret, J.P. Bossavy; Hôpital Rangueil, Toulouse (52) — A. Viguier, V. Larrue, H. Rousseau, P. Arrué, P. Tall, Y. Glock; Hôpital Sainte-
Marguerite, Marseille (47) — B. Denis, S. Cohen, F. Nicoli, J.M. Bartoli, P. Piquet; Hôpital Nord, Hôpital de Bellevue, Saint-Etienne (43)
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
endarterectomy versus stenting in severe carotid stenosis
n engl j med 355;16 www.nejm.org october 19, 2006
1671
— P. Garnier, C. Veyret, F.G. Barral, J.P. Favre, X. Barral; Hôpital Côte de Nacre, Caen (40) — F. Viader, A. Duretête, L. Carluer, J.
Théron, P. Courthéoux, O. Coffin, D. Maïza; Hôpital Sainte-Anne, Hôpital Cochin, Hôpital Georges Pompidou, Paris (29) — E. Touzé,
C. Arquizan, C. Lamy, D. Calvet, V. Domigo, B. Beyssen, J.F. Méder, D. Trystram, P.O. Sarfati, P. Julia, J.N. Fabiani; Hôpital Général,
Hôpital du Bocage, Dijon (28) — M. Giroud, G.V. Osseby, O. Rouaud, I. Benatru, D. Krause, J.P. Cercueil, R. Brenot, M. David; Hôpital
Henri Mondor, Créteil (26) — H. Hosseini, H. Kobeiter, J P. Becquemin, P. Desgranges; Nouvelles Cliniques Nantaises, Nantes (21)
— G. Hinzelin, A. Bouyssou, J C. Pillet; Hôpital Lariboisière, Paris (20) — P. Favrole, K. Berthet, C. Gobron, M.G. Bousser, R. Chapot,
E. Houdart, C. Petitjean; Hôpital Roger Salengro, Lille (20) — C. Lucas, H. Hénon, C. Lefebvre, D. Leys, M.A. Mackowiak-Cordoliani, X.
Leclerc, J P. Pruvo, M. Koussa; Hôpital La Milétrie, Poitiers (17) — J.P. Neau, G. Godenèche, H. Moumy, J. Drouineau, J.B. Ricco;
Hôpital Central, Nancy, Hôpital Brabois, Vandoeuvre les Nancy (15) — X. Ducrocq, J.C. Lacour, S. Bracard, C. Amicabile, O. Hassani, G.
Fiévé; Hôpital Charles Nicolle, Rouen (12) — Y. Onnient, B. Mihout, E. Clavier, J. Thiebot, J. Watelet, D. Plissonnier; Clinique Pasteur,
Toulouse (11) — J.R. Rouane, J.C. Laborde, B. Escude, F. Berthoumieu; Fondation Hôpital Saint-Joseph, Marseille (12) — R. Padovani,
O. Bayle, P. Bergeron, J.M. Jausseran; Hôpital La Timone, Marseille (10) — L. Milandre, J.M. Bartoli, G. Moulin, A. Branchereau, P.E.
Magnan; Hôpital Pellegrin Tripode, Bordeaux (10) — F. Rouanet, J. Berge, X. Barreau, D. Midy, J.C. Baste; Hôpital Privé Beauregard,
Marseille (10) — H. Guinot, P. Commeau, F. Houel; Hôpital Civil, Strasbourg (10) — V. Wolff, J.M. Warter, R. Beaujeux, C. Jahn, J.G.

