Anticoagulation and associated complications in Veno-Arterial Extracorporeal Membrane Oxygenation in Adult Patients: A Protocol for a Systematic Review and Meta-Analysis

1.A Types of studies

We will include all randomized controlled trials (RCTs), quasi-RCTs and controlled and uncontrolled cohort studies that describe anticoagulation and bleeding or thrombotic complications in adult patients on VA-ECMO. We define quasi-randomization as non-true randomization to each arm by means such as alternation, case record number or date of birth. Case series and cohort studies with a sample less than ten patients will be excluded. We will exclude physiological modelling studies or case reports. Studies that include both VA and VV-ECMO will be included if the VA patients’ results are reported separately. There will be no language restrictions to included studies. Abstracts and studies in press will be included. Due to the evolution in ECMO technology and accumulating institutional experience, only studies published after 2010 will be included.

1.B Types of participants

We will include studies involving adults aged 16 years and older admitted to intensive care for any indication who were commenced on VA-ECMO.

1.C Types of interventions

We will include studies that assess any intervention in patients on VA-ECMO. All cannulation configurations will be included, including hybrid configurations, central and peripheral cannulation will be included. Veno-pulmonary arterial ECMO (V-PA) will not be considered VA ECMO and will not be included. All indications for ECMO will be included, including post-cardiotomy and extracorporeal cardiopulmonary resuscitation (E-CPR). Where studies report a mixed cohort of VV and VA-ECMO, only studies that describe outcomes of interest separately for the VA-ECMO patients will be included. In order to be included, a study must report the following:

1. the anticoagulant drug used

   OR

2. the test for monitoring used

   AND

3. at least 1 included outcome of interest.

2.A Selection of studies

Two review authors (RV and one additional author) will independently screen all titles and abstracts of each reference identified by our search and will then independently assess the full text of any potentially relevant studies for eligibility or exclusion. We will use Covidence software to collate search results, remove duplicates and record screening decisions and exclusions at each stage [17]. This will be facilitated by a standardized digital screening, inclusion and exclusion tools, in line with methods outlined in the Cochrane Handbook for Systematic Reviews of Interventions [18]. We will resolve any disagreements by discussion and consensus prior to proceeding at each stage.

We will record the process in sufficient detail to produce a PRISMA flow diagram [15]. For all full-text articles screened and excluded, we will record the reason for exclusion in sufficient detail to present a table of characteristics for excluded studies.

2.B Data extraction and management

Through Covidence, we will use a standardized digital data-extraction sheet, in line with requirements outlined in the Cochrane Handbook for Systematic Reviews of Interventions Version 6.3 [18]. Two review authors (RV and one additional author) will independently extract information regarding trial design, data regarding the anticoagulation strategy, outcomes of interest (as reported below) and information relevant to Risk of Bias grading. Where required, we (RV and one additional author) will contact individual trial authors or organizations to obtain missing data or clarification regarding unclear data. We will resolve any disagreements by discussion and consensus.

2.C Assessment of risk of bias in included studies

Two review authors (RV and one additional author) will independently assess methodological quality. For all included trials we will utilize Cochrane’s Risk of Bias 2 (ROB 2) tool for RCTs [18]. There are five specific measurements covered by this tool, and the Cochrane Handbook for Systematic Reviews of Interventions describes how each domain should be assessed to reach a judgement of low risk, some concerns or high risk of bias. If a study is judged to be low risk in all domains, then it will be deemed to be at low risk of bias overall.

The criteria include:

1. Bias arising from randomisation process;

2. Biase due to deviations from intended intervention;

3. Bias due to missing outcome data;

4. Bias in measurement of outcome data;

5. Bias in selection of the reported data;

For non-randomized controlled trials, the ROBINS-I tool will be used to assess risk of bias [19]. The criteria for the ROBINS-I tool include pre-intervention, at intervention and post-intervention domains with assessment of selection bias, information bias, confounding and reporting bias. Bias surrounding deviation in intended intervention is anticipated, given observational studies are unlikely to report on intentional deviation of anticoagulation management from institutional protocols in patients with increased bleeding or thrombosis.

We will present a ‘Risk of Bias’ summary, as well as ‘Risk of Bias’ judgements for individual studies in the ‘Characteristics of included studies’ tables. Given anticipated heterogeneity in patients and clinical practice surrounding management of VA-ECMO, we will be hesitant drawing firm conclusions from the pooled analyses.

4.A Electronic searches

We will search electronic databases for articles using key words, synonyms and subject headings that relate to ECMO, anticoagulation and relevant complications. We will use controlled vocabulary specific to each database (see ‘Search Strategy’). Two review authors (RV and one additional author) will perform the search independently and no language restrictions will be applied. We will contact individual trial authors for additional information where necessary.

We will search the following databases for published trials:

1. Cochrane Central Register of Controlled Trials (CENTRAL) (1996 to present)

2. MEDLINE Ovid (1946 to present)

3. Elsevier Embase (1947 to present)

4. CINAHL EBSCO (1937 to present)

4.B Searching other resources

We will perform a citation search of all included studies as well as any relevant studies and reviews on anticoagulation for ECMO. If this identifies any additional eligible studies, we will re-examine and update the search strategy. We will also perform a search of Google Scholar for additional trials. We will consider for inclusion studies that are available as abstracts only.

5.A Subgroup analysis and investigation of heterogeneity

Where sufficient data is available, subgroup analyses will be performed for each outcome based on:

1. Anticoagulant choice

2. Anticoagulation target

3. Method of monitoring anticoagulation

4. Cannula configuration (central vs peripheral, percutaneous vs open peripheral)

5. ECPR

6. Post cardiotomy

7. Oxygenator surface area (=<2.0 m²)

5.B Sensitivity analysis

We will perform a sensitivity analysis of studies to explore the influence of ‘Risk of bias’ assessments. This will allow assessment of the robustness of our conclusions. These sensitivity analyses will be limited to the primary outcomes of composite major bleeding events and composite major thrombotic events.

5.C Dealing with missing data

We will contact trial authors to request missing data. We will outline any missing data in the results and explain in the discussion the extent to which the overall results may be affected. In the event of unreported standard deviations, we will make an estimate based on other reports within the meta-analysis.

5.D Assessment of heterogeneity

If there is obvious extensive clinical heterogeneity, we will not perform a meta-analysis. Otherwise, we will investigate clinical heterogeneity via meta-regression, identifying trends within subgroups of the clinical population as detailed above.

We will utilize the I² test to assess statistical heterogeneity for each outcome [17]. We will judge I² values as follows;

• 0% to 40%: might not be important;

• 41% to 60%: may represent moderate heterogeneity.

• 61% to 90%: may represent substantial heterogeneity.