Types of studies

We will include randomized controlled trials (RCTs), including cluster‐controlled trials, cross‐over trials, and quasi‐RCTs (RCTs in which participants are allocated based on data such as date of birth or recruitment or medical record number).

Types of participants

We will include all adult participants (≥ 18 years old) with any clinical characteristic, in any setting, who require a peripheral intravenous line, irrespective of the difficulty of cannulation. We will define a peripheral intravenous line as a catheter placed in a peripheral vein. We will exclude central lines, intraosseous lines, and peripherally inserted central lines. We will exclude children, because the effect of ultrasound guidance will be different for them, due to smaller veins and extremities, as well as a possible lack of cooperation. We will carry out a subgroup analysis to evaluate participants according to the difficulty of peripheral intravenous cannulation.

Types of interventions

We will include studies comparing ultrasound‐guided peripheral intravenous cannulation with the landmark method, irrespective of the profession of the operators, number of operators (one or two), methods (short‐axis or long‐axis, static or dynamic), or the sites of the peripheral veins. We will exclude studies on peripherally inserted central catheters. We will include all studies using ultrasonography, irrespective of the manufacturer or generation of ultrasound machines used.

Types of outcome measures

We will not use outcome measures as a criterion for excluding studies. All outcome measures below will be reported at time of cannulation.

Electronic searches

The Cochrane Vascular Information Specialist aims to identify all relevant RCTs regardless of language or publication status (published, unpublished, in press, or in progress).

The Information Specialist will search the following databases for relevant trials:

• the Cochrane Vascular Specialised Register via the Cochrane Register of Studies (CRS‐Web);

• the Cochrane Central Register of Controlled Trials (CENTRAL) via the Cochrane Register of Studies Online (CRSO);

• MEDLINE (Ovid MEDLINE Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE; 1946 onwards);

• Embase Ovid (from 1974 onwards);

• CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; from 1982 onwards);

• LILACS BIREME (Latin American and Caribbean Health Science Information database; from 1987 onwards).

The Information Specialist has devised a draft search strategy for RCTs for MEDLINE (Appendix 1), and will use this as the basis for search strategies for the other databases listed.

The Information Specialist will search the following trials registries:

• the World Health Organization International Clinical Trials Registry Platform (who.int/trialsearch);

• ClinicalTrials.gov (clinicaltrials.gov).

Searching other resources

Four review authors (MT, TM, CT, NY) will check the reference lists of all identified studies and review articles to find additional studies. We will contact trial authors, experts in this field and manufacturers of ultrasound machines to identify unpublished studies.

Selection of studies

We will use the reference management software Mendeley to collate the results of searches and to remove duplicates (Mendeley). We will use Rayyan software to screen the results of the search (Ouzzani 2016). Four review authors (MT, TM, CT, NY) will independently and in duplicate check titles and abstracts of the results of the search and will identify potentially relevant studies. We will obtain full texts of all potentially relevant studies if any of the authors judge them to be relevant or potentially relevant. We will exclude only the clearly irrelevant articles at this stage. Four review authors (MT, TM, CT, NY) will independently and in duplicate assess the full papers for eligibility using a pre‐designed checklist. We will compare the results and resolve disagreements through discussion. If we are unable to reach a consensus, we will consult the sixth review author (NW). We will record the number of papers retrieved at each stage and will report this information using a PRISMA flow chart.

Data extraction and management

We will use Covidence software to extract data from individual studies, adapting their basic template data extraction form (Covidence). Four review authors (MT, TM, CT, NY) will independently and in duplicate extract the data using the abstraction form. We will resolve disagreements through discussions. If we cannot reach a consensus, we will consult the sixth review author (NW). If additional information is needed, one review author (MT) will contact the corresponding author of the relevant studies. When we have completed data extraction, one review author will enter the data into Review Manager software and another review author will check the data (Review Manager 2014).

Assessment of risk of bias in included studies

Four review authors (MT, TM, CT, NY) will independently assess the methodological quality of each included study using the Cochrane 'Risk of bias' tool (Higgins 2011). We will assess the following domains and rate them as low, unclear, or high risk of bias.

• Random sequence generation (selection bias)

• Allocation concealment (selection bias)

• Blinding of participants and providers (performance bias)

• Blinding of outcome adjudicators/assessment (detection bias)

• Incomplete outcome data (attrition bias)

• Selective outcome reporting (outcome reporting bias)

• Other biases

Due to the nature of the intervention, blinding of the operators will not be possible. This may cause some performance bias, but it is unavoidable and we do not expect it to be serious. It is also not possible to blind the participants, but this should not affect objective outcomes, such as success rate of cannulation. Thus, we will rate the performance bias domain as low risk irrespective of blinding status. Because outcome adjudicators need to observe the procedure on the scene to evaluate the outcomes, it is not possible to blind them to treatment allocation. We will rate a study as being at low risk of bias if a non‐blinded independent outcome adjudicator reports the outcomes. We will rate a study as high risk of bias if the operator and the adjudicator are the same person.

We will define the overall risk of bias as follows.

• Low risk of bias: all seven domains rated as low risk

• Moderate risk of bias: one or more domains rated as being at unclear risk

• High risk of bias: one or more domains rated as being at high risk

We will review the assessments and resolve any disagreements through discussion. If needed, we will consult the sixth review author (NW).

Measures of treatment effect

We will calculate dichotomous data as risk ratios (RR) with 95% confidence intervals (CI). We will calculate continuous data as mean differences (MD) with 95% CI when the outcomes of all studies use the same scale. We will use a standardized mean difference (SMD) with 95% CIs if different scales are used.