Kretz; Hôpital Bretonneau, Tours (9) — D. Saudeau, I. Bonnaud, D. Herbreteau, P. Lermusiaux, R. Martinez; Polyclinique, Essey-les-
Nancy (8) — I. Masson, M. Amor, J.P. Carpena, C. Amicabile; Hôpital Saint-Roch et Hôpital Pasteur, Nice (6) — M.H. Mahagne, J.
Baque, J. Sedat, M. Dib, R. Hassen-Khodja, M. Batt; Hôpital Saint-Jean, Perpignan (5) — D. Sablot, J.L. Bertrand, M. Beaufigeau, G.A.
Pelouze; Hôpital Bichat-Claude Bernard, Paris (4) — P. Amarenco, O. Simon, E. Meseguer, P. Lavallée, H. Abboud, E. Houdart, M.
Mazighi, G. Lesèche; Polyclinique du Bois, Lille (3) — M. Combelles, V. Courteville, G. Gozet, C. Depriester, I. Lambert, J. Pommier;
Hôpital E. Muller, Mulhouse (3) — G. Rodier, D. Weisse, J. Aventin, G. Dalcher; Clinique du Belvédère, Nice (2) — P. Marcel, P. Maillet,
J.M. Gagliardi; Hôpital Jean Minjoz, Besançon (1) — T. Moulin, J F. Bonneville, J.Y. Huart; Fondation Saint-Joseph, Paris (1) — C.
Gauthier, J.M. Pernes, C. Laurian; Scientific Committee — J L. Mas (chair), G. Chatellier (cochair), J P. Becquemin, J F. Bonneville,
A. Branchereau, D. Crochet, J.C. Gaux, V. Larrue, D. Leys, J. Watelet; Events Committee — T. Moulin (chair), S. Bracard, M. Hommel,
J.L. Magne, F. Mounier-Vehier, S. Weber; Accreditation Committee — B. Beyssen (chair), J F. Bonneville, L. Boyer, J.P. Favre, M. Gir-
oud, K. Hassen-Kodja, J.B. Ricco; Imaging Committee — J P. Pruvo (chair), J.F. Meder (cochair), C. Arquizan, F. Becker, F. Cattin, J.M.
Debray, J.M. Jausseran, A. Long, O. Naggara, P.J. Touboul; Safety Committee — M. Lièvre (chair), J.P. Beregi, J. Bogousslavsky, M. Testart.
References
European Carotid Surgery Trialists’
Collaborative Group. Randomised trial of
endarterectomy for recently symptomatic
carotid stenosis: final results of the MRC
European Carotid Surgery Trial (ECST).
Lancet 1998;351:1379-87.
Barnett HJ, Taylor DW, Eliasziw M, et
al. Benefit of carotid endarterectomy in
patients with symptomatic moderate or
severe stenosis. N Engl J Med 1998;339:
1415-25.
Rothwell PM, Eliasziw M, Gutnikov
SA, et al. Analysis of pooled data from the
randomised controlled trials of endarter-
ectomy for symptomatic carotid stenosis.
Lancet 2003;361:107-16.
Coward LJ, Featherstone RL, Brown

MM. Safety and efficacy of endovascular
treatment of carotid artery stenosis com-
pared with carotid endarterectomy: a Coch-
rane systematic review of the randomized
evidence. Stroke 2005;36:905-11.
Endovascular versus surgical treat-
ment in patients with carotid stenosis in
the Carotid and Vertebral Artery Translu-
minal Angioplasty Study (CAVATAS): a ran-
domised trial. Lancet 2001;357:1729-37.
Brooks WH, McClure RR, Jones MR,
Coleman TC, Breathitt L. Carotid angio-
plasty and stenting versus carotid endar-
terectomy: randomized trial in a commu-
nity hospital. J Am Coll Cardiol 2001;38:
1589-95.
Brooks WH, McClure RR, Jones MR,
Coleman TL, Breathitt L. Carotid angio-
plasty and stenting versus carotid endar-
terectomy for treatment of asymptomatic
carotid stenosis: a randomized trial in a
community hospital. Neurosurgery 2004;
54:318-24.
1.
2.
3.
4.
5.
6.
7.

Naylor AR, Bolia A, Abbott RJ, et al.
Randomized study of carotid angioplasty
and stenting versus carotid endarterecto-
my: a stopped trial. J Vasc Surg 1998;28:
326-34.
Alberts MJ. Results of a multicenter
prospective randomized trial of carotid
artery stenting versus carotid endarterec-
tomy. Stroke 2001;32:325. abstract.
Yadav JS, Wholey MH, Kuntz RE, et al.
Protected carotid-artery stenting versus
endarterectomy in high-risk patients.
N Engl J Med 2004;351:1493-501.
EVA-3S Investigators. Endarterectomy
vs. Angioplasty in Patients with Symp-
tomatic Severe Carotid Stenosis (EVA-3S)
trial. Cerebrovasc Dis 2004;18:62-5.
Featherstone RL, Brown MM, Coward
LJ. International Carotid Stenting Study:
protocol for a randomised clinical trial
comparing carotid stenting with endar-
terectomy in symptomatic carotid artery
stenosis. Cerebrovasc Dis 2004;18:69-74.
Ringleb PA, Kunze A, Allenberg JR, et
al. The Stent-Supported Percutaneous An-
gioplasty of the Carotid Artery vs. Endar-
terectomy Trial. Cerebrovasc Dis 2004;18:
66-8.
Hobson RW. CREST (Carotid Revas-
cularization Endarterectomy versus Stent

