Types of studies

We sought to identify all randomised trials of carotid patching. We included quasi‐randomised trials, in which allocation to different treatment regimens was not adequately concealed (e.g. allocation by alternation, date of birth, hospital number, day of the week, or by using an open random number list), and foreknowledge of treatment allocation might lead to biased treatment allocation and exaggerated treatment effects (Schulz 1995).

Types of participants

We included trials that enrolled anybody undergoing CEA, whether the initial indication for endarterectomy was symptomatic or asymptomatic carotid disease.

Types of interventions

We sought to identify all trials comparing routine carotid patch angioplasty (i.e. patching attempted in all participants) with primary closure. Any type of patch material was eligible (e.g. vein graft, Dacron, or polytetrafluoroethylene (PTFE)). We also intended to include trials comparing selective patch angioplasty (i.e. patching attempted only in people thought likely to benefit) with primary closure. Trials that compared one type of patch with another are included in a separate Cochrane Review (Bond 2003).

Types of outcome measures

We aimed to extract from each trial the number of participants originally allocated to each treatment group to allow an intention‐to‐treat analysis. Within each treatment group we then extracted the number of participants.

Electronic searches

The Cochrane Stroke Group's Information Specialist searched the Cochrane Stroke Group's Trials Register (last search 13 September 2021). We also updated electronic searches and handsearched additional issues of relevant journals as follows.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2021, Issue 8), in the Cochrane Library (searched 13 September 2021; Appendix 1)

• MEDLINE Ovid (1946 to 13 September 2021; Appendix 2)

• Embase Ovid (1980 to 13 September 2021; Appendix 3)

• Web of Science Core Collection (last searched 13 September 2021; Appendix 4).

Searching other resources

We searched for ongoing trials in the US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov; searched 13 September 2021), and the World Health Organization (WHO) International Clinical Trials Registry Platform (apps.who.int/trialsearch; searched 13 September 2021), in accordance with Cochrane's MECIR standards (Higgins 2016).

We handsearched the following journals including conference supplements.

• Annals of Surgery (1981 to 13 September 2021)

• Annals of Vascular Surgery (1994 to 13 September 2021)

• Cardiovascular Surgery (now Vascular; 1994 to 13 September 2021)

• European Journal of Vascular Surgery (now European Journal of Vascular and Endovascular Surgery; 1987 to 13 September 2021)

• Journal of Vascular Surgery (1994 to 13 September 2021

• Stroke (1994 to 13 September 2021)

We reviewed the reference lists of all relevant studies, and we contacted experts in the field to identify further published and unpublished studies.

For the previous version of the review, we handsearched the following journals including conference supplements.

• American Journal of Surgery (1994 to 13 September 2021)

• British Journal of Surgery (1985 to 13 September 2021)

• World Journal of Surgery (1978 to 13 September 2021)

We handsearched abstracts of the following meetings for the years 1995 to 13 September 2021.

• AGM of the Vascular Surgical Society (UK)

• AGM of the Association of Surgeons of Great Britain and Ireland

• American Heart Association Stroke Conference

• Annual Meeting of the Society for Vascular Surgery (USA)

• The European Stroke Conference

Selection of studies

Three review authors (SO, TB, and KR) independently read the titles and abstracts of records obtained from the searches, excluded obviously irrelevant studies, and selected those trials that met the inclusion criteria. We obtained the full‐text articles of potentially relevant studies on primary closure and patch angioplasty after CEA. The same three review authors screened all documents and independently extracted data, including details of methods, participants, setting, context, interventions, outcomes, results, publications, and investigators. We resolved all disagreements through discussion and performed meta‐analysis using Review Manager 5 (Review Manager 2020). The same three review authors also assessed the methodological quality of each trial.

Data extraction and management

Two review authors (SO, TB) independently reviewed and assessed all trials (so each trial received two assessments) and double‐checked all data extracted. We recorded data on:

• randomisation method;

• blinding of clinical and Doppler assessments;

• whether outcomes were reported for all participants originally randomised to each group irrespective of whether they received the operation they were allocated to; and

• number of participants lost to follow‐up.

We sought data on the number of outcome events for all participants originally randomised to allow an intention‐to‐treat analysis. For the 11 included trials, we extracted and cross‐checked all data. In addition, we extracted details about:

• the trial participants;

• inclusion and exclusion criteria;

• comparability of the treatment and control groups for important prognostic factors;

• type of patch;

• type of anaesthetic;

• use of shunts; and

• use of antiplatelet therapy during follow‐up.

If any of the above data were not available from the publication, we sought further information by correspondence with the trialists. We resolved all disagreements through discussion with other review authors (BS, KR).

