The efficacy of psychological interventions for child and adolescent PTSD: a network meta-analysis

Pediatric post-traumatic stress disorder (PTSD) is a common and debilitating mental disorder and its effective treatment constitutes a health priority. Numerous randomized controlled trials (RCTs) have examined the efficacy of psychological interventions for pediatric PTSD. Yet, a comprehensive network meta-analysis (NMA) is lacking. The present work addresses this gap. A total of 67 RCTs met the inclusion criteria comprising 5,297 children and adolescents with full or sub-threshold PTSD. Five families of intervention were evaluated: trauma-focused cognitive behavior therapy (TF-CBT), eye movement desensitization and reprocessing (EMDR), other trauma-focused interventions, non-trauma-focused (non-TF) interventions, and multi-disciplinary treatments (MDTs). Most RCTs (73%) examined TF-CBT followed by EMDR. Other trauma-focused interventions had too few trials for analysis. At treatment endpoint, TF-CBT, EMDR, MDTs, and non-TF interventions were all effective in treating pediatric PTSD when compared to passive control conditions in random-effect NMA with large pooled effects (all Hedges’ gs ≥ 0.84, all ps < .001). TF-CBT, EMDR, and MDTs also yielded significant short-term treatment effects compared to active control conditions. In a sensitivity analysis including only high-quality trials, only TF-CBT and EMDR outperformed active control conditions. And in a sensitivity analysis including only trials with ≥ 50% of participants reporting multiple-event-related PTSD, only TF-CBT yielded significant short-term effects. Results for mid-term (up to 5 months posttreatment) and long-term efficacy (beyond 5 months posttreatment) were very similar. TF-CBT consistently yielded the highest treatment effects except being second to EMDR at mid-term. The present NMA is the most comprehensive NMA of psychological interventions for pediatric PTSD to date. Results confirm that TF-CBT can effectively treat PTSD in children and adolescents both in the short and long-term and also for multiple-event-related PTSD. More long-term data and multiple-event-related PTSD data are needed for EMDR, MDTs, and non-TF interventions to draw firmer conclusions regarding their efficacy. Results for TF-CBT are encouraging for clinical practice and may help to reduce common treatment barriers.

