Advertisement
Advertisement
October 2010
Contemporary Randomized Trials Comparing CAS and CEA
A critique of the EVA-3S, SPACE, ICSS, SAPPHIRE, and CREST studies.
“If you think of standardization as the best that you know today, but which is to be improved tomorrow, you get somewhere.”
- Henry Ford, 1863–1947
We now have a wealth of data from randomized trials that sought to compare carotid artery stenting (CAS) and carotid endarterectomy (CEA) as strategies intended to reduce intermediate and long-term risk of stroke ipsilateral to a significant carotid artery stenosis. If the data from these trials were wholly comparable on the basis of trial construct, baseline operator expertise, requirements for equipment used, choice of primary endpoint, and rigor of ascertainment of all chosen endpoints, then we would have one of the most perfect and complete combined datasets the endovascular community has ever seen, and there would probably be no need for ongoing discussion.
This critique seeks to explore the similarities and differences in the design and execution of these trials and attempts to highlight common themes and lessons learned.
TRIAL SIMILARITIES
Independent Review
Regardless of some differences in baseline surgical risk
and majority population in terms of symptom status, and
in common with most randomized trials of carotid intervention,
the trials listed on page 60 (Randomized Trials for
Standard-Risk/High-Risk Population) were all designed as
one-to-one randomized constructs with intended independent
review of outcomes (ie, review by stroke physicians
or vascular neurologists who were not intimately
involved in the performance of the carotid interventions
being compared). The principal investigators (PIs) of the
listed trials were all neurologists, with the exception of
CREST (Carotid Revascularization Endarterectomy Versus
Stenting Trial), which had a vascular surgical co-PI at its
inception. The self-audit of outcome data in this clinical
arena is notoriously unreliable.1 Death is, without doubt, a
hard endpoint and an unarguable outcome, evident to
the nonspecialist. However, minor neurological deficit
that may otherwise affect the primary powered endpoint
of any trial comparing strategies in carotid intervention,
and may very well affect the patients' quality of life, may
be easily overlooked by the operator.
TRIAL DISCREPANCIES
Operator Experience
A cursory glance at the stroke/death outcomes in
Table 1 would suggest substantially better results for
carotid artery stenting (CAS) within CREST than within
the European trials. This is partly explained by more
stringent entry requirements for operators within CREST
than in EVA-3S (Endarterectomy Versus Angioplasty in
Patients With Severe Symptomatic Carotid Stenosis),
ICSS (International Carotid Stenting Study), and, in its
latter half, SPACE, and in part due to inclusion of asymptomatic
patients.
In the European trials, there is evidence to suggest that there were temporal changes in interventionist entry criteria, most likely resulting from attempts to expedite recruitment. In SPACE, after a relaxation of entry requirements, the adverse event rate rose.2 The original sample size for a noninferiority margin of 2.5% was 1,900 patients. A preliminary predefined interim analysis at 3 years (after the inclusion of 369 patients in each arm) demonstrated an overall event rate of < 3.9%, leading to a sample size recalculation of 1,200 patients to show noninferiority. At the second preplanned interim analysis after the introduction of a “preliminary certificate” (allowing interventionists to perform CAS within the trial after performing 10 cases), the sample size required to prove noninferiority was recalculated again (2,500 patients), presumably because of downsizing of absolute risk reduction in the CAS arm. The implication is that there must have been a substantial increase in the endpoint rate in the CAS arm after relaxation of the entry criteria. Notably, there was a linear relationship between numbers of patients recruited into this trial and the event rate for CAS.3 This was not the case for carotid endarterectomy (CEA), suggesting within-trial learning for CAS but not for CEA. The authors of EVA-3S and ICSS stated that there was no demonstrable influence of operator experience on outcome. However, the trials did not allow an adequately powered comparison of experienced and inexperienced operators, and 85% of operators within EVA-3S had performed fewer than 50 cases, the absolute minimal baseline experience for CAS according to expert opinion.4 Assessment of an individual surgeon's CEA performance is thought to require ≥ 200 cases to ensure sufficiently narrow 95% confidence intervals.5 Tellingly, the relationship between trial entry requirement for interventionists within EVA-3S, ICSS and SPACE, and CAS outcomes within these trials is remarkably linear.2
Stroke/Death Ratios
If the all-stroke to all-cause death ratio for CAS is compared across the trials, the figure is reasonably consistent
(see Randomized Trials for Standard-Risk/High-Risk
Population chart). There are two notable outliers; in SAPPHIRE,
the ratio is small (3) perhaps because this trial
ensured a very high level of operator expertise (Table 1) and
recruited a majority asymptomatic population. In EVA-3S,
the ratio is 11.5, higher than in the other trials. Arguably,
expertise in carotid stenting is reflected by the stroke/death
ratio, and we can see from Table 1 that EVA-3S had the
most lax trial entry requirements for CAS (but not for CEA).
