School-based interventions to prevent anxiety, depression and conduct disorder in children and young people: a systematic review and network meta-analysis

Searching and screening

• The number of databases searched was reduced from the original proposal. Instead, we followed Cochrane Methodological Expectations of Cochrane Intervention Reviews (MECIR) conduct guidelines50,51 on the selection of primary databases and applied approaches for optimising search strategies. 52–54

• The protocol stated that the relevant age range for inclusion was 5–25 years. This was difficult to operationalise in practice, and changes were made to the age limits so that the report covers the age range 4–18 years.

• We clarified the intended intervention setting as ‘educational-setting based’. In the original proposal, we stated that the review would be conducted for ‘school and community based . . . prevention interventions’, and defined ‘community’ as ‘school affiliated’ settings. The clarification here pertains to the definition of ‘school affiliated’, which was operationalised as ‘formally attached or linked to a specific school setting’.

• A further clarification relates to the definition of CMDs. In this review, CMDs were defined in reference to their prevalence. The updated protocol clarified that the clinical conditions of interest were anxiety, depressive and conduct disorders, as these are the most common across the included age groups.

Analysis

• The educational setting for each study was categorised as UK-specific primary, secondary and tertiary groupings for the purposes of analysis only. This was not made explicit in the original protocol, which implied that the intervention should be delivered in one of these settings.

• We planned to analyse ‘inequality’ as a main outcome. However, owing to a lack of data, this was not possible; instead, we considered subgroup analyses by socioeconomic status (SES), sex and ethnicity. These characteristics were selected post hoc, based on participant characteristic data that had been extracted.

• We planned to conduct a metaregression of intervention intensity, in which intensity was defined as total session time (number of sessions × duration in minutes). However, we determined that this would not be meaningful in a NMA with differing classes of intervention, and so the analysis was not conducted.

• In response to reviewer comments on Caldwell et al. ,1 we added a post hoc analysis for a composite internalising symptoms outcome (which combined depression and anxiety symptom scores).

• A post hoc NMA of parent-reported child symptoms was conducted.

• In the protocol, we stated that a cost-effectiveness analysis would be done if there was sufficient evidence to build and populate a model. As an alternative, we planned a cost–consequence analysis. We found that there was insufficient evidence for a cost-effectiveness analysis, and so a cost–consequence analysis is reported.

Criteria for considering studies for this review

In the absence of direct head-to-head evidence comparing interventions, indirect comparisons can be made across a connected network of interventions. For example, the effect of intervention C against intervention B (BC) can be obtained indirectly from the effect of C against intervention A (AC), minus the effect of B against A (AB). The combination of indirect and direct evidence across a network of intervention comparisons is known as a NMA.

The following eligibility criteria were specified to address the key consistency assumption required for a valid NMA. In a three-intervention network, the consistency assumption requires the true BC intervention effect estimated in the B versus C trials to be the same as the BC intervention effect estimated by the A versus C and A versus B trials (had they also included the B and C arms). 57 For this to hold, one should check that the populations included across all trials in the analysis are comparable to each other, with respect to any potential effect-modifying characteristics. 58 This requirement has been conceptualised as ‘joint randomisability’ of the interventions for the target population. 59 ‘Joint randomisability’ implies that a hypothetical, multiarm trial of every included intervention would be reasonable, in principle, and that all participants would be randomisable to any of the interventions included. 60 This requires clearly and specifically defined inclusion criteria, to ensure the included studies, populations and interventions are sufficiently comparable. Further details on NMA are provided in Methods for the evidence synthesis of effectiveness studies.

Study design

Parallel-group RCTs and quasi-randomised controlled trials were eligible for inclusion. We defined quasi-randomised trials as those for which allocation was based on a pseudo-random sequence, such as the order in which participants were recruited or their date of birth. Both individually randomised and cluster randomised trials were eligible for inclusion. We did not plan to exclude crossover trials, but only the first period was considered eligible for inclusion.

Population

We followed the NAM’s definition of primary prevention, which refers to universal, selective and indicated populations (see Figure 1). 27 Briefly, universal prevention addresses whole populations not defined on the basis of risk; selective prevention is targeted at subgroups with a higher than average risk of developing a mental disorder; and indicated prevention is targeted at high-risk subgroups and/or individuals with detectable, but subclinical, symptoms of a mental disorder. In the first instance, we used author-reported classifications of the intended prevention level. However, when interventions were delivered to a whole class or school with the same at-risk characteristic (such as schools in low-income areas), they were combined with universal prevention. Studies were excluded if the intervention was described by the author as indicated prevention, but baseline symptoms scores were suggestive of clinically meaningful symptom levels (see Definition of disorder).

Interventions and comparators

Outcomes

According to the NAM classification of primary prevention, the overall, longer-term aim of preventative interventions ‘is the reduction of the occurrence of new cases’26 of mental disorders. However, it also recognises the importance of shorter-term prevention in terms of reducing symptoms, which, in turn, may delay or reduce the risk of the onset of the disorder. All are considered beneficial at a population level and are ‘worthwhile goals of prevention’. 26 In this report, we focus on the effect of prevention interventions on symptoms of anxiety, depression and conduct disorder. The main outcome was prevention or reduction of disorder-specific symptoms for self-reported anxiety, depression and conduct disorder.

All validated disease-specific measurement scales for CYP were eligible for inclusion. When studies reported multiple outcome measures, we applied a prespecified hierarchy to select the most appropriate outcome for analysis from each study (see Appendix 1). We did not exclude studies reporting a composite mental health scale from the systematic review [e.g. the Strengths and Difficulties Questionnaire (SDQ)]; however, they were not combined with disorder-specific scales in the main NMA. In a change from protocol (see Table 1), a post hoc analysis for composite ‘internalising’ symptom scales was conducted. For example, measurement scales reporting a total or combined score across depression and anxiety symptoms, such as a total Revised Children’s Anxiety and Depression Scale (RCADS) score, or a composite outcome such as the SDQ ‘emotional symptoms’ subscale, were included in this post hoc outcome.

The following additional primary and secondary outcomes were also specified a priori. However, in the absence of a core outcome set65 or guidelines for the selection of measurement scales for school-based mental health interventions,66 we determined that an inclusive approach to additional primary and secondary outcomes was appropriate. Therefore, we did not specify how these outcomes should be measured in advance, or which scales should be used. Instead, we extracted outcomes as reported by the study authors.

Additional primary outcomes were as follows:

• self-reported well-being, as defined by study author(s)

• self-reported suicidal ideation, behaviour and self-harm, as reported by study author(s)

• intervention impact on inequalities in health, as defined by study author(s).

The primary time point of interest was immediately post intervention. However, as sustainability of intervention effect is an important question for public mental health,67 we also report results for mid-term (6–12 months) and longer-term (13–24 months) follow-ups. If studies had a follow-up of ≥ 25 months, these results were also extracted.

Secondary outcomes of interest were as follows:

• Mental health-related stigma, as defined by study author(s). During our initial patient and public involvement (PPI) focus groups, reducing the stigma associated with mental health problems was identified as an important outcome for young people.

• Acceptability of an intervention to young people, as reported by the study author(s).

• Parent-reported prevention or reduction of disorder-specific symptoms, as reported by the study author(s).

• Self-reported problem behaviour, such as substance use or involvement in violence.

• Academic attainment, as defined by the study author(s).

Identification of studies

As noted in Chapter 1, the approach to searching was modified from that of the original proposal. Full details and explanation are reported in Table 1.

The revised search strategy involved three stages, which might be considered to combine the ‘known items’ and ‘law of diminishing returns’ approaches described by Booth,52 to optimise searching. First, working with an information specialist (SDa) and following the Cochrane MECIR guidance on conducting searches,50,51 we searched MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL) and PsycInfo electronic databases. The search strategies for each electronic database are described in Appendix 1. The final searches were carried out on 4 April 2018. Searches were not restricted by language, country or date of publication.

Second, in addition to the database searches, we conducted searches of Epistemonikos (www.epistemonikos.org; Epistemonikos Foundation, Santiago, Chile) to identify published systematic reviews of interventions for the prevention of CMDs among CYP. Epistemonikos was searched on 16 November 2016. The reference lists from these reviews were added to the results of the database searches, ready for screening.

Finally, after screening, we conducted an informal scoping search of the Education Resources Information Center (ERIC) database. This was to check whether further relevant studies could be located, cross-referencing with those already identified from the previously mentioned approaches. If scoping revealed further relevant studies, a formal search was planned. In response to reviewer comments on the draft version of this report, the ERIC scoping searches were formalised and extended to the British Education Index (BEI). Further details are reported in Appendix 1.

Screening for study inclusion/exclusion was independently assessed by two reviewers (SRD, JCP, DMC, PC); disagreement was resolved by a third reviewer if necessary (SRD, JCP, DMC, PC, SEH). Owing to the volume of potentially relevant studies retrieved, reasons for study exclusion were recorded at full-text screening only.

We used a standardised data extraction form in Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) to extract information from included studies. Data were extracted by one reviewer and checked by a second (SRD, CF, PC, JCP, DMC). Discrepancies were discussed and a consensus reached. Disagreements were resolved by a third reviewer if necessary. The following information was extracted from the papers:

• Study design and the target CMD of intervention (i.e. anxiety, depression or conduct disorder); whether the intervention was universal or targeted (indicated or selected); and number of participants recruited, randomised and assessed (or clusters, if a cluster RCT).

• Details of participants (country, intervention setting, age, sex, ethnicity).

• Details of intervention as reported by trial author. Narrative description of components and delivery process for experimental and control interventions. This included number of sessions; intervention dose (calculated as intensity of intervention: total session time × duration in minutes); whether the intervention was group or individual, face to face or digital; who facilitated the intervention; and intervention fidelity measures.

• Outcome(s) assessed and all follow-up time points.

• Risk-of-bias assessment, including additional assessment for cluster trials.

• Mean total symptom score and standard deviation (SD) at baseline and follow-up time points for primary measurement scale, change from baseline or mean difference between arms; details on whether results were for completers only or use of methods for handling missing data such as last observation carried forward; and intracluster correlation coefficient (ICC), the statistical model used to account for clustering (if any).

Study authors were contacted for additional data, if necessary.

Classification of interventions and components

Eligible interventions were psychological and psychosocial, educational, or physical interventions. Following previous studies,68–72 and based on author-reported descriptions, the content of psychological and psychosocial interventions was classified into four broad intervention types: cognitive–behavioural, behavioural, third wave and interpersonal. Physical interventions were further classified as exercise or biofeedback. Further categories identified were a combined mindfulness/relaxation intervention, psychoeducational and psychosupportive interventions, and occupational therapy. Control interventions were classified into four categories: no intervention, waiting list, usual curriculum and attention controls. 72–74 A full description of all interventions is given in Chapter 3.

We also conducted an intervention component analysis (ICA)75 to identify features of intervention content and process, influenced by the why, what, who, how, when and how much domains of the Template for Intervention Description and Replication (TIDieR). 76 We applied the constant comparative method,77 whereby the intervention descriptions reported by authors were used to develop a coding scheme to classify the components of each intervention and control. If necessary, descriptions were supplemented by intervention manuals and/or correspondence with the study author. ICA is an iterative and inductive process. One reviewer (SRD) developed and piloted a list of provisional component codes based on published studies. 46,68,72,78 The components were discussed with a second reviewer and a preliminary coding exercise was then undertaken on a sample of interventions by a wider group (SRD, DMC, JK and SEH), and the coding scheme was further refined based on discussion. The refined coding scheme was then applied inductively by one researcher (SDa) and audited by another (DMC). Iterative coding refinements were made until application failed to reveal any new information and the component categories could be described as ‘saturated’.

The working definitions of each intervention and control classification and the classifications of intervention components are provided in Chapter 3.