Most children and adolescents experience at least one potentially traumatic event before reaching adulthood [1][2][3] .Even in countries not affected by war, between one to two thirds of the general child and adolescent population report exposure to at least one traumatic event 4,5 .While most of them remain resilient to trauma and do not develop long-lasting mental health consquences 6 , between one-sixth and one-fourth of traumatized children and adolescents develop post-traumatic stress disorder (PTSD) as a result 4,7 .Risk is amplified for girls 7 and those experiencing interpersonal (vs.non-interpersonal) trauma 4,8 or a higher number and severity of traumatic events 9 .PTSD in children and adolescents -often referred to as pediatric PTSD -is a common, impairing 4 , and often chronic 10 mental disorder characterized by vivid re-experiencing of the trauma (e.g., intrusions or nightmares), avoidance of trauma-related situations, thoughts, and emotions, lasting changes in cognitions (e.g., self-blame) and emotions (e.g., shame), as well as hyperarousal symptoms (e.g., hypervigilance) 11 .Given its high prevalence and chronicity, pediatric PTSD, if left untreated, is a substantial burden to the individual and society [12][13][14][15] .Particularly during sensitive developmental periods in childhood and adolescence, unaddressed PTSD can disrupt crucial developmental processes 16 .This underscores the critical importance of implementing effective treatment approaches and guidelines to mitigate adverse effects and impaired development caused by PTSD.
In the past four decades, numerous of psychological interventions targeting pediatric PTSD have been developed and investigated in randomized controlled trials (RCTs) 17 .
International treatment guidelines for pediatric PTSD list trauma-focused cognitive behavior therapy (TF-CBT) as the first-line recommendation for the treatment of pediatric PTSD [18][19][20][21][22][23] .TF-CBT -in the present work -refers to all interventions based on cognitive behavior therapy (CBT) principles with a direct trauma focus in treatment (e.g., prolonged exposure therapy 24 , cognitive processing therapy 25 ).High efficacy of TF-CBT for the treatment of pediatric PTSD has been reported in multiple pairwise and network meta-analyses 17,[26][27][28][29][30] .Also other psychological interventions have been developed and examined and these can be categorized into the following four families of interventions: eye movement desensitization and reprocessing (EMDR), other trauma-focused interventions (i.e., interventions that explicitly address trauma during the treatment but are not based on CBT or EMDR), nontrauma-focused interventions (non-TF; i.e., interventions that do not address traumatic experiences during the treatment), and multi-disciplinary treatments (MDTs, i.e., interventions that combine techniques from two or more theoretical approaches).However, these four families have produced a) fewer RCTs than TF-CBT and b) lower or less certain efficacy in meta-analyses of RCTs, which is why they are either recommended as second-line treatments (such as EMDR) or not listed in international treatment guidelines.Recent RCTs (i.e., 19 RCTs published since 2019) 25,[31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48] require a new analysis of treatment efficacy of interventions for pediatric PTSD.
Whilst pairwise meta-analyses are restricted to direct treatment comparisons (i.e., comparison of arms within a given RCT), NMAs allow integrating data from both direct comparisons and indirect treatment comparisons (i.e., comparisons of arms across RCTs), providing estimates of the relative effect of any given pair of interventions in a given network 49 .Numerous pairwise and NMAs of psychological interventions for pediatric PTSD have been published, yat all of these omitted the following three aspects.First, a comprehensive summary of the field is lacking.Most previous work solely focused on a particular family of interventions for pediatric PTSD such as TF-CBT 26,50-54 , group-based interventions 55 , interventions with caregiver involvement 56 , on a specific population such as refugees 57 , children and adolescents in low and middle-income countries 29,58,59 or adolescents with interpersonal trauma-related PTSD 60 , or on short-term efficacy only 28 .To this date, only one published NMA 27 comprehensively summarized the literature across interventions and .populations and included both short-term and long-term efficacy (up to 12 months posttreatment).However, this NMA 27 can be considered outdated given that the literature search was conducted in December 2020 and yielded 56 RCTs, compared to 67 RCTs included in the present work.The present work also did not have a restriction with regards to long-term follow-up assessments (up to 24 months posttreatment).Second, no previous NMA included a sensitivity analysis that quantitatively tested whether efficacy estimates might be biased by trials with low quality, despite the fact that treatment efficacy of psychological interventions has been overestimated in some fields due to including low quality evidence in meta-analytic synthesis (e.g., field of adult depression 61 ).Third, no previous NMA performed sensitivity analyses concerning multiple-event-related PTSD.Rather, all previous work has analyzed efficacy across samples irrespective of whether most participants had a single or a multiple trauma history.It is crucial for clinical practice to determine whether treatment efficacy varies in RCTs in which most participants have experienced multiple (rather than single) traumatic events given that clinicians are often reluctant to address trauma, particularly in children and adolescents with multiple-event-related PTSD who are often perceived as particularly vulnerable 62 .In addition to clinicians, also patients and their caregivers must be well-informed about evidence-based approaches to treating multipleevent-related PTSD.The present work addresses all three omissions.That is, the present work a) comprehensively summarizes the short-, mid-, and long-term efficacy of all psychological interventions for pediatric PTSD, b) includes a sensitivity analysis on highquality trials only, and c) includes a sensitivity analysis on multiple-trauma trials only.

Methods
This NMA was conducted and reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2015 guidelines for NMAs 63 .Each step, .including the systematic literature search, data extraction, and quality ratings, were carried out independently by at least two authors.Results were systematically compared and disagreements were discussed between these two authors and in complicated cases amongst at least three of the authors (THH, LW, AK, & NM).In case of missing data, we sent a data request e-mail to corresponding authors and -in case of no response -a reminder a month later.We obtained data for all eligible trials and thus included all trials in the NMAs.The

Search Strategy
For the timespan from inception to April 21 st 2022, we relied on the literature search of our previous work 17 , which included 57 eligible RCTs.To identify RCTs published thereafter, we conducted a new search, which was conducted on April 18 th 2023 and covered the timespan from January 1 st 2022 to April 18 th 2023.The overlap between searches was intentional to not miss relevant studies.THH and LW independently carried out the systematic literature search and results were compared after each step (e.g., title and abstract screening, full-text eligibility check).See Appendix A in the online supplementary material for the full search strategy.In brief, we searched four large bibliometric health sciences databases: PsycINFO and MEDLINE through a combined search via EBSCOhost, Web of Science, and PTSDpubs.We searched for relevant RCTs by means of various search terms for PTSD (ptsd OR ptss OR post-traumatic stress OR posttraumatic stress OR post-traumatic syndrome OR posttraumatic syndrome) and treatment (treatment* OR intervention* OR therap* OR psychotherap* OR exposure OR counse*ing OR trial*) in all-field searches.In line with our previous literature search, no restrictions were applied to languages or publication formats.We also searched two further sources for eligible RCTs: 1) 71 reviews fully or at least partly focusing on psychological treatment of pediatric PTSD (see Appendix B for their references) and 2) the reference lists of included trials.

