Parts of this chapter have been reproduced with permission from Willis et al.1 and Jago et al.2 These are Open Access articles distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.
The main objective of this trial was to assess whether the PLAN-A programme was effective and cost-effective at increasing objectively measured physical activity among Year 9 girls (i.e. those aged 13–14 years). ITT analysis found that there was no intervention effect on the primary outcome of weekday MVPA minutes at T1, with weak evidence of a negative treatment effect in the intervention arm (–2.84 minutes, 95% CI –5.94 to 0.25; p = 0.071). Further adjustment for variables imbalanced at baseline (i.e. free school meals, travel to and from school, and IMD score) slightly changed the magnitude of this difference (–3.57 minutes, 95% CI –6.75 to –0.39; p = 0.028). These results show that there was no evidence that the PLAN-A intervention was effective in increasing adolescent girls’ physical activity. These findings contrasted with the feasibility trial outcomes, for which evidence of promise to increase MVPA at 12 months’ follow-up was observed (a 6.1-minute difference in weekday MVPA favouring the intervention arm over the control arm).51 Because of the relatively small size of the feasibility trial, this definitive trial was conducted to determine whether or not this promise would hold true at scale and therefore merit widespread dissemination. There were also no between-arm differences in the secondary outcomes of weekend MVPA and weekday and weekend sedentary time. These findings are comparable to those in the feasibility trial,52 except for weekday sedentary time, for which, at the 12-month follow-up in the feasibility trial, girls in the intervention arm were sedentary for 23 minutes fewer per weekday than girls in the control arm. The data reported here highlight that dissemination of the PLAN-A programme would not be beneficial to the physical activity levels of adolescent girls.
An embedded process evaluation was conducted during and after the implementation of PLAN-A to provide insight into the fidelity, quality and receipt of the intervention. All 10 schools randomised to the intervention arm received all 3 days of the intervention (i.e. the peer supporter training). However, school 10 received a reduced dose on the top-up training day, which was attributed to a request from the school for the peer supporters to leave early. This school also experienced delivery issues on day 1, which the trainers felt had a significant effect on the engagement of the peer supporters. Training attendance varied between schools, but, on average, was high across all 3 days (84.33%). The target of training 15% of the year group to be peer supporters (the required amount for sufficient peer-group diffusion) was achieved across the schools in the intervention arm.
Fidelity to the peer supporter training session plans and overall intervention design was good. All sessions were delivered in line with underpinning theoretical principles, but trainers had difficulties upholding these at their first sessions, when the group size was large or when peer supporter behaviour was disruptive. Overall, the training content was engaging, albeit slightly repetitive, the peer supporters enjoyed the sessions and had positive feedback for the trainers. The training prepared the peer supporters to carry out their role, in which they used a variety of strategies. These included being active with peers or inviting them to be active together, talking about being active and being subtle in their approach in doing so.
The intervention training and delivery costs were modest, ranging from £20.85 to £48.86 per pupil, with the majority of the cost being borne by the local authority providing the training. In unadjusted analyses, there was some evidence that, at T1, the deterioration in the KIDSCREEN-10 and EQ-5D-Y VAS scores was less in the intervention group than in the control group. However, the differences were small and consistent with no effect of the intervention in adjusted analyses. The primary economic analysis indicated that the intervention was both more costly and less effective at increasing MVPA than usual practice. Exploratory analyses demonstrated a low probability that the intervention was cost-effective at conventional National Institute for Health and Care Excellence (NICE) cost per QALY willingness-to-pay thresholds.
There were two findings of note in the current trial. First, the percentage (31%) of girls in the sample who met the CMO guidelines of an average of 60 minutes of MVPA per day across the week was higher than the national average of 9%, according to Health Survey for England data.98 Second, the subgroup analysis revealed a positive treatment affect among girls who had a high proportion (i.e. > 85%; 4.7% of the sample) of sedentary time at baseline, meaning that ‘high-sedentary’ girls in the intervention arm had higher MVPA than those in the control group at T1. These findings suggest that participants in this trial were more active than the general population and that there is a potential benefit among those who were the most sedentary, but, as this was not the primary analysis, this should be treated with caution and as an interesting observation only.