Trial): background, design, and current
status. Semin Vasc Surg 2000;13:139-43.
North American Symptomatic Carotid
Endarterectomy Trial: methods, patient
characteristics, and progress. Stroke 1991;
22:711-20.
van Swieten JC, Koudstaal PJ, Visser
MC, Schouten HJ, van Gijn J. Interobserver
agreement for the assessment of handicap
in stroke patients. Stroke 1988;19:604-7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
Mas JL, Chatellier G, Beyssen B. Ca-
rotid angioplasty and stenting with and
without cerebral protection: clinical alert
from the Endarterectomy Versus Angio-
plasty in Patients with Symptomatic Se-
vere Carotid Stenosis (EVA-3S) trial. Stroke
2004;35:e18-e20.
Blackwelder WC. “Proving the null
hypothesis” in clinical trials. Control Clin
Trials 1982;3:345-53.
Rothwell PM, Slattery J, Warlow CP.

A systematic review of the risks of stroke
and death due to endarterectomy for symp-
tomatic carotid stenosis. Stroke 1996;27:
260-5.
Wholey MH, Al-Mubarek N, Wholey
MH. Updated review of the global carotid
artery stent registry. Catheter Cardiovasc
Interv 2003;60:259-66.
Kastrup A, Groschel K, Krapf H,
Brehm BR, Dichgans J, Schulz JB. Early
outcome of carotid angioplasty and stent-
ing with and without cerebral protection
devices: a systematic review of the litera-
ture. Stroke 2003;34:813-9.
North American Symptomatic Carotid
Endarterectomy Trial Collaborators. Ben-
eficial effect of carotid endarterectomy in
symptomatic patients with high-grade ca-
rotid stenosis. N Engl J Med 1991;325:445-
53.
Pocock SJ. When (not) to stop a clini-
cal trial for benefit. JAMA 2005;294:2228-
30.
Bond R, Rerkasem K, Rot hwell PM. Sys-
tematic review of the risks of carotid endar-
terectomy in relation to the clinical indica-
tion for and timing of surgery. Stroke
2003;34:2290-301.
Copyright © 2006 Massachusetts Medical Society.
17.

18.
19.
20.
21.
22.
23.
24.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
original article
The
new england journal of medicine
n engl j med 355;10 www.nejm.org september 7, 2006
1006
Enoxaparin versus Unfractionated Heparin
in Elective Percutaneous Coronary Intervention
Gilles Montalescot, M.D., Ph.D., Harvey D. White, M.B., Ch.B., D.Sc.,
Richard Gallo, M.D., Marc Cohen, M.D., P. Gabriel Steg, M.D.,
Philip E.G. Aylward, M.B., Ch.B., Ph.D., Christoph Bode, M.D., Ph.D.,
Massimo Chiariello, M.D., Spencer B. King III, M.D., Robert A. Harrington, M.D.,
Walter J. Desmet, M.D., Carlos Macaya, M.D., Ph.D.,
and Steven R. Steinhubl, M.D., for the STEEPLE Investigators*
From Institut de Cardiologie, Centre
Hospitalier Universitaire Pitié–Sal pêtrière,
Paris (G.M.); the Green Lane Cardiovascular
Service, Auckland City Hospital, Auckland,
New Zealand (H.D.W.); the Montreal
Heart Institute, Université de Montréal,
Montreal (R.G.); the Division of Cardiology,
Newark Beth Israel Medical Center,