Inclusion of Asymptomatic Patients
EVA-3S, SPACE, and ICSS recruited symptomatic
patients exclusively. Of a total of 2,502 patients randomized
within CREST, 1,181 were asymptomatic. This may
have diluted the procedural risks across this trial for
both CAS and CEA, because asymptomatic patients can
expect lower procedural hazards than symptomatic
patients during carotid intervention.6-8
Myocardial Infarction
Myocardial infarction (MI) was included as an outcome
event in EVA-3S, ICSS, and CREST, although it was
not evaluated in SPACE. In ICSS, the primary endpoint
was the difference in fatal or disabling stroke in any territory
at 3 years, and this has yet to be reported (an
interim analysis reflecting 120-day safety data has been
published).9 In EVA-3S, MI was considered a secondary
endpoint. In CREST, MI formed part of a composite primary
endpoint. There is a substantial disparity among
the trials with respect to the magnitude of difference in
MI rates between CAS and CEA limbs (Table 2). This
raises the question of ascertainment bias. Scrutiny of
this bias requires both a thorough evaluation of the definitions
of MI employed within the trials but also an
exploration of the adjudication process around independent
review of this endpoint.
EVA-3S. MI was defined by at least two of the following criteria: typical chest pain lasting 20 minutes or more; serum levels of creatinine kinase MB or troponin at least twice the upper limit of normal range; and new Q waves on at least two adjacent derivations or predominant R waves in V1 (R wave ≥ 1 mm > S wave in V1).
ICSS. MI was defined as the presence of two of the following three criteria: specific cardiac enzymes more than twice the upper limit of normal; history of chest discomfort for at least 30 minutes; or the development of specific abnormalities (eg, Q waves) on a standard 12-lead electrocardiograph.
SAPPHIRE. The definition of MI in this trial was “a creatinine kinase < 2 times the upper limit of normal with a positive MB fraction.”
MI was included in a composite primary endpoint because the investigators understood that patients with carotid artery disease are also likely to be at substantial risk for MI.
CREST MI was defined by a creatinine kinase MB or troponin level that was twice the upper limit of the normal range or higher according to the center's laboratory, in addition to either chest pain or symptoms consistent with ischemia or electrocardiogram (EKG) evidence of ischemia, including new ST-segment depression or elevation of more than 1 mm in two or more contiguous leads according to the core laboratory.10
The CREST criteria appropriately included non—Q wave MI, although the EVA-3S and ICSS protocols accepted only transmural MI as a relevant consideration. The Carotid Stenting Trialists' Collaboration (incorporating the PIs of ICSS, SPACE, and EVA-3S) appears to view non—Q wave MI as irrelevant yet puts great faith in the findings of the ICSS diffusion-weighted imaging study in which much was made of an increase in new hyperintensities on diffusionweighted imaging magnetic resonance imaging of the brain in the CAS limb compared with the CEA limb.11 The fate and clinical relevance of these lesions is not known with any degree of confidence.12 A recent study in which a fairly comprehensive cognitive function test battery was employed to compare outcomes in CAS and CEA patients versus a control population suggests that new brain lesions after carotid revascularization are not associated with cognitive decline.13
The Carotid Stenting Trialists' Collaboration considered that the different rates of MI in the trials could be explained by the “substantially higher baseline prevalence of cardiovascular disease in CREST.” The devil is in the detail. Baseline cardiac comorbidity in CREST was described as “previous cardiovascular disease,” which is an all-encompassing definition. Baseline rates for CAS and CEA were 42.4% and 45%, respectively. In EVA-3S, baseline cardiovascular risk was specified only as “previous MI,” thus explaining the low incident rates (10.7% and 13.1% for CAS and CEA, respectively). In ICSS, a number of cardiovascular risk variables were considered (Table 3).
