Short term and intermediate term comparison of endarterectomy versus stenting for carotid artery stenosis: systematic review and meta-analysis of randomised controlled clinical trialsBMJ 2010; 340 doi: https://doi.org/10.1136/bmj.c467 (Published 12 February 2010) Cite this as: BMJ 2010;340:c467
All rapid responses
Dr Touze and colleagues raise several issues with our meta-analysis
although their supplementary analysis agree with our overall conclusion
that the use of CAS was associated with a higher risk of procedural stroke
while there was no discernable difference in long term outcomes.
We disagree with their statement that using CAS as the reference
somehow changes the significance of the results since a significant
difference will remain significant and a non -significant result will
remain non significant irrespective of the reference chosen. Indeed, while
they cite meta-analysis that have used CEA as the reference, others have
used CAS as the reference(1,2) and there is no valid reason to support
such an arbitrary convention.
Touze and colleagues argue that we should not include certain trials and
only assess the outcomes reported in more recent trials. Further, they
argue for exclusion of SAPPHIRE trial since it included only high risk
patients.The aim of our analysis was not to focus on a very specific
subgroup of patients, whereby statistical power of the analysis and
generalizability of the findings would have been reduced. We aimed to base
our conclusion on the difference between carotid endarterectomy and
carotid stenting, independent of patient characteristics. The recently
presented and carefully designed CREST trial also includes asymptomatic
patients. We regard asymptomatic patients not as group that should be
excluded from studies but one that should receive increased attention. Our
analysis had only moderate heterogeneity and we used random effects model
that account for such heterogeneity. We also have presented a sensitivity
analysis in the paper to demonstrate that exclusion of asymptomatic
patients does not change the overall conclusion.
We agree with Touze’s assertion that one of the challenges for
interpreting a meta-analysis that evaluates a therapy in evolution is that
pooling of prior data may not be valid for guiding contemporary therapy.
This is especially true for a therapy like CAS where there has been
significant improvement in devices and pharmacotherapy and a better
understanding of procedural risk. We do not agree, however, that there is
no value in combining the studies in a meta-analysis. In fact, we, a
priori weighed this decision and performed a cumulative meta-analysis to
precisely address this issue.
We believe that meta-analysis in such rapidly changing fields should be
used to identify opportunities for improvement and defining the optimal
therapy and not to guide an individual patients care. Such decisions are
better guided by the availability of local resources, operator experience
and data that evaluate contemporary therapy. While Touzé and colleagues
argue against a learning curve, some of the studies they cite demonstrate
exactly the opposite. In ICSS, 2 sites had 5 major strokes (of the total
17 in the CAS arm in the entire trial) among 11 treated patients. While
there are many low volume operators that do not have many complications,
the rarity of complications in patients undergoing CAS makes such analysis
underpowered and problematic. The impact of operator experience has been
demonstrated with respect to many other operative procedures(3) and it is
unlikely that CAS would be an exception to the rule.
We agree with the critical view of Touzé et al. regarding the
importance of defining optimal endpoints. However, switching the endpoint
after 30 days to only ipsilateral stroke is problematic. Why should the
switch be done at 30 days and not immediately after the immediate
periprocedural period of hours or a few days? Such a change in convention
should certainly be discussed and implemented prospectively in the design
of new trials. It is not possible to change the definition of endpoints
retrospectively in a study-level meta-analysis. We clearly think that
stroke and death is a reasonable combined endpoint, short-term and long-
term. Subspecifying this endpoint would further reduce objectivity of the
Finally as the authors conclude “although the conclusions to be drawn
are similar” the results of our meta-analysis are extant and inline with
the findings of the CREST trial. This point is elegantly reinforced by the
meta-analysis provided by Bhatt and colleagues in their response to our
paper and their conclusion “The results of CREST are consistent with the
current meta-analysis showing improved outcomes with more recent CAS
studies. Potential reasons for improved outcomes in CAS compared with CEA
are enhanced operator skills and stricter operator requirements especially
in the CREST trial to perform the stenting procedure”