Newark, NJ (M.C.); the Service de Car-
diologie, Hôpital Bichat, Paris (P.G.S.); De-
partment of Cardiology, Flinders Medical
Center, Adelaide, SA, Australia (P.E.G.A.);
Abteilung Innere Medizin III, Universi-
tätsklinikum Freiburg, Freiburg, Germany
(C.B.); the Division of Cardiology, Federico
2nd Uni versity, Naples, Italy (M.C.); Fuqua
Heart Center of Atlanta at Piedmont
Hospital, Atlanta (S.B.K.); the Division of
Cardiology, Duke University Medical Cen-
ter, Durham, NC (R.A.H.); University Hos-
pital Gasthuisberg, Leuven, Belgium
(W.J.D.); Servicio de Cardiología, Hospital
Universitario, Madrid (C.M.); and the
Division of Cardiology, University of Ken-
tucky, Lexington (S.R.S.). Address reprint
requests to Dr. Montalescot at Institut de
Cardiologie, Bureau 2-236, Centre Hospital-
ier Universitaire Pitié–Salpêtrière, 47 Boule-
vard de l’Hôpital, 75013 Paris, France, or at

*Participants in the Safety and Efficacy of
Enoxaparin in Percutaneous Coronary
Intervention Patients, an International
Randomized Evaluation (STEEPLE) trial
are listed in the Appendix.
N Engl J Med 2006;355:1006-17.
Copyright © 2006 Massachusetts Medical Society.
Abstract

Background
Despite its limitations, unfractionated heparin has been the standard anticoagulant
used during percutaneous coronary intervention (PCI). Several small studies have
suggested that intravenous enoxaparin may be a safe and effective alternative. Our
primary aim was to assess the safety of enoxaparin as compared with that of un-
fractionated heparin in elective PCI.
Methods
In this prospective, open-label, multicenter, randomized trial, we randomly assigned
3528 patients with PCI to receive enoxaparin (0.5 or 0.75 mg per kilogram of body
weight) or unfractionated heparin adjusted for activated clotting time, stratified ac-
cording to the use or nonuse of glycoprotein IIb/IIIa inhibitors. The primary end point
was the incidence of major or minor bleeding that was not related to coronary-artery
bypass grafting. The main secondary end point was the percentage of patients in
whom the target anticoagulation levels were reached.
Results
Enoxaparin at a dose of 0.5 mg per kilogram was associated with a significant reduc-
tion in the rate of non–CABG-related bleeding in the f irst 48 hours, as compared wit h
unfractionated heparin (5.9% vs. 8.5%; absolute difference, –2.6; 95% confidence inter-
val [CI], –4.7 to –0.6; P = 0.01), but the higher enoxaparin dose was not (6.5% vs. 8.5%;
absolute difference, –2.0; 95% CI, –4.0 to 0.0; P = 0.051). The incidence of major bleed-
ing was significantly reduced in both enoxaparin groups, as compared with the un-
fractionated heparin group. Target anticoagulation levels were reached in significantly
more patients who received enoxaparin (0.5-mg-per-kilogram dose, 79%; 0.75-mg-
per-kilogram dose, 92%) than who received unfractionated heparin (20%, P<0.001).
Conclusions
In elective PCI, a single intravenous bolus of 0.5 mg of enoxaparin per kilogram is
associated with reduced rates of bleeding, and a dose of 0.75 mg per kilogram
yields rates similar to those for unfractionated heparin, with more predictable an-
ticoagulation levels. The trial was not large enough to provide a definitive com-
parison of efficacy in the prevention of ischemic events. (ClinicalTrials.gov num-

ber, NCT00077844.)
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
enoxaparin in elective percutaneous coronary intervention
n engl j med 355;10 www.nejm.org september 7, 2006
1007
T
he american college of cardiology,
the American Heart Association, and the
European Society of Cardiology recommend
the use of intravenous unfractionated heparin,
with the dose adjusted for the activated clotting
time, during percutaneous coronary intervention
(PCI).
1,2
However, better anticoagulation regimens
are needed for PCI, given the limitations of unfrac-
tionated heparin, which include its sometimes dif-
ficult-to-manage effects on coagulation, the need
for repeated monitoring of coagulation, the nar-
row therapeutic window, the potential induction
of platelet activation, and the risk of thrombocy-
topenia.
3
The use of low-molecular-weight heparins as
anticoagulants is increasing in patients with acute
coronary syndrome who undergo PCI
4-6
and in
those undergoing elective procedures.

