Literature Search

Electronic databases searched were MEDLINE®, Embase®, and the Cochrane Controlled Trials Register. Databases were initially searched without restriction on date, using the search strategy shown in Appendix A. However, during the Topic Refinement phase of this project, the Key Informants strongly recommended limiting study selection to the past 15 years to ensure we identify evidence that is comparable in terms of therapeutic regimens and management protocols. Thus, we reviewed the literature from January 1995 up to November 9, 2009. Literature searches were updated to August 17, 2011, prior to delivery of the final report to ensure the identification of new literature that potentially had an impact on the review.

All search results were compiled into an EndNote® reference manager database with exclusion of duplicates. Additional details on these materials and results of our review are provided in the Results chapter. Search strategies and results are detailed in Appendix A.

Study Selection

Inclusion and exclusion criteria are for all Key Questions.

Inclusion criteria:

• Reports on pediatric patients (age ≤21 years) who have relevant diseases (malignant solid tumors, inherited metabolic diseases, or autoimmune disease).
• Reports on an outcome of interest.
• Reported on HSCT and/or a comparator of interest.
• Intervention and comparator used contemporary regimens with respect to chemotherapy, radiation therapy and supportive care.
• For Key Questions 3 and 4 (inherited metabolic diseases) studies reporting outcomes on the natural history of disease were included as comparators.

Exclusion criteria:

• Studies older than 15 years as they would not represent contemporary regimens except the natural history data for Key Questions 3 and 4.
• Studies where pediatric data could not be separated and abstracted from adult data.
• Duplicate studies or reports with duplicate patients were excluded except the study with the largest number of patients with the longest followup.

Abstract and study selection was performed by a single reviewer for each section of the report. If a reviewer was uncertain whether a study should be selected for inclusion, this was resolved through discussion at team meetings.

Figure 4 shows a PRISMA¹¹ diagram of the studies included in the systematic review. A listing of excluded references with reasons for exclusions is available in Appendix B.

Figure 4

PRISMA diagram of articles included in the systematic review.

Data Abstraction

Data were abstracted by a single reviewer, and fact checked by another reviewer. If there were disagreements, they were resolved through discussion among the review team. The following data elements of primary studies were abstracted from the articles meeting selection criteria:

• Critical features of the study design
  • Patient inclusion/exclusion criteria
  • Number of participants and flow of participants through steps of study
• Patient characteristics, including:
  • Age
  • Sex
  • Race/ethnicity
  • Disease and stage
  • Disease duration
  • Other prognostic characteristics
• Treatment characteristics, including
  • Stem-cell source
  • Chemotherapy versus chemo-radiotherapy
  • Immunosuppressive therapy as prophylaxis for graft versus host disease
  • Supportive care
• Outcome assessment details
  • Identified primary outcome
  • Secondary outcomes
  • Response criteria
  • Use of independent outcome assessor
  • Followup frequency and duration
• Data analysis details
  • Statistical analyses (statistical test/estimation results)

    –

    Test used

    –

    Summary measures

    –

    Sample variability measures

    –

    Precision of estimate

    –

    P values

Full data abstraction tables are available in Appendix C. Evidence tables were generated in Microsoft Excel® and Microsoft Word®.

Study Quality

Evidence consisted largely of case series and case reports; therefore we did not attempt to assess the quality of individual studies. It is well recognized in the study of rare diseases that a common challenge is the preponderance of small, uncontrolled studies.¹² Therefore, because studies tended to be homogenous in design, quality assessment would be unlikely to discriminate between higher and lesser quality studies.

Data Synthesis

Data synthesis was qualitative. We attempted to identify subgroups based on prognostic factors such as tumor stage or location in solid tumors, or disease severity or rate of progression in the inborn metabolic disorders, to see if these subgroups showed patterns of treatment success or failure. The evidence base was considered insufficient and too heterogeneous to use quantitative pooling methods. Where possible we calculated confidence intervals for results and reported ranges of results for studies that addressed the same population and treatment.

Grading the Evidence for Each Key Question

The strength of the body of evidence for each indication was assessed according to the process developed by the AHRQ EPC Program¹³ for the EPC Methods Guide, based on a system developed by the GRADE Working Group.¹⁴ This comprised an iterative, qualitative consensus-driven process among EPC team members familiar with the summarized literature, using the 4 required domains specified in the EPC Methods Guide: risk of bias, consistency, directness, and precision. There were no head-to-head comparative studies for most diseases; in those situations, directness was based on the outcome (e.g., overall survival or other clinically important health outcomes) rather than on the comparison. For small series or a compilation of case reports in which the prognosis absent HSCT is uniformly fatal (e.g., Wolman's disease), the known natural history was considered an indirect comparator. An optional domain, strength of association (SOA, magnitude of effect) was thus ascribed to the body of evidence when there was an apparent benefit or harm, increasing the overall strength beyond what may be normally considered appropriate for such evidence. SOA was deemed not applicable for diseases where there was no clear evidence of benefit or harm with HSCT versus comparators, or if results (e.g., overall survival rates) of individual studies within a body of literature were inconsistent or conflicted. No quantitative scoring method was applied.

Table 3 displays the EPC Methods Guide definitions and applications of GRADE and describes how we applied the domains in this review.

Table 3

Elements of evidence grading for Key Questions.

The overall grade of evidence strength was classified into the following four categories:

• High: Further research is very unlikely to change our confidence in the estimate of effect
• Moderate: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate of effect
• Low: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate
• Insufficient: Any estimate of effect is very uncertain