Eligibility Criteria
In line with our previous work 17 , we included trials that met all of the following inclusion criteria: 1) RCT, 2) investigating the efficacy of a psychological intervention for pediatric PTSD compared to either a passive control condition, an active control condition, or to another psychological intervention from a different treatment family, 3) PTSD or subthreshold PTSD was the primary treatment target, 4) mean age of sample less than 19 years, 5) at least ten participants per arm have outcome data reported, and 6) PTSD was assessed at least once post-intervention with a PTSD-specific outcome measure (i.e., clinicianadministered or self-report measure) based on diagnostic criteria reported in any iteration of the Diagnostic and Statistical Manual of Mental Disorders (DSM 11 ) or the International Statistical Classifications of Diseases (ICD 64 ).In line with inclusion criterion 3), we excluded transdiagnostic approches 65 and school-based interventions with partly preventive focus 66 (i.e., partially including participants without PTSD complaints).The age threshold was set at below 19 years rather than 18 years given that some trials applied developmentally adapted approaches and involved sample with a mean age of just above 18 years 25 . .

Quality Assessment
Risk of bias was independently assessed by two authors (THH & NM) by means of eight dichotomously scored quality criteria reported by Cuijpers et al. 61 .These eight criteria originated partly from the Cochrane Collaboration criteria 67 as well as authoritative criteria for evidence-based psychological interventions 68 and were applied as follows (i.e., a positive score was given when a criterion was met): 1) 100% of the included participants were diagnosed with PTSD at baseline which was assessed with an interview-based/clinicianadministered diagnostic measure, 2) a specific (and identifiable) treatment manual was followed, 3) study therapists were trained to apply this specific treatment manual, 4) treatment integrity of study therapists with regards to this treatment manual was formally checked (e.g., formal fidelity ratings), 5) intention-to-treat (ITT) data (i.e., means, SDs, and ns) were reported to allow meta-analysis of ITT data, 6) large trial (n1 + n2 ≥ 50), 7) random group allocation was performed by an independent party or computerized, and 8) PTSD outcomes post-intervention were assessed either by blinded interviewers (for clinician-based measures) or via self-report (for self-report measures).When insufficient information was reported with regards to a given quality criterion, this given criterion was scored conservatively (i.e., score of zero = not meeting this criterion).Possible quality sum scores ranged from 0 to 8. Note that the score for criterion 6 may differ a) between comparison dyads for multi-arm trials and b) between timepoints (i.e., attrition in completer data).Initial agreement between independent raters was good (91.27% of all ratings).In the present study, RCTs were defined as high-quality trials (i.e., with low risk of bias) when fulfilling at least six of the eight quality criteria.An overview of the quality criteria and their scoring is presented in Appendix C and quality ratings per trial (including the categorization of highquality trials) are presented in Appendix D. .

Data Extraction
Trial characteristics (e.g., country of conduct, specific psychological intervention applied and category/family of interventions, number of treatment sessions applied, treatment delivery format, specific control condition and category of control condition, and risk of bias), sample characteristics (e.g., mean age, percentage females, trauma type(s), and percentage of participants with PTSD relating to one vs.two or more traumatic events), and PTSD outcome data (i.e., PTSD symptom severity means, standard deviations, and sample sizes per arm at a given assessment timepoint to calculate standardized between-group effect sizes) were extracted by at least two authors (THH, LW, AK, & NM).When ITT and completer data concerning PTSD symptom severity were reported, the former was prioritized.Similarly, when PTSD outcomes were assessed via both a clinician-based measure as well as a self-report measure, the former was prioritized.If a trial included two or more arms from the same intervention family (e.g., multiple TF-CBT arms 69 ), the primary intervention arm (e.g., primary TF-CBT arm) as indicated by the authors of the original work was prioritized.

Categorization of Psychological Interventions and Control Conditions
The present study compared four families of psychological interventions: 1) TF-CBT, 2) EMDR, 3) non-TF interventions, and 4) MDTs.Initially, we planned to also analyze other trauma-focused interventions (i.e., interventions with a trauma focus but not belonging to TF-CBT or EMDR), which was not feasible given that the evidence base was too thin.Control conditions were divided into passive control conditions (e.g., wait-list control) and active control conditions (e.g., treatment-as-usual).An overview of all categorizations is depicted in Appendix E. .

