Early Outcomes After Carotid Artery Stenting Compared With Endarterectomy for Asymptomatic Carotid Stenosis
Background and Purpose—Despite the absence of definitive data from randomized clinical trials on the comparative effectiveness of carotid artery stenting (CAS) versus carotid endarterectomy (CEA) for asymptomatic carotid stenosis, the use of CAS has been expanding and seems to be displacing the use of CEA in some parts of the United States.
Methods—We used comprehensive hospital discharge data from January 2010 to December 2012 to identify patients who had CEA or CAS for asymptomatic carotid stenosis at all academic medical centers that participate in the University HealthSystem Consortium. In-hospital death and postoperative stroke after CAS and after CEA were compared using multivariable logistic regression, propensity score matching, and a grouped-treatment approach using multilevel mixed-effects models to adjust for baseline characteristics of patients selected for these procedures.
Results—We identified 17 716 patients with asymptomatic carotid stenosis treated with CEA and 3962 treated with CAS at 186 University HealthSystem Consortium hospitals. Postoperative stroke or in-hospital death was more frequent after CAS (4.0% versus 1.5%; P<0.001), and patients with CAS were more likely to have these adverse outcomes even after adjusting for baseline characteristics using multivariable analysis (odds ratio [OR], 2.5; 95% confidence interval [CI], 2.1–3.1; P<0.001) and propensity score matching (OR, 2.5; 95% CI, 1.9–3.4; P<0.001). In a multilevel mixed-effects model, hospitals that performed a higher proportion of all carotid revascularization cases using CAS had significantly higher rates of adverse outcomes (OR, 3.7; 95% CI, 1.8–7.6; P<0.001) after adjusting for patient-level variables.
Conclusions—For asymptomatic carotid stenosis, CAS is associated with a substantially higher risk of postoperative stroke or in-hospital death than CEA even after adjustment for baseline differences in hospital and patient characteristics.
See related article, p 9.
Asymptomatic carotid stenosis is common, particularly among elderly patients.1 On the basis of several randomized clinical trials, current guidelines recommend prophylactic carotid endarterectomy (CEA) in highly selected patients with severe asymptomatic carotid stenosis when the morbidity and mortality risk from surgery is <3%.2–5 Because the annual risk of stroke is lower for asymptomatic carotid stenosis (≈2%)6 than for symptomatic carotid stenosis, the clinical benefit for carotid intervention for asymptomatic carotid stenosis within the context of improved medical therapies is uncertain.7
Carotid artery stenting (CAS) is a newer option for treating carotid stenosis but has been associated with an increased risk for postoperative stroke for patients with symptomatic carotid stenosis.8–10 A recent pooled analysis of data from the Symptomatic Severe Carotid Stenosis (EVA-3S) trial, the Stent-Protected Angioplasty versus Carotid Endarterectomy (SPACE) trial, and the International Carotid Stenting Study (ICSS) reported a significantly increased risk of periprocedural stroke or death with CAS when compared with CEA for symptomatic carotid stenosis.11 Nevertheless, the use of CAS in the United States has been expanding from <3% of all carotid artery revascularization procedures in 1998 to 13% in 2008.12 Regional variation in the use of CAS has also increased, and CAS seems to be displacing CEA in some parts of the United States.13
Until additional evidence from large-scale randomized clinical trials becomes available, observational data may provide practical information to inform clinical practice, provided that confounding based on baseline differences in patients selected for these procedures can be minimized. Conventional multivariable analysis can adjust for measured confounders, propensity score matching attempts to reduce confounding by matching groups based on the estimated probability of receiving the treatment, and grouped-treatment analysis is a form of instrumental variable analysis that can potentially mitigate the effect of confounding by indication14,15 by taking advantage of practice variation in the proportion of patients treated with CAS when compared with those treated with CEA at the hospital-level. The aim of current study is to compare early outcomes of patients with asymptomatic carotid stenosis treated with CAS and CEA by applying multivariable analysis, propensity score matching, and multilevel analysis techniques to reduce confounding and allow for a comparison of the effectiveness of these 2 approaches.
Study Design and Data Source
We conducted a retrospective cohort study using the University HealthSystem Consortium (UHC)’s Clinical Data Base. This is a comprehensive database with discharge and line-item data on charges from 120 academic centers and 301 affiliated hospitals in United States. The local institutional review board approved this study.
