Types of studies

We searched all randomised controlled trials (RCTs) and quasi‐RCTs (RCTs in which allocation methods were not completely random, for example using alternation). We excluded cross‐over trials as the response to corticosteroid intervention may depend on timing and previous treatment state. We excluded all studies not conforming to the RCT format.

Types of participants

We included all studies of children (less than 19 years old) worldwide diagnosed with KD. The diagnosis of KD should be made according to existing KD guidelines at the time of the study. The KD diagnostic guidelines include the following principal clinical features (Kobayashi 2020).

1. Fever (without requirement for a specific duration).

2. Bilateral conjunctivitis.

3. Changes of lips and oral cavity: reddening of lips, strawberry tongue, diffuse injection of oral and pharyngeal mucosa.

4. Rash (including redness at the site of Bacille Calmette Guerin (BCG) inoculation).

5. Changes in the extremities (reddening, swelling and peeling of the skin).

6. Non‐suppurative cervical lymphadenopathy.

For the Japanese KD research committee guidelines, complete KD is diagnosed with at least five principal clinical features, or with four if there is evidence of coronary artery abnormality found on echocardiography or at angiography. Incomplete KD is diagnosed when there are three principal clinical features plus coronary artery abnormality, or if there are other significant clinical features, which are detailed in the guidelines (Kobayashi 2020). In either case, other febrile illnesses should be excluded before diagnosing KD. The Japanese guidelines also recommend using Z‐scores to evaluate the internal coronary artery diameter (Kobayashi 2020).

The Japanese KD research committee and the American Heart Association guidelines are similar, but the latter specify the duration of fever; classic (complete) KD diagnosed in the presence of fever for five days along with at least four principal clinical features (McCrindle 2017). In the presence of four or more features, especially if redness and swelling of the extremities are present, the diagnosis of KD can be made after four days of fever (or three days if diagnosed by an experienced clinician). A diagnosis of incomplete (or atypical) KD should be considered in a child with prolonged unexplained fever, some of the principal clinical features and compatible laboratory findings (such as raised inflammatory markers e.g. CRP, ESR) or echocardiographic evidence (McCrindle 2017).

Corticosteroids could have been part of the initial treatment for KD or have formed part of the second‐line treatment after failure of first‐line treatment that did not include corticosteroids. The comparison group had to be in parallel. Cross–over trials were not eligible for inclusion.

We excluded participants with positive blood cultures.

Types of interventions

All forms of corticosteroid therapy in conjunction with any combination of no treatment, placebo, immunoglobulin, aspirin or infliximab for the treatment of KD were the intervention of interest. That stated, the use of corticosteroids had to be the only difference in management between trial arms.

Comparator groups included any of:

1. placebo;

2. immunoglobulin only;

3. aspirin only;

4. immunoglobulin and aspirin;

5. infliximab only;

6. infliximab and immunoglobulin;

7. infliximab, immunoglobulin and aspirin.

Types of outcome measures

Electronic searches

The Cochrane Vascular Information Specialist conducted systematic searches of the following databases for RCTs and controlled clinical trials without language, publication year or publication status restrictions.

1. Cochrane Vascular Specialised Register via the Cochrane Register of Studies (CRS‐Web searched on 8 February 2021).

2. Cochrane Central Register of Controlled Trials (CENTRAL) Cochrane Register of Studies Online (CRSO 2021, Issue 1).

3. MEDLINE (Ovid MEDLINE Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE) (searched from 25 November 2016 to 8 February 2021).

4. Embase Ovid (searched from 25 November 2016 to 8 February 2021).

5. CINAHL EBSCO (searched from 25 November 2016 to 8 February 2021).

The Information Specialist modelled search strategies for other databases on the search strategy designed for CENTRAL. Where appropriate, they were combined with adaptations of the highly sensitive search strategy designed by Cochrane for identifying RCTs and controlled clinical trials (as described in the Cochrane Handbook for Systematic Reviews of Interventions Chapter 6, Lefebvre 2011). Search strategies for major databases are provided in Appendix 1.

The Information Specialist searched two trials registries on 8 February 2021.

1. World Health Organization International Clinical Trials Registry Platform (who.int/trialsearch).

2. ClinicalTrials.gov (clinicaltrials.gov).

Searching other resources

We contacted the authors of trials that met the eligibility criteria identified by the searches as ongoing or unpublished trials. We also searched reference lists of relevant trials for further publications.

Selection of studies

For this update, two review authors (JG and RMRT) independently applied the selection criteria to the studies identified by the search strategy. This included independently assessing whether the studies fulfilled the inclusion and exclusion criteria. If there was insufficient information to decide whether a study was truly eligible we contacted the study authors to request further information. A third review author (AJW) resolved disagreements.

We identified and excluded duplicates and collated multiple reports of the same study, so that each study rather than each report was the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram.

Data extraction and management

For this update, two review authors (JG and RMRT) independently extracted data using a modified version of the Cochrane Vascular standard data extraction form. These data were then brought together and monitored for discrepancies by a third review author (AJW), before being entered into Review Manager 5 (Review Manager 2014). We contacted study authors for any required information that was not included in the publications. The key information gathered in the data collection form included the following.

1. General study information: publication type.

2. Fulfilment of eligibility criteria: study type, interventions, outcomes measured and reasons for exclusion.

3. Study methods: allocations methods, study dates, duration, ethical approval and statistical methods.

4. Participants: methods of recruitment, consent, total number, treatment groups, age, sex, race, KD severity, subgroup analyses reported and eligibility criteria.

