Active for Life Year 5: a cluster randomised controlled trial of a primary school-based intervention to increase levels of physical activity, decrease sedentary behaviour and improve diet

Evidence for the effectiveness of school-based interventions to improve diet and physical activity

A systematic review that included 44 school-based randomised controlled trials (RCTs) found beneficial effects on moderate or vigorous physical activity (MVPA) during school hours, but the authors noted that benefits might have been exaggerated as a result of the outcome assessment being self-/parent-reported and not blind to school allocation in most trials, and also that the marked loss to follow-up in several trials might have led to biased findings. 11 Furthermore, it could be argued that it is not surprising, given that the interventions largely included extra compulsory physical activity lessons, that greater time spent in MVPA during school hours was found. Evidence from observational epidemiological studies suggests that compulsory physical activity lessons in school are associated with more school-based activity, but not with more total activity17,18 and that, long term, those who attended schools with more compulsory physical activity have similar levels of physical activity, physical fitness and body mass index (BMI) as young adults. 18 A second systematic review included only studies in which physical activity had been assessed objectively by accelerometers and did not restrict the outcome to activity during school hours; the review included school-based studies, as well as those of interventions in other settings. 12 It reported beneficial effects of interventions, with no evidence that this differed between school-based or family/other setting interventions. The authors commented that the magnitude of the effect was small and unlikely to be of health benefit,12 although modest shifts in risk factors can produce important public health benefit. One systematic review identified five RCTs of school-based intervention to reduce sedentary behaviour and reported that all of them were effective. 15 Results were not pooled formally and the outcome in all of the studies was based on self-/parent-reporting. 15 In a more recent systematic review and meta-analysis, evidence from 34 RCTs suggested that both school-based interventions and interventions in other settings were effective in reducing time spent in sedentary behaviour and, consequently, in reducing mean BMI. 16 Only nine of these studies reported that random allocation was adequately concealed, and only eight reported blinding of the outcome assessment; sedentary behaviour was assessed by self-/parent-report in all studies. 16 Two recent systematic reviews of school-based interventions to increase fruit and vegetable consumption identified 19 RCTs and 27 RCTs or non-randomised trials, respectively. 13,14 One review, which focused solely on primary school interventions, concluded that computer-based interventions were effective (based on pooling of two RCTs), but pooling of other trials did not suggest interventions were effective. 13 The authors noted that the majority of studies did not describe the randomisation method and it was not possible to ascertain if allocation was concealed for most of the 19 RCTs reviewed. The authors also noted that most did not take account of clustering (non-independence between children from the same school) in their analyses, despite all being cluster RCTs. 13 The second review also focused on children in the primary school age range (5–12 years), but they concluded on the basis of pooled results from 21 (out of 27) controlled trials that school-based interventions were effective at increasing fruit but not vegetable consumption. 14 Again, the poor quality of the majority of the trials was noted by the authors.

Rationale for the Active for Life Year 5 intervention

We began work to design, pilot and then fully evaluate a school-based intervention to improve levels of activity and diet and other health outcomes in children in 2006. Consistent with recommendations of the UK Medical Research Council and others for the evaluation of complex interventions, our aim was to develop an intervention that was theory based and built on evidence from appropriate reviews of the literature, and then to test its feasibility and complete pilot work prior to seeking funds for a full-scale RCT. 19,20 Among national and international policy-makers and researchers there has been a strong belief for some decades that simple interventions in schools can change unhealthy behaviours to healthy ones. 21,22 We wanted to test this and, therefore, sought an intervention that would be possible to deliver in schools with minimal disruption to the main aim of educating children and moreover that would be relatively inexpensive. Lastly, we wanted an intervention that focused on children under the age of 11 years because of evidence that persistent overweight/obesity23 and the association of greater adiposity with future coronary heart disease is established by this age. 24

Our original literature search identified a cluster RCT in 11- to 12-year-olds25 and a quasi-randomised trial of 8- to 9-year-olds,26 both completed by the same group of researchers and of a similar intervention that fit with the type of intervention we wanted to develop for use in the UK. The intervention, based on social cognitive theory27 with a particular emphasis on improving child self-efficacy to make behavioural change,28 aimed to reduce childhood obesity and improve health via changes in physical activity, diet and screen viewing. These studies found beneficial effects, including on overweight/obesity in girls in the older age group; in the study of the younger age group, BMI was not assessed.

Between 2006 and 2008 we worked with primary school teachers, the local primary care trust (public health commissioners) and local council (government) in South Gloucestershire, in the south-west of England, to determine whether or not this intervention could be adapted for use in the UK, whether or not it was feasible to deliver the adapted intervention within the national curriculum and whether or not there was evidence of promise for the intervention from a pilot RCT sufficient to justify a full-scale RCT. This work showed that, with minor adaptations, the intervention could be delivered within the UK national curriculum for year 5 children (aged 9–10 years) and the pilot RCT suggested that it might be effective. 29 We had a limited budget for the pilot and so within it were not able to test the use of accelerometers. The process evaluation within the pilot RCT found that the teachers thought the intervention should be extended to include parents if it was to be maximally effective. 29,30 We therefore obtained a further small budget and undertook qualitative work with parents and teachers to develop the intervention in such a way that it involved parents; this showed that parent-interactive homework assignments would be feasible and acceptable to them. 30 We then completed a feasibility study (examining before-and-after intervention changes in the same children) of adding parent-interactive homework assignments to Active for Life Year 5 (AFLY5) and of collecting accelerometer data. 30,31 Results from that work provided further support for going ahead with a full-scale RCT of the AFLY5 intervention that then included parent-interactive homework assignments, as well as lessons. None of the schools or teachers who were involved in the feasibility and pilot work has been included in the main RCT, which is presented in this monograph.

Rationale for our study design

Given the importance of establishing healthy behaviours in relation to physical activity and diet, the potential for effectively doing this via schools, but the lack of high-quality trials to date in this area, the rationale for undertaking the AFLY5 RCT was to address the weaknesses that existed in school-based RCTs to promote healthy diet and levels of physical activity at the time that we began the study. Specifically, in our study design we aimed to ensure that (i) random allocation was concealed; (ii) an objective assessment of time spent in MVPA and sedentary behaviour were collected using accelerometers; (iii) the fieldworkers collecting outcome data from children were blind to school allocation; and (iv) the intervention did not focus solely or primarily on increasing physical activity through physical education (PE) classes in school. 32 The intervention was designed to change the child’s behaviours in a non-compulsory way, and so measurements were concerned with the whole day and both weekdays and weekends, not just during school hours. We also aimed to (v) undertake one of the largest RCTs in this area; (vi) take account of the clustered nature of the design in the sample size calculation and analysis; (vii) work with the schools to ensure loss to follow-up was minimised; and (viii) examine the effect of the intervention both immediately after its completion and again 12 months later in order to determine whether or not any effects were sustained. 32,33

Ethics approval and research governance

We obtained ethics approval from the University of Bristol’s Faculty of Medicine and Dentistry Committee for Ethics (reference number 101115). Once schools agreed to participate in the study, parents/guardians of children in year 4 were sent a letter and information sheet about the study with an opt-out consent form for their child for each of the measurements. They were given the opportunity to contact the research team to discuss the study and also information about being able to withdraw at any stage. An information sheet for the child was sent with the letter that was sent to the parents. The children were given a second copy of this information sheet at the time that measurements were undertaken and they were asked to give signed assent to each of the measurements. Any child whose parents had refused consent for all measurements was not included in the study, although they may have received the AFLY5 lessons and homework assignments, as these were delivered at the school level as part of the school curriculum. Any child whose parents refused consent for one or more (but not all) measurements did not have those measurements undertaken but were included in the study. Any child refusing assent for a measurement at any time did not have that measurement taken (irrespective of whether or not the parent had not refused consent).

Eligibility and recruitment

State primary or junior schools with years 4–6 pupils in the Bristol City and North Somerset administrative areas were eligible for inclusion. Between March and July 2011 all state primary and junior schools with children in years 4–6 (aged 8–11 years) in the areas covered by Bristol City Council (93 schools) and North Somerset Council (55 schools) were invited to participate. Both of these areas are in the south-west of England and include a range of levels of deprivation, as well as urban and rural areas. Special schools (for children whose additional needs cannot be met in a mainstream setting) were excluded because they were unlikely to be teaching the standard UK national curriculum and the children may not have been able to take part in all the measurements. One hundred and forty-eight schools were invited and 63 expressed an interest in taking part, with three schools subsequently withdrawing their interest. Sixty schools were recruited (46 in Bristol City and 14 in North Somerset). Participants were children in year 4 (aged 8–9 years) at the time of recruitment.