Pascal Meier, MD
Hitinder S Gurm, MBBS
Competing interests: No competing interests
Short term Outcomes: An Analysis after CREST
Archit Bhatt, MD1 and Jody Short, DO1
1. Michigan State University, Department of Neurology and Ophthalmology, Division of Stroke, East Lansing, MI.
Corresponding author Information:
Archit Bhatt MD, MPH, FACP,FAHA
Michigan State University,
Department of Neurology and Ophthalmology,
East Lansing, Michigan, 48824
We review the metanalysis(1), which included the recently published multicentric trial- International Carotid Stenting Study (ICSS)(2). The trial published safety and short-term results comparing carotid endarterectomy (CEA) and carotid stenting (CAS) in patients with symptomatic carotid stenosis. The meta-analysis, which included 11 trials, showed that overall periprocedural and short-term outcomes favored CEA, except for myocardial infarction (MI) and cranial neuropathies, which were more common in the CEA group.
The Carotid Revascularization Endarterectomy vs the Stenting trial (CREST trial),(3) which is the largest multicentric trial comparing CEA and CAS was reported in February 2010, at the international stroke conference after the publication of the ICSS trial and your metanalysis. The ICSS and CREST trials included MI in the primary periprocedural outcome less than 30 days.
The CREST trial:
The CREST trial was a multi-centric randomized control trial, which recruited 2502 patients of which 1262 were randomized to CAS and 1240 to CEA. Of the total, 1326 were symptomatic, and rests were asymptomatic. Initial results on the composite primary end point (any stroke, myocardial infarction, or death during periprocedural period or ipsilateral stroke on follow—up), stenting group had 7.2% rate of events vs 6.8% in the CEA group (p = .51). At 30 days, the rate of stroke was significantly higher with stenting group (4.1% vs 2.3%, p = .01). Conversely, MI was higher with the CEA group (2.3% vs. 1.1%, p = .03). Primary endpoint, which included stroke, death or MI within 30 days, was not significantly different (CAS 5.2% and CEA 4.5%) in both groups (P = .38). Symptomatic and asymptomatic patients were not reported separately.
Methods and Results:
All meta-analyses were conducted using CMA (Comprehensive Meta-analysis software). Statistical significance was defined as P<_0.05. heterogeneity="heterogeneity" across="across" studies="studies" was="was" assessed="assessed" using="using" the="the" mantel="mantel" haenzel="haenzel" q="q" statistic="statistic" and="and" i2="i2" statisticthe="statisticthe" latter="latter" being="being" proportion="proportion" of="of" between="between" study="study" variation="variation" not="not" attributable="attributable" to="to" chance="chance" range="range" _0="_0" _100.="_100." we="we" conducted="conducted" a="a" meta-analysis="meta-analysis" _1="_1" _11="_11" trials="trials" reported="reported" in="in" current="current" metanalysis="metanalysis" crest="crest" trial="trial" results="results" where="where" data="data" available="available" compared="compared" summary="summary" estimates="estimates" with="with" present="present" table="table" _1.="_1." effect="effect" size="size" robustly="robustly" showed="showed" that="that" cea="cea" better="better" than="than" cas="cas" preventing="preventing" stroke="stroke" death="death" or="or" mi="mi" myocardial="myocardial" infarction="infarction" within="within" _30="_30" days="days" after="after" procedure="procedure" during="during" peri-procedural="peri-procedural" period.="period." whereas="whereas" had="had" lower="lower" incidence="incidence" cranial="cranial" nerve="nerve" injuries.="injuries." intermediate="intermediate" term="term" risks="risks" were="were" comparable="comparable" cas.="cas." p="p"/>
The CREST trial showed no significant difference in safety or efficacy between CAS and CEA. After including CREST data in our meta-analysis (figure 1 thru 5), the risk of stroke or death within 30 days and stroke within 30 days was still greater in the CAS group. Intermediate term risk for composite stroke or death, myocardial infarction within 30 days and cranial neuropathies did not differ from meta-analysis1 when CREST was included in our analysis with more myocardial infarctions and cranial neuropathies occurring the CEA group (Table 1).