Data preparation

For continuous outcomes, data were extracted for number randomised to each intervention arm at baseline, and baseline mean and SD, and number assessed at follow-up, and follow-up mean and SD (for each time point listed previously). If the mean change from baseline was reported, then this was extracted, together with the standard error (SE) for the mean change from baseline (if reported). Data were extracted for complete cases. However, if authors reported means and SEs from an appropriate model accounting for participant dropout or non-response, this was preferred.

For analysis, we used the standardised mean change from baseline, as a variety of outcome measurement scales were used across the studies. An adjustment for small sample size was applied, following the formula for Hedges’ g. 80 For studies that did not report mean change from baseline, we derived this from reported baseline and follow-up means and SDs. 81 Here we assumed a correlation coefficient of 0.7, based on previous analyses. 82 This value was explored in sensitivity analyses. Results are summarised using SMDs and 95% credible intervals (CrIs).

For dichotomous outcomes, data were extracted for available cases unless authors clearly reported events and number of participants following the intention-to-treat principle. Dichotomous outcomes were summarised using odds ratios (ORs) and 95% CrIs.

If key statistics (e.g. SDs) were not available in the published report, we contacted trial authors for further information. In cases of non-response, or if missing data were not available, we did not impute the data and these studies were excluded from the NMA (but not the systematic review).

For cluster randomised trials that did not account for the effect of clustering, we followed the advice in the Cochrane handbook (section 16.3.4)81 for calculating an approximate sample size. We reviewed reported ICCs from all included papers and used an ICC estimate of 0.03, which is the mean of the values reported and is similar to ICCs used in previous public health systematic reviews. 68,83–85 This value was also explored in sensitivity analyses.

Pairwise and network meta-analyses

Both standard pairwise meta-analyses and NMAs were conducted. A NMA was planned for each primary outcome and for the primary time point only.

A NMA allows data on multiple interventions to be pooled in a single analysis. 86 It is considered a ‘core method’ by Cochrane59 and is routinely used in National Institute for Health and Care Excellence (NICE) technology appraisals and guidelines. 87,88 For the novice, the Cochrane handbook81 provides an accessible introduction; for further depth, we recommend consulting tutorials and introductory papers on NMA. 44,45,60,89–92 However, we describe the fundamentals in the following paragraphs.

A NMA requires that the intervention comparisons made in RCTs can be displayed pictorially as a network of comparisons that are ‘connected’ (i.e. there is a path from any one intervention to another formed by RCT evidence). 59,93 In the absence of direct head-to-head evidence comparing interventions, indirect comparisons can be made via common comparator(s) across the network. For example, the effect of intervention C against intervention B can be obtained from the effect of C against A, minus the effect of B against A. The combination of indirect and direct evidence across a network of intervention comparisons is known as a NMA.

The validity of a NMA assumes that there are no differences between studies in factors that might interact with the intervention effect (effect modification). This is the same assumption made in a pairwise meta-analysis,57 but in a NMA applies across intervention comparisons. It is therefore important to consider separate analyses according to factors that may be potential effect modifiers. In the first instance, separate analyses were conducted by population (universal or targeted) and setting (primary, secondary or tertiary education). Separate analyses were run for the main outcomes of self-reported anxiety, self-reported depression and self-reported conduct disorder symptoms. Following research suggesting common mechanisms and pathways within internalising and externalising disorders (transdiagnostic factors),94–97 we ran analyses across (1) all studies aiming to prevent depression and/or anxiety and (2) studies aiming to prevent conduct disorder. That is, studies contributing to either the depression or anxiety outcome analyses could be studies that aimed to prevent (1) anxiety (2) depression or (3) anxiety and depression. Studies contributing to the analysis of the conduct disorder outcome were only those aiming to prevent conduct disorder. We explore this decision further in a subgroup analysis (see Subgroup, metaregression and sensitivity analyses). A visual check of the inclusion/exclusion characteristics of trials in each network was conducted, to ensure that potential effect modifiers were evenly distributed across studies.

Network plots were drawn in Stata^® version 15 (StataCorp LP, College Station, TX, USA) to allow visual inspection of network connectedness. 98

For the primary time point of post intervention only, we considered three NMA models, each allowing for increasing specificity of intervention detail:46,82

1. Intervention-level model – interventions were analysed as ‘clinically meaningful units’. 48 For example, cognitive–behavioural therapy (CBT) was analysed as a distinct intervention to psychoeducation or third wave-based interventions.

2. Additive model – components nested within intervention. A main intervention effect was estimated (as per the intervention-level model), plus additional effects for specific components within each intervention. For example, we estimated an overall CBT effect, which represents the effect for CBT interventions with components that were common across all the included CBT interventions, and also estimated the additional effect of CBT interventions containing a psychoeducation component, a mindfulness component and so on.

3. Full interaction model – components nested within intervention: under this model, each unique combination of intervention and components was considered as a separate intervention. For example, CBT with cognitive + behavioural + psychoeducation components was considered a distinct intervention to CBT with cognitive + behavioural + psychoeducation + relaxation components.

For follow-up time points, where we anticipated finding fewer studies, we ran the intervention-level model only as prespecified in the protocol. Intervention-level analyses were implemented in a Bayesian framework using OpenBUGS software (version 3.2.3). Component analyses were implemented in WinBUGS99 (version 1.4.3; MRC Biostatistics Unit, Cambridge, UK). Statistical details are reported in Appendix 1. Data and WinBUGS code can be obtained by contacting the corresponding author. Vague prior distributions were specified for intervention effect and heterogeneity parameters (see Appendices 1 and 3). We assessed convergence for the intervention-level NMA based on three chains using the Brooks–Gelman–Rubin diagnostic tool and history plots in OpenBUGS. Specific convergence details for each model and population/setting analysis are reported in the model fit tables in Appendix 3, Table 29–46.

Random-effects models were run for the main outcomes, assuming a common between-study SD (known as ‘homogeneous variance’). 100 However, we assessed both fixed- and random-effects models on the basis of model fit. Heterogeneity was evaluated by examining the posterior median between-study SD (τ) and 95% CrIs from the random-effects model, and by comparing model fit of the fixed- and random-effects models. Model fit was measured by the posterior mean of residual deviance. In addition, we examined the deviance information criterion (DIC), which penalises model fit with model complexity. Differences of ≥ 5 points in posterior mean residual deviance and the DIC were considered meaningful, with lower values preferred. 101 Model fit statistics are reported in Appendix 3.

As described previously, a key assumption for a valid NMA is that of consistency between the direct and indirect evidence. If the effect estimates from the direct and indirect evidence in a network do not agree, this is known as inconsistency. The strict inclusion/exclusion criteria described previously were specified to avoid inconsistency, but they do not guarantee consistency. For this reason, the statistical agreement of the evidence was formally checked. Consistency was assessed by comparing the goodness of fit of a model assuming consistency with that of one allowing for inconsistency (i.e. a model that provides effect estimates based on direct evidence only). A common between-study SD was also assumed for these inconsistency models. 57

Pairwise meta-analyses were also conducted when head-to-head evidence was available. The method of estimation is similar to the NMA, except that the consistency assumption is removed such that intervention effects for separate comparisons are unrelated and separate intervention effects can be estimated. 57 Estimates are reported for the post-intervention time point only and are from a random-effects model that assumes that the heterogeneity parameter is common across intervention comparisons. This better reflects the assumption made in the NMA and, therefore, allows a fair comparison of the intervention effect estimates obtained from both approaches.

Subgroup, metaregression and sensitivity analyses

Metaregression and subgroup analyses were performed in OpenBUGS following the Evidence Synthesis Technical Support Unit code available from the NICE Decision Support Unit’s website and described in Dias et al. 102,103 Subgroup analyses were conducted to assess whether or not intervention effects differed by intended focus of the intervention, for example if interventions addressing anxiety had a larger effect on anxiety outcomes than interventions intended to focus on depression but which also recorded anxiety outcomes.

Metaregression was planned for the intervention-level NMA and main outcomes only, to examine if intervention effects differed by mode of intervention delivery and who facilitated intervention delivery:

• Mode of intervention delivery – interventions were categorised as being delivered face to face or via a computer/internet. To explore whether or not intervention effects were modified by mode of delivery, we fitted a metaregression model for face-to-face (covariate value = 0) and computer/internet (covariate = 1) interventions. A random-effects NMA model was fitted. However, the regression coefficient for the covariate was assumed to be a fixed effect across studies. The between-study SD was assumed to be common for face-to-face and computer/internet interventions. Vague priors were specified.

• Who facilitated intervention delivery – interventions were categorised as being facilitated by a teacher or a mental health professional (MHP). There was considerable variation within the category of ‘MHP’ and it should be regarded as a simplification. Here, MHP included school counsellors, qualified psychotherapists and graduate and post-doctoral psychology students. Graduate and post-doctoral students included those studying general psychology, educational psychology or counselling psychology, when specified. We fitted a metaregression model that enabled us to estimate the intervention effect at each value of the covariate (0 or 1), for each intervention, comparing the effect of each facilitator (e.g. CBT-teacher vs. CBT-MHP vs. usual care). To allow for networks containing multiarm studies with more than two interventions facilitated by a teacher or MHP, a hierarchical model was fitted. In this hierarchical model, a regression coefficient for each intervention was assumed to come from a normal distribution with a common mean and between-intervention SD. The between-intervention SD was assumed to be common for each value of the covariate. Vague priors were specified.

We explored the potential for small-study effects using comparison-adjusted funnel plots. 98

Sensitivity analyses were conducted for the intervention-level NMA, main outcomes and primary time point only. Analyses explored the robustness of results to the following:

• Excluding studies deemed to have a high/unclear risk of bias on the domains of random sequence generation and allocation concealment.

• The ICC value of 0.03 for cluster randomised trials. Sensitivity analyses were conducted assuming an ICC of 0.01 and 0.06.

• The correlation value of 0.7 assumed for calculating change from baseline SD. Sensitivity analyses were conducted assuming a correlation of 0.6 and 0.8.

Interpreting the results and evaluating evidence of effect

As described previously, summary effect estimates and their 95% CrIs from the NMA are reported. In interpreting these statistical findings, we followed the guidance from the Cochrane handbook (section 15.3) that interpretation of results from a meta-analysis should not rely on statements of ‘statistical significance’ or thresholds implying ‘significance’. 104 Instead, we interpret the strength of statistical evidence on a graduated scale from weaker to stronger evidence of an intervention effect. 105,106

In Chapter 9, we also provide a summary of these statistical findings for the primary outcomes of anxiety, depression and conduct disorder symptoms at the primary time point of post intervention. This interpretation forms the basis of the conclusions for the report. The criteria used are based on the considerations outlined in Chapters 14107 and 15104 of the Cochrane handbook. These considerations are informed by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) domains of imprecision, inconsistency (heterogeneity), risk of bias and publication bias. It is not a formal application of the GRADE rating system, which necessitates up- or downgrading of the evidence on the basis of the assessments to form an overall assessment of ‘quality’. 107

Specifically, our interpretation of the evidence is not solely based on the magnitude and direction of the summary point estimate, but also incorporates an evaluation of all of the following:

• the precision of the effect estimate, as described by the 95% CrI

• the extent of the between-study heterogeneity observed in each analysis, as described by τ and its 95% CrI

• the risk-of-bias assessment

• the possibility of non-reporting biases or evidence of small-study effects.

Behavioural therapy

Behavioural therapy is a group of allied techniques that focus on behavioural models of psychology and seek to modify overt maladaptive behaviours. In the current review, we categorised interventions based on behavioural activation, self-monitoring, role-playing, exposure to feared stimuli or scheduling pleasant activities as being behavioural in nature.

Cognitive–behavioural therapy

Cognitive–behavioural therapy can be considered a family of allied techniques, based in both behavioural and cognitive models of psychology, that utilise a set of overlapping cognitive and behavioural techniques. CBT is based on the proposition that a person’s behaviour is influenced by their cognitive activity (and vice versa), and that cognitions can be monitored and altered (cognitive restructuring). In turn, emotions and behaviour can be modified via this cognitive change. CBT interventions for treatment of CMDs typically include a psychoeducation component; however, in preventative interventions, this may not always be present.

