Trial design and methods

Aims and objectives

The study had six research objectives, underpinned by seven research questions.

Objective 1: optimise the intervention to increase activity in boys and girls

1. How can the intervention materials be optimised to increase activity in boys and girls?

Objective 2: identify effective means of recruiting less active children

2. How can recruitment be optimised to recruit less active children?

1. Is it feasible to collect self-reported activity data from Year 4/5 pupils using ‘opt-out’ consent?
2. Is it feasible to collect accelerometer data from all Year 4/5 pupils using an ‘opt-out’ consent procedure at baseline and follow-up in a subgroup of pupils in four schools?
3. Does an ‘enhanced recruitment approach’ facilitate recruiting less active children?
4. Is it feasible and acceptable for schools and pupils to provide opportunities for enrolment at the mid-point of the intervention?

Objective 3: assess intervention fidelity

3. To what extent was Action 3:30 delivered as intended?

1. What proportion of the Action 3:30 sessions were delivered?
2. What was the mean attendance and were there differences by sex?
3. To what extent was the intervention delivery consistent with the underpinning theory?

Objective 4: estimate the effect of allocation to the Action 3:30 intervention on weekday moderate to vigorous physical activity of participants and related physical activity behaviours

4. What is the difference in mean accelerometer-assessed MVPA of boys and girls in the intervention and control arms in the last few weeks of the intervention when the programme was still running (follow-up)?

1. Is there any evidence that intervention participants’ school travel mode or organised club attendance changes as a result of attending Action 3:30?

Objective 5: collect the information needed to assess the feasibility of conducting a definitive trial and assess the implementation potential of the Action 3:30 intervention

5. How many data can be provided for secondary outcomes?

1. Is it feasible to collect information on how costs of implementation are influenced by school infrastructure, staff leadership/participation and overall extracurricular club provision/cost in schools?
2. Is there any evidence of contamination between intervention and control schools?
3. What are the factors that need to be addressed to minimise health inequalities?

6. What would the sample size for a definitive trial be?

7. Is it feasible to collect all data on all dimensions of the RE-AIM framework to estimate the potential population impact of Action 3:30 in a definitive trial?

Objective 6: assess whether or not the five progression criteria for conducting a definitive trial are met

Progression criteria

1. At least 25% of schools that are approached agree to join the study.
2. At least 25% of eligible Year 4/5 pupils express an interest in the study by returning consent forms.
3. At least 40% of participants expressing an interest in the study are girls.
4. At least 50% of the participants in the intervention arm attend 50% of the sessions.
5. At follow-up, at least a small benefit for weekday MVPA is observed for each of boys and girls, comparing intervention schools with control schools, and the upper bound of the 95% CI exceeds 10 minutes.

Research design

The study had two components. Component A, intervention optimisation, is described in Chapter 3. Component B, detailed in this chapter, was a cluster randomised controlled feasibility study in primary schools to compare the Action 3:30 intervention with a usual-practice control. The trial included quantitative, qualitative, process and economic evaluations.

Protocol amendments

The original study protocol (version 1.0) was submitted to NIHR on 3 March 2017. Version 1.1 was submitted on 21 September 2017 with one change, and version 1.2 was submitted on 19 April 2018 with one further change. These revisions are detailed in Table 1.

Ethics

The study was granted ethics approval from the School for Policy Studies, University of Bristol. The trial was registered with Internal Standard Randomised Controlled Trial Register under the reference number ISRCTN34001941. The project was funded by the NIHR Public Health Research (PHR) programme (project number 15/55/09). It was agreed that any adverse events during data collection would be recorded and reported to the chairperson of the Ethics Committee and the chairperson of the Trial Steering Committee (TSC), but there were no such events.

Study population and recruitment

The study sought to recruit 12 primary schools in two local authorities (South Gloucestershire and North Somerset). The number of schools approached and the proportion recruited were recorded, as was the percentage of their pupils claiming free school meals (an indicator of socioeconomic status of the pupils’ families, i.e. higher is more deprived). To participate, schools needed to be willing to allocate space for two after-school sessions per week for 15 weeks between November 2017 and March 2018. To ensure that the sample represented local diversity, half of the schools recruited were above the local authority median for free school meals. Eight schools were recruited in South Gloucestershire and four were recruited in North Somerset (as the latter is a smaller local authority).