Categorization of Assessment Timepoints
To evaluate the short, mid, and long-term efficacy of psychological interventions and in line with previous research 70 , we distinguished between three assessment timepoints: 1) post-treatment, which we defined as assessments at treatment endpoint and which served as an estimate of short-term efficacy, 2) follow-up 1 (FU1), which we defined as assessments of up to or equal to five months after the treatment endpoint and which served as an estimate of mid-term efficacy, and 3) follow-up 2 (FU2), which we defined as assessments longer than five months after the treatment endpoint and which served as an estimate of long-term efficacy.When several assessments fell into the FU1 and FU2 categories, the longest assessment was chosen.

Outcomes
The primary outcome of interest was the standardized mean differences (Hedges' g 71 ) in PTSD symptom severity between comparator groups.

Statistical Analysis
NMAs allow for an estimation of the relative effectiveness of all interventions in a given network, synthesizing evidence across all available RCTs.This is done by including both direct comparisons (i.e., comparisons between two arms in a given RCT) and indirect comparisons (i.e., comparisons arms between RCTs).In light of the high heterogeneity in both included interventions and samples, random effects NMAs were conducted 71 .For all analyses, the applied level of statistical significance was set at α = .05.Data processing and statistical analyses were performed in R (version 4.1.1 72) with the netmeta package 73 .Effect sizes (Hedges' g) were first calculated at the study-level 74 and then pooled and compared between all comparison dyads in NMA 71 .Following Cohen's convention 75 , NMA-pooled gs .were interpreted as small (0.20), medium (0.50), and large effects (0.80).We only included psychological intervention families with minimally sufficient evidence base (i.e., total number of direct comparisons kes ≥ 4 76 ) to ensure that effects were also estimated via a sufficient number of direct comparisons (i.e., effects not solely based on indirect comparisons 77 ).In the present work, k denotes the number of independent RCTs, whereas kes denotes the number of direct comparisons, which may differ in the light of multi-arm trials.
For valid inferences of NMAs transitivity 77 is assumed.That is, the relative efficacy of any given comparison dyad needs to be exchangeable regardless of whether it was derived from direct or indirect comparisons.For this to hold true, a similar distribution of sample and methodological characteristics across comparison dyads is needed.To test whether the transitivity assumption was met, we checked whether the distribution of basic sample characteristics (e.g., mean age, percentage females), clinical sample characteristics (e.g., percentage of total sample meeting full diagnostic criteria for PTSD) and trial characteristics (e.g., trial quality, applied treatment format) was similar across comparison dyads in the networks.To check for (in)consistency 77 in the global network 63,78 and per comparison dyad (i.e., local consistency), we performed the net splitting procedure 79 and examined net heat plots 80 .Whenever the net splitting procedure detected significant inconsistencies, we performed corrected analyses.Further, we calculated outlier-adjusted NMAs whenever (≥ 1) outliers were detected.We defined statistical outliers as direct comparison effect sizes of at least 3.3 standard deviations above or below the pooled g 81 .To check for potential influence of small-study effects, we performed Egger's test 82 and inspected funnel plots.We calculated the  2 and  2 statistics as estimates of overall heterogeneity 83 .We further calculated het and inc as estimates of heterogeneity within and between comparison dyads, respectively 84 .We also calculated surface under the cumulative ranking (SUCRA) rankings, which allow for a ranking of interventions and control conditions based on their efficacy estimates.We performed 50,000 resamples for all SUCRA analyses.Lastly, we performed the following two sensitivity NMAs: 1) NMA including only high-quality trials (as done in previous NMA in the field of adult PTSD 70 ) and 2) NMA including only multiple-trauma trials (as done in a previous pairwise meta-analysis in the field of pediatric PTSD 17 ).In the present study, the term multiple-trauma trial refers to trials involving a majority of participants (i.e., at least 50% of the total sample reported) with multiple-event-related PTSD (i.e., PTSD relating to two or more traumatic events) 17,76 .