Randomisation

Prior to randomisation school heads were asked to complete a brief questionnaire about the school. This included three questions that asked them to list all activities the school was engaged in that related to (1) increasing physical activity, (2) decreasing sedentary behaviour and (3) promoting a healthy diet in pupils. Responses were free text and, on the basis of these responses, each school was classified as either high (one or more initiatives) or low (no initiatives) involvement in health-promoting initiatives relevant to the outcomes of this trial. When heads (or teachers they delegated the task to) reported initiatives that were part of the UK national curriculum or that they had been awarded ‘healthy schools’ or ‘healthy schools plus’ status these were not included as involvement in an initiative, as these are widespread in the south-west of England and we were looking for additional initiatives that varied between schools. Schools were also defined as being in an area of high, medium or low deprivation by splitting them into thirds based on their score on the English Index of Multiple Deprivation (IMD) 2010. 34 Schools were grouped into six mutually exclusive strata by these two characteristics and randomly allocated to control or intervention within these strata. 32,33 Randomisation was undertaken by DAL who was unaware of any characteristics of the schools (school identifications were used to relate them to the two stratifying variables and DAL had no knowledge of which schools these identifications linked to); it was concealed by using the Bristol Randomised Trials Collaboration’s automated (remote) system. After randomisation, one school refused to undertake the intervention; the head reported that they had hoped they would be randomised to control and did not have the time or capacity to accommodate the intervention. The school did agree to participate in all pupil measurement sweeps. This school is included in the randomised group (intervention) for the main intention-to-treat (ITT) analysis and is excluded from the per-protocol analysis.

Intervention

Full details of the intervention have been published in the trial protocol and pilot study. 29,32 It comprised:

1. Training for year 5 classroom teachers and learning support assistants (LSAs), provided by the trial manager, a nutritionist and a PE specialist. The training took place over a whole day (8–9 hours) in a location away from any of the schools and where the teachers/LSAs and those delivering the training would not be interrupted. Teachers/LSAs were given a choice of days to attend the training and schools were financially compensated for the cost of replacement teachers while their staff attended training. At the training days the rationale for the intervention was explained and each lesson and homework assignment were discussed and then taught in interactive ways. Time was provided for questions and discussion. Teachers were instructed to deliver 16 lessons, 10 of which had associated homework assignments to be handed out by the teachers. They were told that they could adapt the teaching plans and materials, as they would with other lessons (e.g., to suit their own style and the range of abilities in their class), but that the aims and knowledge/skills to be imparted should remain the same.

2. Provision of 16 lesson plans and teaching materials, including pictures, compact discs (CDs) and journals for year 5 teachers or LSAs to deliver over two out of the three school terms in year 5 (6–7 months). The 16 lessons included nine lessons that were primarily related to how to be more active and less sedentary and why this was important, six related to healthy nutrition and how to achieve this and one about reducing screen viewing. Each lesson did, however, combine different aspects of healthy behaviour. For example, in the physical activity lessons the children played games based on the food groups using photographs of food, which reinforced the content of the nutrition lessons. Similarly, in the lesson (and associated homework assignment) for reducing screen viewing [called ‘Freeze my TV’ (television)] children were taught how to replace regular TV watching with active play on some days.

3. Provision of 10 parent–child interaction homework activities. The homework assignments were designed to involve parents and other family members in the behaviour change process by reinforcing the messages delivered during lessons. The homework assignments included activities such as ‘Freeze my TV’, in which time normally spent watching TV would be replaced with physically active play, involving the parents and other family members, that the child would write a log about; cooking simple healthy food at home; thinking of different ways to add an extra serving of fruit or vegetables to their diet every day for a week on the ‘5-a-day planner’; and measuring the sugar content of drinks that the family have at home or include in school/work lunch packs.

4. Information was provided for schools to insert (as they wished) in the school newsletters about the importance of increasing physical activity, reducing sedentary behaviour and improving diet. This information was available on the CDs given to all schools with their training packs. Schools were free to edit this information and insert none, all or some of it.

5. Written information for parents on how to encourage their children to eat healthily and be active was delivered via the school children at the start of the intervention in the form of the Change 4 Life: Top Tips for Kids leaflet produced by the Department of Health. 35

The intervention took place when the children were in school year 5 (aged 9–10 years) after baseline assessment. Our previous feasibility work showed that the AFLY5 intervention was aligned to the UK national curriculum for key stage 2 (which is used for all children aged 7–11 years). 29 Schools randomised to the control group continued standard education provision for the school year, including any involvement in additional health-promoting activities, but had no access to the intervention teacher training and no known access to the teaching materials, which have not been published and were not made available by the research team beyond the intervention schools.

Sample size calculation

The sample size calculation was based on the intracluster correlation coefficients for different outcomes and other information collected during pilot/feasibility work. 29,32,33 For each of the three primary outcomes we determined the number of schools required (assuming 25 pupils per school) to detect at least a 0.25- to 0.30-standard deviation (SD) difference between pupils in intervention and control schools with 80–90% power, a two-sided alpha of 0.05 and allowing for 15% loss to follow-up or missing data. 32 For secondary outcomes we took account of multiple testing and aimed to have at least 80% power at an alpha of 0.01 for all of these, including the ability to detect a relative minimum difference of 30% in general or central overweight/obesity. 32 These calculations showed that we needed to recruit 60 schools with a total of at least 1500 children, with 1275 (after allowing for loss to follow-up/missing data) available for the primary analyses.

Outcome measures

All primary and secondary outcome measurements are listed in Box 1.

Participant assessments

Baseline assessments (prior to intervention) were undertaken when the children were in the final term of year 4 or early in the first term of year 5. Outcomes assessment was completed immediately post intervention (end of year 5 and start of year 6), and again at 1 year post intervention (end of year 6). Identical protocols and procedures were used at all three assessments. They were undertaken by trained fieldworkers who had completed enhanced Criminal Records Bureau (CRB)/Disclosure and Barring Service checks. The fieldworkers were blinded to the allocation of schools to the arms of the trial. This was achieved by a clear separation within the research team, whereby only the study project manager and one administrator (neither of whom was involved in data collection or analyses) knew which schools had been allocated to intervention and which to control. These people were told of the importance of not discussing school allocations with anyone. Furthermore, we made every effort for the intervention to be built into normal school lessons in such a way that it did not alert children to the fact that they were in an intervention school. We assumed this would prevent the children from saying things to the fieldworkers that would indicate whether they were in an intervention or control school. Teachers were reminded at the training session not to discuss the intervention with the fieldworkers who collected data on children in schools and these fieldworkers were sent to schools in an order that had no relationship to which arm of the trial the schools were in.

Accelerometer measurements

We used ActiGraph GT3X+ accelerometers (ActiGraph, Pensacola, FL, USA) to assess physical activity and sedentary time. Accelerometers were distributed at the school visit and collected 6 days later from the schools to allow 5 days of data collection (3 weekdays and 2 weekend days). The children were asked to wear them during the day (except when bathing or swimming or participating in contact sports such as karate). Time spent in MVPA was defined as any time spent in activities that generated ≥ 2296 counts per minute and in sedentary behaviour was defined as time spent in activities generating between 0 and 100 counts per minute. 33 Children were included in the analyses of accelerometer-based outcomes if they had at least 3 days with at least 8 hours of valid accelerometer data on each of those days. 33 In these analyses any period of ≥ 60 minutes of continued counts of zero activity were defined as the child not wearing their accelerometer and such periods were removed from analyses. Children (and their parents via written guidance) were told to take the accelerometer off when they went to bed and put it back on as soon as they got out of bed. Our analyses of the raw data suggested that none of the children wore their accelerometer while in bed.

Anthropometric measurements

All anthropometric measurements were completed with children in a private room with two CRB-checked, trained fieldworkers present. Weight was measured without shoes in light clothing to the nearest 0.1 kg using a Seca digital scale (Seca UK, Birmingham, UK). Height was measured, to the nearest 1 mm, without shoes using a portable Harpenden stadiometer (Holtain Anthropometric Instruments Ltd, Wales). Fieldworkers were trained to ensure the correct position for height assessment. WC was measured to the nearest 1 mm at the mid-point between the lower ribs and the pelvic bone with a flexible tape and repeated three times. 38 When BMI and WC were treated as continuous outcome variables, standardised z-scores (also known as SD scores) were derived internally by subtracting the mean BMI/WC for a given sex and age category (in 6-monthly categories) from the observed measure and dividing by the SD for the sex and age category. For binary outcomes, the International Obesity Task Force age (in 6-monthly categories) and sex-specific thresholds for overweight were used to define whether a child was overweight/obese based on BMI. 36 For WC any child above the 90th percentile for age- and sex-specific values derived from UK-relevant centiles37 was defined as having central overweight/obesity, as suggested by the International Diabetes Federation. 39 We did not test inter-rater reliability for these measurements, but given the ways in which we tried to ensure that the fieldworkers were blinded to which arm the schools were randomised we doubt that any measurement error would differ between the randomised arms.