Third-wave interventions

This was a composite category. Third-wave psychotherapies emphasise mindfulness, acceptance and flexibility. They tend to focus on a person’s relationship to their cognitions and emotions, encouraging an acceptance of thoughts, rather than modifying their content. Interventions that described themselves as mindfulness-based CBT, acceptance and commitment therapy or dialectical behavioural therapy were included in this classification. Third-wave preventative interventions were distinguished from mindfulness meditation or relaxation interventions that did not explicitly address cognitions or behaviours.

Interpersonal therapy

From a treatment perspective, interpersonal therapy (IPT) is based on the relationship between mood symptoms and interpersonal relationships. It seeks to relieve symptoms via resolving interpersonal conflict and difficulties. In the preventative context, IPT addresses the relationship between young people significant adults (e.g. teachers, parents), with regard to avoiding/resolving conflict via improved coping communication skills. The techniques used attempt to improve interpersonal skills may include role play, problem-solving exercises and practising effective communication.

Mindfulness meditation and relaxation-based interventions

In the present review, mindfulness/relaxation is a composite category and is distinct from third-wave interventions. Relaxation includes breathing exercises, muscle relaxation and yoga from the Iyengar or Hatha traditions (as opposed to more vigorous traditions). Mindfulness meditation interventions were included in this category if they focused solely on meditation or relaxation without incorporating aspects of traditional ‘talking’ psychotherapeutic approaches.

Biofeedback

Biofeedback is a mind–body intervention that uses physiological monitoring devices or equipment to learn to control physiological responses, such as heart rate. Users may monitor their heart rate variability using pulse oximetry, for example while completing a standard deep-breathing exercise. The feedback received helps the participant learn how to influence the negative, or undesired, response (e.g. a stress response). Smartphone applications and ‘consumer wearables’ have been developed for monitoring stress, anxiety and sleep problems.

Exercise

In this review, we classified an exercise intervention as a cardiovascular intervention designed to raise heart rate and breathing to (at least) a moderate intensity level, for example dancing, running and team sports.

Cognitive bias modification

Cognitive bias modification (CBM) relates to a group of approaches, including attention and interpretation bias training, that aim to retrain cognitive distortions. CBM evolved from a visual attention task (a dot-probe task), in which a participant is exposed to a series of threatening and neutral stimuli via computer, such as angry and neutral faces, and the speed of their response to a ‘probe’ is measured (e.g. where on the screen the angry or neutral face was displayed). In individuals with a CMD, attention tends to be selectively directed towards the negative image and response times to the probe are slower. CBM for the treatment of anxiety disorders seeks to ‘retrain’ this selective attention bias towards the positive stimulus. In preventative interventions for CYP, CBM tasks may be embedded in engaging and user-friendly formats such as interactive video games (‘CBM gamification’).

Occupational therapy

Occupational therapy interventions are based on engaging CYP in meaningful daily activities or ‘occupations’. Interventions are skill based and aim to enable CYP to successfully engage with, and participate in, developmentally appropriate everyday events. For example, an intervention might focus on a favourite activity to increase self-esteem, or schoolwork may be modified to create a positive learning environment and reduce stress.

Psychoeducation

Often a component of CBT-based interventions, psychoeducation can also be used as a distinct intervention. It typically involves a systematic approach to providing background information, for example what the cause or symptoms of a mental disorder are and advice regarding the mental disorder and/or explaining the approaches that can help to mitigate symptoms. Written materials or presentations may be provided.

Psychosupport and counselling

Often a component of other interventions, psychosupportive interventions are also used in a stand-alone format. In the current report, we combined psychosupportive and counselling-based interventions into one category. Here it refers to a non-specific, possibly therapeutic, intervention that could include listening, signposting to further services, or forming an attachment or therapeutic alliance.

Usual curriculum

If an active intervention took place during a regular timetabled class and participants in the control group continued to receive the regular class curriculum, the control intervention was classified as standard provision or ‘usual curriculum’. This included a variety of different classes and could have included a ‘well-being’ or health lesson or a standard timetabled academic lesson (such as history or mathematics).

Waiting list

If participants in the control group were explicitly told (e.g. via informed consent processes) that they would receive the active intervention at a later date, the control condition was categorised as a waiting list. Although participants were also likely to be receiving usual curriculum or a no-intervention control, the use of an explicit waiting list design takes precedence in our categorisation.

No intervention

A no-intervention control categorisation was used to differentiate between a control condition in which participants received something and a control condition in which participants were not involved in any structured activity. This classification was applied when the active intervention was held outside regular timetabled classes (e.g. after school) and the participants were not described as being in a waiting list control.

Attention control

A control was classified as attention control if it was a de novo intervention provided to the participants for the purpose of the research study.

Behavioural

A behavioural component was one in which techniques included helping participants to practise and acquire new skills to cope or manage difficult emotions, moods or behaviours. This component includes strategies used in behavioural activation, social skills exercises (including how to make friends, be a good friend and support your friends), role play, assertiveness training, interpersonal work and activity scheduling and contingency management including goal-setting, planning and decision-making activities, problem-solving and exposure. Following Hetrick et al. ,68 this component was initially subdivided into four further subcomponents: (1) social skills training, (2) problem-solving, (3) exposure and (4) ‘other’ behavioural categories. However, this resulted in unconnected networks, so results are reported for a ‘lumped’ behavioural component only.

Cognitive

This component label was applied when an intervention included strategies or techniques designed to identify and replace cognitive distortions with more accurate and adaptive ones, for example recognising and understanding thoughts and feelings, using positive self-talk and challenging negative self-talk and thoughts.

Third wave

During the ICA, we observed that standard CBT, third wave interventions and mindfulness/relaxation interventions were often based on combinations of the same components. We included a third wave component category to ensure differentiation between these ‘therapy-level’ interventions. The component definition is the same as described previously for the intervention level analysis.

Mindfulness

Mindfulness techniques included guided meditation, colouring and drawing, and exercises to practise being in the moment and being free from judgemental thoughts and distractions. On completion of the component coding, we observed that a mindfulness component was always present in conjunction with a relaxation component.

Relaxation

Separate mindfulness and relaxation components were specified to allow for relaxation techniques that were not defined as meditation or mindfulness. This included strategies such as progressive muscle relaxation, abdominal breathing exercises, cue-controlled relaxation, and identification of physiological arousal (‘body clues’) approaches.

Physiological

A component was coded as physiological if it involved the process of displaying involuntary or subthreshold physiological processes, usually by electronic instrumentation, and learning to voluntarily influence those processes by making changes in cognition.

Bias modification

This component was present only in the main therapy-level intervention, CBM, as described previously. However, on completion of coding, it was retained as a separate component, as the four studies that could be described as evaluating a CBM intervention were assessed as containing different combinations of components:

• study 1114 – bias modification

• study 2115 – behavioural + relaxation + physiological + bias modification

• study 3116 – behavioural + physiological + bias modification

• study 4117 – cognitive + bias modification.

Psychoeducation

Psychoeducation was also included as a component of broader interventions. The definition applied at the component level is the same as described previously for the intervention-level analysis.

Additional process and implementation classifications

We also extracted information on the following implementation and process components of interventions: number of sessions, duration of intervention (minutes), mode of delivery (face to face or digital), group or individual delivery, who the facilitator was and whether or not training was provided, and whether or not intervention fidelity was monitored. The characteristics of interventions, the components and process details are provided in Chapters 4 and 5 and in Appendix 2. We did not include the process components in the NMA owing to concerns regarding a lack of power and network connectedness. 82

Studies included in the review

The overall Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram for the whole review is reported in Figure 2. A total of 11,990 citations were screened, and 1512 full-text articles were retrieved for further screening. Of these, 142 studies were identified as eligible for inclusion in either the anxiety, depression or conduct disorder reviews. Fifty-four studies reported both an anxiety and a depression outcome. The full list of included studies can be found in Appendix 2.

FIGURE 2.

Study selection process: PRISMA flow diagram for whole systematic review. a, Not mutually exclusive. Some studies reported both anxiety and depression outcomes. Of 142 studies, 54 reported both anxiety and depression outcomes. Seven did not report either. Forty-eight of 109 studies contributing to the NMA reported both an anxiety and a depression outcome. b, Study was included in the NMA at any of the follow-up time points. Note that references to the main study publication and articles awaiting classification are listed in Appendix 2.

Of the 142 studies eligible for the review, 79 included a self-reported anxiety outcome; the details are reported in this chapter. Studies reporting a depression or conduct disorder outcome are reported separately in Chapters 5 and 7. Among the studies reporting an anxiety outcome, the primary focus of 38 studies was the prevention of anxiety whereas 13 were focused on the prevention of depression and 28 addressed both anxiety and depression. Subgroup analyses examining whether or not intervention effects differed by intended focus of the intervention are reported in Exploring heterogeneity and small-study effects.

Study characteristics are reported in Appendix 2. Included studies were published between 1982 and 2018, and randomised between 22 and 5030 participants (median 184 participants). There were 43 cluster randomised studies, of which four reported cluster-adjusted means and SDs and 35 reported model-based estimates. Thirty-six were individually randomised trials. Seventy-three studies reported a post-intervention end point, 38 reported a follow-up of between 6 and 12 months and seven reported a follow-up of between 13 and 24 months. Studies could report more than one follow-up time point; details of studies reporting multiple time points are in Appendix 2.

Forty-four studies were classified as universal and 35 were classified as targeted (27 indicated, eight selective). Twenty-seven studies were implemented in primary schools, 45 in secondary schools, five in tertiary education and two across multiple settings (i.e. two or more settings). Seventy studies were conducted in HICs, with eight conducted in MICs and one in a LIC. Of the studies conducted in HICs, five were conducted in lower-income settings, as specified by the trial authors. Categorisation of LICs and MICs was based on 2017 World Bank classifications. 62

Universal population, secondary setting

Post intervention

The analysis-specific network diagram is reported in Figure 4. Twenty-one studies (n = 10,208 participants) contributed to the analysis for the main time point of immediately post intervention. 119,120,125–141,143,144 Most studies in this network were judged to be at unclear risk of bias. The risk of bias was judged to be unclear for 13 studies and low for three studies in both the randomised sequence generation and allocation concealment domains. In four studies, the risk of bias was judged to be low for randomisation but unclear for allocation concealment. In one study, the risk of bias was judged to be unclear for randomisation but low for allocation concealment. Sixteen studies included an intervention based on CBT, three were based on third-wave interventions, and two were mindfulness/relaxation-based interventions; the reference intervention was usual curriculum (see Appendix 2 for details). All reported results are from a random-effects NMA model unless otherwise stated. Model fit and selection statistics suggested that a consistency model was appropriate. Of the three models fitted (intervention, additive and full interaction), the additive model was preferred, suggesting evidence for effect modification by components. All model fit statistics are reported in Appendix 3. Results reported in the following sections are SMDs and 95% CrIs.

FIGURE 4.

Network plot for universal population, secondary setting: post-intervention anxiety outcome.

Intervention-level model

Between-study posterior median SDs (τ) were indicative of moderate heterogeneity (τ 0.11, 95% CrI 0.02 to 0.22). Table 3 reports SMDs (and 95% CrIs) for each active intervention relative to usual curriculum. There was weak evidence of a modest effect of CBT in preventing symptoms of anxiety post intervention (SMD –0.15, 95% CrI –0.34 to 0.04). Mindfulness/relaxation interventions (SMD –0.65, 95% CrI –1.14 to –0.19) reduced symptoms relative to usual curriculum. However, this finding must be interpreted in the context of the possible small-study effects observed in the funnel plot reported in Appendix 6, Figure 17, and the ratings of unclear risk of bias. Only two small mindfulness/relaxation studies (n = 30,130 and n = 79137 participants) contributed to the network, and both were rated as having an unclear risk of bias for random sequence generation and allocation concealment. There was a lack of evidence for the effect of third-wave interventions (SMD 0.03, 95% CrI –0.14 to 0.20).