The results of CREST are consistent with the current meta-analysis (1) showing improved outcomes with more recent CAS studies. Potential reasons for improved outcomes in CAS compared with CEA are enhanced operator skills and stricter operator requirements especially in the CREST trial to perform the stenting procedure. However long term efficacy data from the ICSS trial and subgroup analysis of symptomatic and asymptomatic patients from the CREST trial is still being awaited, which will give us more insights.
1. Meier P, Knapp G, Tamhane U, Chaturvedi S, Gurm HS. Short term and intermediate term comparison of endarterectomy versus stenting for carotid artery stenosis: systematic review and meta-analysis of randomised controlled clinical trials. BMJ;340:c467.
2. Carotid artery stenting compared with endarterectomy in patients with symptomatic carotid stenosis (International Carotid Stenting Study): an interim analysis of a randomised controlled trial. Lancet.epub Feb 25, 2010
3. Brott et. al Be. "The randomized carotid revascularization
endarterectomy vs stenting trial (crest): Primary results" Asa 2010;
abstract a29. Stroke. 2010:1
Competing interests: No competing interests
Short term and intermediate term comparison of endarterectomy versus stenting for carotid artery stenosis: What is the best end point ?
To the Editor
In their meta-analysis of randomised clinical trials, Meier et al
concluded that carotid endarterectomy (CEA) is superior to carotid artery
stenting (CAS) for short-term outcomes but that there is no significant
difference for intermediate term outcomes.1 However, we have major
concerns regarding the presentation of the data they use and their
interpretation of the results which throw doubt on the validity of their
It is confusing that Meier et al have calculated odds ratios and hazard
ratios using CAS as the denominator. When one is comparing two treatments
it is usual to select the standard treatment, in this case CEA, as the
control, i.e. to compare CAS with CEA using CEA as the denominator, which
is how all the previous trials and most meta-analyses have presented the
data. Meier et al have ignored the convention and calculated the ratios
the other way around without any justification. In this letter, we will
present ORs and HRs using CEA as the denominator.
Regarding the 30-day outcomes, a previous meta-analysis has shown that the
risk of stroke or death after CAS is about two times higher in symptomatic
patients than in asymptomatic patients.2 We therefore consider that Meier
et al's approach of combining together the results from treatment of
symptomatic and asymptomatic stenosis to calculate a weighted rate of
periprocedural stroke or death, is not appropriate or helpful in guiding
clinical practice, now that there are several large trials in which only
symptomatic patients were included. Moreover, in combining the results of
different trials together to calculate odds of stroke or death within 30
days, they have taken different outcome measures from different trials.
For example, the data from the 2001 CAVATAS paper which they use, included
only strokes that lasted longer than 7 days. Meier et al should instead
have used the more recent CAVATAS report, which included the incidence of
all 30-day strokes (CAS, 33/251; CEA, 26/253)3. Moreover, the ICSS results
they quoted and used for their calculations as the numbers of 30-day
procedural events, were in fact the numbers of events within 120 days of
randomisation, as clearly stated on the webpage to which they refer. When
considering symptomatic patients only, and using the most recent CAVATAS
data, together with the 30-day ICSS data4 (instead of the 120-day data
used in the Meier et al. paper), the combined risk of stroke or death is
8.8% (95% confidence interval 5.9-12.3) for CAS and 5.4% 3.6-7.6) for CEA,
corresponding to a combined odds ratio (OR) of 1.72 (1.20-2.50, p=0.003)
in favour of CEA, using the correct denominator of CEA with CAS as the
numerator. When considering all studies, as in Meier et al. paper, the
correct combined OR is 1.61 (1.15-2.27, p=0.006).