The study aimed to recruit 30 Year 4 and 5 pupils (aged 8–10 years) in each school, and 40% of the sample was to comprise girls. The pupils were recruited when they were in Years 3 and 4 (aged 7–9 years) during the summer term ahead of the intervention starting the following autumn. The only exclusion criterion was that pupils who were unable to participate in standard PE lessons could not take part. If more than 30 eligible pupils signed up to the study, pupils were randomly selected using a computer-based algorithm.

Two recruitment methods were tested to examine which was more effective in order to inform a recruitment strategy in a definitive trial. The two methods were designed to ensure that we recruited participants with a range of physical activity levels. Recruitment method A (briefing) involved a short briefing in each class to explain the study and to give out pupil and parent information sheets and consent forms. Recruitment method B (briefing plus taster session) involved the same as recruitment method A plus a 20- to 30-minute taster session led by coaches from Bristol City Council. The taster session aimed to give the pupils an idea of what the Action 3:30 after-school intervention would be like. It was also intended to allay any concerns about the level of physical activity that might be required to take part in the study and to highlight that the sessions would be enjoyable. Four schools in South Gloucestershire and two schools in North Somerset were randomly selected to receive recruitment method B. Written parental consent was obtained for each child. A brief ‘reason for non-consent’ questionnaire was sent home to parents who did not give their consent in order to explore why they did not want their child to take part.

Pre-baseline data (opt-out)

Before the main study commenced, all Year 3 and 4 children in all 12 schools were asked to complete the validated Physical Activity Questionnaire for Older Children (PAQ-C).29,30 Data were collected in an opt-out consent process. Parents were informed of the measures on an information sheet and asked to return an opt-out form if they did not want their child to participate. In four of the schools in South Gloucestershire, accelerometer data from all Year 3 and 4 pupils were also collected via this opt-out process. This process was conducted in only four schools to assess the utility of the approach and to provide an indication of the potential additional resources that the research team would need if such an approach were used in a larger trial. These data were used to compare the levels of physical activity between pupils who did and pupils who did not consent to take part in the main study and they were treated as the baseline values for participants who subsequently took part in the randomised controlled trial.

Baseline data

Following recruitment and before randomisation, baseline data were collected from the selected children in each school. A summary of the measures can be found in Box 1. Participant age at baseline was calculated from parent-reported date of birth. Parental ethnicity was self-reported by selecting one of 13 descriptions based on the UK census. Pupils’ socioeconomic status was estimated using two measures: parent-reported highest level of education (i.e. up to General Certificate of Secondary Education or similar; Advanced levels/National Vocational Qualifications; first degree/diploma/Higher National Certificate/Higher National Diploma; or higher degree); and parent-reported postcode, which was used to ascertain Index of Multiple Deprivation (see Appendix 1) using English Indices of Deprivation 2010.31 Accelerometer data that had been collected in the four schools during the opt-out process were used for baseline.

Parent public and patient involvement

Two public and patient involvement meetings were held with parents in two schools following recruitment and randomisation. Both meetings were held in control schools, one of which had received recruitment method A and the other of which had received recruitment method B. The aim of the meetings was to gather more information about why parents had not given consent for their children to join the study (following the completion of the reason for non-consent questionnaire) and how to overcome these issues, as well as the parents’ thoughts about the recruitment methods and materials.

Measures

Data were collected from all of the selected children (intervention and control) at two time points:

1. baseline – June 2017
2. follow-up (weeks 13–15 of the intervention) – February/March 2018.

Accelerometer-determined physical activity

Pupils wore ActiGraph GT3X+ accelerometers (ActiGraph, Pensacola, FL, USA) for 7 consecutive days to assess their physical activity levels. Periods of ≥ 60 minutes of zero counts were recorded as ‘non-wear time’ and were removed.32 Pupils were included in the analysis if they provided ≥ 3 valid days (500 minutes of data between 6 a.m. and 11 p.m.). Minutes in MVPA were estimated for weekdays and weekends using the Evenson cut-off point. 33 In addition, total physical activity was derived from accelerometer counts per minute (c.p.m.) and sedentary time was estimated based on a cut-off point of < 100 c.p.m.