Additive model: components nested within intervention

Component-level models were fitted to evaluate whether or not the observed between-study heterogeneity in the intervention model could be explained by differences in intervention components.

We fitted an additive model with components nested within interventions, such that a main intervention effect was estimated, plus an additional effect for the inclusion of a specific component. Results are reported as regression coefficients (β-values and 95% CrIs) describing the increase or decrease in SMD via the addition of each component to each intervention. The coefficients can be interpreted as the additional effect of each specific component over and above the ‘common’ intervention-level effect.

All CBT interventions in the universal secondary network included a cognitive and a behavioural component. We estimated the additional effect of including a psychoeducational, mindfulness or relaxation component to cognitive and behavioural components. The between-study heterogeneity was reduced compared with that of the intervention-level analysis (τ 0.06, 95% CrI 0.00 to 0.21) (see Appendix 3). The effect of any CBT intervention including a psychoeducation component was to reduce the SMD (β –0.39, 95% CrI –0.78 to 0.01). The effect of including a mindfulness component in a CBT intervention was to increase the SMD (β 0.57, 95% CrI 0.08 to 1.03) (i.e. less effective at reducing anxiety). There was no evidence to suggest an effect of adding a relaxation component to CBT (β 0.07, 95% CrI –0.21 to 0.38).

Table 4 reports the SMDs for all specific additive combinations of intervention components. Under the additive component model, it is possible to estimate an effect for all combinations, even in the absence of directly observable trials. Relative to a usual curriculum control, there is some evidence that the combination of cognitive + behavioural + psychoeducation components is effective at reducing anxiety post intervention in universal secondary settings (SMD –0.30, 95% CrI –0.59 to –0.01). There is a lack of evidence for all other combinations.

TABLE 4 - Results from additive and full interaction component models: universal secondary settings, self-reported anxiety

Notes

Intervention components are nested within the main intervention (CBT). All CBT interventions in the universal secondary analysis contained a cognitive and a behavioural component. The ‘Study arms’ column reports the number of trial arms that include the specific combination of components listed. As there are several multiarm trials, this is not equivalent to the number of studies. For example, there are two study arms that include a CBT intervention, which is defined only by cognitive and behavioural components. The reference intervention is usual curriculum. SMD and 95% CrIs are reported for the additive and full interaction models. For full details of the models, see Chapter 2 and Appendix 1.

Population/setting	Main intervention	Components (within main intervention)	Study arms (n)	Model, SMD (95% CrI)
				Additive	Full interaction
Universal secondary, anxiety	CBT	(Cognitive + behavioural)	2	0.09 (–0.17 to 0.36)	0.09 (–0.22 to 0.40)
		(Cognitive + behavioural) + psychoeducation	6	–0.30 (–0.59 to –0.01)	–0.30 (–0.62 to 0.02)
		(Cognitive + behavioural) + relaxation	2	0.16 (–0.22 to 0.57)	–31.49 (–144.2 to 90.84)
		(Cognitive + behavioural) + psychoeducation + relaxation	6	–0.23 (–0.62 to 0.19)	–0.21 (–0.66 to 0.26)
		(Cognitive + behavioural) + psychoeducation + mindfulness + relaxation	3	0.34 (–0.12 to 0.82)	–31.32 (–144 to 91.01)
		(Cognitive + behavioural) + mindfulness	0	0.66 (–0.02 to 1.31)	–
		(Cognitive + behavioural) + mindfulness + relaxation	0	0.73 (0.03 to 1.45)	–
		(Cognitive + behavioural) + psychoeducation + mindfulness	0	0.27 (–0.10 to 0.64)	–

Full interaction model: components nested within intervention

Under a full interaction model, each different combination of intervention and components is considered as a separate intervention. For example, a CBT intervention with cognitive + behavioural + psychoeducation components is considered to be distinct from CBT with cognitive + behavioural + psychoeducation + relaxation components. However, the power to estimate and detect evidence of interactions is limited by the data available for each distinct combination. For the universal secondary anxiety analysis, there were sufficient data to estimate the effects for CBT-based interventions only, although we note that power was low and convergence was problematic. Model fit statistics suggested that the full interaction model was slightly preferred over the intervention-level model, but that the additive effect model is preferred overall (see Appendix 3). However, for completeness, we also report the full interaction results in Table 4. There was weak evidence that CBT with cognitive + behavioural + psychoeducation components is effective relative to usual curriculum (SMD –0.30, 95% CrI –0.62 to 0.02). The between-study posterior median SD was suggestive of low to moderate heterogeneity (τ 0.09, 95% CrI 0.01 to 0.24). Regression coefficients are reported in Appendix 3, but yielded very imprecise estimates of change in SMD.

Universal, secondary, further time points: intervention-level effects

Full results for the following time points are reported in Appendix 6. Details for studies contributing to each time point are reported in Appendix 2.

Six to 12 months post intervention

Fifteen studies (n = 13,150 participants), comparing seven interventions, were included in the analysis for 6–12 months post intervention. 118,125,126,128,129,131,133–136,138,139,141–143 Twelve studies included an intervention based on CBT, one study included a CBT + IPT intervention and two studies included a third-wave intervention. There was no evidence to suggest that any intervention reduced symptoms of anxiety between 6 and 12 months, relative to usual curriculum (CBT: SMD –0.11, 95% CrI –0.34 to 0.11; third wave: SMD –0.05, 95% CrI –0.32 to 0.22; and CBT + IPT: SMD –0.02, 95% CrI –0.42 to 0.36). Between-study heterogeneity was moderate (τ 0.15, 95% CrI 0.06 to 0.37).

Thirteen to 24 months post intervention

Three studies (n = 1077 participants) contributed to the analysis for 13–24 months post intervention, all of which included an intervention based on CBT. 136,138,139 There was no evidence to suggest that CBT-based interventions prevented symptoms of anxiety between 13 and 24 months, relative to usual curriculum (SMD –0.01, 95% CrI –2.84 to 2.81).

Twenty-five or more months post intervention

One study (n = 92 participants) reported a follow-up time point of 30 months. 139 The SMD for the effect of CBT relative to usual curriculum was –0.23 (95% CrI –0.55 to 0.08). The study was rated as having an unclear risk of bias for random sequence generation and allocation concealment.

Universal population, primary setting

Post intervention

The analysis-specific network diagram for universal primary settings is reported in Figure 5. Fifteen studies from 14 publications (n = 5605 participants) contributed to the analysis for the main time point of immediately post intervention. 145–158 Thirteen studies were deemed to have an unclear risk of bias and one study was deemed to have a low risk of bias for both randomised sequence generation and allocation concealment domains. One study was judged to be at unclear risk of bias for randomisation but at low risk of bias for allocation concealment. All studies included an intervention based on CBT. Model fit and selection statistics suggested that a random-effects consistency model was appropriate. Fit was similar across all three models (intervention, additive and full interaction) but indicated that the intervention-level model was preferred (see Appendix 3). Therefore, we report effect estimates from the intervention-level analysis only. Regression coefficients from the additive and full interaction models are reported in Appendix 3.

FIGURE 5.

Network plot for universal population, primary setting: post-intervention anxiety outcome.

Intervention-level model

Between-study posterior median SDs were indicative of moderate heterogeneity (τ 0.10, 95% CrI 0.01 to 0.26). There was weak evidence of a very small effect of CBT relative to usual curriculum (SMD –0.07, 95% CrI –0.23 to 0.05) (see Table 3).

Universal, primary, further time points: intervention-level effects

Six to 12 months post intervention

Ten studies (n = 4794 participants) contributed to the analysis for 6–12 months post intervention, all of which evaluated a CBT-based intervention. 145,149–152,154,155,157,159 Between-study posterior median SDs were indicative of substantial heterogeneity (τ 0.22, 95% CrI 0.08 to 0.45). There was no evidence that CBT reduced symptoms of anxiety at between 6 and 12 months, relative to usual curriculum (SMD –0.11, 95% CrI –0.35 to 0.11) (see Appendix 5).

Thirteen to 24 months post intervention

Three studies (n = 1603 participants) contributed to the analysis for 13–24 months post intervention, all of which included an intervention based on CBT. 152,157,159 There was no evidence to suggest that CBT-based interventions prevented symptoms of anxiety at between 13 and 24 months, relative to usual curriculum (SMD 0.00, 95% CrI –0.68 to 0.71; τ 0.13).

Twenty-five or more months post intervention

One study (n = 910 participants) reported a follow-up time point of 30 months. 152 This study provided weak evidence of a small effect of CBT relative to usual curriculum (SMD –0.12, 95% CrI –0.26 to 0.02). The study was deemed to be at unclear risk of bias for random sequence generation and allocation concealment.

Targeted population, secondary setting

Post intervention

The analysis-specific network diagram for targeted secondary settings is reported in Figure 6. Fifteen studies (n = 2383 participants) contributed to the analysis for the main time point of immediately post intervention. 114,117,130,141,160–163,165,167,168,170–173 Three studies were deemed to be at low risk of bias and seven were deemed to be at unclear risk of bias for randomised sequence generation and allocation concealment. Four studies were deemed to be at low risk of bias for randomised sequence generation and at unclear risk for allocation concealment. Five studies were multiarm and compared multiple active interventions. Twelve studies included an intervention based on CBT, two studies examined biofeedback interventions, two studies included a CBM intervention, two studies included a mindfulness/relaxation intervention and one study evaluated study an exercise intervention. Model fit and selection statistics suggested that a random-effects consistency model was appropriate. Model fit was similar across all three intervention models (main intervention, additive and full interaction) (see Appendix 3). There was no evidence of effect modification by intervention components in targeted secondary settings. The intervention-level model is preferred, and regression coefficients from the additive and full interaction models are reported in Appendix 3.

FIGURE 6.

Network plot for targeted population, secondary setting: post-intervention anxiety outcome.

Intervention-level effects

There was mild to moderate between-study heterogeneity (τ 0.06, 95% CrI 0.00 to 0.21). Table 3 reports SMDs (and 95% CrIs) for each active intervention, relative to no intervention. There was no evidence of an effect for CBT (SMD 0.03, 95% CrI –0.11 to 0.16), biofeedback (SMD –0.18, 95% CrI –0.55 to 0.21), CBM (SMD –0.17, 95% CrI –0.45 to 0.11) or mindfulness/relaxation (SMD 0.03, 95% CrI –0.42 to 0.48). There was evidence that exercise reduced post-intervention anxiety symptoms, relative to no intervention (SMD –0.47, 95% CrI –0.86 to –0.09). However, exercise was evaluated in only one study, which was judged to be at unclear risk of bias for random sequence generation and allocation concealment.

Targeted, secondary, further time points: intervention-level effects only

Six to 12 months post intervention

Six studies (n = 1284 participants) contributed to the analysis for 6–12 months post intervention, of which all included a CBT-based intervention and one included a CBM intervention (three-arm study). 117,141,160,163,165,173 There was evidence of mild to moderate between-study heterogeneity (τ 0.06, 95% CrI 0.00 to 0.25). There was no evidence that either CBT (SMD 0.05, 95% CrI –0.12 to 0.20) or CBM (SMD –0.14, 95% CrI –0.53 to 0.24) reduced anxiety at between 6 and 12 months, relative to no intervention.

Thirteen to 24 months, and ≥ 25 months, post intervention

One study (n = 260 participants),166 deemed to be at unclear risk of bias, provided evidence for a small effect of CBT, relative to no intervention, for the prevention of anxiety between 13 and 24 months’ follow-up (SMD –0.26, 95% CrI –0.52 to –0.01) and at 48 months’ follow-up (SMD –0.39, 95% CrI –0.65 to –0.14).

