eTable 1. Predictors of Perioperative Stroke/Death After Redo CEA
eTable 2. Predictors of Perioperative Stroke/Death After Primary CEA
eTable 3. Predictors of 1-Year Stroke/Death After Redo CEA
eTable 4. Predictors of 1-Year Stroke/Death After Primary CEA
Customize your JAMA Network experience by selecting one or more topics from the list below.
Arhuidese IJ, Faateh M, Nejim BJ, Locham S, Abularrage CJ, Malas MB. Risks Associated With Primary and Redo Carotid Endarterectomy in the Endovascular Era. JAMA Surg. 2018;153(3):252–259. doi:10.1001/jamasurg.2017.4477
What are the risks associated with redo carotid endarterectomy relative to primary carotid endarterectomy?
In this cohort study of 64 118 carotid endarterectomies, redo carotid endarterectomy was associated with a 2.8 times increase in stroke and a 2.2 times increase in death among asymptomatic patients. There were no differences in outcomes among symptomatic patients.
These redo carotid endarterectomy and primary carotid endarterectomy results should inform patient and clinician expectations at the point of care.
Clinical experience suggests worse outcomes for redo carotid endarterectomy (CEA) relative to primary CEA. Objective quantification of the excess risk attributable to redo CEA in this era of proliferating endovascular therapy remains to be determined.
To evaluate the risks of redo CEA relative to primary CEA.
Design, Setting, and Participants
This study was a retrospective analysis of a prospective cohort of patients maintained by the Society for Vascular Surgery in the Vascular Quality Initiative between January 1, 2003, and April 30, 2016. The setting was consecutive patients from academic and community hospitals across the United States. Participants were patients who underwent primary CEA or redo CEA for symptomatic or asymptomatic carotid stenosis.
Primary CEA and redo CEA.
Main Outcomes and Measures
Stroke, death, myocardial infarction, stroke or death, and stroke, death, or myocardial infarction at 30 days and 1 year.
There were 64 118 CEAs recorded, including 62 749 primary CEAs (97.9%) (median age, 71 years; 39.5% female) and 1369 redo CEAs (2.1%) (median age, 71 years; 42.2% female). Comparing primary CEA vs redo CEA, the incidence of 30-day stroke was 0.7% vs 1.9% (P < .001) for asymptomatic patients and 1.5% vs 1.4% (P = .77) for symptomatic patients for an overall 0.9% vs 1.8% (P = .002). Incidences were 0.8% vs 1.3% (P = .09) for perioperative myocardial infarction, 0.6% vs 1.4% (P = .003) for death, and 1.5% vs 2.6% (P = .001) for stroke or death. Risk-adjusted 30-day stroke (odds ratio [OR], 2.82; 95% CI, 1.69-4.71; P < .001), death (OR, 2.15; 95% CI, 1.21-3.79; P = .009), and stroke or death (OR, 2.06; 95% CI, 1.32-3.23; P = .002) were higher for redo CEA compared with primary CEA among asymptomatic patients but were similar among symptomatic patients for stroke (OR, 0.79; 95% CI, 0.33-1.95; P = .62), death (OR, 1.32; 95% CI, 0.53-3.26; P = .55), and stroke or death (OR, 0.85; 95% CI, 0.39-1.82; P = .68). Stroke or death at 1 year among asymptomatic patients was higher for redo CEA (hazard ratio [HR], 1.36; 95% CI, 1.08-1.69; P = .009) but was similar among symptomatic patients (HR, 0.94; 95% CI, 0.67-1.31; P = .71).
Conclusions and Relevance
The outcomes of redo CEA and primary CEA fall within the professional guidelines for carotid revascularization. However, redo CEA is associated with a 2.8 times increase in the risk of stroke and a 2.2 times increase in the risk of death compared with primary CEA among asymptomatic patients. There is no significant difference in outcomes among symptomatic patients. These redo CEA and primary CEA results should inform patient and clinician expectations at the point of care.
