Long-Term Outcome After Carotid Artery Stenting
A Population-Based Matched Cohort Study
Background and Purpose—Long-term outcome after carotid artery stenting (CAS), a less invasive technique than carotid endarterectomy (CEA), for prevention of stroke, is unclear. The aim was to assess long-term outcomes after CAS, compared with CEA, in a nationwide cohort study.
Methods—All patients registered in the national Swedish Vascular Registry (Swedvasc) treated with primary CAS between 2005 and 2012 were identified. For every CAS, 2 CEA controls, matched for sex, age, procedure year, and indication (symtomatic/asymtomatic), were chosen. Postoperative stroke was identified by cross-matching the cohort with the InPatient Registry and charts review. Primary end point was ipsilateral stroke or death >30 days postoperatively.
Results—A total of 1157 patients were included, 409 CAS and 748 CEA; 73% men with mean age 70 years and 69% were symptomatic. Risk factor profile was similar between the 2 groups. Median follow-up time was 4.1 years. Ipsilateral stroke or death of >30 days postoperatively occurred in 95 of 394 in the CAS group versus 120 of 724 in the CEA group (adjusted hazard ratio, 1.59; 95% confidence interval, 1.15–2.18). The corresponding adjusted rates for death, ipsilateral stroke of >30 days, and any stroke or death of >30 days were 25.7% versus 18.6% (hazard ratio, 1.20; 95% confidence interval, 0.84–1.72), 9.4% versus 2.9% (hazard ratio, 3.40; 95% confidence interval, 1.53–7.53), 34.2% versus 23.6% (hazard ratio, 1.49; 95% confidence interval, 1.10–2.00) for the CAS group versus CEA group, respectively.
Conclusions—In this nationwide cohort study, CAS was associated with an increased long-term risk of ipsilateral stroke and death during after the perioperative phase when compared with CEA.
Carotid artery stenting (CAS) is a less invasive technique than carotid endarterectomy (CEA) for prevention of stroke in patients with carotid artery stenosis. Although reports from high-volume centers have demonstrated acceptable short-term results after CAS, randomized controlled trials (RCT) have demonstrated inferior results, in terms of stroke and death in the perioperative phase, after CAS compared with CEA.1–4 A sustained benefit of CEA, compared with CAS, after long-term follow-up was observed in the RCTs. However, excluding the perioperative period, there has been little difference in outcomes between CEA and CAS in the RCTs.3,5–8
Little is known about the long-term results after CAS outside the RCTs, and consequently the generalizability of these observations has been questioned,9 and a recent Cochrane review pointed out the need for long-term outcome data after CAS.10 Population-based, nonselected cohorts can provide additional evidence.11
Our aim was to assess population-based, nationwide, long-term outcomes after CAS when compared with CEA during the same time period.
We established a nationwide cohort of all patients registered in the National Swedish Vascular Registry (Swedvasc) treated with CAS between 2005 and 2012 and matched controls treated with CEA during the same time period. Postoperative stroke was identified by cross-matching the cohort with the national InPatient Registry (IPR) supplemented with charts review. The study was approved by the regional ethics committee in Stockholm (EPN dnr 2012/1314–31), and written informed consent was not required.
Registry Data and Definitions
The Swedvasc registry has national coverage since 1994 and has been described previously.12 Follow-up includes clinical examination and registration of perioperative complications at 30 days by a vascular surgeon or neurologist, depending on local practice. Mortality data in Swedvasc are retrieved directly from the Swedish National Population Registry and is 100% accurate. A recent independent validation of Swedvasc demonstrated an external validity of 100% and an internal validity of 97.4% for carotid interventions.13
Patients retrieved from Swedvasc were cross-matched with the IPR. Cross-matching of registries is possible because every citizen in Sweden has a unique personal identity code. Hereby, all patients in our cohort and later admitted to hospital with a stroke diagnosis in the IPR during follow-up were identified. We also linked episodes from IPR with a diagnosis of atrial fibrillation.
We defined stroke as any new or worsened focal neurological deficit (both minor and major stroke) lasting for >24 hours. Any stroke was defined as ipsilateral, contralateral, or vertebrobasilar stroke, including intracerebral hemorrhage, whereas subarachnoid and epidural hemorrhages were not included in the analysis. Our definition is narrower than the updated stroke criteria from American Heart Association/American Stroke Association and restricted to the definition of ischemic stroke and stroke caused by intracerebral hemorrhage.15 Consequently, transitory ischemic attacks and asymptomatic lesions on computed tomographic scan or magnetic resonance imaging were not classified as stroke.