Diet and screen viewing questionnaire data

Servings of fruit and vegetables consumed and other dietary outcomes were assessed using the ‘Day in the Life Questionnaire’, which has been previously validated for use in children of the same age as those in this study. 40,41 The method we used for determining servings of different food types from the text responses to this questionnaire, including reliability and validity checks have been previously published. 33,40 Servings of daily fruit and vegetable consumption and other dietary outcomes were assessed using an established and validated scoring scheme. 40,42,43 An abbreviated and updated version of a previously validated screen viewing questionnaire was used to assess self-report sedentary behaviour,44 with details of how we derived variables from this questionnaire previously reported. 29,33

Potential mediators

Potential mediators are listed in Box 2.

Physical activity self-efficacy was assessed using a validated questionnaire that consists of 26 items, each of which was answered by the child indicating their level of agreement on a 5-point scale (scored 1 to 5), where lower scores on this scale indicated lower self-efficacy. 45,46 Fruit and vegetable self-efficacy was assessed using a validated questionnaire consisting of 21 items, each of which was answered by the child indicating their level of agreement on a 5-point scale (scored 1 to 5), where lower scores indicate lower self-efficacy. 47 Parental support for physical activity and reducing sedentary behaviour was assessed using a validated 24-item scale, which provides information on modelling of parental physical activity behaviours (five items for each parent separately), logistical support (three items for each parent separately) and parental support for reduction of screen viewing (four items for each parent separately). 48,49 Each question is scored between 1 and 4, with lower scores indicating low levels of modelling or support for physical activity and low levels of limiting sedentary behaviour. Parental modelling of fruit and vegetable consumption was tested using a 12-item validated questionnaire that asked questions about mothers and fathers (or caregivers) together. 50 We were unable to identify a validated questionnaire for parental logistic support of fruit and vegetable consumption at the time of starting the study. We developed a test to specifically test the knowledge that children in the intervention schools should have gained via the intervention. We piloted the test on children of the same age as participants in AFLY5, but who had no involvement with the study or the participants who were in it, in order to make sure that it was understandable to the age of children in AFLY5. The knowledge test is shown in Appendix 1. It included nine questions each with a three-level multiple-choice response, where just one of the answers was correct and, therefore, the children could score between 0 and 9 on this test.

All potential mediators were assessed by questionnaires that were combined together into one document, together with the diet and screen viewing outcome measurements, and administered in the classroom for the children to complete in the presence of at least one of the trained study fieldworkers who answered any queries and assisted the children with reading and writing according to the study protocol. This instructed them to help with reading and spelling specific words, or understanding the meaning of a particular question, but not to suggest any answers.

Comparison of baseline characteristics

We compared relevant summary statistics of baseline characteristics between participants who were allocated to an intervention school and those allocated to a control school in order to determine whether or not any potentially influential imbalance had occurred (by chance) between these two groups. These comparisons also included accelerometer characteristics, including wear time, time with consecutive zero levels of activity, etc., to test our assumption that the characteristics that are used in criteria for deriving the accelerometer variables do not vary by randomised group. For all continuous and score variables we checked distributions using histograms and normal plots to examine how close to normality these were before deciding which summary statistics to present; mean (SD) or median [interquartile range (IQR)].

We did not compare baseline characteristics between the two groups with a statistical test (p-value) as any low values would simply represent a type 1 error under the assumption that we adequately randomly allocated participants. 51 As described in the general study protocol paper our procedures for randomly allocating schools to control or intervention were adequate. 32

Testing the effect of Active for Life Year 5 on outcomes immediately after the intervention

For all continuously measured primary and secondary outcomes at the immediate follow-up we used multilevel linear regression models to account for the clustering (non-independence) of children within schools; multilevel logistic regression was used for the two binary secondary outcomes (general and central overweight/obesity). The main analyses were ITT, with missing dates dealt with as described below. 33 All analyses included adjustment for the following baseline variables: age, sex, the baseline measure of the outcome being analysed, and the two pre-randomisation stratifying variables (involvement in other healthy behaviour-promoting activities and school-level deprivation). 33

We also did a secondary, per-protocol analysis, in which classes in the intervention arm were included in the analysis only if teachers had taught at least 70% (11/16) of the AFLY5 lessons. 33 As our unit of randomisation was schools and all pupils in any class will have been taught the same number of lessons, this means that whole classes (rather than selected children within intervention classes of schools) for which the teacher did not teach at least 70% of lessons were excluded from the per-protocol analyses. Some of the intervention schools had more than one year 5 class (the maximum was three classes per school). In the main per-protocol analyses presented here, exclusions were made on the basis of classes (i.e., if one out of three classes did not reach the threshold of 70% of lessons taught, only children from that class, and not the whole school, were excluded); repeating the analyses on the basis of whole schools did not materially alter the results (data are available from the authors on request). We assessed the number of lessons taught by reviewing the teacher-completed log, where possible, or by confirming these details with the teacher in person or by telephone. We had information on lessons taught for 29 of the 30 schools allocated to the intervention, including the school noted earlier that refused to do any part of the intervention. For the one school for which we were completely ignorant of how many lessons had been taught, we carried out analyses first assuming that they had taught at least 11 lessons and then repeated assuming that they had taught fewer than 11. The results were identical for these two alternatives.

We did additional analyses to assess whether or not the effect of the intervention on accelerometer-assessed outcomes differed by weekday or weekend day and whether or not the results were affected by implausible values. 33

We undertook a number of sensitivity analyses to examine whether or not our results were robust to any biases caused by missing data (see below).

Dealing with missing data

For the main analyses we used ITT. ITT requires all participants in a clinical trial to be included in the main analyses in the groups to which they were randomised. 51,52 This is straightforward if there is no loss to follow-up or missing data on some outcomes at follow-up among those who have been randomised, but is less straightforward where there is loss to follow-up/missing data. 52,53 A 4-point framework for dealing with missing outcome data has been proposed to deal with this issue and we were guided by that framework. 52,53 It emphasises the fact that all approaches [including complete case analysis (CCA), that is, including only those with observed outcome data] rely on assumptions that in any given situation may be more or less plausible but are always untestable. 52,53

Assumptions of missing data methods

Complete case analyses and several of the common methods for imputing/dealing with missing data assume that missing data are missing at random (MAR). The assumption underlying MAR is that having taken account of observed data in analyses, effects in those with missing data do not differ from those with no missing data. Another way of thinking about this is that the effect of a randomised intervention is the same in those with missing data as in those without missing data. Having similar proportions of participants with missing data in each arm of a trial is reassuring with respect to the MAR assumption being correct, but is not a guarantee, as the plausible reasons for missing data in each arm could be different but result in similar proportions with missing data.

In AFLY5, we minimised the extent of missing data through catch-up data collection, whereby for each participating school at each phase of data collection there was a day for main data collection, but some children may be absent from school on that day; therefore, for each school we had ‘catch-up’ days to obtain data on these children. As a result, the likely reasons for a child having missing follow-up data for all outcomes at one or both of the follow-up assessments are that the child moved school between data collection phases or the child was absent from school for a prolonged period or missed school frequently, such that they missed the main and catch-up data collection days. Missing one or more (but not all) of the specific measurements at follow-up could occur if the child did not give assent or, for the accelerometer-based outcomes, the child did not return the accelerometer or did not wear it for the required period of time. In the case of the AFLY5 RCT, MAR is plausible as randomisation was at the level of the schools, parental opt-out consent was ascertained at the start of the study and relevant for all data collection times, and it was implausible that the delivery of the intervention lessons and homework assignments in the intervention schools, or lack of these in the control groups, would affect the likelihood of a child being absent on days of data collection, declining assent for a particular measure or not returning the accelerometer or wearing it for the required time. Information from the local councils suggested that movement between schools was relatively low, but it is possible that children who moved may differ from those who did not on the basis of unobserved characteristics. Children who moved school might be from families who are relatively unorganised with children often moving school or they could be from families who move their child from state to private school in year 6 in order to attend private secondary school (in Bristol the proportion of children in private secondary school education is higher than than the average for the UK). The possibility that these types of missing data would bias our findings was assessed in sensitivity analyses.

