Editor ’ s Choice – Timing of Carotid Intervention in Symptomatic Carotid Artery Stenosis: A Systematic Review and Meta-Analysis

The evidence from this systematic review and meta-analysis suggests that (at present) carotid endarterectomy (CEA) is safer than carotid artery stenting (CAS) when performed within two or seven days of the index event. Also, considering absolute rates of 30 day stroke, mortality, and death/stroke, CEA performed within two days of the index event complies with the accepted thresholds in international guidelines. The ﬁ ndings of this analysis will guide clinical practice when deciding on the type of intervention in the symptomatic patient with severe carotid stenosis. The ideal timing for performing CAS (when indicated against CEA) is not yet de ﬁ ned. Objective: This review aimed to analyse the timing of carotid endarterectomy (CEA) and carotid artery stenting (CAS) after the index event as well as 30 day outcomes at varying time periods within 14 days of symptom onset. Methods: A systematic review was performed according to the Preferred Reporting Items for Systematic reviews and Meta-analysis statement, comprising an online search of the Medline and Cochrane databases. Methodical quality assessment of the included studies was performed. Endpoints included procedural stroke and/or death strati ﬁ ed by delay from the index event and surgical technique (CEA/CAS). Results: Seventy-one studies with 232 952 symptomatic patients were included. Overall, 34 retrospective analyses of prospective databases, nine prospective, three RCT, three case control, and 22 retrospective studies were included. Compared with CEA, CAS was associated with higher 30 day stroke (OR 0.70; 95% CI 0.58 e 0.85) and mortality rates (OR 0.41; 95% CI 0.31 e 0.53) when performed (cid:1) 2 days of symptom onset. Patients undergoing CEA/CAS were analysed in different time frames ( (cid:1) 2 vs. 3 e 14 and (cid:1) 7 vs. 8 e 14 days). Expedited CEA ( vs. 3 e 14 days) presented a sampled 30 day stroke rate of 1.4%; 95% CI 0.9 e 1.8 vs. 1.8%; 95% CI 1.8 e 2.0, with no statistically signi ﬁ cant difference. Expedited CAS ( vs. 3 e 14 days) was associated with no difference in stroke rate but statistically signi ﬁ cantly higher mortality rate (OR 2.76; 95% CI 1.39 e 5.50). Conclusion: At present, CEA is safer than transfemoral CAS within 2/7 days of symptom onset. Also, considering absolute rates, expedited CEA complies with the accepted thresholds in international guidelines. The ideal timing for performing CAS (when indicated against CEA) is not yet de ﬁ ned. Additional granular data and standard reporting of timing of intervention will facilitate future monitoring.


INTRODUCTION
Carotid revascularisation improves long term stroke free survival in patients with recent ischaemic stroke or transient ischaemic attack (TIA). Recency of the index event has been recognised as a key determinant of the effectiveness of revascularisation, balancing the natural history risk of a EJVES Open Access second (more severe) event against the potential for a higher peri-procedural risk when carotid interventions are performed very early after the onset of symptoms.
The optimal timing for carotid revascularisation, by either carotid endarterectomy (CEA) or carotid artery stenting (CAS), remains a matter for debate. The 2017 European Society for Vascular Surgery (ESVS) guidelines advise that CEA should be performed within 14 days of the index neurological event, as this was the highest risk time period for recurrent stroke. 1 This is particularly true for neurologically stable patients presenting with TIA or minor stroke. However, it remains unclear as to the optimal timing of either CEA or CAS within this 14 day time period (i.e., Is it better for the carotid intervention to be performed < 2 days, < 7 days, or perhaps 8 e 14 days after symptom onset?).
A recent systematic review reported that the risk of recurrent stroke can vary from 6% within 2 e 3 days of the index event, to 20% within 7 days, and up to 26% within 14 days of the index event. 2 Conversely, a meta-analysis of published studies comparing expedited carotid interventions (2 days) vs. early (3 e 14 days) found a significantly higher risk of procedural stroke when CEA was performed within 2 days of the index event. 3 However, this systematic review did not include two large national CEA registries (> 70 000 CEAs), which confounds meaningful interpretation of their data. In the case of CAS, the available data on safety very early after the onset of symptoms appears limited. 4 The lack of high quality evidence and consensus definitions for what constitutes "early" or "urgent" carotid interventions has contributed to conflicting results in the literature. Heterogeneity regarding patient symptoms, medical therapy, and varying surgical approaches have also led to polarised debates about the timing of CEA in patients who present with neurological symptoms. 5 The aims of the current systematic review and metaanalysis were to analyse temporal changes in the timing of carotid interventions after symptom onset and to determine 30 day outcomes following CEA and CAS when performed at varying time periods in the first 14 days after onset of symptoms, to define the optimal timing and carotid intervention (CEA vs. CAS) in recently symptomatic patients.

