Carotid Artery Stenting versus Carotid Artery Endarterectomy in high-risk patients: A systematic review

Search strategy

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines.¹² A comprehensive literature search was performed in February 2025 across six electronic databases: PubMed, Embase, Scopus, Web of Science, the Cochrane Library, and ClinicalTrials.gov. The search was limited to English-language publications, with no restrictions on publication date. Search terms included a combination of Medical Subject Headings (MeSH) and free-text keywords such as “carotid artery stenosis,” “carotid artery stenting” (CAS), “carotid endarterectomy” (CEA), and “high-risk patients.” Filters were applied to identify relevant clinical trials and comparative cohort studies. The complete search strategy for each database is presented in Figure 1.

Figure 1. Database search summary.

Search strategies and total studies identified across six databases: PubMed, Embase, Scopus, Web of Science, Cochrane Library, and ClinicalTrials.gov.

Protocol and registration

No separate protocol was developed for this review beyond the PROSPERO registration (CRD420251162463). The methodological approach, including eligibility criteria, outcomes, and analysis plan, was prespecified in the PROSPERO record.

Eligibility criteria

Studies were eligible for inclusion if they directly compared carotid artery stenting and carotid endarterectomy in adult patients identified as high-risk for surgery based on anatomical, physiological, or comorbid factors, and if they reported at least one relevant clinical outcome such as stroke, death, myocardial infarction, or reintervention. Eligible studies included randomised controlled trials and comparative cohort studies, whether prospective or retrospective. Only studies published in English and available in full-text form were included. Studies were excluded if they lacked a direct comparator group, were case reports, case series, editorials, letters, or narrative reviews, involved paediatric or non-human populations, or assessed mixed-risk populations without separate analysis of a high-risk subgroup.

Study selection

All retrieved records were imported into Rayyan (Rayyan Systems Inc., Qatar), an online systematic review platform that facilitates blinded screening and reviewer collaboration.¹³ Duplicate records were automatically identified and removed. Two reviewers independently screened all titles, abstracts, and full texts according to the predefined eligibility criteria. Disagreements were resolved through discussion or, when necessary, by consulting a third reviewer. The study selection process followed the PRISMA 2020 flow diagram, ensuring transparent documentation of inclusion and exclusion decisions.

Data extraction

Data extraction was performed independently by two reviewers using a standardised data collection form. Extracted information included study characteristics (first author, publication year, study design, and country), patient characteristics (sample size, mean age, sex distribution, and high-risk criteria), intervention details (CAS technique, use of embolic protection, and stent type), comparator details (CEA technique and anaesthetic approach), and clinical outcomes, including peri-procedural and long-term stroke, death, myocardial infarction, cranial nerve injury, restenosis, and reintervention. Follow-up duration and outcome definitions were also recorded as reported in the original studies. Any disagreements in data extraction were resolved by consensus.

Definition of high-risk patients

For this review, “high-risk” was defined as patients with a documented increased risk of perioperative complications or adverse outcomes following surgical treatment. High-risk features included anatomical factors such as high carotid bifurcation, contralateral carotid occlusion, prior neck irradiation, previous carotid surgery or neck dissection, tracheostomy, and significant aortic arch or carotid tortuosity. Comorbid factors comprised severe cardiopulmonary disease such as congestive heart failure, left ventricular ejection fraction below 30%, unstable angina, recent myocardial infarction, severe chronic obstructive pulmonary disease, advanced age (typically ≥80 years), chronic renal insufficiency, uncontrolled diabetes mellitus, and other systemic conditions associated with elevated surgical risk. Only studies that explicitly identified and reported high-risk patients as a discrete cohort undergoing carotid endarterectomy or carotid artery stenting were included, while studies involving mixed-risk populations without separate subgroup analysis were excluded.

Risk of bias assessment

The risk of bias for each included study was assessed independently by two reviewers. Randomised controlled trials were evaluated using the Cochrane Risk of Bias 2.0 (RoB 2) tool, which examines potential bias across five domains: the randomisation process, deviations from intended interventions, missing outcome data, measurement of outcomes, and selection of reported results. The results of this assessment are summarised in Figure 2. Non-randomised comparative studies were assessed using the Risk Of Bias In Non-randomised Studies of Interventions (ROBINS-I) tool, which evaluates bias arising from confounding, participant selection, intervention classification, deviations from intended interventions, missing data, outcome measurement, and selective reporting. The corresponding results are illustrated in Figure 3. Any disagreements between reviewers were resolved through discussion until consensus was achieved. Each study was categorised as having a low, moderate, serious, or critical risk of bias.

Figure 2. Bias assessment using the RoB 2.0 tool.

Summary of risk of bias for randomised controlled trials, evaluated using the Cochrane RoB 2.0 tool across key domains.

Figure 3. Bias assessment using the ROBINS-I tool.

Summary of risk of bias for non-randomised studies, assessed using the ROBINS-I tool across the seven bias domains.