Methods of dealing with missing data and sensitivity analyses

Examining potential mediation effects

Although we did aim to examine some mediators when we first submitted our proposal, these were unspecified and, therefore, we did not take them into account in our sample size calculation. The justification for undertaking these analyses was that we felt that exploring whether the intervention had an effect on mediators that are relevant to the intervention was important for fully understanding the process by which the intervention worked, or why it did not work. For example, we may have found that the intervention was effective and that this was in part mediated by the child’s knowledge, but not by self-efficacy. Or we may have found that the intervention did not work and also that it had no effect on any of the mediators, which would suggest either that it was poorly delivered or that it did not effectively work on the proximal characteristics that it was expected to work on. To balance the importance of looking at mediation with the fact that our original sample size calculation did not take account of this mediation analysis we considered these analyses to be exploratory.

These analyses were largely undertaken in accordance with the published analytical plan that had been approved by the Trial Steering Committee (TSC). 32,33 The analysis plan in relation to mediation was discussed with the TSC (meeting on 1 March 2015) and two changes to the plan for the mediation analyses were agreed. In the original analysis plan we specified that we would explore mediation in relation only to the primary outcomes. Following further discussion, the TSC felt that exploring mediation in relation to secondary outcomes was equally informative and that we should explore whether or not the mediators influenced any of the effects on the three secondary outcomes that were affected by the intervention. In the original analysis plan we indicated that we would take account of multiple testing when exploring the effect of the intervention on mediators and explore whether or not they mediated effects on secondary outcomes only if they passed this ‘multiplicity-corrected’ test. On further reflection, the TSC felt that this was inappropriate because for potential mediation the magnitude of the effect is more important than the p-value.

Mediation was assessed for the effect of the intervention on both primary and secondary outcomes. Mediation analysis assumes that the intervention influences the mediator(s) and through this influence the mediator produces its effect on the outcome(s). Therefore, the first stage in the mediation analyses was to examine the effect of the intervention on the mediators (hence in these analyses mediators are treated as outcomes in the regression analyses).

• First, we determined the effect of the intervention on each of the 10 measured mediators (see Box 2).

  • Each of these mediators was treated as a continuously measured variable, and in the first stage we explored the differences in mean scores of each mediator using the same ITT, as used in the analyses of outcomes secondary and sensitivity analyses.

  • In these analyses we included the following baseline and stratified covariables: age, sex, the baseline measure of each mediator, whether or not the school was involved in other health-promoting activities and school-level deprivation. (Note that knowledge was not assessed at baseline so there was no corresponding measure of this.)

• Second, we explored whether or not mediators explained the effect of the intervention on outcomes. This second stage was conducted only if (1) the intervention had been shown to affect one or more of the outcomes; and (2) the intervention had been shown to affect one or more of the mediators relevant to an outcome that the intervention had affected (first stage of mediation analyses described above).

  • If the two criteria above were fulfilled we used multilevel multivariable linear regression to test the effect of the intervention on outcomes. We then repeated that analysis with any relevant mediator added and compared the effect of the intervention on the outcome before and after adjustment for the mediator.

A relative reduction (change towards the null) of ≥ 10% of the initial effect of the intervention on outcome was considered to indicate some evidence of mediation. In other words, if the effect estimate with adjustment for baseline and stratifying variables moved towards the null by ≥ 10% with the addition of one or more mediating variables, we considered that mediation had occurred as a result of those variables.

Testing differences of effect by sex and by area deprivation

As noted in our detailed statistical analysis plan, this trial had good statistical power for determining main effects, but did not include sufficient participants to examine whether effects differed by participant or school characteristics. 33 However, it was agreed with the TSC that, given the interest in whether or not girls and boys differ in their response to school-based health promotion interventions and also whether or not the effect of such interventions is greater in schools from more affluent areas (where resources to effectively add to the curriculum are greater than in schools in poorer areas), we would present stratified results by these characteristics. The analytical approaches for these stratified analyses were similar to the main ITT analyses but:

1. We examined effects separately in girls and boys and present effect estimates and 95% confidence intervals (CIs) within each of these strata. We also tested for evidence that effects differed by sex by including an interaction term of intervention × sex

2. We examined effects separately in thirds of the IMD34 and present effect estimates and 95% CIs within each of these three strata. We also tested for evidence that effects differed by IMD thirds by including an interaction term of intervention × IMD.

Testing the effect of the Active for Life Year 5 intervention on outcomes 1 year after the end of the intervention

In our published analysis plan, we described examining trajectories of change with time from baseline through to the immediate and the later 12-month follow-up. To do that we planned to use a multilevel model that estimates a trajectory of the three repeat measurements within each individual, with random effects quantifying the estimated person-specific deviation from the study mean in terms of the intercept (baseline measurement) and rate of change (slope between baseline and the 12-month follow-up). We planned to fit an interaction term between the intervention and time since baseline so that we could explore whether or not there were differences between the two randomised groups in the change in outcome between baseline and immediate follow-up and the change in outcome between immediate follow-up and 12-month follow-up. However, when we attempted to run these models, they did not converge. This is likely to be because there were only three measurement occasions, meaning that the model did not have sufficient degrees of freedom. Models converged when only a random effect for the intercept (not the slope) was included, but we did not feel that this was an appropriate way of modelling the data. Therefore, we conducted analyses at a single time point for the immediate follow-up (i.e. we assessed the effect of the intervention on outcomes at the 12-month follow-up) and plotted a graph showing the effect of the intervention at immediate follow-up and again (using the same analysis method) 12 months later (following 12 months with no intervention), so that we could compare effects at the two time points. We used the same ITT multilevel (to account for clustering within schools) analysis approach for the 12-month follow-up as was used for the main analyses with the outcomes assessed immediately after the intervention.

For these later follow-up analyses we focused only on main effects with primary and secondary outcomes. We did not complete secondary per-protocol, mediation or stratified (by sex and school-based deprivation) analyses. This decision was made on the basis of the large number of analyses that were undertaken for the first effectiveness analyses (immediately after intervention) and the importance of not ‘overanalysing’ data. The aim of the analyses of 12-month outcomes was to determine whether or not any immediate effects were sustained and/or any new effects emerged.

Cost–consequences analyses

A cost–consequences approach was used for the economic evaluation. This is when the differences in costs and consequences between the two arms of the trial are given in tabular form, and there is no attempt to estimate a summary score to encapsulate all the costs and benefits [for example, the incremental cost-effectiveness ratio (ICER)]. Decision-makers are left to attribute their own weights to the relative importance of these costs. This approach was chosen because of the number of primary and secondary outcomes assessed in this trial and because it is not known what society would be willing to pay for a unit decrease or increase in the outcomes, such as children’s time spent in MVPA and sedentary behaviour and consumption of fruit and vegetables, that were studied in the AFLY5 intervention. It would therefore have been difficult to produce and interpret ICERs and cost-effectiveness acceptability curves showing the probability of the intervention being cost-effective given society’s willingness to pay for a unit decrease or increase in each of the outcomes evaluated. In addition, this approach is recommended by the UK National Institute for Health and Care Excellence when undertaking public health evaluations. 59 Costs were not discounted as they occurred within a 1-year period.

Primary analysis

The analyses from the combined provider and school perspective and the separate teacher perspective assumed that costs in the control schools would be zero. This is justified on the basis that the control schools neither received teacher training nor delivered the intervention and, therefore, no costs of the intervention can be attributable to these schools. Control schools may have had pre-existing activities that were in addition to the ‘legal’ school requirements and/or introduced new activities during the study period that would have costs and potentially affect our outcomes. However, the randomisation process stratified on pre-existing activities (so these would be equal between the schools) and, as this was adequately concealed, as with all RCTs, it is reasonable to assume that there are no differences between the randomised groups other than the intervention under study. Resources used for global tasks and school-specific tasks were divided by the number of students in the intervention (n = 1064) to obtain a per-pupil amount of resource use. The costs per pupil were estimated by multiplying the units of the resource use for each category by their unit costs. Where only a cost had been given, such as for the trainers’ fee, this was divided by the number of pupils to obtain the cost per pupil. The total cost per pupil was calculated as the sum of the cost of the resource use items. Similarly, the overall cost of the intervention was calculated by summing the intervention costs within each category that had been calculated by multiplying the total resource use by its unit cost. In addition, a total cost of the intervention per school was calculated by multiplying the costs per pupil by the number of pupils in each school (for global resources) and adding this to the specific costs for each school. The opportunity cost of implementing AFLY5 in schools was calculated as the cost difference between AFLY5 and the usual lesson that would be taught in that school, with the cost for the AFLY5 lesson estimated as the time spent delivering the lesson multiplied by the salary of the teacher who taught the AFLY5 lesson; similarly, for the usual lesson but using information from the logs as to who would have taught each lesson. Delivery times were assumed to be the same for both lessons.