Post hoc exploration of mitigating factors
Post hoc analyses of school-by-school outcomes indicated that two schools in the intervention arm (i.e. schools 1 and 2) exhibited significantly greater decreases in MVPA between T0 and T1 than the other schools in the trial (see Figure 4). Examination of the process evaluation data provided no obvious explanation for this marked difference as both schools demonstrated good fidelity to the intervention delivery, and focus groups in these schools did not highlight any issues among the year group. However, further analysis showed that these two schools experienced extreme weather (i.e. sustained storm systems and severe flooding) in their local areas during the T1 data collection period. This could have restricted opportunities for children in these two schools to be active outdoors, such as after-school and community sport, informal leisure time play with peers in and out of school, and active travel to and from school. These weather events may have accounted for an unexpectedly steep decline in observed MVPA in the two schools that exhibited good adherence to the intervention and had school environments that were supportive of physical activity (see Appendices 6–8). Therefore, these data highlight the challenge of using single periods of assessment to capture complex behaviours such as physical activity.99,100
Although this probable explanation may account for the severity of the decline in physical activity in these two intervention schools, it does not account for the fact that no overall intervention effect was observed in the ITT analyses. Moreover, review of Figure 4 would suggest that, although these two schools logged noticeably poorer average MVPA at T1 than the other eight intervention schools, there is no difference in the overall pattern across all intervention schools. Furthermore, post hoc analyses indicate that, in the peer supporter subgroup, the decline in MVPA was more pronounced (although small in magnitude) than in the non-peer supporter group (an adjusted difference in T1 MVPA between peer supporters and non-peer supporters in intervention schools of –2.25 minutes, 95% CI –5.27 to 0.22; p = 0.071), suggesting a possible negative treatment effect. This finding is in contrast with the findings of the feasibility trial,51 as well as early research into peer-delivered interventions, which found that the peer supporters were often more positively affected than the non-peer supporter group.101
Participants in this trial were 1 year older than those in the feasibility trial,52 in which evidence of promise was found. The participant age group in this trial was changed from Year 8 (i.e. aged 12–13 years) to Year 9 (i.e. aged 13–14 years) to allow participants to provide their own consent and meet new ethics requirements that came into force because of the European Union GDPR legislation,97 which was introduced in 2018. Although we cannot be certain what effect this had on the outcomes of the trial, trainers and a researcher who had delivered the intervention in both the feasibility trial and definitive trial believed that engagement was poorer with the older year group, and that the older pupils were more indifferent to the training and the peer supporter role, which could have compromised the DOI delivery model.
Using school-level process, environment and policy data, we explored the convergence with school-by-school change in the primary outcome to further understand the trial findings. In the intervention schools, process data evaluating the delivery quality of peer supporter training sessions showed that perceived delivery quality, pupil engagement and enjoyment scores showed little correlation with MVPA change at a school level (see Table 35). Even on-site delivery, which the process evaluation highlighted as a key factor that negatively affected pupil engagement, enjoyment and delivery quality, did not predicate poorer MVPA outcomes in those schools that were affected by it, such as school 10 (the associated delivery issues are described above).
In the control arm, schools 6, 9, 12 and 13 accrued the smallest reduction in weekday MVPA from T0 to T1 out of all 20 schools (e.g. school 6 saw a 2-minute improvement). School environment and policy audit data indicated that schools 6 and 9 were well provisioned for physical activity, although no more so than some intervention schools that exhibited greater declines in MVPA between time points, but that schools 12 and 13 were not well provisioned. Similarly, the number and range of termly physical activity opportunities, as reported by school contacts, did not change meaningfully between T0 and T1 in any of these schools (see Appendix 4), shedding no further light on what makes these schools different from others in the trial that would account for the reduced decline in physical activity compared with that of other schools in the trial.
One operational difference between schools in the trial arms in PLAN-A was that four control schools changed the focus of their PE curriculum for girls away from competitive sport and towards health and well-being in the time between T0 and T1 measurements, but no intervention schools reported doing so. School 6, which saw a 2-minute increase in mean weekday MVPA between T0 and T1, was one of these schools and underwent a change to become an academy trust during that time. The rebranding of the school came with a shift in school priorities towards pupil well-being. Reportedly, this shift included changing the types of activities undertaken with girls in PE to align more closely with current group-based, health-focused exercise trends that are ‘popular with girls their age’. Collectively, these data provide a range of possible explanations for the differences between the results of the feasibility trial52 and the findings reported here, but there is no clear single factor, and the results could also be a function of something that we did not assess. As a result, there are a number of possible explanations for the project findings, but a degree of uncertainty remains.
Health economics
The direct cost of physical inactivity to global health systems in 2013 was conservatively estimated to be $53.8B (international dollars),102 with an additional $13.7B in productivity losses.102 Effective prevention during adolescence is urgently needed to reduce health system costs, improve health and prevent premature mortality.