Restenosis is a recognized complication of carotid endarterectomy (CEA). Results from the Carotid Revascularization Endarterectomy vs Stenting Trial (CREST) revealed a restenosis rate of 6.3% after CEA at 2 years.1 Other trials and cohort studies have reported post-CEA restenosis rates ranging from 3% to 17%, with some variability in the definition of restenosis and duration of follow-up.2-6 Redo CEA for the treatment of restenosis is often considered a high-risk procedure, largely due to technical challenges relating to fibrous scarring and anatomic distortions. Carotid angioplasty and stenting (CAS) has been offered as an alternative to CEA in this era of proliferating endovascular interventions. The comparison of outcomes between CEA and CAS for the treatment of post-CEA restenosis has attracted significant exploratory attention recently at the expense of evaluations of redo CEA relative to primary CEA. Recent studies7-10 have shown equivalent outcomes for redo CEA vs CAS for the treatment of post-CEA restenosis. The composite outcomes of CEA and CAS have also been shown to be equivalent for de novo carotid disease,11 as well as in patients with restenosis after prior CAS.12 These findings have established equipoise between CEA and CAS on multiple fronts, leaving room for treatment choices based on lesion characteristics, clinician experience, and patient preferences.
Multiple institutional series have shown good outcomes for redo CEA.13-18 However, these results might not be generalizable to a broader population of patients across the United States. Most important, the excess risk attributable to redo CEA deserves objective elucidation for prognostic purposes. In this study, we performed a population-based evaluation of outcomes after redo CEA relative to primary CEA in a large, nationally representative cohort of patients. We also identified predictors of adverse outcomes and potential targets for improvement.
We performed a retrospective analysis of all patients in the Vascular Quality Initiative (VQI) who underwent primary CEA or redo CEA between January 1, 2003, and April 30, 2016. The setting was consecutive patients from academic and community hospitals across the United States. The VQI is a prospectively maintained database approved by the Society for Vascular Surgery (SVS) for patient safety and quality improvement purposes. It stipulates entry of all vascular cases performed at each participating hospital, contains patient-specific and procedure-specific data from multiple sites across all regions of the United States, and incorporates data on mortality from the Social Security Death Index. At the end of the study period, there were more than 370 hospitals and 2800 participating physicians in the VQI. Details of the data collection and validation process have been published previously.19 The present study was approved by the VQI Research Advisory Committee, and the Johns Hopkins Institutional Review Board waived the need for individual patient consent under the provisions for deidentified participant and quality improvement research.
Patients who underwent primary CEA or redo CEA were identified directly from the variables recorded in the VQI database. The relevant patient-related and procedure-related data assessed are listed in Table 1. The presence of a preoperative symptom was denoted as the occurrence of ipsilateral ocular or cortical transient ischemic attack or stroke within 6 months before surgery. The interval between symptom and surgery was also examined. Degree of stenosis was the most severe stenosis obtained from duplex, magnetic resonance angiography, computed tomographic angiography, or angiogram. The total number of CEAs performed by each surgeon was computed, and operators were subsequently grouped into quartiles of case volume.
Postoperative (30-day) outcomes were ipsilateral stroke, death, myocardial infarction (MI), stroke or death, and stroke, death, or MI. One-year outcomes were stroke, death, and the composite of stroke or death. The individual and composite outcomes were chosen a priori based on clinical relevance. The composite outcome was included in this study because it is the counterfactual surrogate for “no adverse event after CEA,” which is the clinical objective of patient-centered care. Stroke was denoted as the occurrence of minor or major, ocular, or cortical stroke after surgery. Myocardial infarction was confirmed by electrocardiogram changes or critical elevation in troponins. Complications examined were cranial nerve injury, wound reexploration after closure, wound infection, return to the operating room for bleeding, neurologic or incision-related events, procedure-related arrhythmia, and reperfusion symptoms.
The variables studied were analyzed as categorical variables except for age, which was analyzed as a continuous variable. Normality was assessed using the Shapiro-Wilk test. Proportions and medians with interquartile ranges are reported; the χ2 test and Wilcoxon rank sum test were applied accordingly. Univariable and multivariable logistic regression analyses adjusting for demographic and operative characteristics were used to evaluate 30-day postoperative outcomes and identify their predictors. Kaplan-Meier estimates were computed to compare the survival function between the treatment groups. The log-rank test and Wilcoxon rank sum test were applied to test the equality of survival functions. Univariable and multivariable Cox proportional hazards regression models were used to analyze 1-year outcomes and identify their predictors. The variables included in the multivariable model were based on prior literature20,21 and the univariable analysis. Interaction terms were included to test for effect modification. Variance inflation factors were applied to test for multicollinearity. Statistical models were built in a stepwise fashion under the guidance of likelihood ratio tests and Akaike information criterion indexes, with a goal to achieve model parsimony. Clinically relevant variables and others known to influence outcomes after carotid revascularization were forced into the models obtained from the stepwise selection process, with no significant influence on the magnitude and statistical significance of the coefficients of interest. To adjust for potential institutional-level differences in our results, we performed post hoc hierarchical or mixed-effects modeling with a random intercept at the facility level. All analyses were performed using statistical software (Stata, version 14.1; StataCorp LLC), and statistical significance was accepted at 2-sided P < .05.