Baseline data shown in Table 1 were collected from Swedvasc. Heart risk was defined as current heart failure, angina pectoris or history of myocardial infarction, coronary bypass or percutaneous coronary intervention. Before May 2008, the Swedvasc definition included atrial fibrillation. Because of this inconsistency data on atrial fibrillation during follow-up were collected; International Classification of Diseases-code I48 was used to identify patients with atrial fibrillation from the IPR. On patients with a stroke event during follow-up, we did a chart review to investigate medical treatment.
Renal insufficiency was defined as s-creatinine of >150 µmol/L. Hypertension was defined as on antihypertensive medication. Diabetes mellitus when treated with insulin or oral medication. Smoking was categorized into either current smoker (including quitted within 4 weeks before operation) or nonsmoker (including never- and ex-smokers). Patients with an ipsilateral neurological event within 180 days before the index date were classified as symptomatic and all others as asymptomatic.
All patients who had CAS performed for symptomatic or asymptomatic internal carotid artery stenosis from January 1, 2005, to December 31, 2012 were included in the cohort (Figure 1). Only primary procedures were included, and a 7-year look-back period was used to identify patients earlier treated with any carotid procedure; all of whom were excluded. Patients operated bilaterally were identified, and only the first operation was used for analysis. Patients operated bilaterally on the same day were excluded. Only patients treated with bare-metal stents were included, whereas covered stent grafts or percutaneous transluminal angioplasties alone were excluded. Thereby, 409 cases with primary CAS were identified. The date of the carotid procedure was defined as index date.
We used patients treated with CEA as controls and aimed for 2 controls per CAS patient, 339 patients had 2 controls and 70 had 1 control resulting in 748 controls. We did exact matching with respect to sex, indication (transitory ischemic attack, amaurosis fugax, minor stroke, or asymptomatic), procedure year, and age (allowing for ±2 years).16 Other risk factors, such as atrial fibrillation and diabetes mellitus, were not used for matching the groups.
Strokes During Follow-Up
International Classification of Diseases-codes I60–I69 were utilized to identify all postoperative strokes. Screening for International Classification of Diseases-codes G45 and G46 was performed to identify any minor stroke misclassified as transitory ischemic attack. Charts and radiological reports from in-hospital episodes were retrieved and scrutinized by 2 independent investigators (M.J. and J.M.) to confirm the stroke diagnosis, to determine the brain territory affected, and to assess the severity of the stroke. We used the modified Rankin Scale (mRS), which was the best estimate of stroke severity we could obtain from the charts. The mRS measures the global disability and need for assistance after a stroke and was categorized from the hospital charts.17
Two authors (D.L. and P.G.) blinded to treatment type, reviewed events with ambiguous diagnosis, and classified the event as a stroke or not (Figure I in the online-only Data Supplement).
The primary end point was ipsilateral stroke and death from 31 days after index date to end of follow-up (December 31, 2012). Secondary outcomes were all-cause mortality, ipsilateral stroke or death, and any stroke or death from index date, and ipsilateral stroke and any stroke or death for >30 days postoperatively. We also designed a competing risk analysis for ipsilateral stroke or death from 31 days postoperatively as competing risk.18,19
Continuous data with normal distribution are presented with mean and SD and non-normally distributed with median and interquartile range. Fishers exact and the χ2 test was used for categorical data. t test was used for continuous data. Wald–Wolfowitz Runs test was used to compare the distribution of mRS. We calculated the 5-year cumulative incidence of outcomes and the corresponding 95% confidence intervals (CIs)20 from the Kaplan–Meier estimates and compared CAS with CEA using the Mantel–Haenszel log-rank test. Pointwise CIs for Kaplan–Meier estimates was calculated by the Kalbfleisch and Prentice21 method. We also calculated 95% CIs for the cumulative incidence at 5 years using bootstrap with 10 000 replications to ascertain the precision of the CIs. To estimate the cause-specific stroke rates, we used the cumulative incidence function, which is the probability of stroke in the presence of no other risks.22 Cumulative incidence function in CAS was compared with CEA using Gray test and calculated pointwise 95% CIs.23
Standard Cox proportional hazards model was used for outcomes, which included death as an event, and an extension of the Cox proportional hazard model (the Fine–Gray competing risk model)24 to estimate the subdistribution hazard ratio (HR) for outcomes including stroke as a single event. Models were adjusted for prespecified covariates known to affect outcome after CAS and CEA; diabetes mellitus, hypertension, cardiac disease, atrial fibrillation, current smoking, pulmonary disease, and renal insufficiency.25,26 All models accounted for matching design by stratification or adjustment for the matching variables. Patients who survived without an event before the end of follow-up were censored. Both visual inspection of log–log plots and including a time-dependent interaction term in the model indicated that the proportional hazard assumption was not violated. All tests were 2 sided, and statistical significance was set at P<0.05. The statistical analysis was performed with IBM SPSS statistics version 22. In addition the software R 3.2.227 was used for the competing risk analysis23 and bootstrap CI calculations.