METHODS
A systematic review was conducted according to the recommendations of the Preferred Reporting Items for Systematic reviews and Meta-analysis (PRISMA) statement. 6 The literature search was from January 1995 to January 2021. Using the Medline and Cochrane databases, the following query ((("Carotid Stenosis"[Mesh]) AND "Stents"[Mesh]) OR "Endarterectomy, Carotid"[Mesh]) AND ("Stroke" [Mesh] OR Symptomatic OR timing of intervention) was used for online search.
Eligibility criteria included any publication regarding the revascularisation of symptomatic carotid artery stenosis by either CAS or CEA. Timing of intervention and impact of delay on procedural risks were documented. Only atherosclerotic stenotic carotid disease was considered, with exclusion of procedures performed for non-atherosclerotic pathologies.
Exclusion criteria were (1) articles published in a language other than English; and (2) case reports and literature reviews.
Endpoints included any stroke and/or death within 30 days of intervention stratified by delay of intervention after the index event and by intervention technique (CEA and  ). An analysis of reporting of timing of CEA and CAS after the index event was also performed. Stroke was defined as a rapidly developing clinical syndrome of focal disturbance of cerebral function lasting more than 24 hours or leading to death with no apparent cause other than that of vascular origin. Stroke was considered procedural if the event occurred at any time between the revascularisation procedure (day 0) and day 30 after revascularisation.
Stroke was classified as disabling if there was an increase in the modified Rankin score (mRS) to ! 3, attributable to the event 30 days after the procedure. Neurological symptomatic status was defined as a transient ischaemic attack or minor disabling ischaemic stroke in the previous six months attributable to the ipsilateral carotid artery territory.
For the purpose of this meta-analysis, "expedited intervention" was used to define any intervention performed within two days of the index event. Index event was defined as the symptom that led the patient to seek medical advice as suggested in the ESVS guidelines. 1 Two reviewers (AC and JP) screened the identified studies independently and were also responsible for data extraction (Fig. 1). Collected data included type of study, year of publication, number of patients and consecutiveness, adjudication of events by a clinical event committee (CEC), age, gender, and criteria for carotid revascularisation (presence and type of neurological symptoms and their timing). The definition of intervention delay regarding the index event was registered in different studies. Neurological events after the index event and before intervention were registered as well as procedural (30 day) events: stroke, myocardial infarction (MI), and death. Comparative data between early and delayed intervention were analysed, especially for interventions performed 2 days vs. between 3 and 14 days and for interventions performed 7 days vs. between 8 and 14 days of the index event.

Quality assessment
The methodology of the studies and risk of bias were systematically assessed by two independent reviewers (AC and JP) using the Methodological Index for Non-Randomized Studies (MINORS) score, 7 with a maximum score of 16 for non-comparative and 24 for comparative studies. A score 8 was considered poor quality, 9 e 14 moderate quality, and 15 e 16 good quality for non-comparative studies. Cut off points were 14, 15 e 22, and 23 e 24, respectively, for comparative studies. Authorship of the studies was unblinded during review. Discrepancies between the reviewers during the search, selection, and quality assessment were resolved by discussion. In case of persisting disagreement, a third reviewer was consulted.

Statistical analysis
The software Review Manager 5.4 (REVMAN) was used to analyse data. Odds ratios (OR) and 95% confidence intervals (CI) were used for dichotomous variables, and mean differences (MDs) with 95% CI for continuous data.
Statistical heterogeneity, defined as a measure of the variability of outcomes between studies, was assessed by the Cochran's Q test: the H 2 test (Higgins and Thompson) was used to quantify the magnitude of heterogeneity. The parameter I 2 retrieved from the H 2 test was used with a cut off of 25% for low, 25% e 50% for intermediate, and above 50% for high heterogeneity. A fixed effects model was used when heterogeneity (I 2 ) was less than 50% and a random effects model was used when heterogeneity (I 2 ) was high.

