SAPPHIRE: Precious Gem or "Fool's Gold"?

No operation has been associated with quite so much controversy as carotid endarterectomy (CEA). Introduced in 1953, it rapidly became the most commonly performed arterial procedure in the western world, largely because the underlying rationale was simple and attractive. By the 1980s, however, concerns about the overall “appropriateness” of endarterectomy became the catalyst for the landmark symptomatic and asymptomatic randomized trials.^1–4

The European Carotid Surgery Trial (ECST), the North American Symptomatic Carotid Endarterectomy Trial (NASCET), the Asymptomatic Carotid Atherosclerosis Study (ACAS) and the Asymptomatic Carotid Surgery Trial (ACST) were, thereafter, responsible for the development of international guidelines which remain to this day. Following publication of the trials, most vascular surgeons probably thought that CEA would progress unchallenged. Few paid serious attention to the emerging threat of carotid angioplasty, primarily because it seemed inevitable (to the surgeons at least) that the high risk of embolic stroke would condemn this alternative therapy to the obituary column. They should, of course, have remembered the successful emergence and technological advances regarding angioplasty in the coronary and peripheral circulations! Early publications in selected patients and the inevitable systematic review rapidly confirmed that carotid angioplasty might become a serious (and attractive) alternative to CEA.⁵

Since those pioneering days, CEA and carotid angioplasty/stenting (CAS) have been compared in 10 randomized trials, 8 of which have been reported. Some are now too historic or were methodologically compromised to influence contemporary practice, but none have aroused quite so much controversy as the Stenting Angioplasty with Protection in Patients at HIgh Risk for Endarterectomy (SAPPHIRE) trial.⁶ SAPPHIRE was unique for two reasons. The first was the inclusion of ECG or biochemical evidence of silent myocardial infarction (MI) within a composite 30-day end-point (death/stroke/MI). Surgeons remain angry about the inclusion of occult MI (inappropriately so to my mind), but this anger served only to deflect attention from the most controversial issue regarding SAPPHIRE. That was the concept of randomizing only the patients considered “high risk for CEA.” To qualify for entry, patients had to be either symptomatic with a 50–99% stenosis or asymptomatic with an 80–99% stenosis AND exhibit one other factor making them “high risk” for CEA. At first sight this seems a not unreasonable strategy, but (as will be seen) meaningful interpretation ultimately depends upon what is meant by “high risk”. SAPPHIRE concluded that in “high-risk for surgery” patients, CAS using an obligatory cerebral protection device (CPD) was at least equivalent to CEA. This was an interesting choice of words because it has now largely been translated to mean “preferable” or “safer” than CEA for all high-risk patients.⁷

So, do the SAPPHIRE results justify these conclusions? The key to understanding the controversy is the definition of “high risk”. The ideal trial should have randomized patients who were high risk for stroke (rather than for surgery), as the two are not the same. Accordingly, the reader must look critically at the cohort of patients who were actually randomized within SAPPHIRE. The vast majority (>70%) were asymptomatic. Moreover, no information was provided regarding the degree of stenosis in the symptomatic patients (50–69%, versus 70–99%), time since the most recent event or gender. In short, everyone randomized within SAPPHIRE seems to have been considered to be of equivalent risk. As will be seen, this is a serious flaw.

Data regarding gender, timing of symptoms and stenosis severity are crucial to interpreting and implementing the conclusions from SAPPHIRE. The Carotid Endarterectomy Trialists Collaboration (CETC) has now combined the ECST and NASCET data, having remeasured the pre-randomization angiograms using the NASCET method.⁸ Table 1 presents a reworking of the outcomes from the 5893 patients in the CETC database relative to: (i) the absolute risk reduction in stroke conferred by CEA at 5 years; (ii) the number needed to treat to prevent one stroke at 5 years; and (iii) the number of strokes prevented at 5 years by treating 1000 patients. In males, CEA continued to confer significant benefit (maximal in the 70–99% stenosis subgroup) irrespective of the time between the last event and surgery. However, while symptomatic women with a 50–69% stenosis gained benefit, provided surgery was performed within 2 weeks of the last clinical event, no benefit was conferred by performing surgery thereafter. Similarly, CEA conferred significant benefit for up to 4 weeks in symptomatic female patients with severe (70–99%) stenoses but none thereafter. Accordingly, while males remain high risk for stroke for weeks/months after their most recent event (and therefore should be considered high risk in the context of SAPPHIRE), the complete opposite was true in females. Symptomatic females with 50–99% stenoses in whom > 4 weeks had elapsed since their last event gained little or no benefit from CEA and therefore could never be considered “high-risk” by anyone’s definition.

Detractors might argue that the ECST and NASCET data are too historic and that the analyses made by the CETC do not reflect contemporary practice. This argument is unsustainable! There is no evidence from any of the subsequent randomized trials involving CEA that the procedural risk has significantly improved over the last 10 years (nor for that matter following CAS). However, the concept of “best medical therapy” has improved considerably over the last two decades. Accordingly, if ECST and NASCET were to be repeated now, it is likely that the overall benefits conferred by surgery would be significantly less! If the definition of “high-risk” in the symptomatic patient is problematical (especially regarding gender), it pales into insignificance when compared to its impact in asymptomatic patients.

The original ACAS alert³ recommended that; patients with asymptomatic carotid artery stenosis of 60% and whose general health makes them good candidates for elective surgery will have a reduced 5-year risk of ipsilateral stroke if CEA performed with less than 3% perioperative morbidity and mortality is added to aggressive management of modifiable risk factors.’ This alert led to an unprecedented increase in CEA numbers with little attention being paid to the ‘small print’ from the study. There were many concerns about this study, but the single most important was that CEA did not confer any benefit in women. This finding has had little impact on contemporary practice, and males and females have generally been treated as if they have the same natural history risk regarding stroke and that CEA confers the same benefit. In SAPPHIRE no discrimination was made regarding gender. ACST (being a much larger study) did, however, observe that women gained significant benefit from immediate CEA.⁴ This information was, thereafter used as justification for intervening. However, once more it is important to read the “small print”. When ACST published their long-term data on stroke prevention in women, they did not include the operative risk. When the operative risk was included, ACST (once again) showed no evidence that women benefited from surgery.⁹ It therefore seems inconceivable that asymptomatic females could ever be included in a trial which only randomized “high-risk” patients! Understandably, SAPPHIRE has largely ignored these issues and simply cites the similar/better outcomes in CAS patients as compared with CEA at 30-days and 1 year. While this may be true (and otherwise acceptable to the “headline” readers), you have to look long and hard to identify the actual 30-day death/stroke rates in the asymptomatic patients who comprised the vast majority of patients within the trial. For those interested, the 30-day death/stroke rate for asymptomatic patients undergoing protected CAS in SAPPHIRE was 5.8%, compared with 6.1% following CEA. What everyone fails to have considered is that this level of procedural risk (approximately 6% for CEA and CAS) is too high for any long-term benefit to accrue regarding stroke prevention.