Assessment of risk of bias in included studies

Two review authors (SO, BS) independently assessed risk of bias (high risk, low risk, unclear risk) using the Cochrane risk of bias tool (RoB 1), and presented the results in risk of bias tables (Higgins 2011). We resolved all disagreements through discussion. Risks of bias included:

• random sequence generation and allocation concealment (selection bias);

• blinding of participants and personnel (performance bias);

• blinding of outcome assessment (detection bias); and

• incomplete outcome data (attrition bias).

Measures of treatment effect

Using Review Manager 5 (Review Manager 2020), we carried out statistical analyses to determine the effect estimates and describe the magnitude of the intervention effect in terms of differences in outcomes between the two groups. To quantify the effect of the intervention on our outcomes, all of which were dichotomous, we used odds ratios (ORs) together with the corresponding 95% confidence intervals (CIs). 

We calculated proportional risk reductions based on weighted estimate of the OR using the Peto method (Deeks 2021). For rare events, the Peto one‐step OR method was found to be the least biased and most powerful method, with the best CI coverage, provided there was no substantial imbalance between treatment and comparator group sizes within studies, and treatment effects were not exceptionally large (Deeks 2021).

Unit of analysis issues

All selected trials included both unilateral and bilateral carotid endarterectomies. It was therefore possible for some participants to have primary closure on one side and carotid patching on the other side (İyigün 2019). Indeed, in one trial, if a participant required bilateral endarterectomies, each artery had to have a different procedure (Myers 1994). In most trials, the artery rather that the participant was randomised to a particular procedure (AbuRahma 1996; Al‐Rawi 2006; De Vleeschauwer 1987; Lord 1989; Mannheim 2005; Myers 1994; Pratesi 1986; Ranaboldo 1993); and only three trials were participant‐randomised (Eikelboom 1988; İyigün 2019; Katz 1994). Arterial randomisation is helpful for measuring outcomes such as haemorrhage, occlusion, ipsilateral events or complications within 30 days of surgery (as most participants waited at least this period between the first and second operation), but not for long‐term clinical outcome events such as death or any stroke. In participants with bilateral endarterectomies who had both patching and primary closure, it would not be possible to relate death or stroke to one particular procedure. Therefore, in trials where it was possible for a participant to have both procedures, we analysed death and any stroke only in those who had unilateral procedures or the same procedure in both arteries. These data were available from the authors in all except two trials (Lord 1989; Myers 1994). In trial, the authors no longer had the original data on participants with unilateral operations, so we excluded it from the analyses of these outcomes (Lord 1989). In the other trial, the number of participants undergoing unilateral endarterectomies was reported, and we were able to estimate the number of clinical events per participant in each group using the number of events per artery and the total number of deaths that were reported (Myers 1994).

We performed a separate analysis of only strokes ipsilateral to the operated artery for all arteries. However, the more clinically relevant outcome is the total number of strokes and not just ipsilateral strokes. We analysed arterial complications – such as occlusion, haemorrhage of the endarterectomy site, restenosis, infection at the operation site, or pseudoaneurysm formation – for all arteries rather than participants. Analyses based on arteries assumed that, in participants who had bilateral endarterectomies, outcome events in each carotid artery were independent. This is unlikely to be true, but as relatively few participants had bilateral procedures (10% overall), we felt it reasonable to perform such analyses. However, their results should be interpreted with caution. 

Outcomes analysed per artery were:

• ipsilateral stroke within 30 days of the operation and during long‐term follow‐up;

• occlusion of the operated artery within 30 days of the operation;

• rupture or haemorrhage of the endarterectomy site;

• infection of the endarterectomy site;

• cranial nerve palsy;

• complications requiring reintervention within 30 days of the operation; and

• restenosis or occlusion of the operated artery during long‐term follow‐up.

The outcomes analysed per participant included:

• any stroke;

• stroke‐related death;

• all‐cause death; and

• any stroke or death, all within 30 days of the operation and during long‐term follow‐up.

Dealing with missing data

When data were missing, we contacted the corresponding author or co‐author at the address given in the publication. If we received no response, we searched for the study group via the internet and contacted group members for missing information. In total, 58 participants were lost to follow‐up (Table 1). For the intention‐to‐treat analyses and the main analyses, we assumed that participants who were lost to follow‐up did not have an outcome event.

Assessment of heterogeneity

We assessed between‐study heterogeneity using the I² statistic, which examines the percentage of total variation across studies due to heterogeneity rather than to chance (Higgins 2020). Thresholds for interpretation of the I² statistic can be misleading, in that the importance of inconsistency depends on several factors. A rough guide to interpretation in the context of meta‐analyses of randomised trials is as follows.