RESULTS
A total of 1 495 potentially relevant articles were identified initially. After reviewing title or abstract, 112 articles were retrieved and 71 judged eligible for inclusion (Fig. 1). Agreement between reviewers was reached for all articles and arbitration by the third reviewer was unnecessary.
Overall, there were 24 retrospective analyses of prospective national databases, 10 retrospective analyses of prospective databases, nine prospective studies, three RCTs, and three case control studies. The remaining 22 studies were retrospective, single centre, or multicentre, analysis of patient data. The total number of symptomatic patients in the constituent studies was 232 952 (Table 1). Methodological quality is reported in Supplementary Table S1. A total of 18 non-comparative studies of moderate quality and 53 comparative studies (50 moderate, two poor quality, and one good quality) were included (Supplementary Table S1).

Definitions
The definitions of "delay" and "index event" were heterogeneous (Table 1). Most studies defined "early intervention" when CEA or CAS were performed within 14 days of the index event, although some studies applied stricter or looser definitions ( Table 1). Stratification of the timing of events within the first 14 days was described in some studies, for example as "acute/urgent/emergency/ultra-early interventions" ( Table 1). One study was identified that defined the timing of intervention as the time from the qualifying event (defined as the most recent neurological event before intervention, rather than the index event). 8  Table 2).

Symptomatic status and timing of intervention
Stratification of demographic data, type of neurological index event, and occurrence of new neurological symptoms, stratified by intervention delay are detailed in Table 2.
Where reported, patients presenting with crescendo TIAs were more likely to undergo an early intervention. 9,10 The remaining presenting events (TIA, amaurosis fugax and stroke) were evenly distributed by intervention delay, with few exceptions (Table 2).

Primary and secondary outcomes
Peri-operative (30 day) outcomes along with data on hospitalisation duration (in days), stratified by intervention delay and by type of revascularisation (CEA vs. CAS) are detailed in Table 3 Carotid endarterectomy vs. carotid artery stenting when performed 2 days after the index event. Two moderate quality studies reported outcomes after CEA vs. CAS (75 917) when performed within two days of the index event, including one retrospective analysis of prospective single centre data (120) and one retrospective analysis of Nationwide Inpatient Sample (NIS) database (72 797). 12,13 Compared with CEA, meta-analysed data revealed significantly higher risks for 30 day stroke when CAS was performed within 2 days (OR 0.70; 95% CI 0.58 e 0.85) as well as significantly higher rates of 30 day death (OR 0.41; 95% CI 0.31 e 0.53) (Fig. 2). One of the above mentioned registries (72 797) analysed patients with and without cerebral infarction separately and concluded that expedited revascularisation in patients with cerebral infarction on admission increased the risk of iatrogenic stroke and death; the increase in mortality was more dramatically seen in patients treated by CAS. No differences were found in stroke/death rates between CEA and CAS if patients presented without infarction. 12 Carotid endarterectomy vs. carotid artery stenting when performed 3 e 14 days after index event. The same large national registry (72 797) cited in the previous section also reported comparative outcomes between CEA vs. CAS when performed 3 e 14 days after the index event (with or without cerebral infarction). There was no statistically significant difference in 30 day stroke after CAS (1.8%) vs. after CEA (1.6%; OR 1.1; 95% CI 0.9 e 1.4). However, 30 day mortality was statistically significantly higher after CAS (1.6%) vs. after CEA (0.8%; OR 1.9; 95% CI 1.4 e 2.5). Again, no differences were found in stroke/death rates between CEA and CAS if patients presented without infarction. 12 Outcomes after carotid endarterectomy 2 days vs. 3 e 14 days after index event. A total of nine moderate quality manuscripts were included in this analysis, three of which were retrospective analyses of national registries, two prospective multicentre studies, and four retrospective studies. CEA performed 3 e 14 days after the index event was associated with a statistically significantly lower 30 (Fig. 3).
Meta-analysis of 30 day stroke, mortality, and MI included the same core studies, while in the analysis of 30 day death/stroke, three studies were excluded as they did not report the composite outcome, 9,12,14 while one study was included that only reported combined stroke/death data, with worse outcomes reported in the expedited cohort (Fig. 3). 15 7 days vs. 8 e 14 days after index event. A total of five moderate quality manuscripts were included in this analysis, two of which were retrospective analyses of national registries and three retrospective studies. Meta-analyses (Fig. 4) (Fig. 5). A single study (n ¼ 323) reported the results of a comparative analysis of 30 day death/stroke and showed no significant difference when CAS was performed in either time period (OR 0.61; 95% CI 0.03 e 11.06). 16 7 days vs. 8 e 14 days after index event. The same national registries that compared outcomes when CAS was performed 2 days vs. 3 e 14 days, also analysed outcomes 7 days vs. 8 e 14 days. 14,16 Forest Plot analyses (Fig. 6)

