Author + information
- Daniel B. Mark, MD, MPH⁎ (, )
- Manesh R. Patel, MD and
- Kevin J. Anstrom, PhD
- ↵⁎Reprint requests and correspondence:
Dr. Daniel B. Mark, Outcomes Research, Duke Clinical Research Institute, 2400 Pratt Street, Room 0311, Durham, North Carolina 27705
Comparative effectiveness research became the focal point of high expectations in 2010 when it was given a starring role in the health care reform bill passed by the U.S. Congress. The goal of comparative effectiveness research from a policy perspective is intuitively very appealing: to identify what works based on evidence to guide the provision of high-quality clinical care. In this modern re-envisioning of medicine, the “evidence” produced by comparative effectiveness research is the engine that powers clinical management and comprises both clinical trial results and high-quality observational research. These data are aggregated into meta-analyses and systematic reviews and are used to inform clinical practice guidelines. What rarely gets mentioned, however, is the critical operational next step in the process: translating the published evidence and guideline recommendations into shared treatment decisions for individual patients.
Traditionally, when the acute risk of death is high, prevention of death tends to dominate treatment decisions. Major disabling stroke is one of the few nonfatal endpoints that patients rank equal to or more important than death in making treatment choices. Moving from acute to chronic disease management decision making means that as the risks of death and major disabling morbidity go down, the other unintended consequences of each therapeutic option become quantitatively more important in decision making. In addition, the nature of the evidence supporting treatment alternatives changes. Assuming common sample sizes and hypothesized relative benefits, clinical trials studying what works in lower risk cohorts have less statistical power to work with and often use composite outcome events to keep the costs of the trial acceptable. Two strong assumptions underlie the use of the most common form of composite endpoint (reflecting treatment efficacy): first, that the components are all equally important to patients, and second, that the individual components all reflect the same relative therapeutic effect (1). In other words, the components are reasonable surrogates for each other. Recently, however, a second form of composite endpoint has been increasingly used in interventional cardiology, in which adverse effects related to both efficacy and safety of each of the treatments being compared are combined into a metric reflecting total net morbidity/mortality. This “net benefits–harms” type of endpoint has also been incorporated into the most recent trials comparing revascularization options for carotid artery disease, yet the result to date has been more controversy than clarity.
The evolution of treatment for carotid atherosclerotic arterial disease has a number of interesting parallels with coronary artery disease. In both cases, the genesis of our modern understanding of the disease can be traced back some 60 years. The first carotid endarterectomy (CEA) was described in 1953, substantially ahead of the first coronary artery bypass procedures more than a decade later, and the first carotid stents were used clinically in 1994, the same year the Food and Drug Administration approved the Palmaz-Schatz stent for coronary artery use. However, unlike coronary artery disease, the increasing degree of carotid atherosclerosis has been linked to the probability of future strokes. In addition, both CEA and carotid artery stenting (CAS) tend to accomplish similar degrees of revascularization. Two clinical trials established CEA as being superior to the medical therapy of the 1990s for symptomatic patients with stenosis >70%. Three additional trials reported between 1993 and 2004 provided evidence for improved outcomes with CEA over medical therapy alone in asymptomatic subjects with >60% stenosis, although the magnitude of benefit was considerably smaller.
As with percutaneous coronary intervention, development of percutaneous carotid revascularization was stimulated by the potential for providing a less invasive, potentially safer alternative to operation. Numerous meta-analyses have been performed over the last 5 years on the growing evidence base of randomized trials comparing carotid stenting with CEA, but they have failed to reach a clear consensus on how patients should be treated (2–4). At the same time, data show that carotid stent use is growing while use of CEA is flat or even falling modestly (5).
Examination of the most recent trial of CEA versus stenting helps to illustrate why more trial data are not likely to resolve the current controversies about treatment superiority. The North American CREST (Carotid Revascularization Endarterectomy Versus Stenting Trial) study found that, in average risk subjects, almost one-half of whom were considered asymptomatic with no preceding transient ischemic attack or stroke symptoms related to the target artery, the composite primary endpoint (stroke, myocardial infarction [MI], or death in the perioperative period and ipsilateral stroke out to 4 years) was similar in the 2 arms over a median follow-up of 2.5 years (hazard ratio for stenting 1.11, p = 0.51) (6). This would suggest that the choice of revascularization procedure is a toss-up. Kassirer and Pauker (7) defined “toss-ups” as “clinical situations in which the consequences of divergent choices are, on the average, virtually identical.” However, the path to equivalence was not symmetrical in the 2 arms of the CREST study. In the periprocedural period, stent arm patients had 4 per 1,000 more deaths (p = 0.18), 18 per 1,000 more strokes (predominantly but not exclusively “minor” or “nondisabling” [p = 0.01]), 11 per 1,000 fewer MIs (a mix of clinically evident and biomarker-only events [p = 0.03]), and 44 per 1,000 fewer cranial nerve palsies (not included in primary endpoint [p < 0.05]). Both the clinical and the biomarker-only MIs confer an adverse long-term prognosis (8). After the periprocedural period, the rates of ipsilateral stroke were equivalent in the 2 arms out to 4 years (2.0% for stenting vs. 2.4% for CEA, p = 0.85).