Objective measures

Pupils’ height and weight were measured to the nearest 0.1 m and 0.1 kg, respectively, with their shoes, coats and jumpers removed, using a portable Seca stadiometer and digital Seca scale (both Seca United Kingdom, Birmingham, UK), respectively. Body mass index (BMI) (kg/m²) was calculated and converted to an age- and sex-specific standard z-score.34

Self-reported measures

Pupils completed a tablet device-based questionnaire assessing their motivation to undertake physical activity and their activity-based perceptions of autonomy, relatedness and competence need satisfaction, using established scales that have been developed and used with a similar age group successfully,24 as well as self-esteem and peer support. Pupil-reported questionnaire data were collected and managed using REDCap35 (https://brtcclinical.bris.ac.uk/redcap/; accessed 20 April 2018) hosted at the University of Bristol.

Physical activity motivation was assessed36 using a five-point Likert scale ranging from 0 (not true for me) to 4 (very true for me). Four types of motivation were assessed via three items each: (1) intrinsic (e.g. because I enjoy being active), (2) identified (e.g. because it is important to me to do active things), (3) introjected (e.g. because when I’m not active, I feel bad) and (4) external (e.g. because other people pressure me to be active). The mean of the three items was calculated for each type of motivation. Composite scores for autonomous (mean of intrinsic and identified) and controlled (mean of introjected and external) motivation were calculated. This scale has been previously validated with primary school-age children.24

Autonomy37 (six items, e.g. I can decide what activities I want to do), competence38 (six items, e.g. I think I do well, compared with other children my age) and relatedness37 (six items, e.g. I am supported by others) were assessed using a 6-point Likert scale, ranging from 1 (not like me at all) to 6 (really like me). A mean score for each variable was calculated.

Self-esteem was assessed using a 10-item Likert scale, ranging from 1 to 5, where pupils had to state how true or false each statement was.39 Eight items were positively worded (e.g. I do lots of important things) and two items were negatively worded (e.g. I can’t do anything right). Negatively worded items were reverse coded and the mean of all items was calculated.

Peer support was assessed using a four-item Likert scale, ranging from 0 to 4 (0 – no, never; 1 – not much; 3 – quite a lot; 4 – all the time). All items started with ‘Thinking about your good friends, do they . . .’, followed by a statement (e.g. ‘encourage you to be active’). The sum of all four items was calculated.

Health-related quality of life (HRQoL) was assessed using the KIDSCREEN-1040,41 and Child Health Utility 9D (CHU9D)42,43 questionnaires as part of the economic evaluation. The economic evaluation is described in Chapter 5.

Parent-reported measures

Pupils’ parents completed a questionnaire to enable the collection of demographic data (baseline only). In addition, data on daily travel mode to and from school and child’s current participation in after-school clubs were collected at baseline and follow-up. Daily travel mode was assessed for weekdays only, with four response options (i.e. walk, cycle/scoot, car, and public transport). Data on usual travel mode and participation in after-school clubs were collected to capture whether or not attending Action 3:30 affected these in intervention pupils.

Randomisation

School was the unit of randomisation. Schools were randomised to the control (n = 6) or intervention (n = 6) arm after baseline data had been collected. Randomisation was stratified by local authority (South Gloucestershire or North Somerset at a ratio of 2 : 1 in each arm) and recruitment method (recruitment method A or recruitment method B at a ratio of 1 : 1 in each arm). Randomisation and allocation were conducted by an independent member of the Bristol Randomised Trial Collaboration using Stata version 15.1 (2017; StataCorp, College Station, TX, USA). Two of the intervention schools were unable to provide staff to attend the intervention training programme and therefore did not deliver the intervention. This is described in more detail in Chapter 4. Therefore, four schools were intervention schools and six schools were control schools.

Blinding

Because of the nature of the study, it was not possible to blind the schools or the pupils, as they had to know whether or not they had been allocated to the intervention. It was also not possible to blind the trial manager or field workers, as they collected the outcome and process evaluation measures. The randomisation processes and primary and secondary analyses were conducted blind by the statistician in the Bristol Randomised Trials collaboration, who was then unblinded to conduct the complier-average causal effect (CACE) analysis.

