Classroom-based cognitive behaviour therapy (FRIENDS): a cluster randomised controlled trial to Prevent Anxiety in Children through Education in Schools (PACES)

Summary Background Anxiety in children is common, impairs everyday functioning, and increases the risk of severe mental health disorders in adulthood. We investigated the eﬀ ect of a classroom-based cognitive behaviour therapy prevention programme (FRIENDS) on anxiety symptoms in children. Methods Preventing Anxiety in Children though Education in Schools (PACES) is a three-group parallel cluster randomised controlled trial. Interventions were given between September, 2011, and July, 2012, with schools as the unit of allocation and individual participants as the unit of analysis. We enrolled state-funded junior schools in southwest England. We sent information to all eligible schools (state-funded junior schools in southwest England) inviting them to enrol in the study. School year groups were assigned by computer-generated randomisation (1:1:1) to receive either school-led FRIENDS (led by teacher or school staﬀ member), health-led FRIENDS (led by two trained health facilitators), or usual school provision. Children were not masked to treatment allocation. The allocated programme was given to all students (aged 9–10 years) in the school year (ie, universal delivery) as part of the school curriculum as nine, 60 min weekly sessions. Outcomes were collected by self-completed questionnaire administered by researchers masked to allocation. Primary outcome was symptoms of anxiety and low mood at 12 months assessed by the Revised Child Anxiety and Depression Scale (RCADS 30). Analyses were intention to treat and accounted for the clustered nature of the design. The study is registered, number ISRCTN23563048. Findings 45 schools were enrolled: 14 (n=497 children) were randomly assigned to school-led FRIENDS, 14 (n=509) to health-led FRIENDS, and 12 (n=442) to usual school provision. 1257 (92%) children completed 12 month assessments (449 in health-led FRIENDS, 436 in school-led FRIENDS, and 372 in usual school provision). We recorded a diﬀ erence at 12 months in adjusted mean child-reported RCADS scores for health-led versus school-led FRIENDS (19·49 [SD 14·81] vs 22·86 [15·24]; adjusted diﬀ erence –3·91, 95% CI –6·48 to –1·35; p=0·0004) and health-led FRIENDS versus usual school provision (19·49 [14·81] vs 22·48 [15·74]; –2·66, –5·22 to –0·09; p=0·043). We noted no diﬀ erences in parent or teacher ratings. Training teachers to deliver mental health programmes was not as eﬀ ective as delivery by health professionals.


Introduction
Anxiety disorders aff ect 10% of children by the age of 16 years. 1 They signifi cantly impair everyday functioning, often persist into adulthood, and increase the risk of other psychiatric disorders in adolescence and young adult hood. [2][3][4][5] The associated health-related burden, and economic and societal costs are large, and the need to improve the mental health of children is being increasingly recognised as a global priority. [6][7][8] Eff ective psychological interventions, especially cognitive behaviour therapy (CBT), are available for children with anxiety disorders. 9,10 However, comparatively few children with anxiety disorders are identifi ed and referred for treatment. 11,12 The poor reach and availability of traditional treatment services has led to interest in more proactive preventive approaches with schools off ering a convenient and natural location to deliver such programmes. 13,14 Findings of systematic reviews show that universal and targeted anxiety prevention programmes are often based on cognitive behaviour therapy. 15 Data from reviews 15,16 suggest that cognitive behaviour therapy prevention programmes can be eff ective, although research methods are poor, adequately powered implementation trials are scarce, results are inconsistent, and eff ect sizes vary greatly. Most studies report overall changes in symptoms, and the preventive benefi ts for less symptomatic participants have seldom been reported.
Prevention programmes can be universally provided to all of an identifi ed population, or targeted towards those at risk of developing a disorder or showing early signs of a disorder, or a combination of both approaches. 17 Universal programmes have good reach, avoid the need for screening, are less stigmatising, and off er the potential to enhance mental health and reduce present symptoms.
Targeted programmes focus scarce resources on individuals with greatest needs, and usually achieve larger treatment eff ects. 18,19 The eff ect of the intervention leader (health vs school professional) has important implications for the method of delivery and sustainability of an intervention but has been directly investigated in only one study. 20 Barrett and Turner noted that a universal anxiety prevention programme (FRIENDS; panel 1) was equally eff ective in the reduction of symptoms of anxiety in children aged 10-12 years when given by a psychologist or teacher. However, systematic reviews have reached diff erent conclusions about who is most eff ective at delivering these programmes. 15,16 Before anxiety prevention programmes can be endorsed and widely provided, independent implementation trials are needed to measure eff ectiveness and cost-eff ectiveness when provided under real-life conditions and to establish the eff ect of the intervention leader on outcome.
We undertook a pragmatic assessment of the eff ectiveness of a classroom-based anxiety prevention programme (FRIENDS 21 ) universally delivered by health and school professionals to school years 4 and 5 (children aged 9-10 years) in UK junior schools.