We identified adults who underwent carotid revascularization procedures (either CEA or CAS) from January 2010 to December 2012 using primary and secondary procedure codes (International Classification of Disease, Clinical Modification-9th Revision code 00.63 for CAS and 38.12 for CEA).16 We then excluded the patients in following sequence: (1) patients aged <18 years, (2) patients undergoing ≥2 same carotid revascularization procedures during the study period, (3) patients who underwent both CEA and CAS during same admission or within 30 days of the index admission, (4) patients with Coronary artery bypass grafting (36.10–36.19),17 cardiac valve surgery (35.20–35.28), or percutaneous coronary intervention (00.66, 36.01, 36.02, 36.05–36.07, 36.09)18 during the same admission, (5) symptomatic presentation, and (6) emergent or urgent admission. We abstracted patient-level information on (1) demographic characteristics, (2) characteristics of the index admission (admission year, symptom status [symptomatic versus asymptomatic], and discharge disposition), (3) 29 comorbidities as defined by the Agency for Healthcare Research and Quality,19 and (4) 4 other comorbidities or conditions that are not included in the Agency for Healthcare Research and Quality measure. Symptom status was defined according to previously validated criteria.16 We also identified patients who met certain high-risk criteria for CEA that would be eligible for CAS based on age >80 years, congestive heart failure, coronary artery disease, or chronic lung disease.20 Detailed information on the methods used to identify comorbidities and symptom status are provided in the online-only Data Supplement.
Using hospital identifiers, we determined the total annual volume of CAS and CEA procedures performed at each hospital. We then calculated the proportion of all carotid revascularization procedures at each hospital that were performed using CAS versus CEA.
We abstracted information on the specialty of the proceduralist for each carotid revascularization procedure. The annual volume of carotid revascularization procedures for each physician and the total procedural volumes and unadjusted outcomes were aggregated by physician specialty.
The primary outcome was a composite of postoperative stroke or in-hospital death during the index admission. The secondary outcome was any postoperative stroke, myocardial infarction (MI) or in-hospital death during the index admission. For a sensitivity analysis, we also included readmissions to UHC hospitals for stroke or MI that occurred within 30 days after discharge from the index admission. The specific methods used to define each outcome are provided in the online-only Data Supplement.
We first calculated unadjusted outcomes after CAS and CEA. We then performed multivariable logistic regression with the primary or secondary outcome as the dependent variable. Because the baseline characteristics of patients with CAS and CEA were expected to be different, we then performed a sensitivity analysis using propensity score matching (online-only Data Supplement). Finally, for the grouped-treatment approach, we used a multilevel mixed-effects model to evaluate the independent effects of hospital-level variables, including the proportion of procedures, that were performed using CAS at each hospital and total volume of carotid revascularization performed per hospital on the outcomes, while also adjusting for patient-level variables including age, sex, race, admission type, and comorbidities.21
We estimated odds ratios (ORs) and 95% confidence intervals for each outcome. For all analyses, a P value of ≤0.05 was defined as statistically significant; all tests were 2-sided. Statistical analyses were performed using Stata 13 (StataCorp LP, College Station, TX).
Patients and Hospital-Level Characteristics
From January 2010 to December 2012, we identified 34 592 discharges where a CAS or CEA was performed, which corresponded to 32 495 individuals treated at 188 UHC hospitals. After applying the exclusion criteria, our final analysis cohort included 21 678 patients (Figure I in the online-only Data Supplement).
Of these 21 678 patients with asymptomatic carotid stenosis, 3962 patients (18.3%) were treated with CAS and 17 716 (81.7%) were treated with CEA. When compared with CEA patients, CAS patients were more likely to be younger, to be black, and to have comorbidities including coronary artery disease, peripheral artery disease, chronic kidney disease, and heart failure. However, CAS patients were less likely to have hypertension, hyperlipidemia, and smoking than CEA patients. Overall, 59% of patients were categorized as high risk and therefore eligible for CAS as defined above. CAS patients were more likely to meet these high-risk criteria (Table 1).