5. Intervention (corticosteroids): number of participants, dosing, frequency, duration, delivery method, providers, compliance and concomitant treatment.

6. Outcomes: coronary diameters (acute), coronary diameter z‐scores (acute), coronary abnormality (long term), duration of clinical symptoms (e.g. fever), adverse effects, duration of laboratory parameter abnormality (e.g. CRP) and duration of hospital stay.

7. Study funding and study author declarations of interest.

Assessment of risk of bias in included studies

We assessed the risk of bias using Cochrane's RoB 1 tool as described in section 8.5 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). In this update, two review authors (JG and RMRT) performed this independently, and resolved any disagreements by discussion with a third review author (AJW). The domains assessed included:

1. sequence generation (selection bias);

2. allocation sequence concealment (selection bias);

3. blinding of participants and personnel (performance bias);

4. blinding of outcome assessment (detection bias);

5. incomplete outcome data (attrition bias);

6. selective outcome reporting (reporting bias);

7. other bias.

Measures of treatment effect

The effect measure of choice for dichotomous data was the odds ratio (OR) with a 95% confidence interval (CI). This is a ratio between the corticosteroid intervention group and its parallel comparator group. We reported continuous data, including time‐to‐event data, using mean differences (MD) and 95% CIs. If necessary, we planned to use a standardised mean difference (SMD) for studies that measured the same outcome but used different methods.

Unit of analysis issues

The unit of analysis was each participant recruited into a trial.

Dealing with missing data

We accounted for all missing data due to dropouts via an intention‐to‐treat (ITT) analysis. We reported if the individual trials conducted ITT analyses. If they did not, then we endeavoured to apply an ITT analysis. In the event that we were unable to do this, we utilised a per protocol analysis. We have explained all post‐allocation dropouts. We have followed up any data missing from the published document with the study's original authors. If data still remained absent, we considered this in the risk of bias assessments.

Assessment of heterogeneity

We measured heterogeneity using the I² statistic for quantification of variability (less than 40% = likely low heterogeneity; 40% to 60% = possible moderate heterogeneity; greater than 60% = possible significant heterogeneity) (Higgins 2011). We also considered the magnitude and direction of effects using the Chi² test (limit = degrees of freedom) and P values (10% significance threshold) (Higgins 2011). Where heterogeneity exceeded generally accepted limits (greater than 60% heterogeneity), we subgrouped the analysis in a logical manner to explain these differences and reduce remaining heterogeneity. These methods were reinforced by visual recognition on forest plots to assess for overlapping CIs (Higgins 2011).

Assessment of reporting biases

We planned to screen for publication and reporting bias using funnel plot asymmetry and measure using tests if there were 10 or more trials, as outlined in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). For smaller studies, we considered 'the small‐study effect', where smaller studies can show larger treatment effects due to poor methodology, heterogeneity, selection bias, chance or artefact. We also searched for eligible studies that had been registered and should have been completed, but were without available published data. It should be noted that the risk of bias assessment has taken into account selective outcome reporting.

Data synthesis

We only undertook meta‐analyses when it was meaningful to do so. That is if the treatments, participants and the underlying clinical question were sufficiently similar for pooling to make sense. Statistical analysis took place using a fixed‐effect model if there was low heterogeneity, and a random‐effects model if there was significant heterogeneity (I² > 60%). We undertook outcome analyses using an ITT model. Two‐sided P values of 0.05 or less were considered significant and all analyses were undertaken using Review Manager 5 (Review Manager 2014). If meta‐analysis was not possible, we planned to report the results using a narrative synthesis.

Subgroup analysis and investigation of heterogeneity

We planned to include the following subgroup analyses (data permitting):

1. first‐line versus second‐line management;

2. type of corticosteroid used;

3. corticosteroid dosing;

4. corticosteroid treatment frequency;

5. total corticosteroid treatment duration;

6. corticosteroid route of administration;

7. geographical distribution of trial participants, ethnicity;

8. KD severity (non‐high risk versus high risk as detailed earlier);

9. recognised concomitant treatments for KD (as detailed in Types of interventions).

We also employed further subgroup analyses if heterogeneity remained significant (I² > 60%) using a Chi² P value threshold of 0.05.

Sensitivity analysis

We planned to perform sensitivity analysis to explore causes of heterogeneity and the robustness of the results if there were sufficient data available. We planned to include the following factors in the sensitivity analysis.

1. Type of study design (RCT versus quasi‐RCT).

2. Low risk of bias trials versus high risk of bias trials. A trial was defined as high risk if any domain in the trial was assessed at high risk of bias.

3. Rates of dropouts for each treatment group. We planned to perform a sensitivity analysis if the rates of dropout were considered sufficient to impact results.

Summary of findings and assessment of the certainty of the evidence

We presented the main findings of the review results concerning the certainty of the evidence, the magnitude of effect of the interventions examined and the sum of available data on the outcomes (Types of outcome measures) in summary of findings Table 1. We assessed the certainty of the evidence for each primary and secondary outcome as high, moderate, low and very low, based on within‐study risk of bias, directness of evidence, heterogeneity, precision of effects estimates, and risk of publication bias, according to the GRADE principles as described by Higgins 2011 and Atkins 2004. We used the GRADEpro GDT software to assist in the preparation of the summary of findings table (www.guidelinedevelopment.org/).