The mean cost of the PLAN-A intervention per school was similar to that estimated in the feasibility trial52 (£2817 vs. £2685, respectively) and in other school-based physical activity interventions (GoActive,91 The Bristol Girls Dance Project103 and Girls Active104) targeted at a similar age group. As the PLAN-A intervention was targeted at girls in Year 9 only, the cost per pupil was higher than that in the Girls Active104 or GoActive91 interventions, which were targeted at all secondary school girls and all Year 9 pupils, respectively. Nevertheless, the mean cost per pupil (£31) of the PLAN-A intervention was considerably lower than that of the Physical Activity 4 Everyone intervention, which involved more intensive classroom-based physical activity and implementation strategies ($394 per pupil in 2014 Australian dollars).105 Providing peer supporter training to encourage physical activity in girls has the potential to be affordable for schools and cost-effective for society, if it results in a sustained improvement in MVPA.
School-based physical activity interventions might demonstrate cost-effectiveness in the short term if the initial intervention costs are justified by improvements in HRQoL or are offset by reduced use of health services. Observational and experimental studies have demonstrated the association between physical activity and higher levels of psychological well-being, and the potential for physical activity interventions to improve adolescents’ mental health.106 Our data highlight the scope for improving the mental health of girls in this age group, with more than half of girls reporting feeling ‘a bit’ or ‘very’ worried, sad or unhappy at the baseline assessment. However, this potential has not been borne out in economic evaluations91,103,104,107 of school-based physical activity interventions that have used preference-based measures of HRQoL, such as the CHU-9D or EQ-5D. In large part, this is because of the lack of a meaningful effect of these interventions on physical activity. There are also concerns that these outcome measures might not be sensitive enough to detect the small improvements in mental health that might be expected from school-based physical activity interventions.108
As with most other evaluations of school-based interventions, we did not assess whether or not the intervention costs were offset by reduced use of health services during the trial. In the Girls Active trial, which elicited this information from girls at 7 and 14 months post baseline, analysis was hampered by high levels of missing data and concerns about the accuracy of recall.104 Given this, the use of routine data from electronic health records might be the most efficient and unbiased way to capture this information in future work.
Most economic evaluations of school-based physical activity interventions that have evaluated short-term cost-effectiveness in terms of cost per QALYs have concluded, as we do, that there is insufficient evidence that the intervention costs are justified by initial improvements in HRQoL. One exception to this is the Daily Mile evaluation in primary school (i.e. aged 4–11 years) children, which found that there was a high probability that the intervention was cost-effective in girls, but not boys.109
A second pathway for school-based physical activity interventions to be cost-effective is through the long-term prevention of activity-related health conditions, such as cardiovascular disease, type 2 diabetes and some types of cancer. This would require that any short-term changes in physical activity and sedentary behaviour are sustained for long enough after the intervention period to have an impact on disease incidence later in life. For many years, work in this area has been restricted by the lack of high-quality models that would allow the intermediate outcomes measured in trials, such as MVPA or body mass index in adolescents, to be extrapolated over the life course to predict disease incidence and associated health-care costs, morbidity and mortality. More recently, decision analysis models have been developed to simulate the lifetime costs and outcomes of changes in body mass index110 and MVPA111 in adolescence. These models demonstrate the potential for interventions with a similar cost to PLAN-A to be cost-effective even if the impact on MVPA is modest (i.e. approximately a 5-minute improvement) and the effect is relatively short-lived (i.e. tapering to no effect by 7 years).111 This is a moot point for PLAN-A, in which we did not observe any short-term effect on MVPA. However, it does demonstrate the importance of conducting long-term follow-up of participants in trials of interventions that do find evidence of short-term improvements in physical activity.
Psychosocial variables
There was no evidence that the intervention affected girls’ self-esteem, which was specified as a secondary psychosocial outcome. This is not dissimilar to the Girls Active intervention,112 which reported a very small increase in self-esteem in favour of the intervention group 7 months post intervention but a null effect at the 14-month follow-up. In both the intervention and control groups, psychosocial variables decreased between Year 9 and Year 10, reflecting a tendency for girls’ motivation, self-efficacy and peer support towards physical activity to decline with age. There was no evidence that the PLAN-A intervention was sufficient to stem this trajectory, although we did not include shorter-term interim measures (e.g. an immediately post-intervention period) so are not able to determine short-term effects. It could be that intervening in Years 9 and 10 is too late into adolescence to meaningfully compete with the multitude of other factors (i.e. biological, social, educational and emotional) that are changing simultaneously and potentially contribute to a decline in girls’ physical activity113 and perceptions, such as physical activity motivation, competence, enjoyment, peer value and acceptance.
