Types of studies

Randomized controlled trials (RCTs).

Types of participants

Types of interventions

Autologous hematopoietic stem cell transplantation (HSCT), stem cells from a peripheral source or the bone marrow, serving as a rescue therapy usually applied after high‐dose chemotherapy (HDCT) versus standard‐dose chemotherapy (SDCT), which is defined as chemotherapy at a lower dose than HDCT without the need for stem cell rescue.

Types of outcome measures

Electronic searches

We conducted an electronic search in the following medical literature databases. We carefully revised the search strategies to improve precision.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 8). The search strategy is shown in Appendix 1. The search dates were limited from 1 January 2012 to 6 September 2016. The search dates of the previous version included references from inception to 5 December 2012.

• PubMed. The search strategy is shown in Appendix 2. The search dates were limited from 1 January 2012 to 6 September 2016. The search dates of the previous version in MEDLINE included references from inception to 5 December 2012.

• Embase. We used the search term "soft tissue sarcoma". The search dates were limited from 01 January 2012 to 29 September 2016. The search dates of the previous version in Embase included references from inception to 05 December 2012.

We searched for ongoing trials by scanning the following online registries on 26 September 2016.

• ClinicalTrials.gov (ClinicalTrials.gov). Search: diagnosis 'soft tissue sarcoma'; intervention 'stem cell transplantation'. Limits: year '2012 to 2016'; study type 'interventional studies'; phase '2' or '3'.

• WHO International Clinical Trials Registry Platform (ICTRP). Search: diagnosis 'sarcoma'; intervention 'transplantation'. Limits: date of registration '1 January 2012 to 26 September 2016'.

We searched abstracts of annual meeting proceedings issued by the following societies on 26 September 2016:

• American Society of Clinical Oncology (ASCO): ASCO meetings in 2012 to 2016.

• American Society of Hematology (ASH): ASH meetings in 2013 to 2015. Search: 'autologous'.

• Bone Marrow Transplantation (BMT) Tandem Meetings of the American Society for Blood and Marrow Transplantation (ASBMT) and the Center for International Blood and Marrow Transplant Research (CIBMTR) (BMT Tandem Meeting 2012; BMT Tandem Meeting 2013; BMT Tandem Meeting 2014; BMT Tandem Meeting 2015).

• European Society for Blood and Marrow Transplantation (EBMT) (EBMT Meeting 2014; EBMT Meeting 2015; EBMT Meeting 2016). Search: 'sarcoma'.

• EBMT current study list (EBMT Studies 2016).

• International Society of Paediatric Oncology (SIOP) (SIOP Meeting 2012; SIOP Meeting 2013). Search: 'sarcoma'.

The search strategies used have been developed and executed by the author team.

Searching other resources

We planned to locate information about trials not registered in electronic databases by searching the reference lists of recently published relevant articles and review articles.

Selection of studies

We endorsed the PRISMA statement, adhered to its principles, and conformed to its checklist (Moher 2009). We retrieved all titles and abstracts by electronic searching, downloaded them, and transferred the bibliographical data into an Excel spreadsheet. We removed duplicates and two review authors (FP, HE) examined the remaining references independently. We excluded those studies that clearly did not meet the inclusion criteria and we documented the reasons for the exclusion of studies. We resolved disagreement by discussion and it was not necessary to consult a third review author (LAS). We considered studies written in languages other than English and asked peers familiar with the particular language and with the principles of study evaluation to translate major methodologic issues. We planned to use the Google Translate 2016 program if required, but this was not necessary.

Data extraction and management

We extracted the following data as in the previous version.

• General information on author, title, source, publication date.

• Study characteristics: trial design, setting, inclusion and exclusion criteria, comparability of participants' characteristics between groups, treatment allocation, blinding, subgroup analysis, length of follow‐up.

• Participant characteristics: age; gender; number of participants recruited, allocated, affected, analyzed; additional diagnoses; participants lost to follow‐up.

• Interventions: type of HDCT, source of stem cells, and type of SDCT.

• Outcomes: OS, TRM, DFS, PFS, EFS including type of event, toxicity, secondary neoplasia, health‐related quality of life.

Assessment of risk of bias in included studies

Two review authors (FP, LAS) independently checked the risk of bias in the included studies using the standard criteria to assess RCTs according to the Cochrane 'Risk of bias' tool (Higgins 2011a). With respect to the previous version, we removed the risk of bias items specifically aimed at checking the risk of bias in nonrandomized studies.

• Random sequence generation (selection bias).

• Allocation concealment (selection bias).

• Blinding of participants and personnel (performance bias).

• Blinding of outcome assessment (detection bias).