Study Selection Process
The new search wave covered 8,845 electronic records and yielded ten additional eligible RCTs.Combined with the 57 previously identified RCTs 17 , the present study summarizes data from 67 independent RCTs. Figure 1 shows a detailed overview of the study synthesis process.
-Figure 1   (13%) included only female participants and four RCTs (6%) included only male participants.In total, 2,667 children and adolescents were randomized to psychological interventions and 2,630 to control conditions.The unweighted mean age across trials was 12.56 years.Across included trials, 85% of the participants met diagnostic criteria for PTSD at the baseline assessment.In 44 RCTs (67% of included RCTs), the majority of participants reported multiple-event-related PTSD and the given trial was therefore categorized as multiple-trauma trial (see Appendix F for this categorization).Psychological interventions comprised an average of 10 sessions (SD = 5 sessions) with a mean total duration of 657 minutes (SD = 383 minutes).A little less than half of the RCTs (30 of 67 RCTs, 45%) involved parents or primary caregivers in the treatment sessions.A little more than half of the trials (38 of 67 RCTs, 57%) assessed follow-up data, with follow-up assessment timepoints ranging from one month to 24 months post-treatment.ITT data concerning PTSD outcomes were reported in most trials RCTs (44 of 67 RCTs, 66%).

Assumption Checks
Assumptions were mostly met.In all but one analysis, no significant inconsistencies were observed.In the NMA on mid-term efficacy, significant inconsistency was found for MDTs and a corrected analysis (i.e., without MDTs) was performed.The transitivity assumption was met.The distribution of sample and methodological characteristics across comparison dyads is presented in Table 1.No outliers were found, with one exception.In the NMA on short-term efficacy, one outlier (concerning TF-CBT) was found and an outlieradjusted analysis was performed.

Network Graphs
.
Figure 2 shows the network graphs for the NMAs on short-term efficacy, mid-term efficacy, and long-term efficacy.Most available trials assessed TF-CBT.Only TF-CBT had enough accumulated evidence to be analyzed across timepoints (i.e., short, mid, and longterm efficacy NMAs).The network graphs for the outlier-adjusted and sensitivity analyses concerning short-term efficacy are provided in Appendix H.An inconsistency-adjusted network graph (with MDTs deleted) for mid-term efficacy is presented in Appendix I.
Network graphs for the NMA on mid-term and long-term efficacy are presented in Appendix J and K, respectively.
-Figure 2 and Table 1 about here -

Network Meta-analysis of Short-term Efficacy
Table 2 provides an overview of the results concerning the short-term efficacy, including the results of the sensitivity analyses.Forest plots for all analyses across timepoints are presented in the Appendices in the online supplementary material.Too few trials were available to include the other trauma-focused interventions category in any NMA.In the short-term, all four therapy families with at least four direct comparisons -TF-CBT, EMDR, MDTs, and non-TF interventions -were efficacious in treating pediatric PTSD compared to passive control conditions with standardized mean differences ranging from g = 0.84 for non-TF interventions to g = 1.08 for TF-CBT (Appendix L).Compared to active control conditions, only TF-CBT (g = 0.56, p < .001)and MDTs (g = 0.40, p = .027)were efficacious in treating PTSD in the short-term with a moderate and small effect size, respectively (Appendix M).Differences in efficacy between treatment families were not statistically significant in any of the analyses, with few or no direct comparisons for most comparison dyads.Heterogeneity in this main analysis of the short-term efficacy was large .within as well as between comparison dyads ( 2 = 0.18,  2 = 73.90%;total = 233.45,df = 61,  < .001;ℎ = 209.28,df = 50,  < .001; = 23.06,df = 11,  = .017).There was high consistency between direct and indirect comparisons for almost all comparison dyads (Appendix N).No significant inconsistencies were detected in the network splitting procedure.Further, no evidence for significant small-study effects was found in the funnel plot (Appendix O including the results of the Egger's test).One outlier 86 investigating group TF-CBT delivered in a sample of former male child soldiers in the DR Congo was detected with an very large effect size (g = 2.71) and an outlier-adjusted analysis excluding this trial produced very similar results (Appendix P).

Sensitivity Analyses.
All results of the sensitivity analyses are presented in Table 2.In the sensitivity analysis of high-quality trials only, the results compared to passive control conditions were confirmed with effect sizes ranging from g = 0.85 for non-TF interventions to g = 1.05 for TF-CBT (Appendix Q).Compared to active control conditions in high-quality trials, however, only TF-CBT (g = 0.53, p < .001)and EMDR (g = 0.43, p = .046)produced significant (moderate-sized) short-term efficacy.In contrast to the results when also lower quality trials were included, MDTs now did not yield significant short-term efficacy, and neither did non-TF interventions (Appendix R).In the sensitivity analysis of multiple--eventrelated PTSD, only TF-CBT and MDTs had sufficient trials available.Both TF-CBT (g = 1.24, p < .001)and MDTs (g = 1.00, p < .001)yielded (large-sized) significant short-term efficacy compared to passive control conditions (Appendix S).Compared to active control conditions in multiple-trauma trials, only TF-CBT yielded significant (moderate-sized) treatment efficacy (g = 0.56, p < .001),whereas MDTs did not produce significant short-term effects (Appendix T).