Study design and participants
We did this three-group parallel cluster randomised controlled trial between September, 2011, and July, 2012, with school as the unit of allocation and individual participants as the unit of analysis. 22 A project information sheet and trial enrolment form was sent to all primary schools in Bath and northeast Somerset, Swindon Borough, and Wiltshire within a 50 mile radius of the University of Bath, UK (n=268). Eligible schools were state-funded junior schools in three Local Education Authorities in southwest England. Such junior schools are mainstream government-funded schools that are attended by 94·5% of children aged 5-10 years in the UK. All children aged 9-10 years (years 4 and 5) in participating schools were eligible, unless they were not attending school (eg, because of long-term sickness or excluded from school) or did not participate in Personal Social and Health Education (PSHE) lessons for religious or other reasons. The allocated intervention was given to all participants in the school year (ie, universal delivery) as part of the PSHE curriculum. The trial protocol is online.
Participation required written consent from the school head teacher, parents' not opting their child out of the study, and signed assent from the child. The study was approved by the University of Bath, Department for Health Research Ethics Committee.

Randomisation and masking
Once all schools had been enrolled, we randomly assigned year groups (1:1:1) to school-led FRIENDS, health-led FRIENDS, or usual school provision. Randomisation was undertaken at school and not the class level to avoid possible contamination within schools. Trial groups were balanced with respect to key characteristics by calculating an imbalance statistic for a large random sample of possible allocation sequences. 23 Children were not masked to treatment allocation. Outcomes were collected by selfcompleted questionnaire administered by researchers masked to allocation. Group allocation was kept in a separate password-protected database. Researchers who analysed data were also masked to allocation-trial groups were numerically coded and data analysis undertaken masked to which code related to each trial group. The variables used for balancing were school size, number of students and classes, number of mixed classes, level of educational attainment, and preferred timetabling. A statistician with no other involvement in  the study randomly selected one sequence from a subset with the most desirable balance properties.

Procedures
Interventions were delivered in the academic year September, 2011, to July, 2012. The anxiety prevention programme we assessed (FRIENDS) is a manualised cognitive behaviour therapy intervention that has been identifi ed as eff ective. 16,24 The programme was developed to be delivered in schools and provided to whole classes of children. FRIENDS is based on the principles of CBT and develops skills to counter the cognitive, emotional, and behavioural aspects of anxiety. Children develop emotional awareness and regulation skills, to identify and replace cognitions that increase anxiety with more balanced and functional ways of thinking and to develop problem-solving skills to confront and cope with situations and events that provoke anxiety. The inter vention trialled in this study consisted of nine, 60 min weekly sessions delivered to whole classes of children. Children had their own workbook and group leaders had a detailed session plan that specifi ed key learning points, objectives, and core activities for each session. The feasibility and viability of delivering FRIENDS in UK schools has previously been established. 25,26 In the health-led FRIENDS programme, each session was led by two trained health facilitators working alongside the class teacher. All facilitators had at least an undergraduate university degree in a relevant discipline (eg, social science or education), appropriate professional backgrounds (eg, psychology, nursing, or education), or experience of working with children or young people. Initial 2 day training and supervision every 2 weeks were provided by accredited FRIENDS trainers. Supervision was in a group format and consisted of review of session plans, the underlying cognitive model, class and behaviour management skills, and any interpersonal diffi culties or communication problems with the class teacher.
In the school-led FRIENDS programme, sessions were led by a teacher or member of the school staff (eg, teaching assistant) who were trained in delivery of the programme and were supported by two facilitators. School staff attended the same 2 day initial training and were off ered ongoing supervision. We assessed treatment fi delity by randomly assessing audiotape recordings of 10% of FRIENDS sessions.
In the usual school provision programme, children participated in the usual PSHE sessions provided by the school. All schools were following a UK National Curriculum programme designed to develop selfawareness, management of feelings, motivation, empathy, and social skills. 27 The sessions were planned and provided solely by the teacher and did not include any external input from the research team.