Our final study population consisted of 1157 patients with a primary procedure for carotid stenosis (409 CAS and 748 CEA; Figure 1). The majority (73%) was men, with a mean age of 70 years and 69% had a symptomatic carotid stenosis. The 2 treatment groups were exact matched with respect to the matching variables (age, sex, indication, and calendar year for carotid procedure; Table 1). The nonsignificant differences in the matched variables are because of the fact that not all patients had 2 controls. Risk factor profile was similar between the 2 groups, but diabetes mellitus was more common in the CAS group and there was a trend of more cardiac and pulmonary disease in the CAS group. Atrial fibrillation at baseline was more common in the CAS group, 15.4% (95% CI, 12.2–19.1) versus 10.2% (95% CI, 8.1–12.5). However, at the time of stroke events of >30 days after surgery, there were no differences in history of atrial fibrillation between the CAS group and the CEA group. Among the CAS patients who had an ipsilateral stroke, 21% (95% CI, 9–39) had a history of atrial fibrillation versus 25% (95% CI, 10–46) in the CEA group.
Also treatment with any antithrombotic agent at the time of stroke event was similar: 93% (95% CI, 79–98) of CAS patients and 95% (95% CI, 79–99) of CEA patients (Table I in the online-only Data Supplement.
The median follow-up time for the primary end point was 4.1 years (interquartile range 2.4–5.8), equivalent to 3994 person-years of observation. The shortest and longest event-free follow-up was 4 and 2912 days, respectively. During follow-up, the study population had 394 hospitalization episodes with a stroke or transitory ischemic attack diagnosis registered in the IPR (from day 1). One new stroke can generate several episodes with a stroke diagnosis in the IPR, as each hospital department reports independently to IPR. For example, first in neurology department, later in internal medicine, and finally in rehabilitation department will generate 3 episodes with a stroke diagnosis in the IPR. We managed to obtain charts for 393 of these 394 episodes (99.7%) among 174 patients. Twelve patients had recurrent strokes with 1 stroke ipsilateral and 1 contralateral in the same patient and were accordingly included in the analysis of both ipsilateral and any stroke.
All strokes registered in the 30-day follow-up in Swedvasc was also found in the IPR, and IPR did not identify any additional early strokes.
Events and Stroke Risk
Within 30 days, there was no difference between the groups with respect to any stroke or death: 14 patients in the CAS group (3.4%) and 24 patients in the CEA group (3.2%), P=0.86. Ipsilateral stroke within 30 days was also similar between the groups: 2.7% and 2.0%, respectively (complete 30-day outcome is available in Table II in the online-only Data Supplement).
After the perioperative phase, 95 patients treated with CAS and 120 with CEA had a new ipsilateral stroke or died, corresponding to a cumulative incidence at 5 years of 30.8% (95% CI, 25.3–37.1) in CAS patients compared with 20.7% (95% CI, 17.2–24.7) in CEA patients.
The majority of strokes were ischemic. The proportion of hemorrhagic strokes was higher in the CEA group (10%; 95% CI, 3–21) than in the CAS group (4%; 95% CI, 1–14) although not significant (Table III in the online-only Data Supplement).
Patients treated with CAS had an overall higher risk (HR, 1.71; 95% CI, 1.26–2.31) for ipsilateral stroke or death after day 30 compared with those treated with CEA in crude analysis. The adjusted risk was essentially unchanged, HR 1.59 (95% CI 1.15–2.18).
After 30 days, 9.4% in the CAS group had a new ipsilateral stroke compared with 2.9% in the CEA group (adjusted HR, 3.40; 95% CI, 1.53–7.53). The combined end point any stroke or death after 30 days was also more common among patients treated with CAS than among patients treated with CEA: 34.2% versus 23.6% (HR, 1.49; 95% CI, 1.10–2.00).
