Efficacy and moderators of efficacy of trauma-focused cognitive behavioural therapies in children and adolescents: protocol for an individual participant data meta-analysis from randomised trials

Introduction Trauma-focused cognitive behavioural therapies are the first-line treatment for posttraumatic stress disorder (PTSD) in children and adolescents. Nevertheless, open questions remain with respect to efficacy: why does this first-line treatment not work for everyone? For whom does it work best? Individual clinical trials often do not provide sufficient statistical power to examine and substantiate moderating factors. To overcome the issue of limited power, an individual participant data meta-analysis of randomised trials evaluating forms of trauma-focused cognitive behavioural therapy in children and adolescents aged 6–18 years will be conducted. Methods and analysis We will update the National Institute for Health and Care Excellence guideline literature search from 2018 with an electronic search in the databases PsycINFO, MEDLINE, Embase, Cochrane Central Register of Controlled Trials and CINAHL with the terms (trauma* OR stress*) AND (cognitive therap* OR psychotherap*) AND (trial* OR review*). Electronic searches will be supplemented by a comprehensive grey literature search in archives and trial registries. Only randomised trials that used any manualised psychological treatment—that is a trauma-focused cognitive behavioural therapy for children and adolescents—will be included. The primary outcome variable will be child-reported posttraumatic stress symptoms (PTSS) post-treatment. Proxy-reports (teacher, parent and caregiver) will be analysed separately. Secondary outcomes will include follow-up assessments of PTSS, PTSD diagnosis and symptoms of comorbid disorders such as depression, anxiety-related and externalising problems. Random-effects models applying restricted maximum likelihood estimation will be used for all analyses. We will use the Revised Cochrane Risk of Bias tool to measure risk of bias. Ethics and dissemination Contributing study authors need to have permission to share anonymised data. Contributing studies will be required to remove patient identifiers before providing their data. Results will be published in a peer-reviewed journal and presented at international conferences. PROSPERO registration number CRD42019151954.

Methodological: The authors fail to provide a clear definition of moderators, ignoring a considerable bulk of theoretical work since the seminal paper of Baron and Kenny (1986). A common definition of a moderator has been provided by Kraemer et al. (2002, p. 879): "To show that M is a moderator of treatment, M must be a baseline or prerandomization characteristic … that can be shown to have an interactive effect with treatment on the outcome." While the authors rightly state that "Factors that might impact the efficacy of trauma-focused cognitive behavioural therapies in children and adolescents form two broad categories: treatment-related and child-related factors.", by definition, treatment-related factors cannot be moderators, but would rather be mediators (which suprisingly the authors explicitely do not set out to analyze: "In future investigations of the obtained data, we plan to conduct mediation analyses…").
Further comments: • To wonder "Why does this first-line treatment not work for everyone?" seems odd, as there is hardly any such treatment for any condition. • The authors should explain better how they plan to make use of the expected results. It is highly unlikely that merely presenting results at conferences and publishing a paper will contribute to "enhance the future provision and development of trauma-focused cognitive behavioural therapies in children and adolescents."

GENERAL COMMENTS
The authors describe a protocol for an individual participant data meta-analysis from randomized trials. I have three main methodological concerns: 1) The primary outcome is not very well defined as no particular instrument is provided (it is good that a time window is provided); maybe the authors can describe how they will deal with this discrepancy.
2) What are the plans to deal with the multiple comparisons situation in case of multiple control groups? 3) What are the plans to deal with imputed data in the analysis?

Reviewer: 1
Reviewer Name: Bernd Puschner This is a protocol for an individual patient data (IPD) meta-analysis of trauma-focused cognitive behavioral treatment (CBT) for children and adolescents. It is worthwhile to focus on moderators of effect of such treatments. However, the planned study has considerable problems.

1.
Conceptual: While arguing that individual RCTs lack power for moderator analyses, primary research question is still the overall efficacy of CBT for young people. This has been already examined by a large number of meta-analyses. The rational for doing this again, even with IPD, remains unclear.  do not set out to analyze: "In future investigations of the obtained data, we plan to conduct mediation analyses…"). We agree that in this protocol we do not make an "explicit" commitment to analyse mediators.
The reason is that it is unclear until we conduct our stated analyses whether the requirements for mediation (as set out in the very same papers the reviewer cites) will be met. We therefore state that these are for future consideration.
Further comments:

3.
To wonder "Why does this first-line treatment not work for everyone?" seems odd, as there is hardly any such treatment for any condition.

4.
The authors should explain better how they plan to make use of the expected results. It is highly unlikely that merely presenting results at conferences and publishing a paper will contribute to "enhance the future provision and development of trauma-focused cognitive behavioural therapies in children and adolescents."

Reviewer: 2
Reviewer Name: André Scherag 1. The primary outcome is not very well defined as no particular instrument is provided (it is good that a time window is provided); maybe the authors can describe how they will deal with this discrepancy.
Our response: As the reviewer notes, we carefully define the time-point and the fact that the primary outcome will be a self-report of symptom severity. Alas, it is not possible to be more precise than this in a protocol. The reason is that for an IPD-MA that collates across many trials, the exact outcome instrument will inevitably vary from trial to trial (it would be lovely if everyone used the same instruments but they do not). We do in fact state this in the paper (see page 4): "A variety of measures for the primary and secondary outcomes will have been used in the individual trials with commensurate methodological and statistical complexities." We describe in the section on synthesis how we will harmonise across different instruments from different trials. We now directly refer to the relevant paragraph to make it easier to follow (see page 9, paragraph 4; see paragraph "Strategy for data synthesis" on page 15): "The primary outcome variable will be child-reported PTSS using a standardised self-report post-treatment (see paragraph "Strategy for data synthesis" for further information)." Moreover, we aligned our wording throughout the manuscript to self-report and proxy-report.
We hope that these generic expressions make clear that we are not able to provide a particular instrument.

2.
What are the plans to deal with the multiple comparisons situation in case of multiple control groups?