Network Meta-analyses of Mid and Long-term Efficacy
See Table 3 for an overview of all mid and long-term efficacy results, including the results of the sensitivity analyses.For non-TF interventions, there were too few direct comparisons available.In the mid-term up to 5 months posttreatment, TF-CBT, EMDR, and MDTs were effective in treating pediatric PTSD as compared to passive control conditions with gs ranging from 0.71 for MDTs to 0.94 for EMDR (Appendix U).Compared to active control conditions, only EMDR (g = 0.51, p = .040)and TF-CBT (g = 0.44, p = .003)produced significant (moderate-sized) mid-term efficacy, whereas MDTs was not found efficacious (Appendix V).Heterogeneity in this main analysis concerning mid-term efficacy was large within as well as between comparison dyads ( 2 = 0.15,  2 = 67.10%;total = 69.88,df = 23,  < .001;ℎ = 45.56,df = 14,  < .001; = 25.45,df = 9,  = .003).There was high consistency between direct and indirect comparisons for nearly all comparison dyads, as illustrated in the net heat graph (Appendix W).However, significant inconsistency was detected for MDTs in the network splitting procedure.Results remained very similar after the exclusion of MDTs from the analysis (Appendices X & Y).There was no evidence for significant small-study effects in the funnel plot (Appendix Z including the results of the Egger's test).
In the long-term, between 6 to 24 months posttreatment, only TF-CBT and non-TF interventions had sufficient evidence for synthesis.Compared to passive control conditions, both TF-CBT (g = 0.75, p = .002)and non-TF interventions (g = 0.74, p = .008)were efficacious in the long-term with moderate-to-large effect sizes (Appendix AA).Compared to active control conditions, both TF-CBT (g = 0.55, p < .001)and non-TF interventions (g = 0.53, p = .007)were efficacious with moderate effect sizes (Appendix AB).Heterogeneity in this main analysis concerning long-term efficacy was large within as well as between comparison dyads ( 2 = 0.10,  2 = 64.80%;total = 42.58,df = 15,  = .002;ℎ = 31.32,df = 10,  = .005; = 11.30,df = 5,  = .046).There was high consistency between direct and indirect comparisons for all comparison dyads as illustrated in the net heat graph (Appendix AC).No significant inconsistencies were found in the network splitting procedure.There was no evidence for significant small-study effects in the funnel plot (Appendix AD including the results of the Egger's test).

Sensitivity Analyses.
Sensitivity analysis on high-quality trials at mid-term could be conducted with TF-CBT and EMDR only.Both EMDR (g = 1.15, p < .001)and TF-CBT (g = 1.06, p < .001)produced large treatment effects compared to passive control conditions (Appendix AE).
Compared to active control conditions in high-quality trials, however, only TF-CBT (g = 0.33, p = .029)produced a significant effect (Appendix AF).With respect to multiple-trauma trials at mid-term, sensitivity analysis could be conducted only for TF-CBT and MDTs and evidence for MDTs was mostly from lower quality trials.Compared to passive control conditions, TF-CBT produced a large pooled treatment effect across multiple-trauma trials (g = 0.84, p < .001)and MDTs a moderate-sized pooled effect (g = 0.67, p = .027,Appendix AG).Compared to active control conditions, only TF-CBT yielded significant mid-term efficacy in lowering multiple-event-related PTSD with a moderate-sized pooled effect (g = 0.56, p < .001,Appendix AH).
In terms of long-term efficacy, only TF-CBT and non-TF interventions had a sufficient number of accumulated trials.Compared to active control conditions, only TF-CBT .produced a significant moderate-sized treatment effect (g = 0.53, p < .001) in high-quality trials (Appendix AI).Compared to passive control conditions, both TF-CBT (g = 0.74, p = .004)and non-TF interventions (g = 0.71, p = .019)produced a moderate-to-large pooled treatment effects in multiple-trauam trials (Appendix AJ).Compared to active control conditions, only TF-CBT (g = 0.45, p = .004)produced a significant long-term effect in multiple-trauma trials (Appendix AK).