Secondary analysis

Resources were grouped into three categories: parental time, household expenditure on intervention-related costs and NHS resource use. Costs were estimated by multiplying the units of resource use by their respective unit costs. The total cost for each subject for each category was calculated as the sum of the costs of the resource use items. The mean resource use and cost was calculated by arm of trial.

Ordinary least squares regression, adjusted for the same baseline variables that were used in the effectiveness analyses (described above), was used to estimate the mean incremental costs of parental time, household expenditure on food and children’s out-of-school activities and NHS resource use. Additionally, the first two groups were combined to estimate the total incremental costs from a parental perspective and further adjustment for the number of household members was undertaken when estimating the costs related to the expenditure on food. Non-parametric bootstrapping was employed to obtain bias-corrected 95% CIs. 60

Missing data and sensitivity analyses

Participant numbers

Figure 1 shows the trial profile. The number of pupils in each class/school year was larger than we had anticipated and, therefore, having recruited 60 schools, the number of pupils included was greater than the required 1500. Of the 2242 potentially eligible students in the 60 participating schools, 10 left the school prior to randomisation and baseline data collection and the parents or carers of 11 pupils did not consent to their child’s participation in any aspect of the study. All other children (n = 2221; 1064 in the schools that were randomised to intervention and 1157 in the control schools), irrespective of whether or not they had all measurements recorded, were included in the trial and used as denominators for baseline comparisons between the two randomised groups.

FIGURE 1.

Trial profile. N refers to number of schools; n is the number of participants (school pupils). None of the schools withdrew from the study and so all of the randomised units are present at baseline and follow-up. The percentage given in brackets for the proportion of children with data at both baseline and follow-up is of the total number of children who were pupils in the randomised schools at baseline. Not all of the pupils with a follow-up measure necessarily have a baseline measure (or vice versa), because of different pupils being absent at both main and catch-up assessments at each time point and because of pupils leaving or moving between schools. In all analyses those who were randomised were analysed in the group (intervention or control) to which they were randomised.

These numbers include a small number of participants (n = 65) whose parent/caregiver refused consent for one or more measurements (most commonly weight and occasionally waist). Up to two catch-up visits were made to schools to obtain data on any pupils who were absent on the day of data collection for their school, but inevitably some pupils will have been absent on the day of both the main and catch-up visits to their school at either baseline or follow-up. No child refused assent to complete the questionnaires, but in a small number of cases it was not possible to code the dietary data because we could not read what was written or identify a food from a brand name. A small number of pupils did not assent to waist or weight measurements, and the proportion of pupils with valid accelerometer data was influenced by the requirement that they had 3 days, each with at least 8 hours of valid wear time. 33 In total, at both baseline and follow-up, between 82% and 96% of participants had data on diet outcomes, BMI and WC, and approximately 60% had valid accelerometer data (see Figure 1). With the exception of valid accelerometer data the number of children included in the main analyses (1825 to 2121) was greater than the 1275 that our sample size calculations showed were required for the main analyses. For accelerometer-based measurements, data were available for 1252 children for the main analyses, 23 (0.02%) fewer than the estimated requirement. 33 Proportions with valid data for each measure were similar at both baseline and follow-up, and in intervention and control schools (see Figure 1).

Baseline characteristics

Baseline characteristics, including for accelerometer return and wear time, were similar in intervention and control schools with the exception of reported screen viewing time on Saturdays, which was 15 minutes greater per day in participants from the control schools than in the intervention schools (Table 3).

Main intention-to-treat analyses

In the main ITT analysis with adjustment for baseline variables, none of the three primary outcomes differed between children in schools allocated to the AFLY5 intervention and those allocated to control schools (Table 4). The intervention was effective on three out of nine of the secondary outcomes after taking account of multiple testing in these analyses: pupils’ self-reported time spent on screen viewing at the weekend (Saturday) and self-reported consumption of snacks and of high-energy drinks were lower in intervention schools than in control schools (see Table 4). There was no strong evidence that the intervention affected the other secondary outcomes in these analyses, especially after taking into account multiple testing.

Per-protocol analyses

Table 5 shows the results of the per-protocol analyses for primary and secondary outcomes. In these analyses children from 16 classes in 12 out of the 30 intervention schools were excluded because their teacher had delivered fewer than 70% of the lessons. The results of the per-protocol analyses were broadly consistent with the ITT analyses results, with no evidence of effect on the three primary outcomes or most of the secondary outcomes. As with the ITT analyses, there was evidence of beneficial effect on self-reported screen viewing on Saturdays and consumption of high-energy drinks. The point estimate for the reduction in self-reported consumption of snacks was similar to that seen in the ITT analysis but the strength of evidence was marginal, particularly after multiple testing was taken into account.

Sensitivity analyses

The sensitivity analyses that we undertook to explore assumptions about missing data produced results that were consistent with the main analyses (Tables 6–9). When we looked separately at time spent in MVPA and time spent in sedentary behaviour by weekday and weekend, the results were consistent with each other and with the main results (both values for the difference between the two estimates p > 0.2; Table 10).

Analyses stratified by sex and school area level of deprivation

Table 11 shows the results presented separately for girls and boys. The effect estimates for the accelerometer-measured primary outcomes looked different in girls and boys, with the effect of the intervention on time spent in MVPA being very close to the null in girls but decreased (opposite to the effect expected) in boys and on time spent in sedentary behaviour decreased in girls but increased in boys (again the opposite of what we expected). However, as noted in the section Testing differences of effect by sex and by area deprivation, these stratified analyses are exploratory and we did not design the study with an aim (or sufficient numbers of participants) to be able to detect sex differences. Furthermore, there was no strong statistical evidence that these effects did differ by sex (both p_interaction = 0.4). For all other outcomes, with the exception of consumption of snacks, the effects were broadly similar in girls and boys. The intervention appeared to have a stronger effect on reducing the consumption of snacks in girls than in boys, with statistical evidence that this difference was unlikely to be caused by chance (p_interaction = 0.006).

Table 12 shows the results according to school area level of deprivation (divided into thirds). For some outcomes, point estimates did seem to suggest that the intervention had different effects within thirds of school area deprivation, but there was no consistency across the thirds for differences in most outcomes (that is no monotonic difference with each category increase in deprivation) and no statistical support for differences by the IMD. 34 The two possible exceptions were time spent in sedentary behaviour assessed by accelerometer and self-reported screen viewing on weekdays. For time spent in sedentary behaviour assessed by accelerometer the intervention appeared to result in greater time spent in sedentary behaviour (opposite to what was expected) in pupils from schools in the middle deprivation third, with some statistical support for this being different from that seen in the lowest third, but the effect was similar in the two extremes (lowest and highest thirds of deprivation). For self-reported screen viewing on weekdays there did appear to be a monotonic increase in effect (reduced time spent screen viewing) as deprivation category increased, but there was no statistical support for differences between categories and this pattern was not seen for reported screen viewing at weekends.

Mediation effects on outcomes

Following the intervention, there was evidence that 3 of the 10 potential mediators (fruit and vegetable self-efficacy, maternal limitation of sedentary behaviour and knowledge) were higher in the intervention group than in the control group (Table 14).

Table 15 shows the main effect of the intervention on the three secondary outcomes found to be affected by the intervention, both before and after adjustment for potential mediators. Adjustment for maternal limitation of sedentary behaviour and child knowledge attenuated the effect of the intervention on time spent screen viewing at the weekend by 22%. There was no notable change in the effect of the intervention on consumption of snacks or high-energy drinks following adjustment for mediators.

Primary analyses

Table 18 shows a breakdown of the resource use and costs incurred during the training days and the delivery of the AFLY5 lessons from the combined school and provider perspective. The main cost drivers for the intervention were the claims by the schools for replacement teachers needed to cover the teachers’ attendance at the training days (£5.00 per pupil, £5095.00 in total); the trainers’ fees (£2.00 per pupil, £2166.00 in total); the time spent by the research staff organising and attending the training days (£2.00 per pupil, £2492.00 in total); and the printing costs of the materials for the AFLY5 lessons and the homework (£6.00 per pupil, £6694.00 in total). We estimated the opportunity cost of implementing AFLY5 in schools (i.e. the cost of teaching AFLY5 minus the cost of teaching the usual curriculum based on data from schools in the control arm) to be £0.05 per pupil. The costs varied by school, ranging from approximately £13.00 to £36.00 per pupil. The variations between the schools were driven by the costs of the teachers’ attendance at the training days.

