Noninferiority was not shown between stenting and endarterectomy in severe symptomatic carotid stenosisPDF
ACP J Club. 2007 Mar-Apr;146:33. doi:10.7326/ACPJC-2007-146-2-033
Related Content in this Issue
• Companion Abstract and Commentary: Stenting led to more stroke or death than did endarterectomy in symptomatic carotid stenosis ≥ 60%
Clinical Impact Ratings
Ringleb PA, Allenberg J, Bruckmann H, et al. 30 day results from the SPACE trial of stent-protected angioplasty versus carotid endarterectomy in symptomatic patients: a randomised non-inferiority trial. Lancet. 2006;368:3239-47. [PubMed ID: 17027729]
In patients with severe symptomatic carotid stenosis, is stenting noninferior to endarterectomy?
Design: Randomized, controlled, noninferiority trial (Stent-Supported Percutaneous Angioplasty of the Carotid Artery verus Endarterectomy [SPACE]).
Blinding: Blinded (data safety and monitoring committee).*
Follow-up period: 24 months.
Setting: 35 centers in Germany, Austria, and Switzerland.
Patients: 1200 patients > 50 years of age who had symptomatic stenosis (amaurosis, transient ischemic attack, or stroke) in the past 180 days and severe ipsilateral carotid artery stenosis (≥ 50% according to North American Symptomatic Carotid Endarterectomy Trial criteria or ≥ 70% according to European Carotid Surgery Trial criteria). Exclusion criteria included intracranial bleeding in the past 90 days; uncontrolled hypertension; intracranial arteriovenous malformation or aneurysm; life expectancy < 2 years; coagulation abnormality; contraindication to heparin, aspirin, clopidogrel, or contrast media; occlusion of the common or internal carotid artery; and other causes of stenosis.
Intervention: Stenting (n= 605) or endarterectomy (n= 595). Neurologists had to be experienced in duplex sonography and stroke treatment, interventionalists had to have done ≥ 25 successful percutaneous transluminal angioplasty or stent procedures, and vascular surgeons had to have done ≥ 25 carotid endarterectomy procedures and provide mortality and morbidity rates.
Outcomes: Any ipsilateral stroke (ischemic stroke, intracerebral bleeding, or both, with symptoms lasting > 24 h) or death from any cause at 30 days. Secondary 30-day outcomes included disabling ipsilateral stroke or death, any stroke, and procedural failure.
Patient follow-up: 1183 patients (98.6%) (mean age 68 y, 72% men) (intention-to-treat analysis).
Stenting and endarterectomy did not differ for any ipsilateral stroke or death from any cause (primary endpoint), and the criterion for noninferiority was not met (Table). Groups did not differ for disabling ipsilateral stroke or death (4.7% vs 3.8%; odds ratio [OR] 1.25, 95% CI 0.71 to 2.22), any stroke (7.5% vs 6.2%; OR 1.24, CI 0.79 to 1.95), or procedural failure (3.2% vs 2.1%; OR 1.56, CI 0.71 to 3.56).
In patients with severe symptomatic carotid stenosis, noninferiority was not shown between stenting and endarterectomy.
Sources of funding: Federal Ministry of Education and Research; German Research Foundation; German Society of Neurology; German Society of Neuroradiology; German Radiological Society; Boston Scientific; Guidant; Sanofi-Aventis.
For correspondence: Professor W. Hacke, University of Heidelberg, Heidelberg, Germany. E-mail firstname.lastname@example.org.
Table. Stenting vs endarterectomy for severe symptomatic carotid stenosis at 30 days†
|Outcome||Stenting||Endarterectomy||Difference (90% CI)|
|Any ipsilateral stroke or death from any cause||6.84%||6.34%||0.51% (−1.89 to 2.91)†|
†The criterion for noninferiority was an upper limit of < 2.5% for the 90% CI, so noninferiority was not shown.
First the good news: Noninferiority trials can (although sometimes don't) ask clinically sensible questions. Once we are convinced that a treatment does more good than harm, if somebody comes up with a new, alternative treatment, there are 2 clinically sensible questions to be asked. First, is the new treatment better than ( “superior to” ) the established treatment? Second, if the new treatment is not better, is it “as good as” the established treatment but preferable in some other way (e.g., safety, comfort, or cost)? To answer these 2 questions, we would carry out a “head-to-head” trial comparing the old and new treatments. At the trial's conclusion, we would want to know how confident we could be in our answers to these 2 questions; we would want any observed differences ( “better than” ) or similarities ( “as good as” ) to be very unlikely to be due to chance.
The bad news is that, in order to answer the “as good as” question, we have to stand our traditional understanding of trials and P values on its head. Alas, P values describe the probability that treatments are different by chance alone, whereas we want to know the probability that treatments are the same by chance alone. Fortunately (albeit confusingly), there is a statistical solution to this dilemma. It begins by documenting the benefit provided by the established treatment and then agreeing on a slightly less favorable ( “inferior” ) outcome that would be acceptable if the new treatment were less risky, less painful, or preferable in some other way. This inferior outcome constitutes a lower limit of tolerability. At any worse outcome, the new treatment would be judged inferior to the established treatment. (The question of who should define this “inferiority” limit of tolerability deserves a commentary of its own; suffice it to say here that patients' definitions should dominate this determination.)
If we pose this “as good as” question in this “worse than” way, the subsequent trial's P value indicates the probability that patients receiving the new treatment have had worse ( “inferior” ) outcomes by chance. When the P value for this inferior outcome is small, we reject the notion that the new treatment produces worse outcomes, consider it “as good as” the established treatment on a clinical level, and then risk confusing everybody by labeling it “noninferior.”
As with other issues in “statistical significance,” sense and comprehension are better conveyed with confidence intervals than P values, and examples of these are shown in the Figure. The dotted line labeled MIH indicates the “minimally important harm” limit of tolerability, to the right of which the new treatment is inferior to the old one. Similarly, the line labeled MIB indicates the “minimally important benefit,” to the left of which the new treatment is superior. Example A describes an inferior treatment (e.g., the EVA-3S trial) and B an “as-good-as,” or noninferior, one. Example C is an all-too-familiar example of an “indeterminate” trial that may have stopped too soon before it could tell us whether the new treatment was worse than (inferior to) or as good as (noninferior to) the established treatment (e.g., the SPACE trial).
For the present, carotid stenting should be recommended only for patients at high risk for cardiovascular complications after endarterectomy based on the results of the Stenting and Angioplasty with Protection in Patients at High Risk for Endarterectomy (SAPPHIRE) trial (1). Based on the results of EVA-3S and SPACE and a recent Cochrane review (2), any other patient should receive maximum medical therapy and have endarterectomy if clinically indicated or be enrolled in ongoing trials comparing stenting with endarterectomy.
David L. Sackett, MD
Trout Research and Education Center at Irish Lake
Markdale, Ontario, Canada
Wieslaw Oczkowski, MD
Hamilton, Ontario, Canada
1. 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. [PubMed ID: 15470212]
2. Coward LJ, Featherstone RL, Brown MM. Safety and efficacy of endovascular treatment of carotid artery stenosis compared with carotid endarterectomy: a Cochrane systematic review of the randomized evidence. Stroke. 2005;36:905-11. [PubMed ID: 15746454]
Figure. Minimally important benefit (MIB) and minimally important harm (MIH) boundaries for trial results in noninferiority trials.
The EVA-3S trial is illustrated by A, and the SPACE trial by C.