In terms of the theoretical model underpinning the intervention’s content (i.e. SDT), the process evaluation confirmed that the peer supporter trainers used an autonomy-supportive style. There was qualitative evidence to suggest that in their role as peer supporters the girls developed confidence/competence; autonomy-supportive techniques, such as empathy, communication and cooperation; and new social bonds. This agrees with the findings of the feasibility trial and previous studies showing that intervention agents (both adults and young people) can be trained to use a motivational style that promotes need satisfaction.51,78,114,115 However, there was limited evidence from peer supporters and non-peer supporters as to whether or not and how peer supporters used positive motivational techniques to influence their peers’ physical activity.
Intercorrelations between psychosocial factors at both time points were as hypothesised in SDT (e.g. the simple pattern of the behavioural regulation subscales and associations between behavioural regulations and autonomy, competence and relatedness need satisfaction, and other psychosocial variables; see Appendix 14). For example, autonomous motivation showed a moderate positive association with physical activity self-efficacy, whereas the equivalent correlation with controlled motivation was near zero. However, like the feasibility trial,52 we found no evidence that the intervention affected the psychosocial variables as hypothesised, and there was no evidence of meaningful associations between the psychosocial variables and MVPA. Furthermore, there was no evidence that the psychosocial variables functioned as mediators. These findings were broadly consistent when run as regression models, which adjusted for the same variables as specified in the primary analysis (see Appendix 15). However, a model that included the five subtypes of motivation showed a small positive association between intrinsic motivation and MVPA.
The data presented in this trial agree with a meta-analysis of 46 studies that reported that children’s and adolescents’ intrinsic motivation for leisure-time (i.e. non-PE) physical activity was positively associated with their leisure-time physical activity.116 This evidence synthesis also reported a null association between identified motivation and physical activity in studies that used objective measures of physical activity. These findings add to the evidence base for the positive role played by adolescent girls’ motivation based on enjoyment and inherent value for physical activity.
The SDT framework has been used extensively to study associations between motivation and physical activity in children and adolescents,116 and numerous school-based physical activity interventions have used SDT as a theoretical foundation.114,117–120 Furthermore, the principles of SDT are integrated into frameworks for the design and delivery of organised physical activity sessions for children and adolescents.121 However, the effectiveness of SDT-based interventions on motivational mediators and/or physical activity is mixed.
Like PLAN-A, the ‘Girls Active’112 and ‘Girls on the Move’118 interventions also showed no effect on girls’ motivation and other psychological mediators, such as enjoyment or self-efficacy towards physical activity. In contrast, Lonsdale et al.120 and Ha et al.115,117 have reported positive effects of SDT-based interventions situated within PE lessons on secondary school-aged children’s MVPA time in PE. However, neither intervention found effects on participants’ broader leisure time MVPA. Although Ha et al.115,117 found a positive effect of the PE-based intervention on girls’ (but not boys’) autonomous motivation, Lonsdale et al.120 found no effect on the hypothesised motivational mediators.
A reason for the differences in the findings of these studies might be the setting in which they were implemented. Despite sharing similar theoretical foundations, PE lesson interventions could be considered more discrete and structured in nature than PLAN-A, Girls Active112 and, to a lesser extent, Girls on the Move,118 which were implemented in more informal and global settings (e.g. changing school or peer culture and expecting girls to develop their own ways to increase their physical activity). It may be easier to explicitly implement SDT principles and measure the effects on motivation and physical activity in discrete contexts (e.g. a PE lesson) than in interventions, such as PLAN-A, that operate and measure constructs at broader levels. For example, Lonsdale et al.120 suggest that their Activity and Motivation in Physical Education (AMPED) intervention would be ‘best implemented as an enhancement component of a comprehensive school PA [physical activity] programme that also includes other “expansion” and “extension” initiatives’.
The efficacy of health improvement interventions based on SDT has been shown in a recent evidence synthesis, with evidence mainly derived from studies promoting physical activity, including school-based trials.122 Furthermore, basing physical activity interventions for young people on theories such as SDT seems intuitive (i.e. it seems logical and morally appropriate to promote motivation based on enjoyment and positive values, rather than guilt and external demands), is recommended121 and aligns with evidence for a positive association between intrinsic and identified motivation and physical activity.116 However, the literature demonstrates that it is challenging to quantify intervention effects on theoretical/psychosocial mediators within school-based physical activity interventions. We have carefully mapped SDT to several interventions,51,78,123 which we have tested in robust trial designs, and used a combination of qualitative and quantitative measures (often from multiple stakeholders) to try to quantify theoretical mechanisms of action. A consistent finding was qualitative results that generally support theoretical fidelity and efficacy, but also quantitative findings that fail to show intervention effects on outcomes or mechanisms of action. Although we endeavour to collect and analyse qualitative process evaluation data with appropriate rigour, confirmation bias is a possibility. Equally, the measures used to quantify psychosocial mechanisms of action may not be sufficiently sensitive to detect change. It is also possible that the intervention simply did not affect the mechanisms of action as proposed.