7,8
As com-
pared with unfractionated heparin, low-molecu-
lar-weight heparins are considered to induce a
more stable and predictable anticoagulant dose
response (obviating the necessity for coagulation
monitoring), and to have a longer half-life and a
greater ratio of anti–factor Xa activity to anti–fac-
tor IIa activity, which reduces the generation and
activation of thrombin.
3,9
Low-molecular-weight
heparins are also less apt to induce platelet acti-
vation, release of the von Willebrand factor, and
inflammation.
10-13
Small or noncomparative trials have evaluated a
single intravenous bolus of enoxaparin — 1 mg,
14-18

0.75 mg,
14,19-21
or 0.5 mg
7,2 2
per kilogram of body
weight — in patients undergoing PCI with or with-
out the administration of glycoprotein IIb/IIIa in-
hibitors. However, these uncontrolled studies have
not allowed definite conclusions to be drawn about
the efficacy of enoxaparin as compared with that

of standard anticoagulation regimens involving
unfractionated heparin. In a meta-analysis of data
from randomized studies comparing intravenous
low-molecular-weight heparins and intravenous
unfractionated heparin in patients undergoing
PCI, there was a nonsignificant trend toward a
reduction in major bleeding with low-molecular-
weight heparins and no difference between groups
in the occurrence of ischemic events.
8
In an ad-
ditional analysis, a dose of less than 1 mg of
enoxaparin per kilogram resulted in fewer ische-
mic and bleeding events than a dose of 1 mg per
kilogram.
We conducted a large-scale, randomized, con-
trolled trial to evaluate whether the safety of in-
travenous low-molecular-weight heparins was su-
perior to that of unfractionated heparin in patients
undergoing elective PCI.
Methods
The Safety and Efficacy of Enoxaparin in PCI Pa-
tients, an International Randomized Evaluation
(STEEPLE) trial was a prospective, open-label, par-
allel-group trial evaluating intravenous enoxapar-
in at a dose of 0.5 mg or 0.75 mg per kilogram,
as compared with intravenous unfractionated hep-
arin, in patients undergoing elective PCI.
The protocol was written by Dr. Montalescot
and modified on the basis of discussions with the

sponsor (Sanofi-Aventis) and the members of the
steering committee (see the Appendix). The data
were gathered by the sponsor and were maintained
and analyzed by Altizem, a contract research or-
ganization. The steering committee vouches for
the integrity and completeness of the data, and
the statistician vouches for the accuracy of the
data analysis. The publication committee pre-
pared the manuscript with suggestions from the
steering committee and the sponsor. The publica-
tion committee had final authority over the con-
tent of the manuscript.
Patients
Patients were enrolled at 124 sites in nine coun-
tries. Patients were eligible for the study if they
were older than 17 years of age, were scheduled
to undergo elective PCI with a femoral approach,
and did not meet any of the exclusion criteria: re-
cent thrombolysis, a planned staged procedure, an
increased risk of bleeding, treatment with a paren-
teral antithrombotic agent before PCI, or a known
hypersensitivity to the drugs used in the study.
The study was conducted according to the Decla-
ration of Helsinki and local regulations. Approv-
al for the trial was obtained from the institutional
review board at each site. All patients gave writ-
ten informed consent.
Study Protocol
Eligible patients were randomly assigned to receive
an intravenous bolus of unfractionated heparin,

adjusted for activated clotting time according to
current guidelines,
1
or intravenous enoxaparin at
a dose of 0.5 or 0.75 mg per kilogram. We assigned
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
The new england journal of medicine
n engl j med 355;10 www.nejm.org september 7, 2006
1008
patients using an interactive voice-response sys-
tem at the central randomization center. Random
permuted blocks were used to make assignments
in a 1:1:1 ratio that were stratified according to cen-
ter and planned use of glycoprotein IIb/IIIa inhibi-
tors (no or yes). All patients received aspirin (75 to
500 mg per day) and thienopyridines, according to
local practice.
Patients who were assigned to either enoxapar-
in group received a single intravenous bolus of
enoxaparin, without anticoagulation monitoring,
after sheath insertion and immediately before PCI.
The assigned dose was used regardless of whether
a patient was also receiving the glycoprotein IIb/
IIIa inhibitor. When procedures were prolonged by
more than 2 hours, an additional bolus of enoxa-
parin (half the original dose) was recommended.
23
Patients who were randomly assigned to receive
unfractionated heparin and who were not receiv-