Outcomes
Child outcomes were collected during class time with selfcompleted questionnaires administered by researchers at baseline, 6 months, and 12 months. As described in the trial protocol, 20 the primary outcome was symptoms of anxiety and low mood 12 months after baseline as established by the Revised Child Anxiety and Depression scale (RCADS 30). 28,29 Secondary outcomes assessed worry (with the Penn State Worry Questionnaire for Children), 30 self-worth and acceptance (with the Rosenberg Self-Esteem Scale), 31 extent of bullying (with the Olweus Bully/Victim Questionnaire), and life satisfaction (with a subjective wellbeing assessment).    32 and the parent version of the Revised Child Anxiety and Depression Scale (RCADS-30-P) 33 at baseline, 6 months, and 12 months.

Health-led FRIENDS (N=489)
Class teachers completed the impact rating of the Strengths and Diffi culties Questionnaire (SDQ) for all children in their class at all three assessment points to assess the presence of an emotional or behavioural problem, chronicity, distress, social impairment, and burden.

Statistical analysis
We powered the study to detect a diff erence of 3·6 points in mean RCADS total scores between FRIENDS and usual PSHE. On the basis of an SD of 12 points and an intracluster correlation coeffi cient of 0·02, 28 pupils per class, 90% consent and 80% retention, eff ect sizes between 0·28 and 0·30, SDs are detectable with 80% power and 5% two-sided α with 1134-1360 assenting pupils.
We used descriptive statistics to assess balance between the trial groups at baseline. The primary outcome was assessed by intention to treat without imputation. To take Separation anxiety Generalised anxiety Data are mean (SD) unless otherwise stated· PSHE=personal social and health education· RCADS=Revised Child Anxiety and Depression Scale· OCD=obsessive compulsive disorder· *RCADS adjusted for sex, school, and baseline RCADS.

Table 2: Analysis of primary outcome (child-completed RCADS) for all children at 12 months for health-led FRIENDS and school-led FRIENDS vs PSHE
appropriate account of the hierarchical nature of the data, we used multivariable mixed eff ects models to compare mean RCADS at 12 months for health-led FRIENDS with school-led FRIENDS and usual school provision, with adjustment for baseline RCADS, sex, and school eff ects. We repeated these analyses for secondary outcomes. For RCADS we undertook a further planned analysis. We used repeated-measures mixed-eff ects analysis of variance models to investigate convergence and divergence between trial groups over time. We did preplanned subgroup analyses with interaction terms in the regression models between the randomised group and the baseline variable (low anxiety RCADS 0-48, high anxiety ≥49). We did sensitivity analyses to assess the potential eff ect of missing data. Completion rates for all groups at 12 months were high (91·8-92·7%), although non-completers tended to be more symptomatic on our primary outcome measure (RCADS) at baseline (data not shown). With multiple imputation methods, we created 20 datasets and showed that imputation for missing data made no material diff erence to the overall results. 34 Therefore, the data we present are based on recorded data only.
The study is registered, number ISRCTN23563048.

Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, interpretation of data, or writing of the report. The corresponding author had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication.

Results
Between September, 2011, and July, 2012, 45 schools were enrolled. 41 consented to participate and were randomly assigned: 14 (n=497 children) to school-led FRIENDS, 14 (n=509) to health-led FRIENDS, and 12 (n=442) to usual school provision. One school from usual school provision withdrew before baseline assessments were undertaken (fi gure). The remaining 40 schools were representative of the UK in terms of academic attainment according to Department of Education performance tables (ie, the percentage of children achieving key stage 2 level 4 in maths and English; data not shown). However, more children had special educational needs (23·2% vs 17·1%), pupil absence rates were lower (4·4% vs 5·1%), and eligibility for free school meals was lower (12·4% vs 18·2%) in the cohort than the national average. Of the 1448 eligible participants, 1362 (94%) consented to participate in the study, of whom 1339 (98%) completed baseline assessments. The proportion of boys in the usual school provision group (42%) was lower than in each of the other two trial groups, but otherwise the groups were well balanced (table 1).
All nine FRIENDS sessions were delivered to classes assigned to both the health-led and school-led conditions. To assess intervention fi delity, we recorded and independently rated 49 sessions (one from each class in the 28 schools delivering FRIENDS). All specifi ed core tasks and home activities were delivered in the 24 health-led sessions. In the school-led sessions, 15 of 25 (60%) delivered all core tasks and the home activity, eight (32%) delivered all except the home activity, and two (8%) did not deliver one core task and the home activity. Session attendance was not recorded although average school absence rates were very low (4·2% in the health-led group and 4·4% in the school-led group).
Analysis of other secondary outcomes and parent and teacher completed measures identifi ed no diff erences between treatment groups at 12 months (table 3).
As specifi ed in the protocol, we did separate subgroup analysis in children with the highest 10% of baseline RCADS scores (high anxiety ≥49) and the remaining 90% (low anxiety ≤48; table 4). 22 We recorded signifi cant within-group reductions for the high-risk group at 12 months but no eff ects between groups. For the low-risk group, we noted between-group diff erences in mean RCADS at 12 months (table 4). Adjusted mean diff erences showed an eff ect for healthled FRIENDS versus school-led FRIENDS (adjusted diff erence -3·78, 95% CI -6·16 to -1·40; p=0·003) and health-led FRIENDS versus usual school provision (-3·13, -5·61 to -0·65; p=0·015). This eff ect relates to a reduction in the health-led FRIENDS group on the social and generalised anxiety subscales (table 2).  In the low anxiety group, the standardised eff ect size of health-led FRIENDS compared with usual school provision (Cohen's d=0·22, 95% CI 0·38-0·07) and school-led FRIENDS (0·25, 0·40-0·11) was small. The economic evaluation will be published separately.