In fact, it is more clinically meaningful not to combine the older studies
with the newer studies of carotid stenting. The early trials used outdated
technology and indeed CAVATAS, which is listed in Meier et al's table of
study characteristics as being a trial of stenting, in fact was a trial of
carotid angioplasty and included very few patients in whom a stent was
deployed. It is also more appropriate not to include SAPPHIRE in the meta-
analysis because this trial included patients deliberately selected to be
at high risk from endarterectomy, whereas the other trials selected
patients considered to be of standard risk. In the meta-analyses of 30-day
risk of CAS versus CEA published in the Lancet recently, all the early
trials, including SAPPHIRE were excluded, leaving three trials (EVA-3S,
SPACE and ICSS) with similar inclusion criteria and treatment strategies.6
The results of the later meta-analysis, an odds ratio of 1.73 (1.29-2.32)
in favour of CEA for the 30-day outcome of stroke, death or myocardial
infarction, are therefore much more robust than the meta-analysis
published by the BMJ, although the conclusions to be drawn are similar.
Meier et al erroneously considered a proven fact that the learning curve
has been a major drawback for some of the largest trials where a
significant proportion of operators had limited experience. They mentioned
that 39% of patients in EVA-3S were treated by physicians in procedural
training, but failed to point out that the 30-day risk of stroke or death
was lower in these patients (7.1%) compared to those treated by
interventionalists who had done ¡Ü50 or >50 procedures (11.0% and
12.2% respectively)5. Similarly, in ICSS, the periprocedural risks of
stroke or death were 6.9% and 8.7% in supervised and more experienced
centres, respectively6 arguing against a learning curve playing a major
role in the recent trials.
Regarding intermediate term outcomes, there is a major error in the
labelling of figure 6 of the BMJ paper regarding the intermediate term
outcomes. The hazard ratios (HR) for any stroke or death presented in
figure 6 have been calculated in the inverse way to the rest of the paper
and correspond in fact to CAS vs. CEA, but are erroneously interpreted as
CEA vs. CAS. The result is actually slightly in favour of CEA (figure 1-
A), not CAS as plotted. Moreover, the interpretation of the combined HR
is not straightforward because Meier et al. combined different outcomes.
Indeed, they used the early rate of any disabling stroke or death for
CAVATAS (rather than the long term follow up data) and ipsilateral
ischaemic stroke or vascular death for SPACE, as the any stroke or death
endpoint was not reported in these papers3 7. In addition, in figure 7,
the number of events cannot be derived from the CAVATAS paper as only
Kaplan-Meier survival rates were reported. Moreover, the number of events
is wrong in the CEA arm of EVA-3S8. When using the most recent CAVATAS
data, the correct number for EVA-3S (54/262) and withdrawing the SPACE
trial in which this outcome was not reported, the combined OR of
intermediate term risk of stroke or death using CEA as the denominator, is
1.22 (0.96-1.55, p=0.10; figure 2-A), favouring CEA.
More importantly, although the any stroke or death endpoint is relevant
when assessing the safety of CAS and CEA, using this endpoint is highly
questionable when assessing the intermediate and long-term efficacy of the
techniques. Indeed, nonprocedural deaths are unlikely to be prevented by
any procedure. As deaths are more frequent than ipsilateral strokes, any
potential difference in risk of stroke between the two arms beyond 30 days
is likely to be diluted. In the long-term, both techniques aim at
preventing ipsilateral ischaemic strokes. Thus, in our opinion, the most
appropriate endpoint is any periprocedural stroke or death and non-
periprocedural ipsilateral stroke. Using this endpoint, when it was
available, the combined HR from EVA-3S and SPACE trials is 1.40 (0.80-
2.45, p=0.24; figure 1-B) (only non-periprocedural ischaemic strokes were
reported in SPACE) and the combined OR from six studies including the most
recent CAVATAS data is 1.33 (0.93-1.90, p=0.12; figure 2-B). Another
relevant endpoint is any stroke or periprocedural death, which includes
potentially ipsilateral strokes that cannot be formally classified as to
side, as well as allowing for the possibility that treatment of one artery
may improve the contralateral circulation. Using this endpoint, the
combined HR from EVA-3S and CAVATAS is 1.47 (1.09-1.98, p=0.01; figure 1-
C) and the combined OR from the same six studies is 1.36 (0.98-1.88,
p=0.07; figure 2-C). The difference is largely accounted for by the higher
periprocedural risk of CAS compared to CEA. In fact, as reported in the
2009 CAVATAS paper, when considering the risk of nonperiprocedural
ipsilateral stroke, the combined OR from the EVA-3S, SPACE and CAVATAS
trials is 1.18 (0.70-1.99).3 Finally, it should also be pointed that,
considering the very small number of strokes occurring during follow-up,
none of the trials had enough power to formally assess non-inferiority of
CAS compared with CEA regarding the long-term prevention of ipsilateral
The upcoming results of CREST9 and the future follow-up data from
ICSS should provide more confidence in the assessment of the risk-benefit
ratio of CAS vs. CEA. A new meta-analysis when these are published will
then certainly add to our knowledge.