Implications for policy and practice
Intervention delivery versus intervention model
Neither our measurable indicators of intervention implementation success (i.e. pupil engagement and enjoyment, fidelity to intervention design and theories) nor our measures of school environment (i.e. physical activity facility provision and policy audit) converged with or could explain a lack of intervention effectiveness (as determined by improvement in weekday MVPA) at the school level. This raises the question ‘If the intervention was delivered as planned, with good fidelity to underpinning theory, did insufficient peer supporting take place to bring about positive improvements in physical activity, or is it realistic to expect a(ny) single intervention to bring about meaningful change?’.
Existing literature in this area suggests that other peer-led interventions delivered at scale had no lasting effect on MVPA levels in adolescent girls112,124 or in a mixed-gender school setting.125 Harrington et al.112 found evidence of a small (i.e. 2–3 additional minutes of MVPA per day), short-term (i.e. 7-month) effect of the ‘Girls Active’ intervention among adolescent girls in English schools, but this was not maintained at the 14-month follow-up. In a randomised controlled trial of an intervention that trained older adolescents (i.e. mentors) and in-class peer leaders to encourage classes to undertake two new weekly activities, Corder et al.125 reported no difference in MVPA between the control and intervention groups at follow-up. Moreover, although a peer-led walking intervention in schools124 resulted in an increase in daily light-intensity physical activity of 8.27 minutes among adolescent girls, it had no significant effect on MVPA.
Many elements of Harrington et al.’s112 Girls Active intervention focused on changing school culture around physical activity, but not adding new opportunities to be active, which was hypothesised to be insufficient to increase MVPA. PLAN-A relies on a similar model, changing peer culture and expecting girls to find their own ways to increase their physical activity. Indeed, in a recent feasibility trial exploring peer-mentoring in conjunction with new physical activity opportunities, Owen et al.126 reported that, when using peer mentoring alone, physical activity declined, whereas, when used in combination with an after-school club, the same peer mentoring intervention increased physical activity by 3.2 minutes per day (p = 0.007); however, this was not a fully powered trial. Collectively, these studies highlight a need to explore alternative intervention approaches.
Peer-led delivery models and associated measure sensitivity
The process evaluation in this trial and in our earlier feasibility trial52 demonstrate that peer-led interventions such as PLAN-A can be delivered with good fidelity to intervention design and theory. However, challenges identified around behaviour management with larger groups and the intervention setting resulted in a ‘drift’ from important design elements, such as the identification of influential peer leaders. There are questions regarding the suitability of the individuals selected to be peer supporters and how well represented social groups are within the cohort,127 which reflect limitations of the peer-nomination process itself. Recent research has tested a social network analysis (SNA) approach to peer nomination using a closeness centrality selection criterion.128 This selection criterion is predicated on establishing how closely each pupil is connected to all of their peers via a SNA and then selecting those needing the fewest intermediaries to reach everyone in the network.129 This approach differs from the approach used in this trial and ASSIST,37 in which the overall number of nominations a person received was the criterion for selection, and presents a more data-driven approach to peer nomination that has been shown to be successful at selecting more effective agents for change, who are capable of influencing physical activity in their peers,130 namely those with higher ‘sociometric popularity’ rather than ‘prestige popularity’.131 Nevertheless, establishing the efficacy of a peer nomination process for selecting pupils with the greatest influence over their peers is only part of the peer-led approach. Once identified, these agents for change must also diffuse new behavioural norms.
In this trial, we experienced difficulties in determining the extent to which trained peer supporters engaged in supportive conversations and other peer support strategies (see Chapter 5, Evidence of peer support: fidelity to diffusion of innovations theory). Participants in this trial and the previous feasibility work52 demonstrated that diaries are not an effective way to capture this information, and qualitative evidence suggests that the subtle nature of much of peer supporting makes it hard to recall, harder still to quantify and, potentially, very difficult for recipients to notice. In a roll-out of the ASSIST programme, it was also a challenge to measure the peer-supporting conversations that took place.132 In contrast to ASSIST, the outcome of interest in PLAN-A (i.e. weekday MVPA minutes) was a much more complex behaviour to measure (comprising type, volume, intensity and context) than a discrete behaviour, such as smoking and, therefore, it was not possible to use a binary variable. Furthermore, the focus of ASSIST for most children was not smoking (for most this meant not starting to smoke), whereas for PLAN-A it was being more active; the latter is, arguably, comparatively amorphous and, perhaps, not as strongly socially linked to peer influence and social status among adolescents. Measurement considerations are twofold: measurement of the exposure (i.e. peer supporting) and of the outcome (i.e. MVPA).