The cost–consequence analysis showed that, for the three secondary outcomes that were affected by the intervention, it cost £18 per child (£18,944 in total) to reduce self-reported time spent on screen viewing at the weekend by 20.86 minutes, self-reported consumption of snacks by 0.22 snacks per day and self-reported consumption of high-energy drinks by 0.26 servings per day.

From the teachers’ perspective, teachers spent, on average, more time travelling to the training day venue than they usually spent travelling to school; this equated to an additional 0.68 minutes’ travel time per pupil, generating an extra cost of £0.19 per child (£206 in total).

Sensitivity analyses

Table 19 shows the results of sensitivity analyses for the primary (perspective of combined provider and school). These results suggest that the main analyses are robust to our assumptions regarding missing data.

Secondary analyses

Only 626 (28%) parental questionnaires [278 (13%) in the intervention and 348 (16%) in the control arm] were returned. In the returned questionnaires the item non-response varied between 0% and 6%. Table 20 shows the results of the cost differences between the intervention and control arms from the perspective of the parents and the NHS, using a CCA and including maximum numbers for each item. After adjustment for the child’s age, sex, area deprivation level for the school and the engagement of the school in health-promoting activities (over and above AFLY5), all of the parental and NHS costs appeared greater in those from schools randomised to intervention compared with those randomised to control. However, all of these results were imprecisely estimated with wide CIs consistent with no difference. With the exception of the NHS costs, adjustment for covariables produced notable differences in comparison with the unadjusted results (although results were all consistent with the null hypothesis, whether or not adjusted). The IMD34 is the covariable that has most effect, and it contributed to change in all of the results that differed with adjustment. Other covariables that contributed to some of the change in results with adjustment were age (total cost from a parental perspective), child’s sex (costs of parental time) and the level of school involvement in healthy activities (expenditure in out-of-school activities). The greater NHS cost is likely to be exaggerated as it was influenced by one participant who was reported to have had surgery for a condition that is unlikely to have been influenced by the intervention (adrenal surgery); this was despite the question that specifically asked about treatments that were related to physical activity injury. Other variables in the cost–consequence analyses were largely normally distributed with no evidence of effect by ‘outliers’.

Sampling and recruitment

The process evaluation for AFLY5 involved collecting data from five groups, with participants in these groups being selected from both intervention and control schools:

• year 5 teaching staff

• head teachers

• parents of year 5 pupils

• year 5 pupils (aged 9–10 years): as the topics covered in the focus groups with these pupils did not relate to key intervention issues these data will be reported separately from this report.

The intervention teaching staff were chosen to provide direct feedback on experiences of delivering the intervention, intervention fidelity and how it could be improved. The other four groups included participants from intervention and control schools to obtain an overview of how both AFLY5 and normal diet, physical activity and healthy living education lessons were delivered and received within the school setting.

Selection of intervention and control schools for the process evaluation

All intervention arm AFLY5 year 5 teaching staff were invited to participate in a one-to-one in-depth interview with the aim that we would continue to interview all who agreed to participate until we had clearly reached saturation in the analysis. For data collection from control arm year 5 teachers, control and intervention arm head teachers, pupils and parents, we planned to sample from 12 purposively selected schools on a first come, first served basis until saturation was reached. These schools were selected to represent different levels of area deprivation (high, medium and low) and different Ofsted (Office for Standards in Education, Children’s Services and Skills) scores at the time of entry into the study (July 2011) (high vs. low). 63 Of these, six would be intervention and six would be control schools. Participation in the process evaluation was not compulsory and, if a school declined, this did not affect its participation in the main part of the study. Initially, if a school declined, a school with a similar profile was approached to replace it. In practice, we were unable to reach our recruitment target this way and eventually opened up the invitation to all AFLY5 schools in order to recruit 12 schools to the process evaluation.

For all types of data collection the final number recruited was guided by the point at which saturation was reached (i.e. the point at which no new data were emerging). If this had clearly been reached before the anticipated number of participants had been interviewed, or focus groups conducted, then the number would be lower. If it had not been reached by the end of the anticipated numbers we would have attempted to recruit more participants. The anticipated numbers are based on our previous experience with similar research and from published research in similar areas. 31,65–67

The numbers of actual schools/participants that participated in interviews/focus groups, along with the anticipated numbers, are indicated in Table 22.

Recruitment of teaching staff

We aimed to recruit as many as possible of the teaching staff who delivered the intervention to allow us to obtain very detailed information about their delivery of the AFLY5 lessons. This included whether they talked about the AFLY5 lessons when asked general questions about teaching health promotion or about the study in general, and also their responses to prompts about the AFLY5 teaching materials, training days, etc. They were invited by e-mail to take part in an interview, with the aim of interviewing one teacher per school. The information sheet and consent form were attached to the e-mail. If they had not replied after 2 weeks, a follow-up call was made or e-mail sent.

We also recruited eight control arm AFLY5 teaching staff members, which included at least one from each of the six process evaluation control schools (selected as described above). We anticipated that with teachers from just these six control arm AFLY5 schools we would obtain sufficient information to understand if the current syllabus (without AFLY5) covered some or a substantial part of the topics covered in the AFLY5 lessons.

Recruitment of head teachers

Intervention and control arm head teachers were invited from the 12 schools (six intervention and six control) as listed above. They were contacted via telephone or e-mail, with a follow-up telephone call around 2 weeks later. Once the head teacher indicated that he or she was willing for the school to participate, a convenient time was arranged for the interview to take place.

Recruitment of parents or guardians and children

All parents or guardians of year 5 children in the 12 selected schools (six intervention and six control) were sent a letter informing them of the parental interviews and child focus groups. In the case of the parental interviews, an information sheet and reply form were included informing the parents about the interviews and asking if they wanted to take part. Parents who replied to say they were interested in participating were contacted by telephone to arrange a convenient time for either a telephone or a face-to-face interview. If they opted for a telephone interview, parents were sent a consent form through the post. If they opted for a face-to-face interview, a consent form was completed in person before the interview started. Consent for parent interviews was on an opt-in basis.

In the same information pack parents were also sent an information sheet relating to the children’s focus group. Parental consent for the children’s participation in the focus group was on an opt-out basis, meaning any child could potentially be approached to participate in the focus group unless their parent had opted them out. Pupils from the selected 12 schools whose parents had not opted their child out of the focus group research were purposively sampled and invited to participate in the focus groups. Pupils were purposively sampled to ensure a balance of sexes, those with high and low levels of reported self-efficacy for physical activity and high and low levels of parental support for physical activity. Information on these characteristics was collected from all pupils participating in AFLY5 as part of the baseline questionnaire. The children were given an information sheet before deciding whether or not to take part in the focus groups. The children were then asked to give written assent to participate in the focus group on the day prior to starting.

School audits, baseline and follow-up

Audits were attempted in all schools (both intervention and control) once per academic year during baseline, intervention and follow-up periods of the study (June 2011–July 2013). They were used to assess physical activity provision, including the number of scheduled PE lessons per week and time allocated for unsupervised outdoor play, school physical activity and nutrition policy (active travel, break-time play, packed lunch policy, etc.) and number and type of school clubs provided. In 2011, these audits were carried out over the telephone by a member of the AFLY5 team using a pro forma. In 2012 and 2013, the audits were filled out by teachers in schools during the measurement process.

Teacher training observation

Observations took place during the teacher training days for the group of teachers randomised to AFLY5. The aim was to observe how effectively the messages of the training day were conveyed by the trainers, and how these messages were received by the teachers. Observations included noting any questions and issues raised by the teachers. Consent to observe the training day was sought from the trainers and the teachers. A copy of the observation guide is in Appendix 3.

Teacher training evaluation forms

The AFLY5 training day was assessed using an evaluation form, which was completed by the teachers who had received the training. A copy of the evaluation form is in Appendix 4. The aim was to ask teachers about their engagement with, and understanding of, the training provided and if they felt they knew enough or required more information.

Lesson observations

We asked teachers for permission to observe the teaching of the AFLY5′ lessons. There are 16 lessons in the syllabus and we wanted to observe each of these lessons twice, which would have allowed us to observe in each school at least once. There were also two pilot observations. The observations did not assess the general teaching ability of staff and no feedback was provided to anyone at the school other than the teacher, and only when specifically requested. This was made very clear to the school (both in the information letter and in person). No recording equipment was used (visual or audio) during the observations, only hand-written notes were taken. Observations took place only in the intervention schools. A copy of the observation guide is in Appendix 5.