Targeted population, primary setting

Post intervention

The analysis-specific network diagram for targeted primary settings is reported in Figure 7. Eleven studies (n = 1314) contributed to the analysis for the post-intervention time point. 115,116,174–182 Three studies were deemed to be at low risk of bias and six studies were deemed to be at unclear risk of bias for both the randomisation and allocation concealment domains. A further two studies were rated as having an unclear risk of bias for allocation concealment, but a low risk of bias for random sequence generation. One study compared two active interventions. Ten studies included an intervention based on CBT, one examined a biofeedback intervention and one included an occupational therapy intervention. Model fit and selection statistics indicated that a random-effects consistency model was appropriate. Model fit was similar across all three intervention models (main intervention, additive and full interaction), but suggested that the intervention-level model was preferred (see Appendix 3). Regression coefficients from the additive and full interaction models are reported in Appendix 3.

FIGURE 7.

Network plot for targeted population, primary setting: post-intervention anxiety outcome. OT, occupational therapy.

Intervention-level effects

There was evidence of substantial between-study heterogeneity (τ 0.42, 95% CrI 0.21 to 0.89). Table 3 reports SMDs (95% CrIs) for each intervention relative to a waiting list. There was weak evidence of an effect for CBT (SMD –0.38, 95% CrI –0.84 to 0.07), but a lack of evidence for biofeedback (SMD –0.38, 95% CrI –1.50 to 0.72) or occupational therapy (SMD 0.11, 95% CrI –0.91 to 1.14).

Targeted, primary, further time points: intervention-level effects only

Six to 12 months post intervention

Five studies (n = 713 participants) contributed to the analysis for 6–12 months post intervention, of which five included a CBT-based intervention and one included a biofeedback intervention (one study compared two active interventions). 116,175–178 The between-study posterior median SD was indicative of substantial heterogeneity (τ 0.52, 95% CrI 0.15 to 2.51). There was no evidence that either CBT (SMD –0.17, 95% CrI –1.37 to 1.06) or biofeedback (SMD –0.28, 95% CrI –2.49 to 1.93) reduced anxiety symptoms at between 6 and 12 months, relative to a waiting list. No studies reported a follow-up of > 12 months post intervention.

Targeted population, tertiary/university setting

The analysis-specific network diagram is reported in Figure 8. Four studies (n = 743 participants) contributed to the analysis for the main post-intervention time point, of which all included an intervention based on CBT. 183,184,186,187 Owing to insufficient data in this network, the component models were not fitted.

FIGURE 8.

Network plot for targeted population, tertiary/university setting: post-intervention anxiety outcome.

Exploring heterogeneity and small-study effects

Subgroup analyses, metaregression and sensitivity analyses were conducted for the intervention-level NMA, for the main outcome and for the primary end point of post intervention only. The comparison-adjusted funnel plots suggest that small study effects are possible in the universal primary and secondary settings analyses. That is, smaller studies may be reporting more beneficial results than larger studies among non-active controlled trials reporting an anxiety outcome (see Appendix 6). This is a potentially important finding. Of all the populations and settings considered in this chapter, the universal analyses provided the strongest evidence for a reduction in anxiety post intervention. There was no evidence to suggest the presence of small-study effects for the analyses of targeted populations in either setting.

A metaregression was conducted for intervention mode of delivery (face to face or via computer), and for intervention facilitator (teacher or a MHP). There was no evidence of effect modification by facilitator or mode of delivery for any population or setting combination (see Appendix 6). However, for the universal primary analysis, there was weak evidence that teacher-delivered CBT interventions (SMD –0.05, 95% CrI –0.21 to 0.08) may be slightly less effective than MHP-delivered CBT interventions (SMD –0.18, 95% CrI –0.42 to 0.00).

Subgroup analyses were conducted to evaluate if intervention effects differed by intended focus of the intervention. For each population and setting combination, intervention estimates were compared across three subgroups: (1) interventions that aimed to prevent anxiety (2) interventions that aimed to prevent only depressive symptoms and (3) interventions that aimed to prevent both anxiety and depression.

For the universal secondary and universal primary anxiety networks, there was very weak evidence that intervention focus was important. Interventions focused on preventing anxiety appeared to have a larger effect on anxiety symptoms than those focusing on depression or combined depression and anxiety. However, CrIs overlapped, and we did not conduct a statistical test to examine subgroup differences. For targeted populations, there was some evidence to suggest that interventions focusing on both anxiety and depression were slightly more effective than interventions focused on anxiety alone. However, we emphasise that these results are to be considered descriptive only. Full results are reported in Appendix 6.

Studies included in the review

The overall PRISMA flow diagram for the project can be seen in Figure 2. There were 105 studies reporting a self-reported depression outcome. The full list of included studies can be found in Appendix 2. Of these 105 studies, the primary focus of 62 studies was the prevention of depression, 29 focused on the prevention of anxiety and depression, and 14 focused on the prevention of anxiety alone. A subgroup analysis to explore whether or not intervention effects varied depending on intended intervention focus is reported subsequently in Exploring heterogeneity and small-study effects.

Study characteristics are reported in Appendix 2. Included studies were published between 1993 and 2018, and randomised between 16 and 8873 participants (median 198). Fifty-three studies were cluster randomised, of which seven reported cluster-adjusted means (SEs) and 29 reported results from appropriate models. The median number of clusters was 13 [interquartile range (IQR) 8–22 clusters]. Fifty-two were individually randomised trials. Ninety-seven studies reported a post-intervention end point, 64 reported a follow-up of between 6 and 12 months, and 18 reported a follow-up of between 13 and 24 months. Six reported a follow-up of ≥ 25 months.

Fifty-seven studies were classified as universal and 48 as targeted (38 indicated, 10 selective). Twenty studies were implemented in a primary school setting, 72 in secondary school, eight in tertiary education and five across multiple settings. A total of 94 studies were conducted in HICs, with 10 conducted in MICs and one in a LIC. Of the studies conducted in HICs, seven were conducted in lower-income settings, as specified by the trial authors.

Universal population, secondary setting

Post intervention

The analysis-specific network diagram is reported in Figure 10. Thirty-four studies (18,094 participants) contributed to the analysis for the main time point of immediately post intervention, of which nine were multiarm trials. 119,125–129,131–136,138,139,141,143,144,190,192–203,205–207 Six studies were deemed to be at low risk of bias, and 18 studies were deemed to be at unclear risk of bias, for both random sequence generation and allocation concealment. Five studies were rated as having an unclear risk of bias for randomisation and a low risk of bias for allocation concealment. Three studies were rated as having a low risk of bias for randomisation and an unclear risk of bias for allocation concealment. Two studies were rated as having an unclear risk of bias for randomised sequence generation and a high risk of bias for allocation concealment. Twenty-five studies included an intervention based on CBT, one included an intervention based on IPT and three included an intervention based on a combination of CBT and IPT. Four interventions were based on third wave and one was based on behavioural therapy. All reported results are from a random-effects NMA model unless otherwise stated. Model fit and selection statistics suggested that a consistency model was appropriate. Of the three component models fitted (intervention, additive and full interaction), the additive model was preferred, suggesting evidence for effect modification by components. All model fit statistics are reported in Appendix 3. Results are reported as SMDs and 95% CrIs.

FIGURE 10.

Network plot for universal population, secondary setting: post-intervention depression outcome.

Intervention-level effects

The between-study posterior median SD (τ) was indicative of moderate heterogeneity (τ 0.15, 95% CrI 0.10 to 0.22). Table 7 reports SMDs (95% CrIs) for each intervention relative to usual curriculum. There was weak evidence of a very small effect of CBT (SMD –0.04, 95% CrI –0.16 to 0.07) in preventing symptoms of depression post intervention. There was weak evidence of a small effect of CBT + IPT (SMD –0.18, 95% CrI –0.46 to 0.08) in preventing symptoms at the post-intervention time point. There was no evidence to suggest that IPT (SMD –0.03, 95% CrI –0.36 to 0.29), third-wave therapies (SMD –0.03, 95% CrI –0.21 to 0.14) or behavioural therapy (SMD –0.02, 95% CrI –0.40 to 0.37) are effective.

Additive model: components nested within intervention

Component-level models were fitted to evaluate whether or not the observed between-study heterogeneity in the intervention model could be explained by differences in intervention components. It was possible to estimate additive component effects in CBT and third-wave interventions.

The between-study posterior median SD was indicative of moderate heterogeneity (τ 0.14, 95% CrI 0.08 to 0.22). All CBT interventions included a cognitive component, and additive effects were estimated for psychoeducation, behavioural, mindful and relaxation components. The regression coefficients are reported in Appendix 3 and suggest that there was no evidence of effect modification by components for CBT interventions. Table 8 reports the SMDs for all specific additive combinations of intervention components. Relative to a usual curriculum control, there was weak evidence that a cognitive plus a behavioural component may be effective at reducing symptoms of depression post intervention in universal secondary settings (SMD –0.11, 95% CrI –0.28 to 0.05). There was a lack of evidence for all other combinations.

TABLE 8 - Results from additive and full interaction component models: universal secondary settings, self-reported depression

Notes

Intervention components are nested within the main intervention (CBT and third wave). All CBT interventions in the universal secondary analysis contained a cognitive component. All third-wave interventions contained a third-wave component. Study arms reports the number of trial arms that included the specific combination of components listed. As there were several multiarm trials, this is not equivalent to the number of studies/trials. For example, one study arm includes a CBT intervention, which is defined by cognitive and psychoeducational components only. The reference intervention is usual curriculum. For full details of model, see Chapter 2 and Appendix 1.

Population/setting/outcome	Main intervention	Components (within main intervention)	Study arms (n)	SMD (95% CrI)
				Additive model	Full interaction model
Universal/secondary/depression	CBT	Cognitive + psychoeducational + behavioural	11	0.01 (–0.12 to 0.13)	–0.03 (–0.15 to 0.09)
		Cognitive + psychoeducational	1	–0.55 (–1.33 to 0.20)	–0.57 (–1.33 to 0.21)
		Cognitive + behavioural + relaxation	4	–0.10 (–0.31 to 0.11)	–0.11 (–0.4 to 0.17)
		Cognitive + behavioural	4	–0.11 (–0.28 to 0.05)	–0.03 (–0.15 to 0.09)
		Cognitive + psychoeducational + behavioural + relaxation	6	0.02 (–0.18 to 0.22)	0.02 (–0.20 to 0.24)
		Cognitive + psychoeducational + behavioural + mindfulness + relaxation	3	–0.01 (–0.4 to 0.38)	–0.02 (–0.46 to 0.41)
		Cognitive + mindfulness	0	–0.70 (–1.6 to 0.18)	–
		Cognitive + relaxation	0	–0.66 (–1.47 to 0.13)	–
		Cognitive + psychoeducational + mindfulness	0	–0.59 (–1.44 to 0.26)	–
		Cognitive + behavioural + relaxation	0	–0.54 (–1.33 to 0.24)	–
		Cognitive + psychoeducational + mindfulness +relaxation	0	–0.57 (–1.44 to 0.29)	–
		Cognitive + psychoeducational + behavioural + mindfulness	0	–0.03 (–0.4 to 0.35)	–
		Cognitive + behavioural + mindfulness	0	–0.14 (–0.59 to 0.30)	–
		Cognitive + behavioural + mindfulness + relaxation	0	–0.13 (–0.58 to 0.32)	–
		Cognitive + mindfulness + relaxation	0	–0.69 (–1.6 to 0.21)	–
	Third wave	Third wave + psychoeducational	1	–0.05 (–0.38 to 0.27)	–0.05 (–0.40 to 0.30)
		Third wave + mindfulness + relaxation	3	0.10 (–0.12 to 0.33)	0.11 (–0.13 to 0.35)
		Third wave + psychoeducational + mindfulness + relaxation	1	–0.35 (–0.70 to 0.00)	–0.35 (–0.72 to 0.02)

All third-wave interventions contained a third-wave component, and additive effects were estimated for psychoeducation and a combined mindfulness + relaxation component. Owing to the data structure, it was not possible to estimate the effects for mindfulness and relaxation components separately. The impact of including a psychoeducation component in third-wave interventions was to reduce the SMD by –0.45 (β –0.45, 95% CrI –0.87 to –0.04). Although this regression coefficient indicates the presence of effect modification, Table 8 shows that there was still only weak evidence that a third-wave intervention (when made up of third wave + mindfulness + relaxation components) is effective at reducing symptoms of depression relative to a usual curriculum (SMD –0.35, 95% CrI –0.70 to 0.00). We note that this is based on evidence from a single study.