Discussion
This is the fi rst large pragmatic randomised trial comparing a universally provided classroom-based cognitive behaviour therapy anxiety prevention pro gramme led by health and school staff with usual school provision. When we transferred health-led FRIENDS to everyday settings, the programme was more eff ective in the reduction of childreported symptoms of anxiety than was school-led FRIENDS or usual school provision (panel 2). Although intervention leaders received the same initial training, our data suggest that a manualised programme might result in diff erent outcomes depending on who delivers it. Although training teachers to deliver mental health programmes off ers a potentially convenient low-cost sustainable option, our results show that this approach is not as eff ective as delivery by health professionals.
Further exploration identifi ed two potentially important diff erences between health and school leaders. First, although treatment fi delity was high, the home assignment was not undertaken in 40% of the school-led sessions that were assessed. Continuing practice of newly acquired skills is an integral part of cognitive behaviour therapy programmes and the absence of such practice might have compromised eff ectiveness. However, this possibility needs to be balanced against the increased opportunities for class teachers to integrate and consolidate programme skills more intensively through their daily contact with children. Second, although both groups had the same initial training, comparatively few teachers attended continuing supervision. Although teachers will be competent in some of the areas addressed in supervision (eg, classroom management and engaging children of diff ering abilities), they will be less familiar with the underlying cognitive model. Therefore, scarce opportunities to consolidate the theoretical model might have compromised programme delivery.
Children with high and low anxiety showed signifi cant reductions in anxiety symptoms over time. The number of children with high anxiety was small and our study was not powered to compare between-group diff erences within this subgroup. Therefore, the absence of betweengroup diff erences might be a power issue or might show regression to the mean. This issue needs to be clarifi ed in further suitably powered studies of highly anxious children. However, our cohort provided suffi cient power to detect diff erences within the low anxiety group in which we found a marked reduction between groups in favour of the health-led FRIENDS group. Universally delivered prevention programmes off er the potential to reduce present symptoms, enhance emotional wellbeing, and potentially shift population means over time. Our fi ndings support this theory and show that children in the low-anxiety health-led FRIENDS group showed markedly lower anxiety symptoms at 12 months than did those in the usual school provision group. Because fears, anxiety, and stress are common in children, anxiety prevention programmes might be especially suited to universal delivery. 16 Further research is needed to establish whether improvements are sustained in the medium term, and to detail the cost-eff ectiveness of such universal interventions delivered by health professionals.
We recorded no diff erences between groups for any secondary outcomes, suggesting that the intervention eff ects were specifi c to anxiety. The FRIENDS programme is designed as an anxiety prevention programme and specifi cally develops skills known to reduce anxiety. Therefore, the specifi c eff ect that we identifi ed is consistent with the underlying theoretical model and programme focus, although in view of the comorbidity between anxiety and depressive disorders and the shared elements of many cognitive behaviour programmes, this result is disappointing. 35 Our fi ndings show that although anxiety prevention programmes can have a positive eff ect on anxiety symptoms, they cannot be assumed to enhance the general emotional wellbeing of children.

Systematic review
Findings of systematic reviews of school-based anxiety prevention programmes show that universally (ie, whole class) provided programmes based on cognitive behavioural therapy can be eff ective. 15,16,24 However, research methods are poor, adequately powered implementation trials are scarce, results are inconsistent, few medium term follow-ups exist, and direct comparison of school versus health-led delivery have seldom been undertaken.
We aimed to undertake a suitably powered implementation trial of a school-based cognitive behaviour therapy anxiety prevention programme (FRIENDS) for children aged 9-10 years. Programme delivery by health and school professionals was compared with treatment as usual at 12 months.