Recurrent events while awaiting a carotid intervention
Recurrent neurological events occurring after a decision to perform CEA but before it was performed were reported rarely. In one single centre study, 42% of patients who waited 0 e 180 days to undergo CEA suffered a recurrent TIA or stroke prior to CEA. 17 The National Norwegian Carotid Study reported that 3.3% suffered recurrent symptoms prior to undergoing CEA within 14 days of the index event (Table 2). 18

Neurological outcome
Surprisingly, few studies used the National Institutes of Health Stroke Scale (NIHSS) to quantify improvements in neurological disability after carotid interventions, stratified for the timing of carotid interventions (Table 3). A single centre study reported improved neurological outcomes for interventions performed within 14 days vs. 15 e 30 days of the index event (NIHSS range 0.9 AE 0.4 vs. 0.5 AE 0.2; p ¼ .011). 19 Other studies report NIHSS range but with no discriminative data concerning carotid intervention delay from index event. 13,18,20 Hospital stay Hospital stay analysis presented a trend towards prolonged stay in patients undergoing CEA between 3 e 14 days after the index event vs. 2 days, with a mean difference (MD) of À1.28 (95% CI À6.96 e 4.40) (Fig. 7).

DISCUSSION
The ESVS guidelines advise that CEA (CAS) should be performed within 14 days of symptom onset. 1 Evidence suggests that there has been a major drive towards performing interventions < 14 days (especially in Europe), where the median delay to CEA is now 11 days in the Netherlands, 21 7 days in Sweden, 22 9 days in Germany, 23 and 11 days in the UK. 24 A temporal trend towards a progressive decrease in delays from index event to undergoing CEA (or CAS) has been reported by several national registries. 24e26 The proportion of Danish Stroke Registry patients 17 undergoing carotid interventions within two weeks of the index event increased from 13% in 2007 to 47% by 2010 (OR 5.8; 95% CI 4.3 e 10.1). 25 Similar findings were reported by the UK National Vascular Registry. 24 However, uncertainty persists regarding the ideal timing for either CEA or CAS within the 14 day time frame to balance the dichotomy between recurrent stroke prevention and minimising peri-operative risks. 25 The Swedish Vascular Registry (Swedvasc) were the first to highlight concerns about intervening within 48 hours of the index event, as they observed an 11.5% rate of 30 day death/stroke, compared with 3.6% (3 e 7 days), 4% (8 e 14 days), and 5.4% (> 14 days) for CEA. However, only a small proportion of Swedvasc patients were treated 48 hours (5.7%), which may have limited the generalisability of the Swedish registry data. 27 Other (much larger) national registries have not corroborated the Swedvasc findings. In the German CEA registry (56 000 CEAs), there was no difference in 30 day death/stroke between patients treated 48 hours by CEA (3%) vs. later time periods (2.5% between 3 e 7 days; 2.6% between 8 e 14 days; 2.3% for CEA thereafter). 23 In the UK national registry involving 20 000 patients, conclusions were that the pathway from most recent symptom to surgery for patients with symptomatic carotid stenosis, could be shortened to maximise the benefit of intervention, without increased peri-operative risk in the period. However, they admitted a slight increase in peri-operative risk of stroke and death in the first 48 hours. 24 In this systematic review, 44 410 (29.9%) carotid interventions were undertaken within 2 days of the index event with no significant difference in 30 day stroke, mortality, and MI, while CEA performed 3 e 14 days after the index event was associated with a significantly lower risk of the composite outcome 30 day death/stroke. On the other hand, CEA within 7 days was associated with a significantly lower risk of stroke (vs. 8 e 14 days).
These contradictory results may be explained by the differences in included studies in each analysis, as already shown. Compared with the analysis of the outcomes stroke and mortality, analysis of the composite outcome stroke/ death did not include data from two national registries and one prospective multicentre study, 9,12,14 while it included data from one retrospective single centre study that only reported combined stroke/death data. 15 Therefore, studies included in the analysis of 30 day stroke, death, and MI are of better study design compared with the studies in the 30 day death/stroke analysis, even though quality assessment is similar.