There were 64 118 CEAs performed between January 1, 2003, and April 30, 2016, in the VQI database. Of these procedures, 62 749 (97.9%) were primary CEAs, while 1369 (2.1%) were redo CEAs. Detailed patient characteristics are listed in Table 1. The proportion of patients who underwent redo CEA remained largely similar between the first 4 years of the study (2.1%, 2.8%, 2.0%, and 1.4%) and the last 4 years of the study (2.2%, 1.9%, 2.1%, and 1.8%), indicating minimal change in the prevalence of redo CEA over time (P = .16). Carotid endarterectomies were performed most commonly in asymptomatic patients (70.0% primary CEA vs 68.1% redo CEA, P = .11) for stenosis greater than 80% (61.7% primary CEA vs 65.1% redo CEA, P = .001). The mean follow-up duration was 206 (95% CI, 204-207) days for primary CEA and 217 (95% CI, 205-229) days for redo CEA (P = .79).
Overall, 617 patients (1.0%) had ipsilateral stroke within 30 days of their procedure (1.0% primary CEA vs 1.8% redo CEA, P = .002). For asymptomatic patients, the incidence of 30-day stroke was 0.7% for primary CEA and 1.9% for redo CEA (P < .001) (Table 2). Within the stratum of symptomatic patients, the incidence of 30-day stroke was 1.5% for primary CEA and 1.4% for redo CEA (P = .77). Overall, there was no statistical difference in the incidence of perioperative MI within 30 days for primary CEA vs redo CEA (0.8% vs 1.3%, P = .09). However, death (0.7% vs 1.4%, P = .003) and the composites of stroke or death (1.5% vs 2.6%, P = .001) and stroke, death, or MI (2.2% vs 3.6%, P = .001) within 30 days were significantly higher after redo CEA compared with primary CEA in the combined cohort. Cranial nerve injuries occurred in 3.8% of primary CEAs compared with 4.4% of redo CEAs (P = .24), although the transient or permanent nature of these injuries was not ascertained due to insufficient data. Comparing primary CEA with redo CEA, the incidence of other complications was 1.9% vs 1.5% (P = .29) for procedure-related arrhythmias, 1.8% vs 2.4% (P = .11) for artery reexploration after closure, 0.2% vs 0.1% (P = .36) for reperfusion symptoms, 0.1% vs 0.3% (P = .01) for wound infections, and 2.2% vs 3.5% (P = .001) for return to the operating room for bleeding, neurologic, or incision-related events.
The logistic regression analyses adjusting for patient characteristics showed a 75.0% increase in the odds of stroke within 30 days after redo CEA compared with primary CEA (odds ratio [OR], 1.75; 95% CI, 1.13-2.73; P = .01) for all patients (Table 3). The adjusted OR (aOR) of stroke within 30 days was significantly higher for redo CEA compared with primary CEA among asymptomatic patients (aOR, 2.82; 95% CI, 1.69-4.71; P < .001) but was similar among symptomatic patients (aOR, 0.79; 95% CI, 0.33-1.95; P = .62). There was no difference between redo CEA and primary CEA after excluding patients having amaurosis fugax (aOR, 1.44; 95% CI, 0.19-11.0; P = .73) compared with those having other symptoms (aOR, 0.73; 95% CI, 0.27-1.96; P = .53). The pattern of difference between redo CEA and primary CEA among asymptomatic patients and similarity among symptomatic patients was sustained for 30-day death, stroke or death, and stroke, death, or MI, with MI remaining similar irrespective of symptomatic status. Preoperative use of β-blockers was the only significant predictor of perioperative stroke or death after redo CEA (eTable 1 in the Supplement). In contrast, older age, female sex, preoperative symptoms, contralateral occlusion, active smoking, history of congestive heart failure (CHF), dialysis dependence, and American Society of Anesthesiologists class IV status were significant predictors of perioperative stroke or death after primary CEA (eTable 2 in the Supplement). There was a progression suggestive of higher odds of 30-day stroke or death among lower-volume operators compared with higher-volume operators who performed redo CEA. However, this association did not reach statistical significance (Table 4).