Interpretation
Our results are consistent with those from systematic reviews that show that school-based anxiety prevention programmes informed by cognitive behaviour therapy are eff ective in the reduction of anxiety in children. The programme leader is important because the programme we assessed was only eff ective when delivered by health staff . Our fi nding that children with low symptoms also benefi ted from the programme supports the use of universal approaches. This study supports the implementation of a universal anxiety prevention programme delivered by health professionals in regular schools in everyday settings.
The eff ect was also specifi c to child report because although parent and teacher ratings reduced over time, these changes were not signifi cant. This fi nding might suggest that although children are feeling better, adults have not noticed any improvement in functioning or reduction in distress. However, only 42% of parents returned baseline questionnaires and as such these fi ndings might not be representative of the total cohort. Similarly, the teacher assessment was global and might be insensitive to specifi c changes in anxiety symptoms. Alternatively, the absence of a parent and teacher eff ect might show the diffi culty of assessment of changes in internal emotional symptoms and cognitions that are not directly observable.
The results of our study are consistent with those of reviews in which the intervention leader aff ected programme eff ectiveness, 16 and with other implementation trials in which teacher-led FRIENDS was not eff ective. 36,37 Our fi ndings diff er to those of the study by Barrett and Turner,20 which directly compared teacher and psychologist delivery of FRIENDS, although their study lacked statistical power. Overall, our study supports the growing evidence base for the use of the FRIENDS programme as an eff ective school-based anxiety prevention programme. 16,24 The intervention that we assessed was based on cognitive behaviour therapy, which is typically used to treat mental health disorders. However, for schoolbased universal emotional health programmes, most children will be healthy and will not need treatment. Furthermore, provision of individual treatment in a classroom context would not be appropriate. Therefore, it is important to emphasise that we did not provide treatment but used the cognitive behaviour therapy framework to help children to develop emotional, cognitive, and behavioural skills. Provision of life skills development within the school context fi ts with the growing reognition that schools are not just concerned with the development of academic skills, but also have an important role in the enhancement of emotional development in children.
Our study has many strengths. We used a manualised anxiety prevention programme that has been shown to be eff ective in schools and has been piloted in UK schools. 16,25,26 Recruitment and retention were high, absenteeism was low, and programme fi delity was good. In terms of limitations, we relied on self-report measures and did not undertake any diagnostic interviews. Although anxiety symptoms were reduced in the health-led FRIENDS group, whether this reduction was indicative of changes in diagnostic status or impairment is unclear. Second, although our regional cohort included a representative sample of UK schools on several key dimensions, our group was less socially disadvantaged and had more white British participants than did the average UK state school. Therefore, whether similar results would be obtained with a more disadvantaged or ethnically diverse population is unclear. Similarly, although our schools were well matched across trial groups and our exploratory qualitative analysis identifi ed few diff erences in school culture and ethos, these diff erences could have aff ected outcomes. Third, although the health-led and school-led FRIENDS facilitators had the same initial training and treatment fi delity was good, we did not directly assess how the intervention was delivered. Therefore, the diff erences between these groups might be indicative of diff erences in leader enthusiasm, confi dence, and ability to engage and motivate students, which are skills that could be developed in school staff with additional training and supervision. 16 Finally, although we reported symptoms at 12 months, we were unable to estabish whether the improvements reported would be sustained over time. This point is especially important for prevention programmes in which the full extent of the preventive eff ects might take several years to emerge.
In summary, our results are encouraging and show that anxiety prevention programmes delivered in schools to children aged 9-10 years do reduce anxiety symptoms at 12 months. The fi nding that children with low symptoms benefi ted from the programme supports a universal approach, which also fi ts well with school timetables and organisational structures. 18 However, our data suggest that the same programme can result in diff erent eff ects depending on who delivers it. Further research is needed to explore the longer term eff ect, and the potential mediators and moderators of anxiety prevention programmes, to assess cost-eff ectiveness, and to establish whether anxiety prevention programmes are eff ective with more diverse and disadvantaged groups.

Contributors
PS, GT, RP, HD, RA, and NS conceived and designed the study. ES managed the trial and supervised data collection. GT undertook the statistical analysis. All authors had access to all study data and participated in interpretation of the fi ndings, contributed core ideas and were involved in critically revising the paper for important intellectual content. All authors read and approved the fi nal manuscript. PS was principal investigator and will act as guarantor for the paper.

Declaration of interests
We declare no competing interests.