Regarding the exposure, ecological momentary assessment (EMA) has been used in various settings and fields to capture real-time data through repeated measures of behaviour or behavioural contexts. Using EMA dramatically reduces the period of recall, meaning biases associated with retrospective recall are reduced and the need for participants to summarise events or experiences over extended periods of time is removed.133 Thus, EMA could, theoretically, be used to improve measurement of peer-supporting actions, or at least to reduce the recall bias experienced in this and other peer-based trials.
However, as highlighted in a recent review of EMA studies on physical activity and sedentary behaviour,134 the methodology is vulnerable to several risks of bias that would present challenges to its use in capturing peer-supporting activity. The primary risk is the sampling approach, time-based versus event-based sampling, as either approach would be imperfect. If a time-based sampling approach was used, then peer supporters would be asked to recall peer-supporting activity and context at scheduled intervals during the day, which may lose the benefit of momentary assessment and be subject to recall bias. On the other hand, an event-based approach – which is usually self-initiated – would rely on participants logging details of peer-supporting activities at the time, which, based on feedback from peer supporters in this trial about the use of their diaries, is something that adolescents would not engage with reliably.
Furthermore, despite advances in technology that mean that EMA can be performed using participants’ smartphones, the method is burdensome for participants and is likely to blur the lines between measurement and providing an additional intervention, which would need to be replicated if the intervention were scaled up. Although many studies using EMA report high compliance,134–136 the measurement periods are usually short (between 1 day and 2 weeks) and the feasibility of using EMA continuously for extended periods of time (e.g. a 10-week diffusion period, as used in PLAN-A and ASSIST37) has not been tested.
Considering the outcome measure’s sensitivity, the use of objective means of measurement (i.e. accelerometers) has been widely accepted as a more robust way of capturing physical activity than self-report, which is prone to overestimation of higher-intensity physical activity and underestimation of sedentary time.137,138 However, there are known limitations of using accelerometers, including an inability to identify time spent in different activities139 and the risk that the brief measurement period (7 days is typical) may be too short to reliably observe changes in ‘normal’ behaviour patterns, or it may be subject to confounding by external factors. In this trial, we hypothesised that participants in two schools exhibited lower-than-predicted physical activity because of adverse weather, reinforcing this concern. In addition, using hip placement of the accelerometer, as was used in this trial, and asking that the devices be removed when in water or when playing contact sports risked underestimating or missing certain types of physical activity, such as cycling,140 swimming and contact sports, that can account for significant amounts of MVPA. It is possible that the challenges associated with measurement of the exposure (i.e. peer supporting) or nuanced and imperfect measurement of the outcome (i.e. change in MVPA) relative to the exposure reduced our ability to accurately quantify the efficacy of a peer-led approach to changing physical activity.
Likewise, if such interventions can be delivered at scale, but are limited in their effectiveness at changing physical activity, then we may naturally question whether the reasons for that lie in the intervention model itself (using peer support to diffuse new norms) or in social factors that are hard to measure (e.g. pre-existing local cultural norms) that undermine the assumptions inherent in a peer-based DOI approach. One such social factor pertinent to adolescents today is the use of social media to interact and influence.141 There is uncertainty about whether or not peer-led interventions in schools can fully capitalise on how adolescents interact and influence each other using social media because of challenges in measuring engagement and impact.142 Thus, there are major challenges to designing an intervention that includes substantial smartphone and social media components (whether delivering messages, support, ideas to peer supporters and encouraging peer supporters to influence their peers through their usual social media channels, or in the assessment of peer-supporting activity) and is delivered during school time. It would be worth exploring how social media and mobile technology could be incorporated into peer-based interventions and their acceptability to schools, pupils and parents.
Issues with scaling up interventions
The data presented in this report are consistent with recent evidence on the challenges of scaling up physical activity interventions from feasibility studies to definitive trials. In a recent meta-analysis, Beets et al.143 identified nine forms of bias in this area, of which two were relevant to this trial. We found that there were changes in intervention support between the feasibility trial52 and this trial as the group sizes were larger than those in the feasibility trial and, therefore, trainer–pupil contact was more diluted, and there was a target audience bias as we had to change the year group from Year 8 to Year 9. The data reported here reinforce the need to focus on external validity and the potential implementation of intervention results from the design stage.144
Alternative approaches to interventions targeting individual-level change and the need to explore whole-school approaches
The ultimate target of public health interventions is behaviour change at a population level, which is logically driven by behaviour change among the individual members of a population. However, as advocated in recent papers,145,146 individual-level interventions are not sufficient to bring about individual- or population-level change because the determinants of our behaviour are diverse, complex and often beyond individual control (e.g. policy, physical environment, SEP). PLAN-A attempted to advance beyond a simple individual-level education-based intervention; instead, it was designed to harness the power of naturally existing peer groups and used informal (e.g. peer diffusion) rather than formal (e.g. educational lessons, additional structured physical activity time in school) methods to promote the uptake of physical activity. However, despite this, the intervention model is relatively constrained and one-dimensional: it does not embrace a broader system or ‘whole-school’ approach,146 which would intervene at multiple levels including, but not limited to, peer influence. Indeed, from a systems perspective, it seems somewhat unrealistic to expect a short-term intervention focused on a single determinant of physical activity (i.e. peers) to have a meaningful impact on physical activity behaviour. It is important to note that this was a relatively high-agency intervention, and it may be the case that an intervention that required lower agency would have been more successful. Low-agency interventions are consistent with the concept of whole-school interventions147 and, thus, in the future we suggest that lower-agency interventions are considered.