Overall, there were 746 ipsilateral stroke events (1.2%) over the study period, including 718 after primary CEA and 28 after redo CEA (P = .002). The incidence of absolute ipsilateral stroke was 0.9% (0.9% for primary CEA vs 2.4% for redo CEA, P < .001) for asymptomatic patients and 1.8% (1.8% for primary CEA vs 1.4% for redo CEA, P = .49) for symptomatic patients. Unadjusted estimates of freedom from stroke at 1 year obtained from Kaplan-Meier analyses were 98.1% (95% CI, 98.0%-98.2%) for primary CEA and 96.6% (95% CI, 95.1%-97.7%) for redo CEA (P = .004 by log-rank test). The multivariable Cox proportional hazards regression models showed that the incidence of ipsilateral stroke within 1 year was higher for redo CEA compared with primary CEA (adjusted hazard ratio [aHR], 1.53; 95% CI,1.01-2.29; P = .04) in the combined cohort of symptomatic and asymptomatic patients (Table 5). The incidence of ipsilateral stroke within 1 year was significantly higher for redo CEA compared with primary CEA among asymptomatic patients (aHR, 2.61; 95% CI, 1.64-4.15; P < .001) but was similar among symptomatic patients (aHR, 0.59; 95% CI, 0.25-1.45; P = .26).
Absolute all-cause mortality was 8.4% (8.3% for primary CEA vs 10.5% for redo CEA, P = .01) over the 13-year study period. Kaplan-Meier estimates of mortality at 1 year were 3.3% (95% CI, 3.1%-3.4%) after primary CEA vs 4.6% (95% CI, 3.6%-5.9%) after redo CEA (P = .02 by log-rank test). Per symptomatic status, 1-year mortality Kaplan-Meier estimates for primary CEA vs redo CEA were 3.0% (95% CI, 2.8%-3.1%) vs 4.3% (95% CI, 3.2%-5.1%) for asymptomatic patients (P = .04 by log-rank test) and 3.9% (95% CI, 3.6%-4.2%) vs 5.3% (95% CI, 3.5%-7.9%) for symptomatic patients (P = .23 by log-rank test). Risk-adjusted Cox proportional hazards regression models showed that mortality within 1 year was higher for redo CEA compared with primary CEA among asymptomatic patients (aHR, 1.28; 95% CI, 1.01-1.62; P = .04) but was similar among symptomatic patients (aHR, 0.98; 95% CI, 0.69-1.40; P = .93).
The incidence of absolute stroke or death over the study period was 9.5% (9.4% for primary CEA vs 12.2% for redo CEA, P < .001). Unadjusted Kaplan-Meier estimates of stroke or death at 1 year were 4.2% (95% CI, 4.1%-4.4%) after primary CEA vs 6.1% (95% CI, 4.9%-7.6%) after redo CEA (P = .002 by log-rank test). After stratifying by symptomatic status, Kaplan-Meier estimates of stroke or death at 1 year for primary CEA vs redo CEA were 3.7% (95% CI, 3.5%-3.9%) vs 6.2% (95% CI, 4.8%-8.0%) for asymptomatic patients (P = .001) and 5.4% (95% CI, 5.1%-5.8%) vs 6.0% (95% CI, 4.1%-8.7%) for symptomatic patients (P = .43). Adjusted hazards of stroke or death within 1 year were higher for redo CEA compared with primary CEA among asymptomatic patients (aHR, 1.36; 95% CI, 1.08-1.69; P = .009) but similar among symptomatic patients (aHR, 0.94; 95% CI, 0.67-1.31; P = .71). Significant predictors of stroke or death in the long term after redo CEA were older age, active smoking, history of CHF, and dialysis dependence (eTable 3 in the Supplement). Significant predictors of stroke or death in the long term after primary CEA were older age, preoperative symptoms, contralateral occlusion, diabetes, active smoking, history of coronary artery disease, history of CHF, history of chronic obstructive pulmonary disease, dialysis dependence, and American Society of Anesthesiologists class IV status (eTable 4 in the Supplement).