Outcomes assessed

The primary outcomes of interest were stroke (any and disabling), all-cause mortality, myocardial infarction, and reintervention, assessed during both the peri-procedural period (30 days) and at long-term follow-up, where available. Secondary outcomes included procedural complications such as cranial nerve injury and restenosis. Outcome definitions and follow-up durations were recorded as reported in the original studies to maintain consistency with source data.

Effect measures

For dichotomous outcomes, effect measures such as odds ratios (ORs) or relative risks (RRs) with corresponding 95% confidence intervals (CIs) were planned to be reported where data permitted. Continuous variables were summarised as means with standard deviations or medians with interquartile ranges, depending on data distribution.

Data synthesis and analysis

Extracted data were reviewed to determine whether a meta-analysis could be performed. Owing to substantial heterogeneity among study designs, populations, and reported outcomes, a quantitative synthesis was not deemed appropriate. Instead, a narrative synthesis approach was employed to summarise findings across studies. Key outcomes, including stroke, death, restenosis, and reintervention, were compared descriptively to identify consistent trends and highlight inter-study differences. Any disagreements in interpretation were resolved through discussion among reviewers.

Reporting bias and certainty of evidence

Due to heterogeneity and the narrative synthesis approach, a formal assessment of reporting bias and certainty of evidence using funnel plots or the GRADE framework was not performed. However, the potential for publication bias and selective reporting was considered when interpreting results. Study quality, consistency of findings, and precision of outcome estimates were evaluated qualitatively to inform the overall certainty of the evidence.

Study selection

The study selection process followed the PRISMA 2020 guidelines, as illustrated in the PRISMA flow diagram ( Figure 4). Initially, 187 records were identified from six databases: PubMed (n = 29), Embase (n = 31), Scopus (n = 28), Web of Science (n = 32), Cochrane Library (n = 26), and ClinicalTrials.gov (n = 33). After removing 28 duplicates, 159 unique records remained for screening. Full-text retrieval was attempted for all 159 records, resulting in 141 full-text reports being successfully retrieved, while 18 reports were inaccessible. Of the 141 reports screened, 49 were excluded as animal studies, 58 were case reports, and 25 were literature reviews. Following this process, nine studies were included in the systematic review.

Figure 4. PRISMA 2020 flow diagram.

Flowchart depicting the identification, screening, eligibility, and inclusion process for studies in the systematic review.

The characteristics of the included studies are summarised in Table 1, while outcomes and conclusions are presented in Table 2.

Study characteristics

This systematic review consolidates evidence from both randomised and non-randomised studies comparing outcomes between carotid artery stenting (CAS) and carotid endarterectomy (CEA) in high-risk patients. The definition of high-risk varied across studies but generally encompassed individuals with severe comorbidities, prior neck surgery, or complex anatomical features such as contralateral carotid occlusion or restenosis. Key studies contributing to this analysis include the SAPPHIRE trial¹⁹ and investigations by Shawl et al.,¹⁴ Yadav et al.,¹⁵ Halabi et al.,¹⁶ Marine et al.,¹⁷ Gurm et al.,¹⁸ Giles et al.,¹⁹ Wang et al.,²⁰ and Mo et al.,²¹ each of which evaluated procedural safety, efficacy, and long-term outcomes of CAS and CEA in high-risk surgical populations.

The stent designs and materials used in CAS varied, with most studies employing self-expanding nitinol stents such as Smart and Precise (Cordis),^15,20 although stainless-steel biliary stents (Palmaz) and Integra stents were also used in some cases. Technical success rates for CAS ranged from 95.6%¹⁵ to 99%,¹⁴ indicating high procedural success in high-risk patients. However, the lack of consistent reporting on procedural duration prevented comparative analysis of procedural time.

The use of embolic protection devices varied widely across studies, ranging from 44%¹⁶ to 91.4%,¹⁷ introducing heterogeneity in procedural technique and potentially influencing outcomes. Post-procedure antiplatelet regimens were generally consistent, with most studies prescribing dual therapy with aspirin and clopidogrel or ticlopidine, though the duration varied between 2–4 weeks^15,18 and 3 months.²¹

Risk of bias in studies

In the cohort of randomised controlled trials, Gurm et al.,¹⁸ Yadav et al.,¹⁵ and Mo et al.²¹ were judged as having an overall risk of bias categorised as “some concerns”, primarily due to potential deviations from intended interventions (Domain 3) and uncertainty regarding outcome measurement (Domain 4). The summary of bias assessment for randomised studies is shown in Figure 2.

For non-randomised studies evaluated using the ROBINS-I tool, the overall risk of bias was generally low, suggesting acceptable methodological quality. Giles et al.,¹⁹ Marine et al.,¹⁷ and Wang et al.²⁰ exhibited moderate risk in certain domains, particularly for confounding (D1) and participant selection (D2, D4). Shawl et al.¹⁴ had moderate concerns related to outcome measurement and reporting (D6, D7), while Halabi et al.¹⁶ showed moderate risk for selective reporting (D7). The ROBINS-I assessment summary is illustrated in Figure 3. Despite these issues, all non-randomised studies were deemed methodologically sound and appropriate for inclusion.