Patients may clearly have had more than one risk factor, rendering a summation of percentage risk overly simplistic. Even allowing for the fact that patients with atrial fibrillation were excluded from CREST, the overall cardiovascular risk profile for ICSS looks similar to CREST.
Relevance of Periprocedural MI
A procedure-related MI (whether Q wave or non—Q
wave) is a very relevant consideration for any patient. A
non—Q wave MI will confer an increase in the risk of
death by a factor of 6 and an increase in the 6-month
risk of further MI by a factor of 27.14
Ascertainment and Adjudication
The CREST protocol mandated postprocedure screening
for MI with EKG and enzyme assay. In EVA-3S and
ICSS, it would appear that only patients with cardiac
symptoms postprocedure were evaluated (by EKG and
cardiac enzyme assay). A specific MI adjudicating committee
oversaw outcome events in CREST; this was not
the case for EVA-3S and ICSS.
Mandatory CAS equipment. Tightly proscribed trials ensure standardized practice within them but compromise generalizability. The regulatory and reimbursement environment in the United States dictates mandatory use of approved carotid stents and embolic protection devices (EPDs), and this often results in use of a particular manufacturer's stent and EPD in combination. Although these constraints ensure majority operator experience of particular stent/EPD combinations within trials such as CREST, thereby mitigating against learning curve issues with particular devices, there are two important counter arguments.
First, lesion-specific stenting is becoming a firmly entrenched belief in the European Union. This means that operators select the stent system and EPD according to factors such as the patient's anatomy, lesion characteristics, symptoms, and cardiovascular status. Of course, this presupposes a level of baseline experience that supports this degree of complex decision-making. Many interventionists performing CAS, certainly within EVA-3S and ICSS, simply did not have the baseline experience to be able to make these sorts of value judgements about CAS equipment. Second, restricted use of stents/EPDs may limit the validity of inferences regarding generic CAS outcomes in the real world.
WEAKNESSES AS EXECUTED
Both EVA-3S and SPACE were stopped early (the former
on the grounds of safety, the latter for reasons of
insufficient funding and futility). Generic problems of
truncated trials include credibility issues, imprecision
(wide confidence intervals), and bias (suspended trials
may stop on a “random high”). When noninferiority trials
(such as SPACE) are terminated early, the hazards of
treatment in the “experimental” limb may be systematically
overestimated, especially when the number of outcome
events is small. The SPACE investigators concluded
“that noninferiority of CAS was not proven.” The
upper limits of the confidence intervals of both the
intention-to-treat (ITT) and the per-protocol (PP)
analyses exceeded the preset delta of 2.5% (the margin
of difference between CAS and CEA outcomes thought
to be clinically relevant); however, because the lower
limits of both analyses crossed zero, the superiority of
CEA cannot be assumed. Note the value of both ITT
and PP analyses in noninferiority trials: In superiority
trials, the ITT analysis tends to be conservative, underestimating
differences in the treatment effect and protecting
against type 1 errors. In noninferiority trials,
underestimates of the treatment effect can bias toward
noninferiority, inflating the false positive effect, hence
the additive value of the PP analysis. SPACE failed to
prove the noninferiority of CAS due to lack of statistical
power and insufficient numbers. There were no significant
differences in the primary endpoint of ipsilateral
stroke/death between CAS and CEA; the absolute difference
was four outcome events in 600 patients per group.