Results from the post hoc hierarchical model adjusting for center were not different from those reported previously comparing redo CEA with primary CEA. This was true with regard to 30-day stroke (aOR, 2.78; 95% CI, 1.66-4.65; P < .001) and 30-day death (aOR, 2.16; 95% CI, 1.22-3.84; P = .01) in asymptomatic patients and with regard to 30-day stroke (aOR, 0.80; 95% CI, 0.33-1.95; P = .62) and 30-day death (aOR, 1.30; 95% CI, 0.53-3.24; P = .57) in symptomatic patients.
In this large cohort of patients drawn from multiple institutions, symptomatic status modified the risk of adverse outcomes within 30 days and at 1 year for redo CEA relative to primary CEA. There was a 2.8 times increase in stroke and a 2.2 times increase in death and the composite of stroke or death within 30 days for redo CEA compared with primary CEA among asymptomatic patients but no significant difference among symptomatic patients. These results show that the excess risk associated with redo CEA relative to primary CEA is similar to the excess risk associated with primary CEA treatment for symptomatic disease relative to primary CEA for asymptomatic disease.
Most patients who underwent primary CEA and redo CEA in this study were asymptomatic and had ipsilateral stenosis exceeding 70%. This finding is similar to the proportion of asymptomatic patients who undergo treatment for carotid artery stenosis in the general and high-risk populations in the United States.22-26 Carotid endarterectomy is performed for stroke prevention, and the consensus for treatment in asymptomatic patients with ipsilateral stenosis greater than 70% is resolute. There is some concern about performing redo CEA in asymptomatic patients because the procedure is not without risk, as we have shown in this study, and unpredictable patterns of restenotic lesion regression, stability, and progression have been described, with low risk of embolic events for lesions due to neointimal hyperplasia,27,28 prompting a surveillance approach of watchful waiting by some clinicians. Unfortunately, we do not have details of the interval between the index surgery and prior CEA in this study. Another area of concern is the safety of the procedure for patients with a coexisting risk factor, such as renal failure, which was associated with worse outcomes in this study. However, a study comparing redo CEA vs medical management for restenosis is better suited to support this conclusion. Nonetheless, it is generally accepted that patients with hemodynamically significant lesions should be offered appropriate treatment given the significant increase in risk of stroke associated with greater than 70% stenosis reported from CREST.1 The concern about the treatment of asymptomatic patients must be balanced with the fact that symptoms of carotid stenosis are not entirely benign. More than 40% of the symptomatic patients in this study had a stroke as their initial symptom, suggesting the need for prophylactic treatment. This concern also underscores the ongoing drive by professional bodies to develop innovative methods for differentiating asymptomatic patients who have low risk vs high risk of stroke so as to limit the exposure of low-risk patients to surgery at population levels.
The higher incidence of stroke and stroke or death in the asymptomatic redo CEA patients compared with the symptomatic redo CEA patients contradicts conventional results from patients with de novo disease. Prior investigators of redo CEA did not report outcomes based on preoperative symptoms29 or could not identify differences between symptomatic and asymptomatic patients because there were no event rates to enable such discrimination.13 Other studies8,30 (among 212 patients and 64 patients, respectively) showed worse outcomes in symptomatic patients relative to asymptomatic patients. Results from a multi-institutional study of comparable sample size to ours are largely nonexistent, leaving room for validation of this unusual finding in subsequent studies. Differences in results between single-institutional series with small numbers of patients and large population-based studies underscore the importance of validating institutional findings at population levels due to the potential to reveal new knowledge that might contradict conventionally held beliefs. Redo CEA poses a unique surgical milieu, and the pathophysiologic underpinnings of our finding of more events in the asymptomatic patients compared with the symptomatic patients deserve further detailed evaluation. One possible reason for this finding is that the procedural risk associated with high-risk plaque in symptomatic patients obscures the difference between patients undergoing symptomatic redo CEA and patients undergoing primary CEA but not among asymptomatic patients. It is also possible that the association between symptomatic status and adverse events during the first CEA extends to the redo CEA even if the patient is classically asymptomatic for the redo operation. Unfortunately, we do not have information on the distribution of symptoms for the preceding CEA, leaving this theory to be proved or refuted in a study capable of such discrimination.