• Incomplete outcome data (attrition bias).

• Selective reporting such as not reporting prespecified outcomes (reporting bias).

• Other bias.

We applied Cochrane criteria for judging risk of bias (Higgins 2011a). In general, there was a 'low risk' of bias if the bias was unlikely to seriously alter the results, for example, participants and investigators enrolling participants could not have foreseen assignment. There was a 'high risk' of bias if the bias seriously weakened confidence in the results, for example, participants or investigators enrolling participants could possibly have foreseen assignments. There was 'unclear' risk of bias if the bias raised some doubt about the results, for example, the method of concealment was not described or not described in sufficient detail to allow a definite judgment.

Measures of treatment effect

The primary effect measure was the hazard ratio (HR) for time‐to‐event data. If the HR was not directly given in the publication, we planned to estimate HRs according to methods proposed by Tierney 2007, but this was not necessary. We planned to calculate odds ratios (ORs) with 95% confidence intervals (CIs) for dichotomous outcomes, but this was not applicable. In the case of rare events, we planned to use Peto OR instead, but this was not applicable. We planned to analyze continuous data and to present them as mean differences, if all results were measured on the same scale (e.g. length of hospital stay), but this was not applicable. If this was not the case (e.g. pain or quality of life), we planned to use standardized mean differences, but this was not applicable.

Dealing with missing data

We conformed to Cochrane's principal options for dealing with missing data and analyzed only the available data (Higgins 2011b). If data were missing or only imputed data were reported, we planned to contact trial authors to request data on the outcomes among participants who were assessed.

In the previous version of the review, we contacted the authors of the study by Bui‐Nguyen 2012 by e‐mail (1 December 2012) to ask for missing data about the histologic types that were combined as 'others'. The authors responded and as a consequence we could base the inclusion or exclusion of participant data on the additional data (Table 1). In the current version of the review (September 2016), we sent e‐mail inquiries as shown in Appendix 3 to two authors (Binh Bui‐Nguyen, 3 October 2016; Jean‐Yves Blay, 5 October 2016) of the included study (Bui‐Nguyen 2012) regarding clarification of survival data and Food and Drug Administration (FDA) warning letter on objectionable conditions and inadequate responses (FDA 2015). We did not receive any reply by 7 March 2017.

Assessment of heterogeneity

We had planned to assess heterogeneity between studies by visual inspection of forest plots; by estimation of the percentage heterogeneity between trials which cannot be ascribed to sampling variation (I² statistic) (Higgins 2003); by a formal statistical test of the significance of the heterogeneity (Cochran's Q) (Deeks 2011); and, if possible, by subgroup analyses (see Subgroup analysis and investigation of heterogeneity). We had planned to investigate and report possible reasons if there was evidence of substantial heterogeneity. We had planned to use the random‐effects model with inverse variance weighting for statistical pooling (DerSimonian 1986). We did not pool estimates.

Assessment of reporting biases

In addition to the evaluation of reporting bias as described in the Assessment of risk of bias in included studies section, we had planned to assess reporting bias (such as publication bias, time lag bias, multiple (duplicate) publication bias, location bias, citation bias, language bias) by constructing a funnel plot if there were a sufficient number of included studies (i.e. at least 10 studies included in a meta‐analysis otherwise the power of the tests would be too low to distinguish chance from real asymmetry (Sterne 2011)). We did not assess reporting bias because of the low number of identified studies.

Data synthesis

We analyzed data using Review Manager 5 (RevMan 2014). This was done by one review author (FP) and checked by another review author (LAS). If sufficient, clinically similar studies were available, we had planned to pool their results in meta‐analyses if they used comparable outcome definitions. As we included one RCT, we presented the results descriptively. We had planned to use random‐effects models with inverse variance weighting for all meta‐analyses (DerSimonian 1986), but this was not applicable.

For each comparison, we prepared a 'Summary of findings' table using the GRADE profiler software (GRADEpro 2014), in which we presented the following primary outcomes: OS, TRM, DFS, PFS, non‐hematologic toxicity grade 3 to 4 and health‐related quality of life . For each outcome, two review authors (FP, LAS) independently assessed the quality of the evidence by using the five GRADE considerations, that is, study limitations, inconsistency, indirectness, imprecision, and publication bias as described in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011).

Subgroup analysis and investigation of heterogeneity

We had planned subgroup analyses based on age, stage, and time period of treatment. However, we found no appropriate data to conduct these analyses.

Sensitivity analysis

We had planned sensitivity analyses to compare the results of studies with low versus high risk of bias. As we included only one study, a sensitivity analysis was not applicable.