Teacher log

We asked all year 5 teachers in intervention schools to complete a log of session delivery during the intervention period (September 2011–June 2012) (n = 48). The log asked teachers to report the date that the sessions were delivered, whether homework assignments were set and returned and any problems or positive comments about the lesson. A copy of the teacher log is given in Appendix 6.

Interviews

Intervention and control arm AFLY5 teaching staff were interviewed to obtain information about their views on what contributes to healthy lifestyles in general and for children specifically, on teaching health promotion, whether they believe school- or education-based health promotion is effective in changing children’s behaviours and what teaching activities they are involved in that are concerned with health promotion (and specifically promoting healthy diet and physical activity). They were also asked how they felt about delivering these lessons and their likely effectiveness. Information was sought on the teachers’ views of the children wearing the accelerometers and any barriers they perceived to them doing this for the required time and returning the accelerometers. There were also some questions about the intervention for the teaching staff in intervention schools, mainly related to how they found the experience. A copy of the topic guide is provided in Appendix 7. The interviews were held post intervention either at school or at a location chosen by the participant, at a time convenient for them. The interviews took up to 60 minutes.

Intervention and control arm AFLY5 head teachers were interviewed to obtain information about their views on healthy lifestyles in general and for children specifically, on teaching health promotion in schools and if they believe school- or education-based health promotion is effective in changing children’s behaviours. Information on what health-promoting activities (in any area of health promotion) were taking place in their school at the time and their views on schools participating in research were also sought. The head teachers from schools that had taken part in AFLY5 were also asked about their views of the school taking part in this study. A copy of the topic guide is provided in Appendix 8. The interviews were held post intervention at school, during the school day, and took up to 60 minutes.

Intervention and control arm AFLY5 parents were interviewed to obtain information about their knowledge of healthy lifestyles, whether they are aware of their children learning about healthy lifestyles in school, their involvement in any homework assignments concerned with healthy lifestyles or health promotion and what they remembered about their children being measured in school. The topic guide can be seen in Appendix 9. The interviews were held post intervention either face to face, at their child’s school or the university, or over the telephone, between 8 a.m. and 9 p.m. The interviews took up to 60 minutes.

Focus groups

Focus groups with intervention and control arm AFLY5 pupils were conducted to obtain information about their views on what things make them healthy, how easy or hard they find it to do things that make/keep them healthy. The topic guide is presented in Appendix 10. These questions were asked to provide broader contextual data on the children’s understanding of issues related to diet and physical activity and the teaching of these subjects in schools, irrespective of whether students were in the intervention or control arm. These focus groups do not provide information in relation to the three research aims of the process evaluation and are therefore not reported in this monograph. Data from them may be used in separate future publications.

Types of data

Training and preparation

A total of 44 teachers from 29 schools delivered the intervention. Overall, 43 participants attended the training; 42 of these were teachers and one was a LSA. Data from the teacher training evaluation questionnaires indicated that 42 of the 43 (98%) teachers/teaching assistants who attended training agreed or strongly agreed with the statements ‘I feel confident that I can teach the nutrition sessions as per the lesson plans’ and ‘I feel confident that I can teach the physical activity sessions as per the lesson plans’. Teacher interview data indicated that, on the whole, they appreciated having the opportunity to work through the programme during the training and, in particular, the opportunity to receive instruction on the physical activity component:

I think the training we got when we came for the Active for Life was really, really helpful ‘cause it certainly pointed out a few things to us . . . about like how easy it was to run different activities.

School 15, teacher 2 interview

. . . I really liked the physical exercise training. And the activities that were supplied. Thought that was really good and gave me lots of ideas. I still use them, even though I’ve moved on to a different class . . . I thought the lady who ran the course was quite inspirational.

School 50, teacher 3 interview

Using data reported by the 39 teachers who noted at least one lesson preparation time in the teacher log, the median length of lesson preparation time calculated across lessons delivered was 10 minutes (IQR 10–20 minutes). Also from the teacher logs, 38 teachers noted at least once whether the lesson preparation time was more than, less than, or the same as usual (n = 450 lessons with data for this question). More preparation than usual was required for 46% of lessons, less time than usual for 15% and the same preparation as usual for 38% of lessons. During the interviews several teachers indicated negative feelings about the extra preparation time required and noted that it was often required for the physical activity sessions:

I mean once the resources were made it was fine . . . I don’t spend a lot of time preparing PE lessons. So I mean it might have been 15 minutes of reading through notes whilst it would have taken me 5 minutes normally so we are not talking a lot of time but . . . if it takes you 15 minutes and you are teaching it for 45 minutes how much, how long does each lesson take to prepare?

School 10, teacher 2 interview

Dose

Of the 44 teacher logs, 41 were completed and returned to the study team. The remaining three teachers were contacted by telephone and asked to provide only the number of lessons delivered and (if possible) the dates of delivery. Data from these 44 teachers showed that the mean number of lessons delivered was 12.3 out of all 16 lessons (SD 3.7 lessons, median 13.5 lessons, range 1–16 lessons) which equates to 77% of the intervention. Of the 41 teachers that returned teacher logs and indicated that they had delivered some of the intervention, all of them delivered lesson 1, but delivery declined over the intervention period such that only 46% delivered lesson 16. Seven teachers out of 41 (17%) delivered all 16 of the lessons. The data from the teacher logs and interviews revealed that by far the most commonly mentioned reason for not delivering all the lessons was lack of time to fit all the lessons into an already full curriculum. This is explored further using data from the teacher interviews (see Chapter 4, Amendments).

The mean number of homework assignments delivered, calculated from data given in the teacher logs, was 6.2 out of a total of 10 (SD 2.6, range 2–10) equating to 62%. Teachers who did not hand out the homework assignments stated in both the teacher logs and interviews that they had to prioritise core skills homework above those from the intervention:

All our homework is literacy + numeracy at the moment, building up to end of year tests.

School 51, teacher 1, lesson 11 – written extract from teacher’s log

In interviews with the parents, 5 out of 13 stated with certainty that their children had received AFLY5 lessons, and could remember homework items that were definitely part of the AFLY5 programme. Other parents were unsure about whether or not their children had received the lessons and homework assignments.

Reach

The reach, or percentage of children receiving all the lessons taught, calculated from teacher log data, was 95%. The time frame over which the lessons were delivered was a median of 17.7 weeks (IQR 9.1–23.3 weeks). The teacher logs indicated that there were two main patterns of delivery: (1) a regular dose, fairly evenly spread out; and (2) a varied dose that changed in response to lack of time, curriculum or engagement issues. As one teacher explained during an interview, they delivered AFLY5 in a variable dose because of the length of time required to deliver the AFLY5 programme and the potential for diminishing engagement over time:

. . . it went over a term or well over one term, normally every term’s like a fresh start, something completely different . . . so they need that chopping and changing ‘cause otherwise . . . they’d hate it and that’s with anything, that’s not just with Active for Life.

School 56, teacher 2 interview

Teaching staff

Based on the information reported in the teacher logs, among the 44 classes, all AFLY5 lessons over the full period were delivered by the same teacher in 51%, by two teachers in 20%, by three teachers in 22% and by four or more teachers in the remaining 7%. Out of the 80 teachers or teaching assistants listed as delivering at least one AFLY5 lesson, 49% had received the training. Of the 494 lessons for which data on who delivered the lesson were available, 386 (78%) were delivered by someone who had received the training. Of the 108 sessions recorded as taken by staff not trained in the intervention, 25 (23%) were covered by a main class teacher, 20 (19%) were taken by Preparation, Planning and Assessment (PPA) cover staff, who enabled teachers to work away from the classroom, 13 (12%) were delivered by supply teachers, nine (8%) by student teachers, nine (8%) by teachers whose status was not recorded, one (1%) by a LSA and the remaining 31 sessions (29%) were taken by people whose status was not recorded. AFLY5 lessons were considered suitable for handing over, since the lesson plan, worksheets and homework assignments were prepared. As data from the teacher logs and interviews revealed, the lessons that were handed over to these staff members, who were not trained in the use of AFLY5, were often the physical activity or PE lessons:

I only taught a few lessons my PPA cover took the majority . . . He did more the physical the activities, because he was taking PE.

School 45, teacher 1, interview

Amendments

The written materials provided no guidance about amending or adapting the lessons, but teachers were told at the teacher training sessions that they should teach the lessons in the order that they were listed and that they could amend content as long as the message and learning outcomes remained the same. Observations of the teacher training sessions indicated that teachers were already considering, at that stage, how the lessons could be adapted.

As they sit back down the teachers discuss how they might need to adapt a lesson for their own classes. One teacher is heard to say: ‘My kids can’t read so those work cards won’t work’.