Unit of analysis issues

The unit of analysis is the individual participant. If we include any cluster‐randomized trials, we will adjust the sample size by the trial's intracluster correlation coefficient, using the method described in the Cochrane Handbook for Systematic Reviews of Intervention (Higgins 2011). We will exclude any cluster‐randomized trials which do not report the intracluster correlation coefficient.

Dealing with missing data

We will contact the study authors when possible to obtain missing data. We will perform an intention‐to‐treat analysis when possible. We will impute data for binary outcomes using various scenarios such as 'best‐case' and 'worst‐case' scenarios. For continuous outcomes, we will use available‐case analysis. We will calculate the standard deviation from P values, standard errors or CI according to the instructions given in the Cochrane Handbook for Systematic Reviews of Intervention (Higgins 2011). Otherwise, we will impute them from other studies in the meta‐analysis according to the validated method (Furukawa 2006).

Assessment of heterogeneity

We will assess heterogeneity by inspecting forest plots visually, and will examine statistical heterogeneity by Chi² and I² statistic. We will use P = 0.10 as the predefined significance level of heterogeneity for the Chi² test. We will consider I² statistics of 25% or lower to indicate low heterogeneity, between 25% and 50% to indicate moderate heterogeneity, and 50% or more to indicate substantial heterogeneity. However, we will interpret this value in light of the size and direction of effect and the strength of the evidence for heterogeneity, based on the P value from the Chi² test (Higgins 2011). If we identify substantial heterogeneity, we will investigate and report potential reasons for this. We will define substantial heterogeneity as I² over 50%.

Assessment of reporting biases

We will try to minimize the effect of publication bias by performing well‐designed comprehensive literature searches, by using trial registries, such as ClinicalTrials.gov, and by contacting the manufacturers of ultrasound machines. If we include a sufficient number of studies in a meta‐analysis (i.e. more than 10 studies; Higgins 2011), we will visually inspect funnel plots to evaluate small study effects and use contour‐enhanced funnel plots to evaluate publication bias. We will evaluate reporting bias by checking the protocol of the study, if we can identify one from searching trial registries.

Data synthesis

We will review the data from the included studies and, if possible, synthesize and analyze data using Review Manager software (Review Manager 2014). We plan to use the random‐effects model to pool data, because we expect the definitions of participants and operators to vary to some extent between studies, and also because the random‐effects model is more conservative than the fixed‐effect model. If it is not possible to pool data, we will provide clear reasons for this and report results narratively.

Subgroup analysis and investigation of heterogeneity

We plan to perform subgroup analyses for the following parameters, if we find sufficient data from the included studies.

• Difficulty of obtaining intravenous access: 'difficult' versus 'moderately difficult' versus 'easy'

Because the effectiveness of ultrasound guidance will vary depending on the difficulty of obtaining intravenous access in each participant, we will evaluate participants separately for each difficulty level, according to the following criteria:

• • we will use the definition of the difficulty of peripheral intravenous cannulation adopted by original studies;

  • we will define the difficulty based on the first‐pass success rate or the overall success rate of cannulation using the landmark method.

Because we expect the definition of difficulty of peripheral intravenous cannulation to differ between studies (Egan 2013; Liu 2014), we will also define the difficulty based on the first‐pass success rate or the overall success rate of cannulation using the landmark method. As in previous studies, we will categorize a success rate of less than 60% as 'difficult', a success rate of 60% to 80% as 'moderately difficult', and a success rate of over 80% as 'easy' (McCarthy 2016; Sebbane 2013; van Loon 2016).

• Practical difficulties of obtaining intravenous access: difficult versus not difficult

We will analyze participants separately where he or she satisfies the definition of a difficult case. A difficult case is defined as any of the following:

• • the operator could not see and palpate the targeted vein;

  • the operator identified a participant as a difficult case;

  • the participant had a history of difficult intravenous access;

  • the participant had multiple failed attempts.

• Experience of ultrasound‐guided cannulation: training versus clinical experience versus training plus clinical experience

We will analyze separately if operators satisfy the following criteria:

• • having finished any kind of training program for ultrasound‐guided peripheral venous cannulation;

  • having clinical experience of ultrasound‐guided peripheral intravenous cannulation;

  • having finished any kind of training program for ultrasound‐guided peripheral venous cannulation plus having clinical experience of ultrasound‐guided peripheral intravenous cannulation.

• Study settings: emergency departments or intensive care units (ICUs) versus operating rooms

Compared to participants in operating rooms, those in emergency departments or ICUs are more likely to be in shock or dehydrated. Since these factors are associated with difficult intravenous cannulation, the ultrasound‐guided method may be more effective in the setting of emergency departments or ICUs than in operating rooms.

• Date of publication: 1999 to 2008 versus 2009 to 2019

Advances in machine technology may lead to improved ultrasound image quality, improving the effectiveness of ultrasound guidance. Therefore, we will stratify the studies by publication year into two groups: 1999 to 2008, and 2009 to 2019. If we include more than 10 studies, we will also perform univariate meta‐regression with Stata software, using publication year as a continuous covariate (Stata).

• Types of ultrasound guidance

We will analyze studies separately according to type of ultrasound guidance:

• • short‐axis technique versus long‐axis technique;

  • dynamic method versus static method.

Sensitivity analysis

We plan to perform sensitivity analyses for the following factors, if applicable.

• We will limit the analysis to studies with low risk of bias. We define low risk of bias as satisfying all of the following domains: adequate allocation concealment, blinding of outcome assessment, and data analysis performed according to the intention‐to‐treat principle.

• We will limit the analysis to RCTs only.