LESSONS LEARNED
Apart from emerging trends with respect to lower procedural
hazards following CAS than CEA in women (in the
European trials), patients presenting with ocular symptoms,
and patients with contralateral occlusion (explored at the
roundtable discussion), the issue of age at treatment is an
intriguing and consistent finding.
Octogenarians
The matter of the safety record for CAS in octogenarians is the subject of ongoing debate. Higher rates of procedural
hazards were noted in the CREST lead-in phase;
thereafter, octogenarians were excluded from the leadin
phase, although they remained eligible for the randomized
phase. A recent meta-analysis of pooled data
from EVA-3S, SPACE, and ICSS demonstrated that in
patients < 70 years of age, the procedural risks of
stroke/death were 5.8% for CAS and 5.7% for CEA (relative
risk, 1.00; confidence interval, 0.68–1.47).15 The
excess risk of CAS appeared to be in patients 70 years or
older, where the risk of CAS versus CEA was 12% versus
5.9%, (relative risk, 2.04; confidence interval, 1.48–2.82). A
comparable pattern was seen in CREST; prespecified analyses
demonstrated an interaction between age and treatment
effect, with a crossover at approximately 70 years.
It is important to appreciate that the influence of age at
treatment on outcome for CAS is not a stochastic phenomenon,
and outcomes can be plotted against age as a
continuous variable, providing a predictive model that can
be used to help direct decision making.6
There have been recent reports, particularly from experienced operators working with proximal protection devices, suggesting very acceptable outcomes in this population. It is not clear whether baseline experience of the operator, type of EPD, or a combination of both was ultimately responsible for improved results.16,17 A point worth considering is the influence of the EPD on microembolic burden, which may vary from one type of device (proximal vs distal filter, etc.) to another. The elderly, with reduced cerebrovascular reserve, may fare better with devices that not only entrap macroemboli but also reduce the procedural microembolic burden.18
CONCLUSION
Unfortunately, there seem to be irreconcilable differences
between the current randomized trials, and it remains an
inconvenient truth that the differences are perhaps most
marked between the American and European trials. The
intertrial variability for the European trials appears superficial,
and the trials are similar enough to have allowed a (preplanned)
meta-analysis of their datasets, which has recently
been published. There are some common themes, to
include lower procedural hazards in patients with ocular
rather than cortical symptoms, in females (certainly in the
European trials), and in those with contralateral occlusion.
These factors will require further exploration as they were
not a priori endpoints (and the trials were not powered
to definitively answer questions about these subsets).
The impact of age on outcome is unmistakable but there
is an increasing body of evidence to suggest that octogenarians
can be treated safely by CAS in experienced centers
employing modified CAS technique.
Sumaira Macdonald, MBChB (Comm.), FRCP, FRCR, PhD, of the Steering and Technical Management Committees of the Asymptomatic Carotid Surgery Trial (ACST-2), is a Consultant Vascular Radiologist and Honorary Clinical Senior Lecturer at Freeman Hospital in Newcastle upon Tyne, England. She was a collaborator and proctor (physician instructor) in the International Carotid Stenting Study (ICSS). She has disclosed that she receives grant/research funding from Abbott Vascular, Cordis Corporation, Bard Peripheral Vascular, ev3 Inc., Medtronic Invatec, Pyramed, and W. L. Gore & Associates, and that she is a paid consultant to Abbott Vascular, ev3 Inc., Medtronic Invatec, and W. L. Gore & Associates. Dr. Macdonald may be reached at sumaira.macdonald@nuth.nhs.uk.
Advertisement
Advertisement