Sample size

As this was a feasibility trial, no formal sample size calculations were performed. For safety reasons, the maximum number of pupils that could be enrolled in the main study was 30 per school, giving a maximum sample of 360 pupils. Potential sample sizes for a definitive trial were estimated using the derived intraclass correlation coefficient (ICC) for MVPA from this study and published ICCs from comparable studies and based on combinations of key parameters (type I and type II error).

The Action 3:30 intervention

Teaching assistant training

Nine TAs from the four intervention schools (two from three schools and three from one school) attended a 25-hour (5-day) training programme in September/October 2017. TA time to attend the training was covered by the schools. The training programme also met the requirement of the Community Sports Leader award. The training programme was delivered by the coach development manager at Bristol City Council, a very experienced senior coach who has led training courses for a number of sports. Action 3:30 is based on the principles of SDT22 and therefore the programme focused on promoting children’s autonomy, relatedness and competence in relation to being physically active. To promote autonomy, TAs were taught to empower pupils by offering choice within activities, such as leading warm-ups, adapting games and controlling the speed at which activities progressed. They were encouraged to invite ideas for designing games and to help children to set their own goals. They were discouraged from using prizes or promised rewards to incentivise club members. A section of the training also focused on how to run child-led sessions in which the pupils choose the activities. TAs were encouraged to create relatedness through empathetic TA–child interactions, and through emphasis on co-operation rather than competition. TAs were trained to support competence by setting progressive activities targeting success balanced by optimal challenge, and by using specific praise to reinforce a message of quality over quantity. A learning point from the previous Action 3:30 pilot study was that TAs needed more support to manage disruptive behaviour.28 In response to this need, a section of the training was tailored with input from the coach development manager at Bristol City Council, who co-designed the original training materials, to prepare TAs for how to best manage disruptive behaviour in different scenarios. TAs also received a training guide, a comprehensive reference document that reinforced all of the content in the training.

The after-school club

Once trained, the TAs delivered the Action 3:30 after-school club twice per week for 15 weeks between November 2017 and March 2018. To mirror usual school provision, each session lasted 1 hour, commenced immediately after school and was available to all pupils in the year group. TAs received a leader’s manual, which included 30 detailed session plans, a range of games and activities, and reminders about how to embed the key principles of SDT. TAs were asked to deliver the sessions in the prescribed order. Video recordings of model delivery were supplied for 22 games/activities within the sessions using a secure online platform, and links to online resources were given for 19 of the sessions to aid delivery. In addition, pupils were provided with home activity cards every three sessions. These cards aimed to reinforce session content and advise pupils about how they could practise the activities at home with friends and family.

The intervention was refined and improved following the original Action 3:30 study based on findings from the primary analysis18 and process evaluation.28 Table 2 highlights the changes made to the intervention content.

School/pupil appreciation in the intervention and control arms

Intervention schools received £200 to spend on equipment, which they used to cover any school-specific gaps in the equipment that was necessary to deliver the Action 3:30 sessions. Schools were reimbursed for 2 hours per week for two TAs at the TAs’ usual pay rate to cover their time spent delivering the club. Control schools and the two former intervention schools received a £300 donation in recognition of the time spent by school staff to accommodate data collection. Each child (control and intervention) received a small gift at each data collection. The gift was used to encourage the prompt return of accelerometers and was selected to promote physical activity.

Re-enrolment

Data from the original Action 3:30 study and a recent evaluation of an extracurricular dance programme18,44 highlighted that schools usually provide an opportunity for pupils to re-enrol in after-school programmes every term to allow more children to join the programme if others drop out. To enhance external validity, attendance rates in each intervention school were assessed after session 12 (December 2017/January 2018). In each club, if there were spaces available, information sheets and consent forms were sent home with all remaining eligible Year 4/5 pupils. Demographic data were collected from pupils who joined the study at the enrolment point and participated in the follow-up measures.

Process evaluation

Overview

A process evaluation using the RE-AIM framework, reporting on recruitment, dose, intervention effectiveness, fidelity and adoption, was conducted in the four intervention schools. It had both quantitative and qualitative components, which are described below.