It has been shown that CAS is the more prevalent option for the treatment of post-CEA restenosis.7,9,10 It is likely that some of the patients who underwent redo CEA in the present study had anatomic or lesion characteristics that would have made CAS technically challenging or impossible. We are of the opinion that the treatment of patients with restenosis should be based on evidence of progressive disease over a surveillance period. Much remains to be elucidated on the progression of restenotic disease, in addition to considerations of contralateral disease and the influence of optimal medical therapy, in a large sample of patients, as has been done for patients with de novo carotid disease.31,32
Despite the worse outcomes associated with redo CEA, absolute event rates for this treatment fall within the limits stipulated by professional guidelines.33 It is the recommendation of the SVS that the composite of perioperative stroke or death should be less than 3% to ensure benefit of this preventive procedure in asymptomatic patients.33 The finding of a 2.9% incidence of composite stroke or death for redo CEA in the present study falls within this range, albeit at its upper limit. The SVS recommendation is based on the natural history of de novo carotid disease.31,32 Best-practice guidelines for the treatment of asymptomatic recurrent stenosis will be helpful as more robust data on the natural history of recurrent stenosis become available. When the decision to perform redo CEA is reached, the estimates from our study provide useful information for patients and their surgeons. Although the comparison of adverse outcomes across quartiles of redo case volume did not reach statistical significance, the association suggests that a learning curve might exist in this regard.
In the present study, there were more active smokers in the redo CEA cohort vs the primary CEA cohort (36.4% vs 27.9%, P < .001), and active smoking was associated with a 67.0% (OR, 1.67; 0.67-1.70; P = .04) increase for redo CEA and a 30.0% (OR, 1.30; 1.21-1.41; P < .001) increase for primary CEA in the hazards of stroke or death at 1 year. These results show that smoking cessation remains a viable target for the improvement of outcomes after primary CEA and redo CEA. Patients with CHF and those dependent on dialysis had significantly higher risk of stroke or death at 1 year. The predictive value of these patient characteristics is in tandem with findings from other studies.20,25 We found no benefit of preoperative use of β-blockers, antiplatelets, or statins on 30-day postoperative outcomes in contrast to findings from carotid artery stenting investigations.34
We acknowledge that this study is limited in its retrospective design. We did not have information on the interval between the primary CEA and the redo CEA. Hence, we were unable to evaluate the underlying etiology and the influence of early or late restenosis on outcomes after redo CEA. The study contains data up to 1 year; as such, our findings might not be generalizable beyond that time point, which leaves room for further investigation. Despite these limitations, our study is novel in its estimation of the excess risk associated with redo CEA relative to primary CEA in this era of proliferating endovascular treatments.
Redo CEA is associated with a 2.8 times increase in the risk of stroke and a 2.2 times increase in the risk of death and the composite of stroke or death compared with primary CEA among asymptomatic patients. There is no significant difference in the outcomes of redo CEA vs primary CEA among symptomatic patients. These results should inform patient and clinician expectations at the point of care.
Accepted for Publication: July 23, 2017.
Corresponding Author: Mahmoud B. Malas, MD, MHS, Division of Vascular and Endovascular Surgery, Department of Surgery, Johns Hopkins Medical Institutions, Johns Hopkins Bayview Medical Center, 4940 Eastern Ave, Building A/5, Baltimore, MD 21224 (email@example.com).
Published Online: November 8, 2017. doi:10.1001/jamasurg.2017.4477
Author Contributions: Drs Arhuidese and Malas had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Arhuidese, Faateh, Nejim, Malas.
Acquisition, analysis, or interpretation of data: Arhuidese, Faateh, Locham, Abularrage, Malas.
Drafting of the manuscript: Arhuidese, Faateh, Nejim, Locham, Malas.
Critical revision of the manuscript for important intellectual content: Arhuidese, Faateh, Nejim, Abularrage, Malas.
Statistical analysis: Arhuidese, Faateh, Nejim.
Administrative, technical, or material support: Locham, Abularrage.
Study supervision: Malas.
Conflict of Interest Disclosures: None reported.