ing concurrent glycoprotein IIb/IIIa inhibitors were
given an initial intravenous bolus of 70 to 100 IU
per kilogram to achieve a target activated clotting
time of 300 to 350 seconds. Patients who received
concurrent glycoprotein IIb/IIIa inhibitors were
given an initial bolus of 50 to 70 IU of unfraction-
ated heparin per kilogram to achieve a target ac-
tivated clotting time of 200 to 300 seconds. Ad-
ditional boluses of unfractionated heparin were
given before the PCI if the lower limit of the tar-
get activated clotting time was not reached. Un-
fractionated heparin was readministered during
the procedure, at the discretion of the investiga-
tor, when measurements of activated clotting time
dropped below the recommended range. In all
centers, activated clotting time was measured with
a standardized Hemochron device (ITC).
Arterial closure devices were permitted accord-
ing to the practice at each institution. Sheath re-
moval was authorized at an activated clotting time
between 150 and 180 seconds in the unfraction-
ated heparin group,
24
4 to 6 hours after the end
of the PCI in the group given 0.75 mg of enoxa-
parin per kilogram, and immediately after the end
of the PCI in the group given 0.5 mg of enoxa-
parin per kilogram. No monitoring of anticoagu-
lation was required before sheath removal in pa-
tients receiving enoxaparin.

End Points
The primary end point of the trial was the occur-
rence of major or minor bleeding not related to
coronary-artery bypass grafting (CABG) during the
first 48 hours after the index PCI, according to
prespecified definitions (
Table 1
). The main sec-
ondary efficacy end point was the achievement of
therapeutic anticoagulation at the beginning and
end of PCI. Specifically, we compared the propor-
tion of patients receiving enoxaparin in whom the
target anti–factor Xa levels of 0.5 to 1.8 IU per
milliliter (analyzed centrally) were achieved
25,26

with the proportion of patients receiving unfrac-
tionated heparin in whom the target activated clot-
ting time (200 to 300 seconds with glycoprotein
IIb/IIIa inhibitors or 300 to 350 seconds without)
1

was achieved, at the start and the end of PCI.
We also studied other secondary end points.
The first was a composite of non–CABG-related
major bleeding up to 48 hours after the index PCI,
death from any cause, nonfatal myocardial infarc-
tion (defined by a new Q wave in two or more leads
or a total creatine kinase level or creatine kinase
MB fraction that was ≥3 times the upper limit of

the normal range during hospitalization for the
index PCI or that was ≥2 times the upper limit
of the normal range after discharge), or urgent
target-vessel revascularization during the first 30
days after the index PCI. The second was a com-
posite of death from any cause or nonfatal myo-
cardial infarction during the first 30 days after
the index PCI, whichever occurred first. The third
was a composite of death from any cause, non-
fatal myocardial infarction, or urgent target-ves-
sel revascularization during the first 30 days after
the index PCI, whichever occurred first. The fourth
consisted of each of the individual end points dur-
ing the first 30 days after the index PCI. All events
were adjudicated by an independent clinical-events
committee whose members were unaware of the
treatment assignments.
Statistical Analysis
We initially estimated that we would need to en-
roll 2700 patients, given a 7% incidence of any type
of bleeding up to 48 hours after PCI in the un-
fractionated heparin group, a statistical power of
80% to detect a relative risk reduction of 47%
with enoxaparin, and a type I, two-sided error rate
of 2.5% for each comparison of the enoxaparin
groups with the unfractionated heparin group. At a
planned interim evaluation, the sample size was
reevaluated, and because the overall bleeding rate
was lower than the anticipated rate (3.5% vs. 4.8%),
a final enrollment goal of 3690 patients was set.