An important question that arises from this project is ‘What next?’. Whole-school approaches to physical activity have been advocated for some time,148 and the recently developed Creating Active Schools framework advances this work using insight from researchers, teachers and sport practitioners.146 However, like PLAN-A, school-based interventions often focus on a single or a few components of the whole-school approach only. This is probably driven by a methodological need to:
- define what the intervention is and design it well
- estimate the effect of particular intervention strategies to inform future investment (e.g. active travel, enhanced PE or after-school activity clubs)
- design interventions that could be attractive to schools (i.e. interventions that are clear, discrete, simple and not resource intensive)
- design interventions that can be rolled out at scale.
However, these elements create a conflict between intervention depth and breadth. Many facets of good intervention design (e.g. detailed planning, iterative co-development, piloting, pre-testing, the use of logic models, theory, attention to behaviour change techniques) take considerable time and resources to perform well and often necessitate a narrowing of the intervention scope. This was the approach used in PLAN-A, and it arguably resulted in good intervention ‘depth’. However, a systems or whole-school approach requires greater ‘breadth’, with interventions operating across multiple domains simultaneously (e.g. active travel, school culture and ethos, high-quality PE, active lessons, active breaks, active extracurricular clubs). Nevertheless, it is not yet clear how multidimensional whole-school approaches can be designed in a way that balances and optimises depth and breadth.
The balance between breadth and depth is also an issue that affects intervention evaluation. It is, relatively speaking, straightforward to evaluate a well-defined, one-dimensional, school-based physical activity intervention such as those conducted in recent years by our group and others.51,78,123,125 The intervention can be clearly described, fidelity can be assessed, and its effect compared using a control group within a traditional cluster-randomised controlled trial design; again, we can achieve good depth, but limited breadth. It is not clear how best to evaluate broader whole-school approaches, which act on multiple domains of physical activity, nor how to determine which elements are and are not effective, or how interventions may interact with pupil-level factors (e.g. age, gender, SEP), cohort effects and school-level factors (e.g. deprivation, facilities, size or location). Even if researchers can determine that a whole-school approach is effective, it may be unrealistic to expect the approach to be generalisable. Schools are like ecosystems in that no two are the same: they are in a state of constant flux; serve diverse communities with different needs; and have shifting priorities, variable budgets and resources, and changeable internal and external contexts. Effective interventions are likely to be those that can be designed and modified to accommodate these factors; therefore, an intervention that works in one school may not suit, work or be needed in another.
One approach is to focus further on depth by designing specific and discrete interventions (e.g. increasing the amount of MVPA time in PE) and very closely measuring the outcomes that are matched to the behaviour (e.g. MVPA time in PE), repeating this across multiple domains (e.g. active travel, classes, miles and clubs) and being content with the assumed additive effect of the small effects on MVPA across multiple domains, summing to a meaningful or health-enhancing level of physical activity. If multiple interventions are needed as part of a whole-school approach, testing discrete interventions would also lack external validity and would not provide insight as to whether or not schools are able to implement multiple interventions simultaneously. If we focus more on breadth, we risk developing weaker or more diluted interventions and it may be very challenging to identify which intervention actions are effective. It may be harder to convince schools and/or commissioners to adopt more resource-intensive approaches. This is a challenge that needs to be addressed in the next generation of system-focused public health interventions.
Specific research recommendations
Based on the data presented in this report and the issues covered throughout this chapter, there are several specific research issues that we recommend as the focus of future research. There is a need for a better understanding of how changes in physical activity track over time, how change can be achieved more effectively, and how change can be measured more accurately and reliably. The research needs are:
- the need for better evidence on the sustainability of improvements in physical activity in adolescence into adulthood
- the need to evaluate the utility of whole-school approaches to promote physical activity in schools and explore implementation options that would make such an approach feasible in a range of settings
- the need to develop and test EMA methods for capturing peer supporting and assessing the potential impacts of peer-led interventions on behavioural outcomes
- the need for new or revised psychosocial scales that can better capture change in physical activity motivation.