There were inconsistent findings regarding timing and outcomes in CAS patients. In patients undergoing CAS 2 days of the index event (vs. 3 e 14), there was no apparent difference in 30 day stroke or MI but there was a statistically significantly higher risk of death. Conversely, there were no differences in 30 day outcomes between CAS performed 7 days (vs. 8 e 14). The pathophysiology of procedural stroke may differ with expedited (vs. delayed) interventions in line with acute changes in atherosclerotic plaque vulnerability, which have been associated with an increased risk of embolism and neurological events after CAS. 28 The systematic review also addressed the question of whether CEA or CAS was safer (or equivalent) when performed in the first 14 days after symptom onset. Compared with CEA, CAS was associated with significantly higher 30 day stroke and death rates when performed within 2 days  of symptom onset. In an individual patient meta-analysis of data from the four largest RCTs comparing CEA with CAS (4 138 patients), CAS was associated with significantly higher risks of 30 day stroke, mortality, and death/stroke when performed within 7 days of the index event. 12 These data suggest that, at the current time, CEA is probably safer than CAS both when performed 2 days and 7 days after symptom onset. However, virtually all of the CAS procedures in the current meta-analyses were performed via the transfemoral route. Registry data suggest that TCAR can be performed with 30 day outcomes similar to CEA in symptomatic patients. 29 Unfortunately, no studies have published outcome data for TCAR when used in the first 14 days after symptom onset, 11 and these data are keenly awaited.
There are relatively few data published on the incidence of recurrent events prior to expedited interventions. A prospective cohort study concluded that the risk was about  12% with modern best medical therapy, but that half of all recurrent events occurred within two days of the index event. 30 On the other hand, a recent meta-analysis revealed a cumulative 120 day risk of recurrent stroke of 1.97% (95% CI 0.75 e 3.17) in recent large RCTs, which was statistically significantly lower than in historical controls. 31 Historically, vascular surgeons have not really considered the prevention of recurrent stroke in the time period between initiating investigation and initial management and undergoing CEA as being their primary responsibility. However, this attitude is likely to change as more symptomatic patients are started on dual antiplatelet therapy (DAPT) within 24 hours of symptom onset. The 2017 ESVS guidelines recommended that early treatment with DAPT "may be considered" to prevent recurrent events (prior to CEA) in patients with TIA or minor ischaemic stroke and an ipsilateral 50% e 99% stenosis awaiting CEA (Evidence IIb, Level C). 1 At the time, the ESVS Writing Group were unable to recommend routine DAPT in all symptomatic patients because there was no compelling evidence that this strategy conferred additional benefit over antiplatelet monotherapy.
However, based on a meta-analysis of three recent RCTs (CHANCE, POINT, and FASTER) in which 10 447 patients were randomised within 24 hours of experiencing a minor ischaemic stroke (NIHSS 3) or "high risk TIA" (ABCD 2 score ! 4) to aspirin monotherapy or short term aspirin and clopidogrel DAPT, there is now compelling evidence to support short term treatment with DAPT in these patient subgroups. 32 A recently published RCT also proved that in the subgroup of stroke patients with carotid artery stenosis, ticagrelor added to aspirin in the first 24 hours after the      event, had greater absolute risk reduction of stroke or death at 30 days than stroke patients without carotid artery stenosis with a clinically meaningful benefit with a number needed to treat of 34 (95% CI 19 e 171). 33 Methodological quality assessment revealed that the included studies are moderate to low quality, with a single high quality study in this analysis. Only a small number of studies was eligible for quantitative analysis, hindering conclusions. Also, heterogeneity of quantitative synthesis is significant, as determined by the I 2 test. Risk of bias is therefore significant. Probably one of the main biases was introduced in the election for CAS/CEA (selection bias), with fit patients treated by CEA while high risk patients were treated by CAS. Also, with the inclusion of mainly prospective cohort studies the risk of confounding is inherent.

Study or Subgroup
In conclusion, the predicted magnitude of procedural risks will ultimately determine whether CEA or CAS is safer in the early time period after onset of symptoms. 34 The evidence from the current systematic review and metaanalysis suggests that (at present) CEA is still safer than transfemoral CAS when performed 2 days of the index event. Also, considering absolute rates, expedited CEA complies with the accepted thresholds in international guidelines. The ideal timing for performing CAS (when indicated against CEA) is not yet defined and it remains to be seen whether newer CAS technologies (such as TCAR) can provide outcomes similar to CEA when performed in the first 2 e 7 days after symptom onset. Additional granular data and standard reporting of timing of intervention will facilitate future clinical decisions.