• Values of I² over 75% indicate a high level of heterogeneity.

• Values of I² equal or below 75% indicate a low‐to‐moderate level of heterogeneity.

• An I² value of 0% indicates no heterogeneity.

Assessment of reporting biases

We performed an extensive literature search, and we are confident that we have identified all major relevant trials. We also contacted experts in this field. We searched for trials published in all languages, and we arranged translation of all possibly relevant publications when required. In addition, we searched all relevant ongoing clinical trials from ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP) portal, and we handsearched relevant journals and reference lists. We had planned to compare study protocols with final study reports to evaluate selective reporting of outcomes. We used funnel plots to assess publication bias because more than 10 studies were included. However, none of the trials reported limits outside the 95% CI.

Data synthesis

We calculated proportional risk reductions based on a weighted estimate of the OR using the Peto method (Deeks 2021). Since all the outcome events assessed were rare, the ORs quoted will be similar to the relative risks. The Peto one‐step OR method was found to be the least biased and most powerful method for the rare event rate study (Deeks 2021). We calculated absolute risk reductions from the crude risks of each outcome in all trials combined (Deeks 2021).

Subgroup analysis and investigation of heterogeneity

We planned to explore heterogeneity by conducting the following subgroup analyses.

• Age (younger than 65 years versus 65 to 74 years versus 75 years or older)

• Sex (men versus women)

• Diabetes versus no diabetes

• Hypertension versus no hypertension

• Previous myocardial infarction or angina versus no coronary artery disease

• Peripheral arterial disease (PAD) versus no PAD

• Current smoker versus non‐smoker

• Asymptomatic versus symptomatic carotid stenosis

• Contralateral carotid stenosis versus unilateral carotid stenosis

• Contralateral carotid occlusion versus no occlusion

• Preoperative antiplatelet therapy versus no antiplatelet therapy

• Intraoperative shunt versus no shunt

• Irregular or ulcerated symptomatic carotid plaque versus smooth plaque on the pre‐randomisation angiogram

We planned to use an established method for subgroup analyses (Deeks 2001). We will fulfil planned subgroup analyses when more studies are included in a single analysis, with sufficient information for each subgroup.

Analyses were stratified by patch type within the patch angioplasty group. Tests for overall effect and subgroup differences by patch type included:

• venous patch versus primary closure;

• synthetic patch versus primary closure; and

• synthetic or venous patch versus primary closure.

Each subgroup was analysed for 10 perioperative and six long‐term outcomes; therefore, we examined a total of 16 groups of data. We did not investigate potential effect modifiers via subgroup analysis.

Sensitivity analysis

Before we had clearly defined the methods for this systematic review, we planned to undertake different sensitivity analyses to explore effects of certain methodological features, excluding, in turn, studies at high risk of selection bias, detection bias, attrition bias, reporting bias, and other bias (not peer reviewed).

Given that foreknowledge of treatment allocation might lead to biased treatment allocation and exaggerated treatment effects (Schulz 1995), in the original version of this review we performed separate sensitivity analyses of trials with adequate allocation concealment and those with inadequate allocation concealment. For the intention‐to‐treat analyses and the main analyses, we assumed that participants who were lost to follow‐up did not have an outcome event.

However, the current review authors considered these analysis to be unreasonably critical, so did not carry out any sensitivity analyses for this update.

Summary of findings and assessment of the certainty of the evidence

We created a summary of findings table for the seven main comparisons with GRADE Profiler 3.6 (GRADEpro GDT 2015), which imports data from RevMan 5 (Review Manager 2020). This table presents the results and the certainty of the evidence (high, moderate, low or very low) on the main outcomes according to the GRADE criteria (Schünemann 2011). We included the following important outcomes.

• Any stroke (fatal or non‐fatal; contralateral, ipsilateral or brainstem; haemorrhage or infarct) within 30 days of the operation

• Ipsilateral stroke (fatal or non‐fatal; haemorrhage or infarct) within 30 days of the operation

• Occlusion of the operated artery within 30 days of the operation

• Any stroke (fatal or non‐fatal; contralateral, ipsilateral or brainstem; haemorrhage or infarct) during long‐term follow‐up, including events during the first 30 days

• Ipsilateral stroke (fatal or non‐fatal; haemorrhage or infarct) during long‐term follow‐up, including events during the first 30 days

• Any stroke or death during long‐term follow‐up, including events during the first 30 days

• Restenosis (greater than 50%) or occlusion of operated artery during long‐term follow‐up, including events during the first 30 days