The median annual volume of CAS procedures per physician was 1.5 (interquartile range, 1–3) compared with a median of 3 (1–7.3) for CEA. CAS was performed most often by vascular surgeons (24.9%), followed by cardiologists (22.8%), neurosurgeons (16.4%), and radiologists (12.3%). For CEA, vascular surgeons (61.2%) also performed most of these procedures, followed by general surgeons (15.0%), neurosurgeons (7.1%), and cardiothoracic surgeons (4.5%; Table I in the online-only Data Supplement).
The median volume of CAS, CEA, and total carotid revascularization procedures per hospital per year was 4.2 (interquartile range, 0–10.7), 20.2 (8.0–44.0), and 26.7 (9.7–58.3), respectively. The median proportion of carotid revascularization procedures performed with CAS per hospital was 14.0% (interquartile range, 0%–26.2%). Hospitals that performed a higher annual volume of carotid revascularization procedures also tended to perform a higher proportion of these procedures with CAS (ρ=0.34; P<0.001).
Patients admitted to hospitals that performed a higher proportion of all carotid revascularization procedures with CAS were more likely to have coronary artery disease, smoking, congestive heart failure, and renal failure, but were less likely to be older, white, and to have hypertension and hyperlipidemia (Table II in the online-only Data Supplement). Hospitals with a higher annual volume of carotid revascularization procedures per hospital (P<0.001) tended to have a high proportion of high-risk patients. However, the proportion of high-risk patients treated at each hospital was not correlated with a higher proportion of carotid revascularization procedures done with CAS at that hospital (P=0.88).
In the unadjusted analysis, postoperative stroke or in-hospital death was more frequent after CAS than after CEA (4.0% CAS versus 1.5% CEA; P<0.001). Both postoperative stroke (3.7% CAS versus 1.5% CEA; P<0.001) and readmission for stroke within 30 days (0.8% CAS versus 0.2% CEA; P<0.001) were more frequent with patients with CAS than with CEA. In-hospital mortality was also higher for CAS (0.3% CAS versus 0.1% CEA; P<0.001).
The postoperative MI rate or early readmission for MI rate was not significantly different in the 2 groups (Table 2). High-risk patients did not have significantly higher rates of in-hospital death or postoperative stroke, but did have higher rates of postoperative MI or 30-day readmission for MI (Table III in the online-only Data Supplement).
After adjustment for age, sex, race, and multiple comorbidites (hypertension, coronary artery disease, atrial fibrillation, peripheral artery disease, smoking status, congestive heart failure, hyperlipidemia, diabetes mellitus, chronic kidney disease, and chronic lung disease), the odds of postoperative stroke or in-hospital death after CAS remained significantly higher than after CEA (OR, 2.5; 95% CI, 2.1–3.1; P<0.001). A sensitivity analysis using propensity score matching produced similar results (OR, 2.5; 95% CI, 1.9–3.4; P<0.001).
In a similar series of analyses using a secondary outcome that includes MI and, postoperative stroke, or in-hospital death, adverse outcomes were significantly more likely after CAS when compared with CEA using both multivariable and propensity score matching analysis (Table 3).
Hospital-Level and Physician-Level Analyses
In an unadjusted analysis, hospitals that performed a higher proportion of carotid revascularization cases with CAS also had a higher rate of postoperative stroke or in-hospital death by trend analysis (Figure [A]; P=0.001). Hospitals with a lower volume of carotid revascularization procedures had an increased risk of postoperative stroke or in-hospital death (Figure [B]; P=0.002).
At the physician-level, there seemed to be evidence of a volume–outcome relationship for CEA but not for CAS: a higher annual volume of CEA procedures was associated with a lower risk of postoperative stroke or in-hospital death (P<0.001), but a higher annual volume of CAS procedures was not associated with lower postoperative stroke or in-hospital death rates (P=0.51; Figure II in the online-only Data Supplement).
Postoperative stroke or in-hospital death rates after CEA performed by vascular surgeons, general surgeons, neurosurgeons, and cardiothoracic surgeons were all <3%. In contrast, the adverse outcome rate after CAS for vascular surgeons, neurosurgeons, radiologists, general surgeons, and neurologists were all >3%. Cardiologists had an adverse event after CAS of 2.6%. Postoperative stroke or death rates varied by physician specialty for CEA (P<0.001, χ2 test) and for CAS (P=0.013; Table I in the online-only Data Supplement).