Surprisingly few of the clinical trials of stenting versus CEA before the CREST study assessed patient-reported outcomes, specifically health-related quality of life (QOL). The preference among trialists for objective, event-type outcomes is not unique to carotid revascularization trials (9). However, as the difference between treatments in rates of major morbid outcomes gets smaller and smaller, the importance of understanding the patient's experience of the different treatments becomes greater. Can health-related QOL information provide the clarity that equally weighted composite endpoints have failed to generate? Using the Medical Outcomes Study Short Form-36 health status assessment, the CREST study investigators found that a major stroke was about 3 times worse in its effect on QOL than a minor stroke was, whereas an endpoint MI had about two-thirds the QOL effect of a minor stroke (6).
In this issue of the Journal, Cohen et al. (10) report additional details from the QOL portion of the CREST study. Their study reported 2 important findings. First, the major differences in QOL between CAS and CEA occurred early and were related to the recovery after the procedures. No differences were seen at 12 months. Second, an endpoint stroke had large adverse effects on 1-year QOL, particularly in the areas of physical function, role functioning, and vitality (energy/fatigue). In contrast, endpoint MI and cranial nerve palsies had quantitatively smaller, and not statistically significant, effects on QOL. At first glance, the 2 findings of this study might seem at odds. If strokes cause persistent decrements in QOL whereas MI and cranial nerve palsies do not, should not the QOL scores for the stent arm reflect this? The resolution is found by focusing on the absolute number of excess perioperative strokes caused by stenting: 18 per 1,000. Even assuming that all 18 had a major disabling stroke (in contrast to the CREST study data, which showed that most of the strokes were not disabling), such a small number has an imperceptible effect on the population mean QOL values. Thus, our evaluation of the QOL data supports the view of the choice between stenting and CEA as a toss-up.
Finding that the choice between these 2 therapeutic options is a toss-up considering both clinical and QOL outcomes means primarily that, for the clinician, whose primary role is to provide expert advice to the patient, there is no “wrong decision.” That can be particularly difficult to convey to patients who want the “best therapy,” and it is important to recognize that we should not be asking patients to choose whether they would rather face a slightly higher risk of having a stroke or an MI and pick their treatment on that basis. Neither risk can be completely avoided as both events occur after each procedure, but the incremental risks are too low to affect the expected QOL results of either treatment. Patients may still not be indifferent to the choice of therapy, due to preferences for the less invasive nature of stenting with less pain and a shorter hospital stay or concerns regarding unresolved questions about the long-term cognitive effects of microembolic events during stenting, even with protection devices (11). And payers may not be indifferent to the modestly higher costs of stenting (12). In the end, patients need a balanced presentation of risks and benefits framed in a neutral way. Knowledge of local experience (procedure volumes and complication rates) can be particularly helpful, but accurate data are often difficult to obtain and to interpret properly.
The largest area of uncertainty in the contemporary management of significant carotid disease is actually not which revascularization procedure to select but whether patients who are asymptomatic need revascularization at all. Such patients comprise >50% of current carotid revascularization procedures each year, and the modest outcome benefits seen previously in this cohort may have been eliminated by interval advances in medical therapy (13). Given that new comparative effectiveness evidence will likely continue to employ various forms of composite clinical endpoint variables for efficiency purposes, also including the patient's perspective in the form of patient-reported outcomes will be crucial in being prepared to take that next step: translating evidence into practice.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
↵⁎ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
- American College of Cardiology Foundation
- Blackshear J.L.,
- Cutlip D.E.,
- Roubin G.S.,
- et al.
- Rahimi K.,
- Malhotra A.,
- Banning A.P.,
- Jenkinson C.
- Cohen D.J.,
- Stolker J.M.,
- Wang K.,
- et al.,
- CREST Investigators
- De Rango P.,
- Caso V.,
- Leys D.,
- Paciaroni M.,
- Lenti M.,
- Cao P.
- Abbott A.L.