Quantitative process evaluation measures

The quantitative component was conducted using self-report questionnaires and observations. TAs were asked to complete a log book, which recorded attendance and the dose of the intervention (whether each session was delivered fully, partially or not at all to plan). A member of the research team observed three randomly selected sessions (one per half-term) in each intervention school during the 15 weeks. The visits were spaced evenly in each school: approximately 4 school-term weeks apart, starting at or near week 5 of the intervention. During the three observation visits, pupils completed the Sport Climate Questionnaire45 to assess their perceived autonomy support of the TAs, which consisted of six questions on a seven-point Likert scale. Pupils also completed an enjoyment and exertion questionnaire at these observation time points, which consisted of two items referring to their enjoyment of that day’s session and how tiring they perceived that session to have been. Pupils responded using a five-point Likert-type scale ranging from 1 (‘not at all’) to 5 (‘a lot’) for the enjoyment item and an 11-point Likert-type scale ranging from 0 (‘not tired at all’) to 10 (‘very, very tired’) for the exertion item. The dose of the intervention was also assessed by the researcher. TAs were asked to complete a validated self-efficacy questionnaire, made up of an adapted version of the Physical Education Teacher’s Physical Activity Self-Efficacy Scale46 and a vignette about autonomy-supportive teaching style and eight items referring to this.47 The questionnaire was completed before and after the intervention training and during the three observation visits. A researcher assessed school context in all 12 schools using a validated school physical activity environment scale.48 Data on school physical activity policy context were collected by soliciting information from a school contact about physical activity policies and curriculum. These data were used to examine whether or not differences in the social/physical environment and in school policy strategies could have an impact on the delivery of the intervention. Adoption of the intervention was assessed by recording the number and proportion of schools and TAs who stated that they would continue to deliver a version of Action 3:30 once the intervention period had ended.

Qualitative process evaluation measures

Following completion of the intervention training programme, the coach development manager at Bristol City Council (lead trainer) took part in an interview with the trial manager, which focused on the delivery of the training and whether or not it could be improved. He also completed a checklist to assess whether or not training had been delivered as planned.

Once the 15-week intervention was finished, all TAs (n = 9) and intervention school key contacts (staff members who act as the primary liaison between the school and the study team; n = 4) were asked to take part in semistructured interviews, and selected children (six boys and six girls in each intervention school) were asked to take part in focus groups. Children were purposely selected based on their attendance at the club. The mean number of sessions attended was calculated for all pupils. Pupils with low, medium and high attendance rates were then selected. One pupil who joined at the second enrolment in each intervention school was also purposively selected. School key contacts (n = 4) and Year 3/4 classroom teachers (n = 2) in schools who took part in the accelerometer opt-out process were asked to take part in an interview to discuss the burden associated with this process, and how to mitigate it in future studies. One intervention school also took part in the accelerometer opt-out process, and so that school’s key contact was asked about this process in the same interview. Eight external stakeholders were also asked to take part in an interview. After two schools withdrew from the intervention, a school key contact from each school was asked to take part in an interview to explain the reasons for withdrawal and how this could be prevented in future trials. Two researchers with master’s-level qualitative experience (BT, male; and AP, female) conducted all of the interviews and focus groups. All informants received information about the research aims and the researchers and provided written consent prior to interview. Field notes were made during and after each interview or focus group by the researcher present.

Interview topic guides

Interview topic guides were developed for each informant group (five guides in total). Child focus groups were conducted in school with just the researcher present. The focus groups explored recruitment, attendance, delivery, enjoyment and possible improvements to the intervention. Pupils were asked what they enjoyed the most and least about the intervention and what motivated them to attend. Pupils were also asked to comment about their physical activity context and environment in school.

Face-to-face paired in-depth interviews49 were conducted with TAs in each school, which involved interviewing TAs from the same school together50 for the purpose of gathering information about how the pair/trio perceived the same event(s),51 in this case the training programme and the intervention delivery. The topic guide explored the content and delivery of the training programme, with a focus on the managing disruptive behaviour section. Regarding the intervention, TAs were asked about attendance, delivery, theoretical fidelity and potential improvements. They were also asked to comment on recruitment (TAs and pupils), the re-enrolment strategy and potential contamination between schools.

Face-to-face interviews were conducted with intervention school key contacts in school. The topic guide addressed their involvement in the study, thoughts about the recruitment, content, delivery and enjoyment of the intervention, potential to continue the Action 3:30 programme, potential contamination between schools, and suggested improvements. If applicable, they were also asked to comment on the accelerometer opt-out process.