Furthermore, future research would benefit from investigating the pattern of accumulation and the potential compensatory effect or composition of daily physical activity to ascertain whether interventions simply have no effect or whether they change the manner in which physical activity is accumulated and, therefore, have an effect on associated health factors.
Strengths and limitations
The main strengths of this trial were functions of the robust cluster-randomised design. Cluster-randomised controlled trials enable the testing of an intervention’s real-world efficacy and rely on large sample sizes. Schools were the unit of allocation and the intervention was delivered to an entire year group at each school, rather than to a subgroup of pupils. Our trial successfully recruited 20 schools from across South West England; the schools varied in size and demographics, providing ample statistical power to address our research questions. Physical activity was measured using accelerometers, and a comprehensive mixed-methods process evaluation was conducted alongside the trial to gather views from each key stakeholder group that was involved in the delivery and receipt of the intervention.
The intervention delivered in this trial was a refined version of an intervention (which was itself based on a successful smoking cessation intervention) that was co-developed with adolescent girls in a robustly conducted feasibility trial,51 and has been reported in line with the TIDieR framework.60 Refinements included an improvement to the peer-nomination process to build in the flexibility to accommodate different year group structures, and changes to the underlying programme theory by incorporating SDT. We have also ensured that our reporting of the main trial adheres to the CONSORT extension for cluster-randomised trials guidelines149 and the reporting of the qualitative process evaluation data follows the COREQ59 checklist.
Although schools were sampled to reflect a range of SEPs, this trial was limited by a geographical focus that included schools in south-west England only. In addition, although schools were assessed for their suitability for physical activity using environment audits and changes in school policies were captured at a superficial level, greater detail on school-level environment changes between time points and trial arms would have enabled greater exploration of the potential impact of school environment (including changes in school policies) on the stated trial outcomes. Several schools in the control group changed their approach to physical activity provision in the curriculum, which may have yielded changes in the year group’s physical activity for which we were unable to account. This could have been better controlled for if we had asked schools to defer any such changes for the time that they were in the trial, but this would have raised concerns over the external validity of the trial results.
This trial presented qualitative and qualitative data largely separately; however, this was a decision taken to retain the independence of the researchers analysing each type of data and to ease readability. The process evaluation was restricted to topics that were not already covered in detail in the feasibility trial;52 however, the findings were based on both the feasibility trial and this trial. Despite this, the current trial was unable to quantify the impact of delivering the intervention in an older year group than that in the feasibility trial. Additional work to understand changes in social and behavioural norms between Year 8 and Year 9 would have improved our ability to describe, if not quantify, this factor.
Furthermore, capturing fidelity to the principles of DOI, namely peer-supporting activity, was unreliable and had questionable validity because of the juxtaposition of the behaviour in question (i.e. subtle, ambiguous and intuitive) and the measures available to capture it (i.e. quantitative and definitive). We have suggested alternative approaches that warrant further exploration.
We also note that there are many options available for collecting and processing accelerometer data. It may have been the case that greater compliance would have been obtained if we had opted for a wrist-worn device. However, we opted for the waist-worn ActiGraph GT3X+ as this device has been widely used and, therefore, its use facilitates comparison with other data sets.
Finally, the economic analysis did not measure health service use and only indirectly estimated the CHU-9D scores from the KIDSCREEN-10 responses; however, given the lack of effect on MVPA, these limitations are unlikely to have altered the interpretation of the economic results.
Conclusions
This report found no evidence that the PLAN-A intervention led to increased MVPA among adolescent girls and, consequently, the intervention was not cost-effective. The process data showed that this was despite the intervention having high acceptance and being delivered with high fidelity. Challenges included the accurate measurement of peer-supporting behaviour and connecting that exposure to changes in physical activity in the face of myriad confounding factors.
Possible explanations for the lack of a positive effect of the intervention include a change in year group from the feasibility trial and the diluted intervention dose; however, it is difficult to accurately estimate the impact of these factors on the complex intersection between environment, motivation and behaviour. Therefore, the findings provide strong evidence that the PLAN-A intervention is neither an effective nor cost-effective means of increasing physical activity among adolescent girls; thus, it should not be implemented or disseminated. It has been argued that single individual-level interventions implemented in schools are insufficient to bring about increases in physical activity for the masses and further system-wide approaches may be needed. This report recommends further research exploring the implementation considerations and utility of using a whole-school approach to promote physical activity in young people.