Selection of studies

After applying the search strategy, two review authors will independently identify studies meeting the inclusion criteria for this review. We will resolve discrepancies between review authors by discussion. If we cannot reach consensus, we will achieve final resolution using a third‐party arbitrator. We will obtain the complete article of any study that seems to meet the inclusion criteria, in accordance with the title or the abstract, or both. We will produce a Characteristics of included studies table, and include detailed information for each study. We will clearly state details of the reasons for exclusion of any study considered for the review in the Characteristics of excluded studies table. We will include a PRISMA flow diagram of the selection of studies in the review. If there are multiple reports of the same study, we will use the most recent report as the primary publication; we will check the other available reports for data not reported in the primary publication.

Data extraction and management

Two review authors will independently extract data, using standardised forms. We will resolve discrepancies between review authors by discussion. If we cannot reach consensus, we will achieve final resolution using a third‐party arbitrator. We will extract data on the characteristics of participants (e.g. age, gender, and other known risk factors in participants (tumour volume, primary metastatic disease, time of relapse after primary diagnosis, type of relapse and sites of metastases, response to second‐line therapy)), interventions, outcome measures, study design, length of follow‐up, details of funding sources, and declaration of interests for each included study.

Assessment of risk of bias in included studies

Two review authors will independently assess the risk of bias in included studies (i.e. selection bias, performance bias, detection bias (for each outcome separately), attrition bias (for each outcome separately), reporting bias, and other bias). We will use the risk of bias items and definitions of low risk of bias, unclear risk of bias, and high risk of bias as described in the module of the Childhood Cancer Group (Module CCG), which are based on the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will resolve discrepancies between review authors by discussion. If we cannot reach consensus, we will achieve final resolution using a third‐party arbitrator. We will take into account the risk of bias in included studies in the interpretation of the review's results.

Measures of treatment effect

We will analyse dichotomous variables using risk ratios (RR). We will analyse survival using hazard ratios (HR). We will use Parmar's method if HRs have not been explicitly presented in the study (Parmar 1998). We will present all results with the corresponding 95% confidence interval (CI).

Dealing with missing data

When relevant data regarding study selection, data extraction, and risk of bias assessment are missing, we will attempt to contact the study authors to retrieve the missing data. We will extract data by the allocated intervention, irrespective of compliance with the allocated intervention, in order to allow an intention‐to‐treat analysis. If this is not possible, we will state this, and we will perform an as treated analysis.

Assessment of heterogeneity

We will assess heterogeneity both by visual inspection of the forest plots, and by a formal statistical test for heterogeneity (i.e. the I² statistic). In the absence of significant heterogeneity (I² less than 50%), we will use a fixed‐effect model for the estimation of treatment effects (Higgins 2011). Otherwise, we will explore possible reasons for the occurrence of heterogeneity, and take appropriate measures, such as using a random‐effects model.

Assessment of reporting biases

In addition to evaluating reporting bias as described in the Assessment of risk of bias in included studies section, we will assess reporting bias by constructing a funnel plot, when there are a sufficient number of included studies (i.e. at least 10 studies included in a meta‐analysis). When there are fewer studies, the power of the test is too low to distinguish chance from real asymmetry (Higgins 2011).

Data synthesis

We will enter data into Cochrane's Review Manager 5 software, and undertake analyses according to the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011; Review Manager 2020). We will include outcome measures only if it was the intention of the study to perform the necessary assessments in all randomised participants (i.e. not only optional, or only performed in some centres). When the results of a particular outcome measure are available for less than 50% of the participants of a study, due to the associated high risk of attrition bias, we will not report the results of this outcome measure. We will pool results only if both treatment groups are comparable, including the definition of outcomes used. We will provide a descriptive summary for studies for which pooling of results is not possible. We do not expect multi‐arm studies (i.e. including more than two treatment groups); however, if we include these studies, we will take appropriate measures as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will analyse historical controlled clinical trials separately. We will analyse studies that compare immunotherapy plus high‐dose chemotherapy with autologous haematopoietic cell transplantation.

For each comparison, we will prepare a summary of findings table, using GRADEpro GDT software to calculate absolute effects, and the certainty of the evidence for each outcome (GRADEpro GDT). Two review authors will independently assess the certainty of the evidence, which we will rate as very low, low, moderate, or high according to the five GRADE considerations (i.e. study limitations, inconsistency, indirectness, imprecision, and publication bias), described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). For each comparison, we will include evidence for the following outcomes: event‐free survival, overall survival, quality‐adjusted survival, toxicity, and progression‐free survival.

Subgroup analysis and investigation of heterogeneity

We will not perform subgroup analyses.

Sensitivity analysis

For all outcomes for which pooling is possible, we will perform sensitivity analyses for all risk of bias criteria separately. We will exclude studies with a high risk of bias or unclear risk of bias, and compare the results of studies with a low risk of bias, with the results of all available studies.