School key contacts and classroom teachers involved in the accelerometer opt-out process were interviewed face-to-face in school. The topic guide explored the logistics and potential burden of the process.

Face-to-face (n = 3) or telephone (n = 5) interviews were conducted with external impact stakeholders, including regional public health leads, school sport co-ordinators and directors of public health non-profit organisations. The topic guide examined the sustainability, future commissioning and dissemination opportunities of Action 3:30, as well as potential improvements.

Qualitative analysis

To synthesise the wide range of qualitative data collected, the framework method52 was used, as it enables the production of a matrix of the data (from different participant groups), which allows for constant comparison between groups. The framework method also allowed a combined approach to the analysis, enabling themes to emerge inductively from participant accounts and specific issues to be explored deductively. Data were analysed using the following steps:

• Researchers from the study team (BT and AP) thoroughly read and re-read each transcript and listened back to audio-recordings to become familiar with the data set. Their initial impressions of the data were recorded.
• Initial codes were created from the central tenets of the underlying intervention theory (SDT22) to interpret the data. These codes examined whether or not the intervention supported autonomous motivation among the participants and the three needs support constructs (autonomy or feelings of choice; relatedness – a feeling of connection between the participants and the instructor; and competence). In addition, a predefined code, ‘school context’, was included to assess the differences between schools in intervention experience and implementation linked to their physical/social/policy environment. For the deductive analysis, the predefined codes were broad, the primary purpose being to categorise relevant information to be further interrogated to elicit more refined codes and interpretations. The deductive analysis codes are presented in Appendix 2.
• An analytical framework was developed to fit each informant group to avoid losing any key data. Four researchers (BT, AP, SS and RJ) met to discuss and refine a set of codes that could be applied to all transcripts using the following process:
  • Two researchers (BT and AP) independently read and analysed two transcripts from each informant group, met, and discussed and created draft frameworks for each informant group.
  • A third researcher (SS) read two transcripts from half of the informant groups (e.g. boy pupils, TAs) and a fourth (RJ) read the other half (e.g. girl pupils, school contacts) and made notes of the key codes/themes.
  • All four researchers met to discuss and refine draft frameworks.
• The two primary researchers then applied frameworks to remaining transcripts (with some level of double coding, i.e. each researcher coded two more transcripts from each stakeholder group) using NVivo (QSR International Pty Ltd, Warrington, UK). Any new codes that emerged were discussed, and amendments were made to each framework.
• Coded data were then charted into a framework matrix in NVivo. This summarised the data for each informant by category and included representative quotations. To ensure consistency within the research team, summarising techniques were compared.
• Data were interpreted and themes were generated in frequent meetings to review the matrix. The two primary researchers agreed on illustrative quotations to show the nature of each theme.
• The frameworks were triangulated to compare the codes for each informant group to assess the degree of convergence.

The qualitative analysis is reported in accordance with the Consolidated Criteria for Reporting Qualitative Research (COREQ) guidelines53 (Table 3) to optimise the credibility and transparency of the research.

Statistical analysis

Stata statistical software was used for all statistical analyses. Summary statistics were presented comparing the control and intervention arms at baseline and follow-up on demographics [age, ethnicity, socioeconomic status (using Index of Multiple Deprivation)] (baseline only), psychosocial variables and accelerometer variables, including MVPA. Where the distribution of the outcomes was approximately normal, mean values and standard deviations (SDs) were presented; otherwise, medians and interquartile ranges (IQRs) were presented. For binary/categorical variables, a number and percentage were presented. As this was a feasibility trial, the primary and secondary outcomes were reported using basic statistics to describe the recruitment, session attendance, accelerometer and questionnaire data. To avoid losing power when calculating average scores for the measure of autonomy-supportive teaching experienced by pupils (Sport Climate Questionnaire collected at three process evaluation time points), a multiple imputation approach was taken, which provides a more realistic representation of what would have been provided had all of the children provided complete data at all time points (see Table 17). Non-imputed data are also presented in Appendix 4.