Outcomes assessed

Cerebral hyperperfusion syndrome (CHS) was explicitly defined in one study, occurring in 22.6 % of patients undergoing standard CAS procedures without staged angioplasty²² (p < 0.01). Contrast volume and radiation exposure were inconsistently reported, precluding comparative evaluation of procedural imaging burden.

Reintervention rates were reported infrequently; one study found lower reintervention after CAS (0.6%) compared with CEA (4.3 %)¹⁵ (p = 0.03). Only one study¹⁴ reported residual stenosis post-stenting, averaging 2 ± 3 %. Restenosis rates ranged from 2%¹⁴ to 4.3%¹⁵ in CAS groups, with lower rates in CEA groups (p = 0.08–0.12).

Peri-procedural stroke rates for CAS ranged from 1.1–9.7%, while those for CEA ranged from 1.4–4.2%, indicating comparable outcomes between the two. Myocardial infarction (MI) incidence was rarely reported separately, though one study¹⁵ found a lower incidence of MI following CAS (2.4 %) than CEA (6.1 %) (p = 0.04). Cranial nerve injury occurred in 4.9% of CEA cases but in none of the CAS cases¹⁵ (p < 0.01).

Statistical findings and synthesis

Giles et al.,¹⁹ using Nationwide Inpatient Sample data, reported that CAS was associated with greater odds of stroke or mortality than CEA among both high-risk and non-high-risk patients (adjusted OR 2.4, 95% CI 2.1–2.8; p < 0.001). Conversely, Gurm et al.¹⁸ found no statistically significant difference between CAS and CEA regarding long-term stroke, MI, or death (3-year composite endpoint 24.6 % vs 26.9 %; p = 0.62). Halabi et al.¹⁶ reported an in-hospital major adverse event rate of 2.6%, with a non-significant reduction in distal protection subgroups (p = 0.08). Marine et al.¹⁷ reported 30-day stroke rates of 1.1% for CAS and 2.1% for CEA (p = 0.45). Mo et al.²¹ demonstrated a significant reduction in hyperperfusion events with staged angioplasty (0 % vs 22.6%; p = 0.04). Wang et al.²⁰ found no significant difference in safety or efficacy between CAS and CEA. Yadav et al.¹⁵ reported a relative risk of 1.39 (95 % CI 0.90–2.14; p > 0.05) for the composite endpoint of stroke, MI, or death between CAS and CEA.

No formal meta-analysis was performed because of substantial heterogeneity in study designs, populations, and reported outcomes. Therefore, results were synthesised narratively.

Reporting bias and certainty of evidence

No formal assessment of publication or reporting bias was performed because of the small number and heterogeneity of included studies. Certainty of evidence was not formally evaluated using the GRADE approach, given the limited number of randomised trials and the observational nature of most included studies. However, potential sources of bias were considered qualitatively when interpreting the results, and findings were discussed considering study quality, consistency, and precision.

Key findings

The SAPPHIRE trial¹⁸ and Yadav et al.¹⁵ both demonstrated non-inferiority of CAS to CEA, with comparable composite endpoints at short- and long-term follow-up. Both highlighted reduced cranial nerve injury and myocardial infarction with CAS, while stroke rates remained comparable under optimal procedural conditions with embolic protection. The SAPPHIRE trial¹⁸ also supported the equivalence of CAS and CEA, reporting similar periprocedural stroke risks, though without formal non-inferiority testing.

Conversely, Giles et al.¹⁹ observed higher rates of death and stroke with CAS, particularly in symptomatic high-risk patients (13.1% vs 5.9%). This contrasts with lower rates reported in randomised trials, suggesting that operator experience, centre volume, and patient selection may account for observed outcome differences.

Limitations

This synthesis was constrained by several factors. First, considerable heterogeneity in follow-up duration across studies hindered the ability to effectively compare long-term outcomes. Additionally, differences in procedural practices, stent selection and device usage, and anaesthesia types may have influenced results. Study design heterogeneity, with both randomised controlled trials (RCTs) and non-randomised studies included, further complicated direct comparisons between the two study types.

Moreover, some studies did not report detailed procedural features or outcome definitions, limiting the accuracy and comparability of the findings. A lack of consistent high-risk patient definitions across studies also hindered meaningful comparisons. Furthermore, the underreporting of patient subgroups, such as those with multiple comorbidities or advanced age, restricted the generalisability of findings to the broader high-risk population.

None of the included studies reported data on procedural factors like contrast volume, procedural time, or radiation exposure, which are particularly important in assessing the safety and tolerability of interventions in high-risk patients. As a result, conclusions regarding these aspects could not be drawn.

A major limitation of this systematic review is the high heterogeneity in outcome definitions, reporting methods, and follow-up durations across the included studies. This variability significantly limits the ability to perform pooled analyses and undermines the reliability and generalisability of the conclusions. It should be acknowledged as a central methodological constraint rather than a minor limitation.

Lastly, the limited number of high-quality, large-scale RCTs in this field reduced the strength of the evidence synthesis. The small number of robust RCTs means that some conclusions were based on lower-quality evidence. Although data extraction and risk of bias assessment were performed independently by two reviewers, some degree of subjectivity in interpretation cannot be excluded.

Clinical recommendations