Full interaction model: components nested within intervention

Table 8 also reports the number of studies comparing each unique combination of components for CBT and third-wave interventions for the full interaction component model. Model fit and regression coefficients are reported in Appendix 3. The between-study posterior median SD for the full interaction model was suggestive of moderate heterogeneity (τ 0.15, 95% CrI 0.10 to 0.23). There was no evidence of a differential effect of CBT or third wave by intervention components in universal secondary settings.

Universal, secondary, further time points: intervention-level effects

Six to 12 months post intervention

Twenty-eight studies (19,817 participants) contributed to the analysis for 6–12 months post intervention, eight of which were multiarm trials. 118,125,126,128,129,131,133–136,138,139,141–143,194–203,205–207 Twenty-one studies included an intervention based on CBT, four studies included a CBT + IPT intervention and one study evaluated an IPT-based intervention. Three studies included a third-wave-based intervention. The between-study posterior median SD was indicative of low heterogeneity (τ 0.08, 95% CrI 0.02 to 0.15). There was weak evidence, of small effects, to suggest that CBT (SMD –0.02, 95% CrI –0.10 to 0.06), CBT + IPT (SMD –0.10, 95% CrI –0.26 to 0.05) and third-wave interventions (SMD –0.13, 95% CrI –0.27 to 0.01) could prevent symptoms of depression, compared with a usual curriculum comparator. The third-wave studies were judged to be at low risk of bias. The CBT and CBT + IPT studies were judged to be at mostly unclear risk of selection bias (see Appendix 2). There was no evidence to support IPT (SMD 0.11, 95% CrI –0.13 to 0.35) reducing symptoms at 6–12 months, relative to usual curriculum.

Thirteen to 24 months post intervention

Eight studies (7584 participants) contributed to the analysis for 13–24 months post intervention, seven of which included an intervention based on CBT and one that included an intervention based on CBT + IPT. 136,138,139,194,197,204,206,207 There was no evidence to suggest that CBT-based (SMD –0.04, 95% CrI –0.20 to 0.14) or CBT + IPT (SMD –0.10, 95% CrI –0.57 to 0.39) interventions prevented symptoms of depression at 13–24 months, relative to usual curriculum. The between-study posterior median SD was indicative of low heterogeneity (τ 0.07, 95% CrI 0.00 to 0.35).

Twenty-five or more months post intervention

We combined studies reporting time points closest to 36 months post intervention. Three studies (1303 participants) reported time points between 30 and 36 months post intervention. 138,194,206 All evaluated a CBT intervention. There was no evidence of an effect for preventing symptoms (SMD –0.14, 95% CrI –2.89 to 2.63), compared with usual curriculum.

Universal population, primary setting

Post intervention

The analysis-specific network diagram is reported in Figure 11. Twelve studies (4116 participants) contributed to the analysis for the main time point of immediately post intervention. 145,147,150–152,155–157,208,209,211,212 Eleven included an intervention based on CBT and one study evaluated a behavioural intervention. Model fit and selection statistics were suggestive of slight lack of fit, but a consistency model was preferred. Model fit was similar across all three component models fitted (intervention, additive and full interaction), but suggests that the intervention-level model was appropriate (see Appendix 3). Therefore, we report effect estimates from the random-effects intervention-level NMA only. Regression coefficients from the additive and full interaction models are reported in Appendix 3.

FIGURE 11.

Network plot for universal population, primary setting: post-intervention depression outcome.

Intervention-level effects

The between-study posterior median SD was indicative of moderate to substantial heterogeneity (τ 0.32, 95% CrI 0.18 to 0.59). Table 7 reports SMDs (95% CrIs) for each intervention, relative to usual curriculum. There was a lack of evidence that CBT (SMD –0.13, 95% CrI –0.44 to 0.17) or behavioural therapy (SMD –0.10, 95% CrI –1.04 to 0.80) reduced self-reported symptoms of depression post intervention.

Additive model: components nested within intervention

The between-study posterior median SD was indicative of high heterogeneity (τ 0.37, 95% CrI 0.20 to 0.70). All CBT interventions included a cognitive and a behavioural component, and additive effects could be estimated for psychoeducation and relaxation components only. However, there was no evidence for effect modification; regression coefficients are reported in Appendix 3, Table 40.

Full interaction model: components nested within intervention

The between-study posterior median SD was suggestive of high heterogeneity (τ 0.39, 95% CrI 0.21 to 0.78). There was no evidence of effect modification by intervention components for CBT in universal primary settings; regression coefficients are reported in Appendix 3, Table 40.

Universal, primary, further time points: intervention-level effects

Six to 12 months post intervention

Nine studies (4134 participants) contributed to the analysis for 6–12 months post intervention, all of which evaluated a CBT-based intervention. 145,150–152,155,157,159,208,211 The between-study posterior median SDs were indicative of moderate to substantial heterogeneity (τ 0.21, 95% CrI 0.06 to 0.56). There was weak evidence, of a small effect, that CBT prevents symptoms of depression at 6–12 months, relative to usual curriculum (SMD –0.15, 95% CrI –0.43 to 0.09).

Thirteen to 24 months post intervention

Three studies (1602 participants) contributed to the analysis for 13–24 months post intervention, all of which included an intervention based on CBT. 152,157,159 There was no evidence to suggest that CBT-based interventions prevented symptoms of depression at 13–24 months, relative to usual curriculum (SMD –0.03, 95% CrI –0.62 to 0.55).

Twenty-five or more months post intervention

One study (910 participants) reported a follow-up time point of 30 months post intervention. 152 There was evidence that CBT prevented symptoms of depression at 30 months’ follow-up (SMD –0.27, 95% CrI –0.42 to –0.13).

Targeted population, secondary setting

Post intervention

The analysis-specific network diagram for targeted secondary settings is reported in Figure 12. Twenty-four studies (3669 participants) contributed to the analysis for the main time point of immediately post intervention. 114,141,160–163,165,167,170,173,214,215,217–219,221–224,226–230 Four studies were deemed to be at low risk of bias and seven studies were deemed to be at unclear risk of bias for both random sequence generation and allocation concealment. Eleven studies were judged to be at low risk of bias for random sequence generation but at unclear risk of bias for allocation concealment, and two studies were judged to be at unclear risk of bias for random sequence generation but at low risk of bias for allocation concealment. Eighteen studies included an intervention based on CBT, three studies included IPT-based interventions, one study evaluated a third-wave intervention, one included a CBM intervention and one evaluated an exercise intervention. Model fit and selection statistics were suggestive of slight lack of fit, but a consistency model was considered reasonable. Model fit was similar across all three component models (intervention, additive and full interaction). Reported results are from a random-effects, consistency intervention-level NMA. Full model fit details are provided in Appendix 3.

FIGURE 12.

Network plot for targeted population, secondary setting: post-intervention depression outcome.

Intervention-level effects

There was evidence of moderate to substantial between-study heterogeneity (τ 0.38, 95% CrI 0.25 to 0.58). Table 7 reports SMDs (95% CrIs) for each intervention relative to no intervention. There was no evidence of an effect for any intervention: CBT (SMD –0.22, 95% CrI –0.58 to 0.13); CBM (SMD –0.90, 95% CrI –2.20 to 0.40); IPT (SMD –0.65, 95% CrI –1.50 to 0.16); exercise (SMD –0.28, 95% CrI –1.13. to 0.58); and third wave (SMD –0.68, 95% CrI –1.83 to 0.47).

Component models: additive and full interaction

Regression coefficients are reported for both the additive and full interaction models in Appendix 3, Table 42. There was no evidence of effect modification according to components for CBT interventions in a targeted secondary setting.

Targeted, secondary, further time points: intervention-level effects only

Six to 12 months post intervention

Seventeen studies (2728 participants) contributed to the analysis for 6–12 months post intervention, of which 14 included a CBT-based intervention and three included an IPT intervention. 141,160,163,165,173,214,218,219,221–224,226–230 There was evidence of high levels of between-study heterogeneity (τ 0.44, 95% CrI 0.27 to 0.71). There was no evidence that either CBT (SMD –0.04, 95% CrI –0.51 to 0.41) or IPT (SMD –0.49, 95% CrI –1.49 to 0.48) prevented symptoms of depression at 6–12 months, relative to no intervention.

Thirteen to 24 months post intervention

Five studies (1089 participants) provided data for the NMA of 13–24 months’ follow-up. 166,218,223,224,229 Four evaluated a CBT intervention and one evaluated an IPT-based intervention. Between-study heterogeneity was high (τ 0.58, 95% CrI 0.12 to 3.08). There was no evidence that CBT (SMD –0.18, 95% CrI –2.56 to 2.16) or IPT (SMD 0.09, 95% CrI 3.81 to 3.93) reduced symptoms of depression at 13–24 months post intervention.

Twenty-five or more months post intervention

One study (260 participants), judged to be at unclear risk of bias, provided no evidence to suggest that CBT prevented symptoms at 48 months’ follow-up (SMD –0.27, 95% CrI –1.05 to 0.50). 166

Targeted population, primary setting

Post intervention

The analysis-specific network diagram for targeted primary settings is reported in Figure 13. Five studies (497 participants) contributed to the analysis for the post-intervention time point, of which four included an intervention based on CBT, and one examined an occupational therapy-based intervention. 175,179–181,232 One study was deemed to be at low risk of bias for both random sequence generation and allocation concealment, and three studies were deemed to be at unclear risk of bias. One study was deemed to be at low risk of bias for random sequence generation but at unclear risk of bias for allocation concealment. Model fit statistics are reported in Appendix 3. There were limited data available for the component-level models; however, model fit was similar across the three models. The between-study posterior median SD was lowest for the intervention-level model; it is on this basis that the intervention-level model is preferred here (see Appendix 3). All reported results are from a consistency random-effects NMA model unless otherwise stated.

FIGURE 13.

Network plot for targeted population, primary setting: post-intervention depression outcome. OT, occupational therapy.

Intervention-level effects

There was evidence of substantial between-study heterogeneity (τ 0.60, 95% CrI 0.08 to 3.80). Table 7 reports SMDs (95% CrIs) for each intervention relative to a waiting list. There was no evidence of an effect for CBT (SMD –0.48, 95% CrI –2.49 to 1.50) or occupational therapy (SMD –0.10, 95% CrI –2.94 to 2.71) at the immediate post-intervention time point.

Component models: additive and full interaction

For the additive component model, the between-study posterior median SD tended to the prior (τ 2.48, 95% CrI 0.12 to 4.87), as did the regression coefficient for occupational therapy (β –8.97, 95% CrI –144.7 to 144.5). As a result, we conclude that data were insufficient to estimate additive or full interaction effects for targeted primary settings.

Targeted, primary, further time points: intervention-level effects only

Two studies (230 participants) reported follow-up at 6–12 months, both of which included a CBT-based intervention. 175,232 A fixed-effects NMA was conducted. There was weak evidence that CBT (SMD –0.34, 95% CrI –0.72 to 0.05) prevented symptoms of depression at 6–12 months, relative to a waiting list. There was weak evidence from one study (83 participants), judged to be at unclear risk of bias, that CBT prevented symptoms of depression at 24 months’ follow-up, compared with a waiting list (SMD –0.50, 95% CrI –0.96 to 0.05). 232

Targeted population, tertiary/university setting

The analysis-specific network diagram is reported in Figure 14. Five studies (789 participants) contributed to the analysis for the main time point of post intervention, four of which included an intervention based on CBT and one of which included an intervention based on behavioural therapy. 183,184,186,187,235 Model fit and selection statistics suggested that a random-effects consistency model was reasonable (see Appendix 3); however, the between-study posterior median SD was indicative of substantial heterogeneity (τ 0.51, 95% CrI 0.12 to 2.50). This was considerably reduced in the unrelated intervention effects model (τ 0.26, 95% CrI 0.02 to 2.48), and may indicate the presence of inconsistency. Therefore, results for the tertiary setting are not reported. This possible inconsistency should also be interpreted in the light of the inclusion criteria adopted in this review, that interventions needed to be delivered in the educational institution itself, and the unanticipated limitations it caused for the tertiary/university setting analyses. We consider the limitations in Chapter 10.