Emmanuel Touzé ; Université Paris Descartes, INSERM UMR S894, Hôpital Sainte-Anne, Service de Neurologie, Paris, France.
Ludovic Trinquart ; INSERM CIE4, Assistance Publique–Hôpitaux de Paris, Unité de recherche clinique, Hôpital Européen Georges Pompidou, Paris, France.
Martin M Brown ; Institute of Neurology, University College London, Box 6, The National Hospital for Neurology and Neurosurgery, Queen Square, London WC1N 3BG, UK.
Jean-Louis Mas ; Université Paris Descartes, INSERM UMR S894, Hôpital Sainte-Anne, Service de Neurologie, Paris, France.
1. Meier P, Knapp G, Tamhane U, Chaturvedi S, Gurm HS. Short term and
intermediate term comparison of endarterectomy versus stenting for carotid
artery stenosis: systematic review and meta-analysis of randomised
controlled clinical trials. BMJ 2010;340:c467.
2. Touze E, Trinquart L, Chatellier G, Mas JL. Systematic review of the
perioperative risks of stroke or death after carotid angioplasty and
stenting. Stroke 2009;40:e683-93.
3. Ederle J, Bonati LH, Dobson J, Featherstone RL, Gaines PA, Beard JD, et
al. Endovascular treatment with angioplasty or stenting versus
endarterectomy in patients with carotid artery stenosis in the Carotid and
Vertebral Artery Transluminal Angioplasty Study (CAVATAS): long-term
follow-up of a randomised trial. Lancet Neurol 2009;8:898-907.
4. Safety results of the International Carotid Stenting Study (ICSS):
early outcome of patients randomised between carotid stenting and
endarterectomy for symptomatic carotid stenosis. European Stroke
Conference; 2009 26-29 May; Stockholm, Sweden.
5. Mas JL, Chatellier G. Recent carotid stenting trials. Lancet Neurol
6. Ederle J, Dobson J, Featherstone RL. Safety of carotid artery stenting
compared with endarterectomy in patients with symptomatic carotid stenosis
in the International Carotid Stenting Study (ICSS): Short-term results of
a randomised trial. Lancet. 2010. Early Online Publication, 26 February
7. Eckstein HH, Ringleb P, Allenberg JR, Berger J, Fraedrich G, Hacke W,
et al. Results of the Stent-Protected Angioplasty versus Carotid
Endarterectomy (SPACE) study to treat symptomatic stenoses at 2 years: a
multinational, prospective, randomised trial. Lancet Neurol 2008;7:893-
8. Mas JL, Trinquart L, Leys D, Albucher JF, Rousseau H, Viguier A, et al.
Endarterectomy Versus Angioplasty in Patients with Symptomatic Severe
Carotid Stenosis (EVA-3S) trial: results up to 4 years from a randomised,
multicentre trial. Lancet Neurol 2008;7:885-92.
9. Lal BK, Brott TG. The Carotid Revascularization Endarterectomy vs.
Stenting Trial completes randomization: lessons learned and anticipated
results. J Vasc Surg 2009;50:1224-31.
Competing interests: No competing interests
There is a typo in the corresponding author's email address,
firstname.lastname@example.org. With a tenure of 6 years at University of Michigan
Medical School, I am confident such an email address is invalid. It should
read email@example.com, or firstname.lastname@example.org.
The public profile of Dr. Gurm at UMich directory:
Competing interests: No competing interests