Observation notes, teacher training session on ‘Physical Activity games’ held on 27/09/2011

The participants (teachers) engage well with trainer 4, making comments or asking questions throughout the session about how particular activities might work in their classes or how they might adapt the games.

Observation notes, teacher training session on ‘A Safe Workout’ held on 03/10/2011

Data from 39 teacher logs, when cross-referenced with 30 lesson observations, revealed that whether or not the lesson was amended was recorded for a total of 468 sessions, and that 28% were amended. A majority (89%) of the teachers amended the resources or lesson content on at least one occasion, and each of the 16 lessons was amended by at least one of the teachers. Comparisons between lesson observations and teacher intervention delivery logs revealed that some teachers did not record amendments that were noted during the lesson observation. On 20 occasions the teacher stated that they had not amended the session, but in nine (45%) of these sessions the observer indicated that amendments had been made, suggesting that teachers may have under-reported amending the lessons or had a different understanding of what constituted an amendment.

During interviews with 20 teachers from intervention schools (nine of whom were from schools included in the process evaluation), those who reported amending lessons said that they did so because they felt that the lessons or resource materials did not fully meet their needs. The reasons for their adaptations fell into four main categories: adjusting length of lessons to suit the overall ability level of a particular class; a need to differentiate for differing pupil ability; conversion for use with new technology; and making the lessons more appealing to pupils to ensure their engagement.

Teachers’ views of the intervention

Although the quantitative fidelity of implementation data indicate that the AFLY5 intervention was well implemented, the interviews and teacher logs revealed a mixed view of the intervention. Teachers often noted the lack of time that they had to fit the lessons into an already full curriculum. This reasoning allowed teachers to present an acceptable ‘public’ explanation for not always implementing the intervention in full, which pointed to a structural constraint and, thus, did not involve them in overt criticism of the intervention programme. There was a sense, however, both during the interviews and in the analysis of the transcripts, that sometimes lack of time really meant lack of enthusiasm to make time or that lessons were to be fitted in only when there was extra time. Teachers were not wholly negative or positive about the intervention; the vast majority of responses were mixed:

So if anything this year we sort of almost missed it in a way because it was quite good at sort of, you know, filling, when we had little bits of time, pockets of time, we could, we could squeeze it in.

School 50, teacher 1a interview

This did not mean, however, that teachers disliked the overall purpose of AFLY5; on the contrary, they often mentioned how the messages behind the lessons were laudable, but that there were presentational issues. As these teachers explained:

It’s an amazing initiative, I think it was really, really important but it was just a huge amount to get through.

School 56, teacher 2 interview

So we did, a lot of the ideas were very good. But I just felt that the whole programme needs updating.

School 51, teacher 2 interview

Problems or concerns with the resources provided as part of AFLY5 were mentioned by many of the teachers, either for not being suited to their class, as this teacher explains:

Yeah I didn’t use any of your worksheets, I think I adapted every one of your worksheets.

School 10, teacher 1 interview

or for being rather old-fashioned when compared with other available resources:

I would suggest a DVD [digital versatile disc] or website resource to support the learning . . . Although good, the resource does seem unambitious and rather old-fashioned.

School 46, teacher 1, written extract from teacher’s log

The fact that teachers felt they had to alter the materials and that guidance and training on differentiation for ability was not provided as part of AFLY5 meant that there was a good deal of preparation for some teachers and this could also have contributed to the narrative regarding their lack of time. The results presented earlier revealed that 46% of teachers felt that, on average, they needed more time to prepare the AFLY5 lessons than regular lessons. This is perhaps not surprising given that these were completely new lessons. One limitation of this evaluation was that we did not determine how the preparation time for these lessons compared with that for any other completely new lesson. The trend towards more preparation time for PE lessons than nutrition lessons, for some of the teachers, could also reflect a general lack of enthusiasm for PE among some of the teachers. As this teacher reveals, this meant that, when they were running out of time, PE components were often dropped:

And I have to admit if there are any bits that I skipped it was the PE bits because we were doing PE anyway, but those required more preparation for me than a normal PE lesson.

School 10, teacher 1 interview

This could be seen as part of a wider issue relating to the lack of training and lack of confidence in delivering PE experienced by some primary school teachers. As this teacher explains when describing why the AFLY5 training was so helpful:

I am fairly keen on sport and PE in general but perhaps not the most confident in being able to teach it to children and stuff. So sort of taking it on board and being positive about it and seeing a sequence of lessons come about from it was actually very, very good.

School 36, teacher 1 interview

Some schools have found that one way to address this problem was to employ dedicated staff responsible for delivering PE lessons across the school years. Teachers in some schools handed over all their PE lessons to these staff and AFLY5 PE lessons were no exception. Again there was a tendency among some teachers to hand over the PE lessons in particular:

I mean the handbook is quite straightforward and he is a bit of a sports, more of a sports expert so he brought his sports expertise to it and what he tended to do was, he’d do the Active for Life lesson and then he’d finish it up with a game or something so they actually had sort of like extra PE.

School 15, teacher 2 interview

Main trial design

The study design was carefully developed to take account of known sources of bias in other RCTs in this area. A protocol was published before recruitment started and a detailed analysis plan (to which we have adhered) was written prior to any access to the study data. We developed an intervention according to guidelines for complex interventions (see Rationale for the Active For Life Year 5 intervention)19,20 and we have shown that it was both feasible to deliver and promising in our pilot RCT. 29 Our sample size calculation, which took account of the likely degree of clustering from our pilot and feasibility studies and the number of outcomes that we planned to assess, indicated that we required a total of 1500 participants from 60 schools to be randomised and 1275 included in the primary analyses. 32,33 For all outcomes, except those related to accelerometer data, we achieved considerably higher numbers than this. The number included in the main analyses for accelerometer-based data was very slightly lower than this at both follow-up assessments (n = 1252 at immediate follow-up and n = 1066 at long-term follow-up). Sample size calculations are an approximation of the numbers required, and we doubt that such a small difference will have had a major impact on our conclusions. Participants were instructed to remove the accelerometer during swimming and contact sports; waist-worn accelerometers provide poor assessments of cycling and, as such, it may be the case that some activity in which the participants engaged was not captured. These issues would, however, apply equally to both intervention and control arms and thus should not have affected any assessment of intervention effectiveness. Consistent with widely adopted procedures, we applied an 8 hour per day minimum wear time criterion for our accelerometer data. 33 This may have meant that some participants who were active but wore the monitor for < 8 hours were excluded. We have previously shown that girls, children reporting that their parents restricted their sedentary behaviour and those from schools with larger year group sizes were more likely to wear accelerometers for the required time. 80 Those from more deprived schools and who reported more weekday activity were less likely to meet these requirements. However, the same criteria for how to wear the accelerometers were applied to both intervention arms. Furthermore, wear time was similar in children in intervention and control schools, and in sensitivity analyses at both the immediate and long-term follow-up assessments, using different approaches to deal with missing data, the results were essentially the same as in the main analysis. One school refused to deliver any of the intervention and others did not deliver all of the lessons. However, the per-protocol analyses, which do not differ from the main ITT analysis for either the immediate or the long-term follow-up, show that this does not explain the results.

At the time of applying for funding for the study we did not anticipate assessing effects of the intervention on number of potential mediators that we ultimately assessed. However, these are plausible mediators for our intervention and all have been assessed using questionnaires that have been developed and validated for use in children of the age of those in our trial. 45–50 Given that the original study design did not anticipate all of these analyses, we did not take account of these mediators in our sample size calculation. However, our effect estimates are precisely estimated with narrow 95% CIs, suggesting that we have reasonable power to detect effects if they were present. All of the mediators were reported by the children; thus, in relation to parental mediators, our analyses represent the effect of the intervention on the children’s perception of these. Such perceptions might differ from what the parents actually did, but we would argue that, for both the extent to which the parents provided logistic support and the role modelling of healthy behaviours, the child’s perception is important. For example, parents might undertake most of their physical activity during the day when the child is at school but that is not likely a useful model for the child.

We do not have information on schools that were invited to participate but declined, and this study was undertaken in the south-west of England. Although the participating schools included a range of levels of socioeconomic deprivation, class sizes and both rural and urban settings, we cannot assume that these results are generalisable to all primary school settings.

One of the reviewers of this report expressed concerns that the BMI and WC variables were skewed, both in their original units and when presented as z-scores, and suggested that we should have used lambda-mu-sigma (LMS) or other appropriate methods to deal with the skew of these variables. We can confirm that the residuals in the regression models for all analyses were normally distributed and our conclusions would not be changed by using LMS or other transformations, which might make the results less easy to interpret.