PDF: SAPPHIRE
(2008)Objective
Compare long-term outcomes of carotid artery stenting versus endarterectomy in patients at high surgical risk.
Study Summary
• In high-surgical-risk patients with significant carotid stenosis, carotid stenting with embolic protection was not inferior to endarterectomy over 3 years in terms of stroke, myocardial infarction, or death. Stroke rates were similar between groups, and long-term freedom from target-vessel revascularization was high in both arms.
Intervention
Randomized trial of 334 patients with symptomatic ≥50% or asymptomatic ≥80% carotid stenosis and high surgical risk. Intervention was carotid artery stenting with embolic protection vs. endarterectomy. Median follow-up: 3 years.
Study Design
Arms: Array
Outcome
• Major adverse event (death, stroke, or MI within 30 days or death/ipsilateral stroke through 3 years): 24.6% (stenting) vs. 26.9% (endarterectomy); absolute difference −2.3% (95% CI −11.8 to 7.0); P=0.71
• Stroke: 9.0% in both groups; 11 ipsilateral in stenting, 9 in endarterectomy
• Death (any cause): 18.6% vs. 21.0%
• MI: 5.4% vs. 8.4%
• Target-vessel revascularization: 2.4% (stenting) vs. 5.4% (endarterectomy); P=0.26
• Kaplan–Meier 3-year freedom from stroke: 92.0% (stenting) vs. 93.3% (endarterectomy); P=0.80
• Stroke: 9.0% in both groups; 11 ipsilateral in stenting, 9 in endarterectomy
• Death (any cause): 18.6% vs. 21.0%
• MI: 5.4% vs. 8.4%
• Target-vessel revascularization: 2.4% (stenting) vs. 5.4% (endarterectomy); P=0.26
• Kaplan–Meier 3-year freedom from stroke: 92.0% (stenting) vs. 93.3% (endarterectomy); P=0.80
Bottom Line
In high-risk patients with severe carotid artery stenosis, there was no significant difference in the 3-year composite risk of death, stroke, or myocardial infarction between patients treated with carotid stenting with emboli protection and those who underwent carotid endarterectomy.
Major Points
- SAPPHIRE was the FIRST randomized trial to demonstrate noninferiority of carotid artery stenting (CAS) vs carotid endarterectomy (CEA) — published initially in 2004 (NEJM), with 3-year follow-up in 2008.
- Enrolled 334 HIGH-RISK surgical patients — those typically excluded from earlier CEA trials (NASCET, ECST). High-risk criteria: significant cardiac disease, severe pulmonary disease, contralateral carotid occlusion, recurrent stenosis, prior neck radiation, or age >80.
- 30-day primary composite (death, stroke, MI): 4.8% CAS vs 9.8% CEA (p=0.09) — stenting numerically favored, driven primarily by lower MI rates with CAS.
- 3-year composite (death, stroke, MI at 30 days + death or ipsilateral stroke at 31 days–3 years): 24.6% CAS vs 26.9% CEA (p=0.71) — no significant difference, confirming durability of noninferiority.
- The trial included a nonrandomized registry of 406 patients deemed too high-risk for one treatment — these had higher event rates, validating the high-risk classification.
- Used the Cordis PRECISE nitinol self-expanding stent with the Angioguard emboli-protection device — the FIRST trial to mandate distal embolic protection during CAS.
- Led directly to FDA approval of the Cordis stent system for high-risk patients in 2004 — transformed carotid revascularization practice.
- Criticized for industry sponsorship (Cordis) and the inclusion of MI in the primary endpoint, which favored stenting (MI more common with CEA due to general anesthesia/cardiac stress).
- Subsequent trials in AVERAGE-risk patients (CREST, ACT-1, SPACE, EVA-3S, ICSS) showed more nuanced results — CAS had higher stroke rates and CEA had higher MI rates, but long-term outcomes were similar.
- 3-year follow-up was incomplete in 22.2% of patients, raising concerns about ascertainment bias in the long-term results.
Study Design
- Study Type
- Prospective, randomized, multicenter trial.
- Randomization
- Yes
- Blinding
- Unblinded (Open-label), typical for procedural trials.
- Sample Size
- 334
- Follow-up
- 3 years (1080 days).
- Centers
- 29
- Countries
- United States
Primary Outcome
Definition: Composite of death, stroke, or MI within 30 days or death/ipsilateral stroke 31 days-3 years
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 26.9% (CEA) | 24.6% (CAS) | - (-11.8 to 7.0 (absolute difference)) | 0.71 |
Limitations & Criticisms
- Small randomized cohort (n=334) limits statistical power for subgroup analyses and increases susceptibility to type II error for individual endpoints.
- 22.2% of patients lost to follow-up at 3 years — this high attrition rate could introduce significant ascertainment bias, particularly if losses were differential between arms.
- Results apply ONLY to high-surgical-risk patients — cannot be extrapolated to average-risk patients (later addressed by CREST, ACT-1, SPACE, EVA-3S, ICSS).
- No medical-therapy-only arm — cannot determine whether either revascularization strategy is superior to modern optimal medical therapy alone (a question raised by CREST-2).
- Industry-sponsored (Cordis, the stent manufacturer) — potential for bias in trial design, conduct, and reporting. Cordis employees were involved in data analysis.
- Inclusion of MI in the primary composite endpoint favored CAS — MI was more common with CEA (due to general anesthesia, hemodynamic stress), inflating the apparent benefit of stenting.
- Embolic protection device (Angioguard) was first-generation — modern devices and stent designs have improved, making SAPPHIRE results less reflective of current CAS outcomes.
- The definition of 'high surgical risk' was broad and somewhat subjective — included heterogeneous conditions (cardiac disease, pulmonary disease, contralateral occlusion, prior neck surgery) that carry different risk profiles.
- Open-label design (inherent in procedural trials) introduces performance bias — surgeons and interventionalists may have managed patients differently based on treatment assignment.
Citation
N Engl J Med 2008;358:1572-9.