Ranking of Efficacy
See Table 4 for all SUCRA rankings.TF-CBT was the highest ranking psychological intervention at all timepoints in both the main analyses and sensitivity analyses, except at mid-term when it was second to EMDR.

Discussion
In this network meta-analysis, we synthesized the findings of 67 RCTs on the efficacy of psychological interventions for PTSD in children and adolescents.The present review extends previous NMAs 27,28 and conventional meta-analyses 26,30,53,60,87 in the field.With 67 RCTs on pediatric PTSD, the present NMA is by far the largest to date.The main finding is that TF-CBT is currently the most evaluated and most effective treatment for PTSD in children and adolescents, followed by (in this order) EMDR, MDTs and non-TF intervention.This result supports the recommendations of international treatment guidelines, such as the .recommendations by the International Society for Traumatic Stress Studies 22 , the World Health Organization 19 , the UK National Institute of Clinical Excellence (NICE) 21 , the World Federation of Societies of Biological Psychiatry 18 , the American Academy of Child and Adolescent Psychiatry 20 , or the Australian National Health and Research 23 for the treatment of PTSD in children and adolescents.Notably, the effectiveness of TF-CBT was robust across different comparison groups (passive and active control conditions), across timepoints (short, mid, and long-term follow-up), and when restricting analyses to high-quality trials or multiple-trauma trials.These results are important for the training of therapist as well as for implementation of treatment in clinical practice.The findings regarding the treatment efficacy of TF-CBT for children and adolescents with a history of multiple traumatic events should be taught in therapist training, helping therapists to overcome fears about applying a trauma-focused treatment approach when working with individuals with a multiple-trauma history 62 .This challenge may be particularly pronounced when working with children and adolescents, who are often seen as especially vulnerable.Based on our robust evidence, training programs can equip therapists to confidently guide their patients through traumafocused interventions.Further, young people themselves may also harbor fears and reservations about engaging in trauma-focused therapy.Addressing these concerns from an empirically informed perspective within the therapeutic relationship is crucial for mitigating barriers to treatment implementation.
The present review, however, also details remaining gaps in the literature.Data on EMDR, MDTs, and non-TF interventions are emerging, but many analyses of these interventions were not possible due to a paucity of available trials.EMDR, while showing promise in several trials with significant treatment effects at short and mid-term follow-up, has not been evaluated sufficiently at long-term follow-up.Further, multiple-trauma trials and trials comparing to active control conditions are also largely lacking for EMDR.There are also insufficient data on MDTs, non-TF interventions, and other trauma-focused interventions (i.e., not belonging to TF-CBT or EMDR).More long-term follow-up, high-quality, and multiple-event-related PTSD data are needed for all of these families of interventions to draw firmer conclusions about their efficacy.As more RCTs accumulate, more fine-grained (adequately powered) NMAs will become feasible.At present, there is insufficient data to allow a sub-categorization of the heterogeneous umbrella-categories of MDTs, non-TF interventions, and other trauma-focused interventions.
The short and mid-term efficacy of EMDR is supported by the present review.However, our review highlights the lack of evidence in relation to multiple traumatic exposures, long-term follow-up data, as well as comparisons with active control conditions.
The present results therefore add credibility to international treatment guidelines 21,23   exposure and cognitive processing therapy separately rather than clustering these as TF-CBT).Clustering different manuals may mask efficacy differences between manuals.However, to date there is no evidence for this notion, but rather evidence to suggest a noninferiority between different TF-CBT manuals 88,89 .TF-CBT interventions -while being labeled differently in manuals -have substantial overlap in terms of theoretical grounding, assumed underlying mechanisms (e.g., habituation, extinction learning, and reappraisal of trauma-associated maladaptive beliefs), applied techniques (e.g., exposure) and all involve a direct trauma focus during treatment.Notably, other NMAs for common mental disorders have also clustered psychological therapies [90][91][92] including one in the field of adult PTSD 70 .Second, we found evidence for inconsistency in the NMA regarding the mid-term efficacy data of MDTs.However, an analysis excluding MDTs produced similar results to the main analysis.The third limitation is inherent in the field and concerns the rather slim evidence base with regards to mid and long-term efficacy data for all intervention families apart from TF-CBT.While robust evidence on the mid and long-term efficacy of TF-CBT was found in the main and sensitivity analyses, other families of interventions had either just enough or insufficient evidence available for analysis.More research is needed.