For the accelerometer-assessed mean minutes of weekday MVPA outcome, the following analysis procedures were undertaken. All pupils providing measurements were included, as per their allocation, in an ITT primary analysis to assess the effectiveness of the Action 3:30 intervention to improve the mean minutes of MVPA per weekday among Year 4 and 5 pupils at follow-up using multivariable mixed-effects linear regression. Multivariable mixed-effects linear regression considers the hierarchical nature (i.e. the clustering) of participants within schools. The model included covariates for intervention arm, local authority (South Gloucestershire or North Somerset), recruitment strategy (taster session or not) and the baseline measurement of mean minutes of weekday MVPA. The difference in mean weekday MVPA between the two allocated arms was presented with its 95% CI. This was repeated with stratification by sex, with the differences in mean weekday MVPA between the two allocated arms being presented with 95% CIs for boys and girls separately. This model was adapted to secondary measures based on accelerometer estimates such as sedentary time. The total proportion of pupils meeting the UK recommendations of 60 minutes of MVPA per weekday was calculated based on the average minutes of MVPA across all valid weekdays.

The large majority of trial participants had indicated a willingness to join the Action 3:30 programme at the outset, before the allocation of their school to intervention or control, thereby ensuring allocation concealment. The ITT estimate was based on a comparison of these participating children between those whose school was allocated to intervention and those at schools allocated to control. Children who later joined their school’s ongoing programme at ‘re-enrolment’ made this decision knowing that they were at an intervention school; this fundamental difference prompted us to prespecify a separate analysis of the outcomes in these children, which, if justified, would then be combined using meta-analysis techniques with the ITT estimate.

A key issue when estimating the effect of the intervention in those children joining at re-enrolment is identifying an appropriate comparison group. It is not known which children in the control schools would have joined the intervention at the later opportunity, so the only unbiased comparison based on randomly allocated groups is between all children who did not indicate a willingness to join the Action 3:30 programme at the outset, comparing those at schools allocated to intervention with those allocated to control. This estimate is not equivalent to our primary ITT estimate; here we pilot a CACE method to obtain an unbiased estimate of the effect of the intervention in those who joined the intervention at re-enrolment. This CACE estimate is comparable with the ITT estimate from those children indicating a willingness to enrol at the outset. It is worth noting that both the ITT estimates and the CACE estimates disregard the number of sessions each child attends: the intervention effect is calculated for those children intending to join the programme, whether this is at the outset or the later opportunity.

The CACE estimate and its CI were obtained using the two-stage least squares approach. This method needs all eligible children at each school to provide the necessary measurements; in this pilot, we did this for four schools. The key aim here was to specify the method and consider the feasibility of the approach, as the estimates produced at this stage would not be usefully precise.

Sample size

The sample size for a future definitive trial was estimated using the ICC for weekday MVPA. The school-related ICC was estimated for daily weekday MVPA to power a definitive trial to detect a difference in 10 minutes MVPA per weekday between the two arms. A variance component random-effects model was used to estimate the ICC. As the ICC for MVPA was estimated using data from only 12 schools, the school-related ICC was compared with that observed in other studies.

Compliance and missing data

Accelerometer compliance was measured by the number of valid days of accelerometer data provided by each pupil enrolled in the study at each time point. At minimum of 3 valid days of accelerometer data were required for compliance. Questionnaire compliance was measured by the number and percentage of pupils enrolled in the study who provide questionnaire data. Data provision rates for accelerometer (missing, invalid, valid) and questionnaire data (missing, not missing) were recorded for baseline and follow-up. There were missing data because some participants did not wear their accelerometer for enough time for data to be valid and some did not return their accelerometer (see Table 11 for data provision at each time point).

Governance

A local advisory group was formed, which included local authority physical activity staff from the two local authorities. The group met three times during the study: once during baseline data collection, once during the intervention delivery period to advise on school-related/intervention issues and once at the end of the project to provide feedback about the interpretation of the findings and the future sustainability and dissemination of Action 3:30. A Children’s Advisory Group was formed with children in the four intervention schools. Four children in each school (two boys and two girls) were recruited, one high active and one less active of each sex, to provide feedback about the intervention and raise any logistical issues. Three Children’s Advisory Group meetings were held during the intervention. A Trial Management Group was formed, comprising all of the applicants, the trial manager and the field worker. The Trial Management Group met once per month to discuss key issues. A TSC was formed, comprising an independent chairperson, two independent members and the study team. The TSC met three times during the study and provided independent scientific scrutiny of the study, support to the study team and guidance on how to progress to a definitive trial. A member of the local advisory group was invited to attend the TSC meetings to provide input.