Using a multilevel mixed-effects model, we found that hospitals that performed a higher proportion of carotid revascularizations procedures with CAS had significantly higher postoperative stroke or in-hospital death rates (OR, 3.7; 95% CI, 1.8–7.6; P<0.001) even after adjustment for patient-level variables including age, sex, race, and comorbidities. The OR reported refers to the expected increase in the odds of an adverse outcome for a hospital that converted from exclusive use of CEA to exclusive use of CAS. Including MI as an adverse outcome produced similar results (Table 4). The volume–outcome relationship seen at the physician-level was not demonstrated at the hospital-level: A lower annual volume of carotid procedures at the hospital-level was not significantly associated with an increased risk in postoperative stroke or in-hospital death (Table 4).
Inclusion of 30-Day Readmissions for Stroke or MI
To account for the possibility of delayed MI and stroke after the index hospitalization, we conducted a sensitivity analysis that included readmissions because of stroke or MI that occurred within 30 days after discharge from the index hospitalization to UHC hospitals. The unadjusted multivariable analysis, propensity score matching, and grouped-treatment analyses all produced similar results (Tables IV and V in the online-only Data Supplement).
In this analysis of 21 678 patients with asymptomatic carotid stenosis treated at 186 US academic centers and their affiliated teaching hospitals, we found that CAS was associated with a >2-fold greater odds of postoperative stroke or death than CEA, an effect that remained robust after applying multivariable adjustment and propensity score matching which are designed to adjust for baseline differences in the patients selected for these 2 procedures. In addition, using a grouped-treatment analysis approach that takes advantage of variations in practice among the UHC hospitals and which is designed to mitigate the effect of confounding by indication, we found more frequent adverse outcomes after CAS may not be fully explained by differences in the baseline characteristics of patients chosen for these procedures.
Despite the fact that >90% of carotid revascularization procedures are performed for patients with asymptomatic carotid stenosis,22,23 and that a substantial proportion of these procedures are currently done using CAS, there is no high-level evidence that supports the use of CAS for these patients. Approximately one fifth of all carotid revascularization procedures were performed with CAS in our study, which mirrors an overall trend of toward increasing numbers of CAS procedures performed at US academic centers.
The absolute perioperative risks we observed are particularly important when considering whether CAS should be applied for asymptomatic carotid stenosis because there is a lower absolute risk of stroke than for symptomatic carotid stenosis,6 and previous studies have suggested that prophylactic revascularization for asymptomatic carotid stenosis may not have clear advantages over medical management when the procedural risk exceeds 3%.5 We found an absolute postoperative risk of stroke or in-hospital death for CAS of 4.0%, which is significantly higher than for CEA and would exceed the typical threshold recommended for a treating asymptomatic carotid stenosis.5
Accordingly, we also found that hospitals that performed more carotid procedures using CAS had significantly increased risk of adverse outcomes, which is consistent with previous reports.24,25 The effect was rather large, with an expected 3.7-fold increase in the odds of postoperative stroke or in-hospital death if a hospital were to move from exclusive use of CEA to exclusive use of CAS and which again reflects adjustment and steps to mitigate confounding by indication.
Our study applied a grouped-treatment analysis approach, which has the potential to produce more reliable estimates of the differences in outcomes after CEA and CAS than standard models which only provide for adjustment at the patient-level and do not directly address confounding by indication. Because we found remarkably consistent effect sizes using multivariable analysis, propensity score matching, and grouped-treatment analyses, the increased risk of CAS found in this study may more reliably reflect the true comparative risk of postprocedural stroke or death for asymptomatic carotid stenosis in real-world practice.
Although we found that hospitals where fewer carotid revascularization procedures were done did not have a significantly increased risk of postoperative stroke or death after adjustment, previous data have suggested that experience or procedural volume for both CEA and CAS is associated with improved outcomes.26–29 A recent pooled analysis examined the association between operator experience and 30-day risk of stroke or death in the Carotid Stenting Trialists’ Collaboration database and found that the 30-day risk of stroke or death was significantly higher in patients treated by operators with low (mean ≤3.2 procedures/y; risk 10.1%) and intermediate annual in-trial volumes (3.2–5.6 procedures/y; 8.4%) compared with patients treated by high annual in-trial volume operators (>5.6 procedures/y; 5.1%).30 That we did not find a similar effect in our multilevel analysis could be explained by the positive correlation between the annual volume of carotid revascularization procedures and the proportion of CAS procedures at a given hospital. Because the median annual volume of CAS procedures per physician was low, the increased risk of adverse outcome in patients undergoing CAS might be at least partially explained by the low procedural volume for many interventionists that perform CAS. Overall, because the annual caseload for CAS was low in this study, so the failure to define a clear volume–outcomes relationship should be interpreted with caution.