All outcomes on stroke (both crude and adjusted) showed higher risk for the CAS group (Table 2). In contrast, all-cause mortality after 30 days was similar in both groups after adjustment for confounders (HR, 1.20; 95% CI, 0.84–1.72). The Kaplan–Meier plots for the risks of stroke and death up to 6 years are shown in Figure 2.
Competing Risk Analysis
The cause-specific risk for ipsilateral stroke from day 31 during follow-up (with death as competing event) was increased after CAS compared with CEA (adjusted subdistribution HR 2.72 (95% CI 1.54–4.81). The corresponding risk for death, (with stroke as competing risk), was not elevated after CAS compared to CEA (adjusted subdistribution HR, 1.16; 95% CI, 0.85–1.59; Kaplan–Meier plots for competing risk for ipsilateral stroke and death after 30 days are available in Table IV and Figure II in the online-only Data Supplement).
Severity of Stroke
The functional outcome at admission to hospital, measured by the mRS was mainly in the range 1 to 3 (corresponding to minor stroke) in both groups (Figure III in the online-only Data Supplement). The distribution of the mRS did not differ between the 2 groups.
In this nationwide cohort study, including all primary CAS procedures in Sweden, a substantially (59%) increased risk for late (>30 days) stroke or death after CAS compared with CEA was observed. The increased risk was valid for both ipsilateral and any stroke. There was no difference in all-cause mortality between the groups. Both standard and competing risk analysis consistently showed an elevated risk specifically for stroke. In fact, the cause-specific stroke risk was even higher in competing risk analysis (subdistribution HR, 2.14). The population-based design with controls matched for several confounders minimized the risk for bias, and the similar mortality rate (HR, 1.20; 95% CI, 0.84–1.72) and the small change in HR after additional adjustment in regression analyses indicate a low degree of bias.
The 5-year results of the EVA-3 S study (Endarterectomy Versus Angioplasty in Patients with Symptomatic Severe Carotid Stenosis) showed a higher incidence of any procedural stroke and ipsilateral stroke or death, but similar rates of ipsilateral stroke after the procedural period.5 In the CREST trial (Carotid Revascularization Endarterectomy Versus Stenting Trial) with a median follow-up of 7.4 years, there was no significant difference between CAS and CEA with respect to ipsilateral stroke in the postprocedural period.8 However, similar to our results, the ICSS (International Carotid Stenting Study) long-term follow-up, where patients were followed up for a median of 4.2 years, also had a higher incidence of total stroke after 30 days in the CAS group (HR, 1.53; 95% CI, 1.02–2.31). In the ICSS, the difference was mainly driven by strokes occurring in the contralateral or vertebrobasilar territory.6
In the SPACE trial (Stent-Protected Angioplasty Versus Carotid Endarterectomy), there were no significant difference between patients randomized to stenting or endarterectomy with respect to ipsilateral stroke between day 31 and 2 years.28
A recurrent concern for CAS is whether it could be associated with higher rates of late restenosis with a possible increased risk for subsequent late stroke. However, in the ICSS long-term follow-up study, no difference in the long-term rates of severe restenosis or occlusion was found.6 Although restenosis in general was a significant risk factor for recurrent ipsilateral stroke in the CREST trial, there was no difference in rate of restenosis between CAS and CEA (12.2% and 9.7%) after 7.2 years.8,29 By contrast, the CAVATAS trial (Carotid and Vertebral Artery Transluminal Angioplasty Study) reported 31% restenosis at 5 years after CAS compared with 10% after CEA.30 Also the SPACE trial and the EVA-3 S trial reported higher incidence of restenosis in the stented group than in the endarterectomy group (11% versus 5% at 2 years and 12% versus 5% at 3 years).28,31
There is no uniform protocol for follow-up after CEA or CAS in Sweden, and the Swedvasc registry does not include information about surveillance details after treatment. Accordingly, the natural course of recurrent stenosis after CAS is difficult to study in this setting. Nevertheless, restenosis may be one explanation for the inferior results of stenting shown here.
Our results differ from the long-term results in the RCTs and one can only speculate about mechanisms for this. One concern with the RCTs is that patients and centers are selected, causing questionable generalizability. A systematic review of registry data found higher stroke/death rates after CAS than after CEA in the perioperative period. The adverse events after CEA were similar in the registries as in the RCTs, whereas CAS had a significantly higher proportion of adverse events than RCTs.32
The present study has some limitations. The 2 groups, although matched, cannot compete with randomized data. The baseline characteristics were fairly similar in the 2 groups, but unmeasured differences between the groups may have affected the results, to some extent.