All analyses were performed according to the
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .
enoxaparin in elective percutaneous coronary intervention
n engl j med 355;10 www.nejm.org september 7, 2006
1009
intention-to-treat principle and included all ran-
domized patients analyzed according to the treat-
ment assigned. The primary end point was also
analyzed in the safety population, which included
all patients who received at least one dose of study
drug, analyzed according to the treatment actually
received. For all end-point analyses, each enoxa-
parin dose was compared with unfractionated
heparin. The Simes adjustment for multiplicity was
applied to ensure a global type I error of 0.05: if
both P values were 0.05 or less, both were con-
sidered to indicate statistical significance; if the
highest P value was greater than 0.05, the other
P value had to be 0.025 or less to be considered to
indicate statistical significance.
27
Analyses were
performed with SAS statistical software, version
8.2 (SAS Institute).
Logistic-regression analysis was used to com-
pare the incidence of the primary end point be-
tween the enoxaparin and unfractionated hepa-
rin groups, with adjustment for the use or nonuse
of glycoprotein IIb/IIIa inhibitors. If enoxaparin

was not found to be superior to unfractionated
heparin, we evaluated whether enoxaparin was
noninferior to unfractionated heparin with the use
of a two-sided, adjusted confidence interval (CI)
for the differences in event rates as planned; the
noninferiority margin was set at 30% of the ob-
served bleeding rates in the unfractionated hepa-
rin group.
For secondary end points, target anti–factor Xa
levels or activated clotting times were analyzed
with a logistic-regression model, with adjustment
for the use or nonuse of glycoprotein IIb/IIIa in-
hibitors. The objective for the composite quadru-
ple end point was to test for the noninferiority of
the enoxaparin doses with the use of a two-sided
CI of the difference in event rates; the noninferi-
ority margin was set at 39% of the observed rates
in the unfractionated heparin group. Time-to-
event analyses for the other secondary end points
up to day 30 were performed with Cox propor-
tional-hazards models.
An independent data-monitoring committee
followed the progress of the trial to ensure that
patient safety was not compromised. The commit-
tee met four times during the trial; the fourth in-
terim analysis, which included 3089 patients, in-
dicated that there were more deaths from any
cause in the group given 0.5 mg of enoxaparin per
kilogram (9 patients) than in either the unfrac-
tionated heparin group (3 patients, P = 0.1477) or

the group given 0.75 mg of enoxaparin per kilo-
gram (1 patient, P = 0.0265). On the basis of the
Pocock boundary of 0.0266, the committee rec-
ommended that randomization to the group giv-
en 0.5 mg of enoxaparin per kilogram be discon-
tinued. Enrollment was suspended in that group
on November 22, 2004, just before the end of the
enrollment period in December 2004. Because the
decision to stop enrollment in this group could
have altered decisions about the inclusion of a pa-
tient or the conduct of the study, the final analyses
were adjusted for whether a patient underwent ran-
domization before or after November 22, 2004.
Results
Characteristics of the patients
Between January 2004 and December 2004, 3528
patients were enrolled: 1070 were randomly as-
signed to receive 0.5 mg of enoxaparin per kilo-
Table 1. Definitions of Major and Minor Bleeding.
Major bleeding*
Fatal bleeding
Retroperitoneal, intracranial, or intraocular bleeding
Bleeding that causes hemodynamic compromise requiring specific treatment
Bleeding that requires intervention (surgical or endoscopic) or decompres-
sion of a closed space to stop or control the event
Clinically overt bleeding, requiring any transfusion of ≥1 unit of packed red
cells or whole blood
Clinically overt bleeding, causing a decrease in hemoglobin of ≥3 g/dl
(or, if hemoglobin level not available, a decrease in hematocrit of ≥10%)
Minor bleeding†

Gross hematuria not associated with trauma (e.g., from instrumentation)
Epistaxis that is prolonged, repeated, or requires plugging or intervention
Gastrointestinal hemorrhage
Hemoptysis
Subconjunctival hemorrhage
Hematoma >5 cm or leading to prolonged or new hospitalization
Clinically overt bleeding, causing a decrease in hemoglobin of 2 to 3 g/dl
Uncontrolled bleeding requiring protamine sulfate administration
* Major bleeding was defined as bleeding that met at least one of the criteria
listed.
† Minor bleeding was defined as bleeding that did not meet any of the criteria
for major bleeding and that met at least one of the criteria for minor bleeding.
Copyright © 2006 Massachusetts Medical Society. All rights reserved.
Downloaded from www.nejm.org at RIKSHOSPITALET HF on February 18, 2008 .