Conclusion
There is robust evidence that PTSD in children and adolescents can be effectively treated by psychological therapies, in particular TF-CBT.A large evidence base on the efficacy of TF-CBT supports efficacy at short, mid, and long-term, in comparison to both passive and active control conditions, in high-quality trials, and in multiple-trauma contexts.
A comparably thin evidence base also supports the short and mid-term efficacy of EMDR and to a lesser extent MDTs and non-TF interventions.For these latter three interventions families, more high-quality, long-term, and multiple-trauma context data are needed to draw firmer conclusions regarding their efficacy.
objectives and methods of the present NMA were defined a priori and pre-registered in the PROSPERO database (ID: CRD42020206290).We defined the main research question in line with the recommended format (Population, Intervention, Comparison, Outcome, and Study; PICOS) as follows: In children and adolescents with full or subthreshold PTSD (P), how do psychological interventions (I), compared to either passive control conditions, active control conditions, or amongst different families of interventions (C), perform in terms of lowering PTSD symptom severity (O) in randomized controlled trials (S)?
about here -Study Characteristics An overview of the characteristics of included trials is presented in Appendix F and their references are given in Appendix G. Apart for one dissertation 85 , all RCTs were published in peer-reviewed journals.Most RCTs (40 of 67 RCTs, 60%) were conducted in high-income countries and the rest (27 of 67 RCTs, 40%) were conducted in low and middleincome countries.The 67 RCTs included 5,297 children and adolescents in total (ns of first assessment after treatment termination added up), 66% of whom were females.Nine RCTs

.
that list EMDR as second-line treatment recommendation (e.g., recommended when TF-CBT has proven ineffective for an individual or when TF-CBT is not available).The data presented here provide tentative support for other forms of intervention.The preliminary findings for MDTs and non-TF interventions suggests that these interventions may become second-line treatment recommendations in the unforeseeable future.However, more data are needed for that.As long as high-quality data, long-term data, and data from multiple trauma contexts are scarce, one is well-advised to not overstate conclusions regarding the efficacy of these interventions families until sufficient data has accumulated.Limitations Three limitations should be noted.First, the categorizations of psychological interventions can be critically scrutinized.Other NMAs in this field such as Mavranezouli et al.'s 28 and Xiang et al.'s 27 decided to analyze comparisons by manuals (e.g., examining prolonged

.
childhood sexual and physical abuse: a randomized clinical trial.JAMA Psychiatry.2019;76(5):484-491.26.Huang T, Li H, Tan S, et al.The efficacy and acceptability of exposure therapy for the treatment of post-traumatic stress disorder in children and adolescents: a systematic review and meta-analysis.BMC Psychiatry.2022;22(1):259.27.Xiang Y, Cipriani A, Teng T, et al.Comparative efficacy and acceptability of psychotherapies for post-traumatic stress disorder in children and adolescents: a systematic review and network meta-analysis.BMJ Ment Health.2021;24(4):153-160.28.Mavranezouli I, Megnin-Viggars O, Daly C, et al.Research Review: Psychological and psychosocial treatments for children and young people with post-traumatic stress disorder: a network meta-analysis.J Child Psychol Psychiatry.2020;61(1):18-29.29.Alozkan-Sever C, Uppendahl JR, Cuijpers P, et al.Research Review: Psychological and psychosocial interventions for children and adolescents with depression, anxiety, and post-traumatic stress disorder in low-and middle-income countries-a systematic review and meta-analysis.J Child Psychol Psychiatry.2023;64(12):1776-1788.30.Bastien RJ-B, Jongsma HE, Kabadayi M, Billings J.The effectiveness of psychological interventions for post-traumatic stress disorder in children, adolescents and young adults: A systematic review and meta-analysis.Psychol Med.2020;50(10):1598-1612.31.Auslander W, Edmond T, Foster A, et al.Cognitive behavioral intervention for trauma in adolescent girls in child welfare: A randomized controlled trial.Child Youth Serv Rev. 2020;119.32.Barron I, Freitas F, Bosch CA.Pilot randomized control trial: Efficacy of a group-based psychosocial program for youth with PTSD in the Brazilian favelas.J Child Adolesc Trauma.2021;14(3):335-345.

Figure 1 .
Figure 1.PRISMA flow chart depicting the study selection process.

Figure 2
Figure 2Network graphs for main analyses concerning short-term (left), mid-term (middle), and long-term (right) efficacy

Table 1
Trial and sample characteristics across comparison dyads