This study should be interpreted in the context of several limitations. Despite our multiple analytic methods, the potential for residual confounding by indication remains. Coding errors and differences in coding between hospitals on diagnosis, comorbidities, procedures, and outcomes could affect our results, particularly if they are differences for patients who are selected for the 2 carotid revascularization procedures. Our ability to establishing symptom status based on diagnosis codes may also be limited. For outcome measures, readmissions to non-UHC hospitals were not captured and could bias the results if patients with CAS were more or less likely to be readmitted to a UHC hospital than patients with CEA, for instance, although including readmissions as an outcome in our sensitivity did not seem to affect the overall study results. Because detailed angiographic information was not available with the discharge data, we used clinical criteria for identifying high-risk patients, which could lead to residual confounding although recent subgroup analysis has suggested that angiographic risk factors were not associated with increased outcome in SPACE trial.31 Finally, our study is based on data from the academic medical centers and their affiliated hospitals in United States, and these results may not necessarily generalize to all other practice settings.
Despite these limitations, we found a substantially higher risk of postoperative stroke or in-hospital death for CAS when compared with CEA for asymptomatic carotid stenosis, with rates exceeding those that justify any intervention and despite multiple methods intended to adjust for confounding by indication. These data are consistent with the view that widespread prophylactic use of CAS for asymptomatic carotid stenosis is not justified without additional evidence of clear benefit and provides further justification for additional randomized clinical trial data for this indication.
Sources of Funding
This work was supported by the 2014 scientific promotion program from Jeju National University (J.C.C.).
Guest Editor for this article was Bo Norrving, MD, PhD, FESO, FAHA.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.006209/-/DC1.
- Received May 20, 2014.
- Revision received September 16, 2014.
- Accepted September 29, 2014.
- © 2014 American Heart Association, Inc.
- de Weerd M,
- Greving JP,
- de Jong AW,
- Buskens E,
- Bots ML
- Goldstein LB,
- Bushnell CD,
- Adams RJ,
- Appel LJ,
- Braun LT,
- Chaturvedi S,
- et al
- International Carotid Stenting Study Investigators,
- Ederle J,
- Dobson J,
- Featherstone RL,
- Bonati LH,
- van der Worp HB,
- et al
- Goodney PP,
- Travis LL,
- Malenka D,
- Bronner KK,
- Lucas FL,
- Cronenwett JL,
- et al
- Johnston SC
- Johnston SC,
- Henneman T,
- McCulloch CE,
- van der Laan M
- Shahian DM,
- Silverstein T,
- Lovett AF,
- Wolf RE,
- Normand SL
- Nallamothu BK,
- Lu X,
- Vaughan-Sarrazin MS,
- Cram P
- 20.↵Centers for Medicare and Medicaid Services. National Coverage Determination (NCD) for Percutaneous Transluminal Angioplasty (PTA). Concurrent with Carotid Stent Placement in Patients at High Risk for Carotid Endarterectomy (CEA). http://www.cms.gov/medicare-coverage-database/details/ncd-details.aspx?NCDId=201&ncdver=9&NCAId=62&NcaName=Percutaneous+Transluminal+Angioplasty+(PTA)+of+the+Carotid+Artery+Concurrent+with+Stenting&IsPopup=y&bc=AAAAAAAAEAAA&. Accessed June 25. 2014.
- Wang FW,
- Esterbrooks D,
- Kuo YF,
- Mooss A,
- Mohiuddin SM,
- Uretsky BF
- Calvet D,
- Mas JL,
- Algra A,
- Becquemin JP,
- Bonati LH,
- Dobson J,
- et al
- Stingele R,
- Berger J,
- Alfke K,
- Eckstein HH,
- Fraedrich G,
- Allenberg J,
- et al