Another limitation is the lack of information on the rationale behind the chosen treatment modality. Indications, such as hostile neck or anatomically high stenosis, may confound our results. On the contrary, it is likely that such confounding factors would affect the short-term more than the long-term outcomes. Furthermore, there is the possibility that some patients having stroke during follow-up were not admitted to a hospital, and thereby missed. If so, it should be a low number and not related to type of treatment (ie, nondifferential misclassification).
The strengths of this study are the nationwide, real-world nature of the data combined with high reliability of mortality data and low number of missing data on stroke outcome (1/394). In addition, the size of the study makes the risk of a type II statistical error less likely. Adjusting for possible confounders at baseline, such as diabetes mellitus, heart disease, pulmonary disease, and atrial fibrillation did not alter the hazards substantially. The all-cause mortality did not differ between the groups, which supports that CAS patients were comparable with CEA patients at baseline. Moreover, there were no differences in antithrombotic medication or history of atrial fibrillation between the CAS and the CEA group at the time of stroke events of >30 days after surgery, thus a difference in strokes caused by cardioembolism is not likely.
This nationwide cohort study shows that CAS confers an increased long-term risk of stroke and death when compared to CEA. This increased risk is mainly explained by an increased rate of ipsilateral stroke after the periprocedural period, indicating that CAS is not as durable as CEA for treatment of carotid artery stenosis.
We thank Kristian Smidfelt Sahlgrenska, Samuel Ersryd Gävle, Lars Isaksson Linköping, Håkan Pärsson Helsingborg, Claes Skiöldebrand Södersjukhuset, Manne Andersson, Ryhov, Katarina Björses, Malmö, Marcus Lingman, Hallands, Pierre Cherfan, Eksjö, Håkan Lindvall, Karlskoga, Stefan Nilsson, Kristianstad, Görel Sundbäck, Hallands Kungsbacka, Ann Törnell, Kungälv, Jan Staaf, Köping, Ann-Sofi Ericsson Lindesberg, Martin Gunnarsson Örebro, and Sven Erik Persson Umeå for help with data collection.
Sources of Funding
The project was partly funded by the Swedvasc registry.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.116.013018/-/DC1.
- Received February 8, 2016.
- Revision received June 1, 2016.
- Accepted June 3, 2016.
- © 2016 American Heart Association, Inc.
- Ringleb PA,
- Allenberg J,
- Brückmann H,
- Eckstein HH,
- Fraedrich G,
- Hartmann M,
- et al
- Ederle J,
- Dobson J,
- Featherstone RL,
- Bonati LH,
- van der Worp HB,
- de Borst GJ,
- et al
- Mas JL,
- Arquizan C,
- Calvet D,
- Viguier A,
- Albucher JF,
- Piquet P,
- et al
- Bonati LH,
- Dobson J,
- Featherstone RL,
- Ederle J,
- van der Worp HB,
- de Borst GJ,
- et al
- Sacco RL,
- Kasner SE,
- Broderick JP,
- Caplan LR,
- Connors JJ,
- Culebras A,
- et al
- Peck J
- Banks JL,
- Marotta CA
- Andersen PK,
- Geskus RB,
- de Witte T,
- Putter H
- Wolkewitz M,
- Cooper BS,
- Bonten MJ,
- Barnett AG,
- Schumacher M
- Rothman KJ,
- Boice JD
- Kalbfleisch JD,
- Prentice RL
- Hoke M,
- Ljubuncic E,
- Steinwender C,
- Huber K,
- Minar E,
- Koppensteiner R,
- et al
- Cunningham EJ,
- Bond R,
- Mehta Z,
- Mayberg MR,
- Warlow CP,
- Rothwell PM
- 27.↵R Development Core Team. R: A language and environment for statistical computing [computer program]. Vienna, Austria: R Foundation for Statistical Computing, 2013. ISBN 3-900051-07-0, http://www.R-project.org/.
- Eckstein HH,
- Ringleb P,
- Allenberg JR,
- Berger J,
- Fraedrich G,
- Hacke W,
- et al
- Bonati LH,
- Ederle J,
- McCabe DJ,
- Dobson J,
- Featherstone RL,
- Gaines PA,
- et al
- Arquizan C,
- Trinquart L,
- Touboul PJ,
- Long A,
- Feasson S,
- Terriat B,
- et al