FIGURE 14.

Network plot for targeted population, tertiary/university setting: post-intervention depression outcome.

Exploring heterogeneity and small-study effects

Subgroup analyses, metaregression and sensitivity analyses were conducted for the intervention-level NMA, for the main outcome and primary end point of post intervention only. Comparison-adjusted funnel plots did not provide evidence of small-study effects (see Appendix 6). However, the small number of studies available for the primary settings makes interpretation difficult.

Metaregression was conducted for intervention mode of delivery (face to face or via computer) and for intervention facilitator (teacher or a MHP). There was no evidence of effect modification by facilitator or mode of delivery for any population or setting combination (see Appendix 6).

Subgroup analyses were conducted to assess whether or not intervention effects differed by intended focus of the intervention, for example whether or not interventions addressing depression had a larger effect on anxiety outcomes than interventions intended to focus on anxiety but which also recorded depression outcomes. For each population and setting combination, intervention estimates were compared across three subgroups: (1) interventions that aimed to prevent anxiety (2) interventions that aimed to prevent depression only and (3) interventions that aimed to prevent both anxiety and depression. The results are reported in Appendix 6 but should be considered descriptive only. For the universal secondary network, there was some evidence that intervention focus was important. Interventions focused on preventing depression appear to have a larger effect on self-reported symptoms of depression than those focusing on anxiety or combined depression and anxiety. However, CrIs overlap, and we did not conduct a statistical test to examine subgroup differences.

Self-reported psychological well-being

Of the 137 studies focused on preventing anxiety, depression, or both depression and anxiety, 15 reported an outcome of self-reported psychological well-being, life satisfaction or quality of life. Five studies reported measuring well-being using the Warwick–Edinburgh Mental Wellbeing Scale,124,126,127,132,133 seven studies reported a measure of life satisfaction124,131,159,181,187,203,239 and two studies reported a quality-of-life measure. 129,235 One study reported a measure of social functioning206 and one used the Ryff Scales of Psychological Well-being. 143 Planned NMAs by population and setting were not possible; data are reported by outcome measure and study in Table 10. From their model-based analysis, Calear et al. 16 reported that there were no differences between interventions for school-led CBT and a waiting list (Cohen’s d –0.08, 95% CI –0.29 to 0.13) or between health service-led CBT and a waiting list (Cohen’s d –0.06, 95% CI –0.21 to 0.09). Using the data reported in the paper, we calculated the difference in mean change from baseline at post intervention, which suggested a small drop in well-being at post intervention. For each of the remaining 12 studies, summary intervention effects were compatible with both an increase and a decrease in well-being/life satisfaction.

Self-reported suicidal ideation, behaviour and self-harm

Author-defined suicidal ideation, behaviour or self-harm was referenced in 34 studies narratively or quantitatively. However, of these, the majority (n = 21) excluded participants reporting suicidal thoughts or behaviours and/or removed questions asking about suicide and self-harm from the baseline questionnaires. Some studies reported that the removal of questions was requested by participating schools or education authorities. Eleven studies reported that suicidal thoughts and behaviours were measured at baseline and that participants were referred to further services when necessary. However, nine did not then provide details on whether or not these students continued in the study, nor did they provide follow-up measures. The full details of these studies are reported in Appendix 7.

Seven studies reported participants experiencing suicidal thoughts or self-harm at post intervention. 123,136,142,162,221,235,240 Details are reported in Table 11. Three studies were conducted in a universal secondary setting and formed a connected network via attention control. However, NMA model fit was suggestive of inconsistency, and combined results are not reported. Two studies were conducted in a targeted secondary setting, one in an indicated tertiary setting and one in a universal primary setting. There was no evidence to suggest that educational setting-based interventions to prevent anxiety and/or depression had an impact on suicidal ideation or thoughts of self-harm for CYP.

Inequalities in health

None of the included studies reported the impact of the intervention on inequalities in health;241 therefore, we conducted post hoc subgroup analyses on the basis of available data. Data had been extracted on participant characteristics, including SES, sex and ethnicity (see Chapter 2). Descriptions of SES, sex and ethnicity, as defined by study authors, are reported in Appendix 7. Owing to insufficient data, subgroup analyses could not be conducted by sex or ethnicity for any population or setting. Subgroup analyses for studies conducted in lower socioeconomic settings (as described by the author) are reported.

Eleven studies reported being conducted in lower SES settings, of which three were conducted in MICs118,122,189 and eight in HICs. 139,140,152,156,157,196,208,210 Interventions evaluated in lower SES settings were CBT relative to usual curriculum, no intervention or a waiting list. Unfortunately, for the primary time point of post intervention, data were available for only seven studies conducted in HICs. 139,140,152,156,157,196,208 The results suggest that, in the case of interventions in secondary school settings, those delivered in lower SES settings may be less effective than those delivered in higher/mixed SES settings in reducing self-reported symptoms of anxiety. However, owing to the number of studies available, it was necessary to conduct fixed-effects analyses for the lower SES subgroup. This was not observed for self-reported depression, although it is of interest that the SMD for CBT compared with usual curriculum in higher/mixed SES settings was –0.07 (95% CrI –0.20 to 0.06); in lower SES settings, the SMD was 0.04 (95% CrI –0.06 to 0.15).

Table 12 reports the subgroup analysis findings. There was no evidence of a difference by SES for either self-reported depression or anxiety in primary educational settings. However, for self-reported anxiety, the SMD of CBT compared with usual curriculum in higher/mixed SES settings was 0.15 (95% CrI –0.37 to 0.02); in lower SES settings, the SMD was 0.05 (95% CrI –0.08 to 0.18).

Combined depression and anxiety scores and other ‘internalising’ outcomes

As described in Chapter 1, Changes to protocol, clarifications and additional analyses, and Table 1, a post hoc decision was made to assess composite ‘internalising’ outcomes, for example outcome scales that reported combined anxiety and depression scores (such as total Depression, Anxiety and Stress Scale or RCADS scores) or ‘internalising’ subscales of broader psychological functioning measures such as the SDQ. Six studies reported an internalising outcome, with useable data at post intervention, and are reported in Table 13 by population and setting. 119,123,163,171,216,219 Four studies in targeted secondary settings provided sufficient data for inclusion in a random-effects NMA. 163,171,216,219 Model fit statistics suggested that a consistency model was reasonable; however, the posterior median between-study SD was indicative of substantial heterogeneity (τ 0.97, 95% CrI 0.03 to 9.03) and tended to the prior distribution [uniform(0,10)]. Consequently, we report only study-specific SMDs in Table 13.

A fixed-effects NMA was possible for a further two studies142,206 reporting a 12-month follow-up in a universal secondary setting, as they formed a connected network via usual curriculum.. However, there was no evidence that interventions based on CBT (SMD 0.03, 95% CrI –0.14 to 0.20) or CBT + IPT (SMD 0.01, 95% CrI –0.16 to 0.17) were more effective than usual curriculum for improving composite internalising outcomes at a 12-month follow-up.

Acceptability and attendance

Intervention acceptability was defined by study authors. We anticipated considerable variability in the reporting of acceptability outcomes, but have combined them here under loose categories relating to (1) attitudes towards and satisfaction with intervention, (2) enjoyment and utility, and (3) attendance. Thirty-three studies reported a summary of acceptability of the intervention to study participants in terms of satisfaction or enjoyment and are summarised in Tables 14 and 15 by population and setting. Attendance data, as reported by authors, are reported in Appendix 7, Table 72.

Parent-reported child anxiety, depression or internalising outcomes

A post hoc decision was made to analyse parent-reported child anxiety, depression or internalising symptoms. For some of the studies, participants were considered too young for self-reported measures and parents were the primary respondents. In universal primary settings, five studies145,146,156,158,242 provided baseline and follow-up data on parent-reported child anxiety outcomes. From a random-effects NMA, there was no evidence that CBT-based interventions reduced parent-reported child anxiety symptoms relative to usual curriculum (SMD –0.10, 95% CrI –2.13 to 1.83).

Six studies115,116,176,177,181,182 provided both baseline and follow-up data to enable a NMA for targeted primary settings. There was no evidence that CBT-based interventions (SMD –0.43, 95% CrI –1.08 to 0.14), biofeedback (SMD –0.48, 95% CrI –1.76 to 0.72) or occupational therapy (SMD 0.05, 95% CrI –1.07 to 1.19) reduced parent-reported child anxiety symptoms relative to a waiting list.

Academic attainment

Seven studies reported academic attainment data. 131,136,137,145,159,168,174 Various measures of attainment were used; we report study-specific results in Table 16. Across all studies, there was no evidence of an effect of intervention on academic outcomes.

Problem behaviours

Few studies reported problem behaviours, and those that did reported substance use. The following results are reported as per the original publications. In a universal secondary setting, Stallard et al. 142 reported that there was evidence of a beneficial effect of CBT + IPT on cannabis use at 6 months (OR 0.56, 95% CI 0.38 to 0.82) and at 12 months (OR 0.70, 95% CI 0.48 to 0.93), but not on alcohol or ‘street drug’ use. In targeted secondary settings, Stice et al. 224 reported that CBT reduced ‘substance use’ (F[6,674] 3.60; p = 0.002) relative to control and Topper et al. 173 reported that there was ‘no significant difference’ between the intervention and control groups for binge drinking (Cohen’s d 0.22). In a tertiary/university setting, Reynolds et al. 233 reported results from the Alcohol Use Disorders Identification Test (AUDIT) and concluded that there was ‘no significant difference’ observed between the intervention and control groups for alcohol consumption or alcohol problems, although they also report that:

. . . a significant interaction of the orientation group effect with the linear effect of time was found, suggesting differential change for the [intervention] and [control] groups across the three time points. The overall percentage of youth above the clinical cut off for the BATD [behavioural activation treatment for depression] group generally decreased over time.

Mental health-related stigma

During our initial PPI work, reducing the stigma associated with mental health problems was identified as an important outcome for young people. Only one study reported a mental health-related stigma outcome. In Perry et al. ,136 stigma was measured using a nine-item subscale of the Depression Stigma Scale, an 18-item scale that assesses personal and perceived stigma towards depression, on which higher scores indicate greater stigma. At the immediate post-intervention time point, there was no evidence of a reduction in personal or perceived stigma in the SPARX (smart, positive, active, realistic, X-factor thoughts) computerised CBT group, compared with the attention control intervention (mean difference –0.50, 95% CI –2.90 to 1.90).

Studies included in the review

The overall PRISMA flow diagram for the project is reported in Figure 2. Twenty-seven papers relating to five studies met our eligibility criteria for inclusion in the conduct disorder review. Included studies were published between 1999 and 2018, and randomised between 225 and 891 participants (median 245 participants). All studies were cluster randomised, of which four reported results from appropriate models incorporating the multilevel nature of the data.

One study was classified as universal243 and four were classified as targeted244–247 (all indicated). All studies were conducted in primary school settings, with age at baseline ranging from 4.2 to 7.91 years. The majority of study participants were boys. In the four indicated studies, the proportion of boys ranged from 69% to 74% of participants. Four studies were conducted in HICs,243,244,246,247 and one was conducted in a MIC. 245 Of the studies conducted in HICs, all were conducted in lower-income settings, as specified by the trial authors.

Owing to intervention complexity and the flexible nature of the intervention implementation, we were not able to calculate an average ‘dose’, as was done for anxiety and depression outcomes.

Risk-of-bias assessments are reported in Figure 15 by individual study. Three studies were judged as having unclear risk of bias for random sequence generation and allocation concealment. 243,244,247 All studies were judged to be at high risk of bias for participant and outcome assessor blinding. Prospective trial registrations were available for one study. 245 For cluster randomised trials we also considered how recruitment, randomisation and analysis were conducted under the Cochrane Risk of Bias tool heading of ‘other bias’. Four studies were judged to be at high or unclear risk for ‘other bias’.

FIGURE 15.

Conduct disorder risk-of-bias assessments by domain and study. +, Low risk of bias; –, high risk of bias; ?, unclear risk of bias.

Owing to the diversity of interventions, outcome measures and time points reported by the studies, results are reported narratively by trial, and by date of publication. Owing to this diversity, intervention components are also described narratively, by study, where these could be identified from trial reports. Unless otherwise stated, statistical summaries are reported as described in the original publications.

Narrative review

Intervention costing

The NMA results (see Tables 3, 4, 7 and 8) suggested that CBT-based interventions were the most likely to produce positive outcomes, with CBT + IPT-based interventions also showing some indication of effect for preventing depression in a universal secondary-age population. Results from the component NMA indicated that the only component for which there was evidence of a beneficial effect in universal secondary CBT interventions was a psychoeducation component, with CBT interventions with psychoeducational components reducing the standardised anxiety score by –0.39 (95% CrI –0.78 to 0.01). Therefore, our intervention costing analysis (and subsequent component breakdown) focused on CBT interventions that incorporated identifiable psychoeducation components and CBT + IPT interventions. As there are considerable similarities between interventions designed to prevent depression and those focusing on anxiety (with some interventions explicitly targeting both), we included both depression and anxiety interventions in the costing estimate without distinguishing between them.

Studies describing universal secondary CBT interventions with a psychoeducation component were identified from the systematic review described in Chapter 2. Details of the interventions were extracted (i.e. year of publication, number of sessions offered, average session duration, size of group, number of group leaders, professional background of leaders, provision of a manual, training, materials, provision of parent sessions and other costs). Further details of the interventions were sought from intervention websites and from linked papers describing the branded interventions in more detail. Through scrutinising the extracted data, an ‘indicative’ intervention was developed. Unit costs for the individual elements for a universal intervention in a secondary population were identified in the UK context, and cost estimates were calculated for the ‘indicative’ intervention. The CBT cost estimates were extrapolated to primary populations and targeted interventions. A similar extraction process was carried out for CBT + IPT interventions.

Cost–consequence analysis

The intervention costs and consequences (SMD relative to usual curriculum post intervention) that could be ascribed to the ‘indicative’ interventions and psychoeducation components are presented as a cost–consequence table (Table 19). If no effectiveness estimates are available relative to usual curriculum, the results relative to a waiting list are reported; if these are also not available, then the results relative to no intervention are presented.

Narrative review of previous economic evaluations

Article selection

A flow diagram summarising the identification and selection of articles is given in Figure 16. A total of 434 titles and abstracts were screened for inclusion; full texts were obtained for 36 articles. Eight articles, reporting on six study time points, were deemed to meet the inclusion criteria.

FIGURE 16.

Study selection process: flow diagram for review of economic evaluations.

Only 3 of the 137 depression and anxiety effectiveness articles (2.2%) included in Chapters 4 and 5 were found to have published economic evaluations. Three further authors of effectiveness papers were contacted to request details of potential economic evaluations. All three authors confirmed that no economic study had been published, although one is still planning to complete and publish the study in future. Of the five included conduct disorder studies (see Chapter 7), only one (i.e. Fast Track247) had undergone economic evaluation. The additional scoping search carried out in MEDLINE to identify economic decision model papers returned 186 potentially relevant articles. The search identified both of the modelling papers that had already been identified in the original review,259,260 but did not identify any additional articles. Based on this finding, we did not conduct further searches for economic decision model papers in the other databases searched in the systematic review. Although there is a possibility that we may have missed some decision model papers, we think that this is unlikely.

Intervention costs

Extracted elements of the interventions described in 20 papers evaluating universal CBT interventions with a psychoeducation component identified via the ICA (see Chapters 4 and 5) are given in Appendix 8. The number of sessions for children ranged from 3 to 30, with a median of 10 sessions. Session duration ranged from 35 to 120 minutes (median 50 minutes). Group sizes ranged from 4 to 30 participants, with a median of 25. Parent sessions formed part of the intervention in only five studies, with low attendance cited as an issue in two of these studies. The group leaders included teachers (n = 8), psychologists (n = 11) and students (n = 5) (sometimes in combination). Groups were led by either one (n = 12) or two (n = 7) individuals (the exact leadership was unclear in one study). Of the 12 groups led by one individual, four were led by psychologists and eight by teachers. Training for delivering the intervention (when described) took a number of forms, including self-participation in the intervention, workshops varying from 90 minutes to 5 days (n = 9, median 1 day) and co-leading a cohort. A facilitator manual was provided in 16 out of 20 interventions, and workbooks or worksheets were supplied in 13 out of 20 interventions.

A ‘typical’ universal secondary CBT intervention might, therefore, comprise 10 sessions of 50 minutes, each session each delivered to 25 children at a time by one teacher who had received 1 day of specific intervention training. Typically a manual would be provided for facilitators, and workbooks for the participants. Unit costs associated with these elements in the UK context are given in Appendix 8. This leads to an estimated overall intervention cost of £43 per student from the school budget perspective. At the school level, this would represent approximately £1825 per (small) two-form entry school, or £7300 per larger eight-form entry school in the first year. A two-form entry school is one that has an annual intake of two classes of approximately 30 students each, who then progress through the school as a stable cohort. An eight-form entry school is approximately four times as large. If this universal intervention were delivered in a primary school, the teacher costs would be slightly lower, leading to an approximate cost of £42 per student. Assuming that a targeted (indicated or selective) intervention would be delivered to a smaller group of students (e.g. 10), the cost per student would be £95 in a secondary setting or £91 in a primary setting. Similarly, based on three articles describing universal secondary CBT + IPT interventions, a typical CBT + IPT intervention might have both workbooks and a manual, and consist of 11 sessions of 60 minutes each, delivered by one teacher to 10 students after 2 days’ training. This leads to an estimated cost of £157 per student.

Cost–consequence analysis

Intervention costs and consequences in SMDs that could be ascribed to the ‘indicative’ CBT interventions for a universal secondary population are presented in Table 19, alongside extrapolated intervention costs for similar interventions in primary and targeted populations. Adding IPT to CBT in a universal secondary population is more costly than CBT alone, driven mainly by additional teacher training and delivery to smaller groups of students. Although the estimated SMD shows a bigger reduction in depression score, compared with usual curriculum, than for CBT alone, the estimate is very uncertain. There is no evidence for the effect of CBT + IPT on anxiety in the universal secondary population. Further evidence on the relative efficacy of CBT and CBT + IPT would be required to justify the additional intervention cost. No studies used a usual curriculum control in either secondary or primary targeted populations. There is some evidence that CBT is effective compared with a waiting list for reducing anxiety in a targeted primary population, but this difference may not extend to comparisons with usual curriculum.

In Table 20, we report the intervention costs and consequences (SMD) for CBT with or without a psychoeducation component in a universal secondary population. Because all of the CBT interventions in the NMA for this population contained cognitive and behavioural components, the comparison is between CBT with just cognitive and behavioural components and CBT with cognitive, behavioural and psychoeducation components. We give results for both anxiety and depression outcomes, although the NMA results only show evidence of a difference for the anxiety outcome. We also report estimated intervention costs, based on our microcosting of typical interventions with these components as described in the RCTs. Adding a psychoeducation component increases intervention costs, but there is evidence that anxiety symptom scores improve post intervention.

Review of previous economic studies

The body of evidence described in this review is both small and heterogeneous. Conditions being targeted, interventions and cost-effectiveness analyses were all variable. It was, therefore, not possible to synthesise the results numerically. Multiple perspectives for the analyses were taken, leading to the resources included in the analyses varying widely. The potential long-term costs associated with missing school15,16 (such as lifetime earning capacity) were not estimated in any of the studies. All included studies appropriately compared the intervention with usual provision, or with no intervention (i.e. the model-based studies). Although an active control comparator is appropriate (and necessary) for investigating effectiveness mechanisms, economic analyses are most informative for decision-makers when compared with usual care/curriculum. The two model-based analyses suggested that hypothetical interventions could be cost-effective in the Australian context. However, the trial-based analyses suggested that the interventions were unlikely to be cost-effective in the UK or the USA. For the one intervention that was considered potentially cost-effective in a high-risk subgroup at an interim point of the study, the effect disappeared with longer-term follow-up. There was little empirical costing evidence to inform decisions on the implementation of preventative interventions.

Cost–consequences analysis

The figures used to obtain intervention costs were based on a highly stylised, ‘indicative’ universal intervention that was assumed to be delivered as specified, and there are many ways in which differences might occur in the real world. Although it was assumed that there would be no room hire costs if the intervention were delivered on school premises, there could be heating, lighting or security overheads if the intervention were delivered outside school hours. It is possible that a project manager or administrator would be required to oversee the intervention if it were rolled out across the country. The salary of the teacher delivering the intervention could vary according to experience. Delivery by a psychologist instead of a teacher would affect the salary costs, and psychologists additionally undergo supervision meetings as part of their professional conduct. The intervention would be more costly if two individuals led it instead of one. The detailed intervention cost breakdown by Stallard et al. 262 illustrated the variability in individual components that go into making up an intervention, with small changes in delivery method (e.g. the requirement for travel) leading to big differences in costs between methods. Stallard et al. ’s262 work suggested that it is a fallacy to assume that delivery by schoolteachers is less costly than delivery by health-care specialists. None of the interventions described here explicitly mentioned a licence fee, and potential administration costs have not been considered. Deriving a value for the opportunity cost of time diverted from other learning activities in school is currently considered problematic and has not been taken into account here. Despite these caveats, this simplistic model serves to give an idea of the costs that might accrue to a school budget in the first year of implementation. Subsequent years would incur lower training outlay, assuming that the same teachers deliver the course. The limited details of interventions described in the published reports meant that it was challenging to assign accurate proportions to psychoeducation components.

Implementation might be more attractive if it could be demonstrated that the intervention has a positive effect on educational outcomes or reduces the need to pay for educational psychologists in the future. However, the systematic review found that few studies of effectiveness had considered educational attainment outcomes (see Chapter 6). Future work should consider developing methodologies for evaluating both effectiveness and cost-effectiveness in terms of school outcomes, as well as the outcomes more typically encountered in health-care studies.

The cost–consequence analysis was based only on the intervention costs (i.e. the effect on health-care use was not addressed). It is worth noting that the beneficial effects of an intervention that aims to prevent or reduce symptoms of anxiety may also extend to reducing depression, and vice versa, which would increase the value of such an intervention. A comparison of costs and standardised outcomes in a cost–consequence analysis is difficult to interpret; ideally, a full cost-effectiveness analysis based on a recognised clinical outcome would be conducted. However, we did not find sufficient evidence for the efficacy of these interventions at 6 months’ post-intervention (and longer) follow-up for the development of a cost-effectiveness model to be useful at the present time. An economic model needs to capture long-term costs and benefits, which may be substantial, as issues with mental health and conduct disorder during school years have been shown to be associated with a range of health and behavioural problems as adults,4 which are costly to society. 265 However, further research is needed to develop and evaluate effective interventions for the prevention of mental health and conduct disorders in school-age children before the longer-term consequences of such interventions can be fully assessed.

To ensure high-quality information for decision-makers, it is imperative that future reports of school-based interventions to prevent anxiety, depression or conduct disorder are described in some detail and that the cost implications of interventions are adequately measured.