Recruitment

Figure 9 presents the CONSORT diagram and detail on recruitment and retention throughout the study.

FIGURE 9

Trial profile for PLAN-A (CONSORT flow diagram). a, Returned at T2; b, took part in T2; and c, present at T2.

Participant (student) recruitment

The recruitment briefing was delivered to 451 Year 8 girls. Four girls (0.89%) returned opt-out forms and 20 girls (4.43%), despite not handing in an opt-out form, chose not to take part resulting in a 94.68% participation rate across the six schools (Table 10). A total of 427 girls from six secondary schools were recruited (PLAN-A arm, n = 269; control arm, n = 158).

TABLE 10

Student and PS recruitment rates across study schools

Trainer recruitment

Expressions of interest were received from 10 potential trainers with a range of experience in youth work, PA or both. Five female trainers were recruited: two had experience in drama and working with young people and three had experience in PA and working with young people (Table 11).

TABLE 11

The PLAN-A trainer characteristics

Data provision

Accelerometer return and data provision rates are shown in Table 12. Return rates were high (> 85%) at each time point and only differed markedly between arms at T1, in which return was approximately 10% lower in the control than in the PLAN-A arm. The wear-time criteria was met by 82.63%, 71.13% and 62.21% of participants at T0, T1 and T2, respectively. A lower proportion of participants in the control arm, than in the PLAN-A arm, met the wear-time criteria at T1 (5%) and T2 (7%). A total of 64% of participants provided valid accelerometer data at both T0 and T1, and 57% participants provided valid data at T0 and T2.

TABLE 12

Accelerometer return and data provision rates at each time point, by trial arm

Questionnaire data provision is shown in Table 13. Completion of the psychosocial questionnaire using the tablets exceeded 90% at each time point and was very similar between trial arms. Completion of the quality-of-life measures (i.e. KIDSCREEN-10 and EQ-5D-Y), which were administered on paper, was slightly lower but still high, with completion rates of > 87% at all time points.

TABLE 13

Psychosocial questionnaire and quality-of-life measure data provision

Data linkage

The three variables needed to perform data linkage (i.e. full name, date of birth and home postcode) were collected for 88.76% (i.e. 379 out of 427) of students. This was solely limited by the provision of full postcode information, as full name and date of birth information were collected for all students. Three (50%) of the schools returned the completed data linkage questionnaire. Responses from one school indicated that it had concerns with the concept of data linkage that would need to be addressed before children and parents were involved, and that it would not provide the Unique Student Number identifier; however, no further details of these concerns were provided in the free-text fields. The other two schools indicated that they had no concerns and would provide the Unique Student Number identifier, assuming that parental consent was sought.

Baseline data

Baseline data are described in Table 14. Medians and lower and upper quartiles are presented for IMD and accelerometer variables, as these data were skewed. Overall, the trial arms were well balanced. IMD score (i.e. level of deprivation) was slightly higher in the intervention than in the control group, suggesting increased levels of deprivation in the PLAN-A arm, although both median scores were in the second quintile range (i.e. 8.5–13.8) of the least deprived households in England. Accelerometer-assessed MVPA was similar between study arms. Participants in the PLAN-A arm recorded more minutes of light PA and weekday sedentary time. Psychosocial variables were similar. KIDSCREEN-10 t-scores, indicating health-related quality of life, approximated European and UK norm values for girls aged 12–18 years (i.e. were within ± half a SD of the threshold of the reference group mean).⁶²

TABLE 14

Baseline descriptive statistics

Characteristics of peer supporters and non-peer supporters

The baseline characteristics of the PSs and the NPSs are presented in Table 15. For this analysis, students who were nominated as PSs in PLAN-A and control schools were included. PSs had higher family affluence, although similar IMD to NPSs, and a greater proportion of PSs than NPSs were of white ethnicity. PSs recorded approximately 12 minutes more MVPA on weekdays and weekends and less sedentary time (mainly because of weekday activity) than NPSs. A total of 59% of PSs versus 35% of NPSs met the PA guidelines, and more PSs than NPSs walked to and from school. PSs also reported higher self-esteem, PA-based autonomous motivation, need satisfaction, stronger social support and peer norms for activity.

TABLE 15

Baseline descriptive statistics of PSs and NPSs

Attendance at peer supporter training

Attendance rates at the PS training was very high across all training days, with 94% (i.e. 53 out of 55) of PSs attending all of the training. See Chapter 6, Peer supporter training: attendance at peer supporter training for complete attendance data.

Evidence of promise

Physical activity variables

Table 16 shows that at T1 (immediately after the 10-week intervention period, i.e. the end of Year 8) there was no evidence of a difference between PLAN-A and control groups for weekday MVPA or weekend day MVPA. There was some evidence that the girls in the PLAN-A arm were less sedentary than the control arm on weekdays at T1 (–31.8 minutes, 95% CI –57.44 to –6.18 minutes). In addition, there was some evidence that the control arm performed a higher volume of activity (i.e. CPM) at weekends, although the CI was wide (–124.24 CPM, 95% CI –213.48 CPM to –35.01 CPM).

TABLE 16

Evidence of promise for intervention effects on accelerometer outcomes

At T2 there was evidence for a between-group difference in weekday MVPA in favour of the PLAN-A arm (6.09 minutes), with the 95% CI exceeding 10 minutes of MVPA per day (95% CI 1.43 to 10.76 minutes). The data suggested that the between-group difference was based on the PLAN-A arm girls maintaining their baseline level of MVPA, while the MVPA of the girls in the control group decreased over time. Similar to T1, there was some evidence that the girls in the PLAN-A arm were also less sedentary on weekdays at T2 (–23.26 minutes, 95% CI –43.73 to –2.79 minutes).

Sensitivity analysis

Data imputation

Table 17 shows the results of the adjusted regression analysis on the likely primary outcome at T1 and T2 in a future definitive trial based on the imputed data set. This sensitivity analysis explored whether or not there was comparable evidence of promise under the assumption of accelerometer data being ‘missing at random’ (i.e. < 2 valid days). The results were very similar to the complete-case analysis in both point estimate and 95% CIs (see Table 16).

TABLE 17

Sensitivity analysis of evidence of promise for intervention effects on weekday MVPA at T1 and T2 using the imputed data set and complete data for which accelerometer wear was ≥ 1 valid day

Complete data using ≥ 1 valid days of accelerometer data

A second sensitivity analysis (see Table 17) relaxed the criterion for participant inclusion based on their accelerometer data from ≥ 2 valid days to ≥ 1 valid days of data. Again, this analysis led to identical conclusions regarding the potential effect of the intervention on MVPA.

Intervention effects for peer supporters and non-peer supporters (exploratory)

Comparing the intervention effect between PSs and NPSs (Table 18), there is evidence that PSs benefit more from the intervention at T1 than NPSs, but there is no evidence of a difference between the two subgroups at T2.

TABLE 18

Exploratory comparison of the intervention effect on MVPA at T1 and T2, between those nominated as PSs and NPSs

Psychosocial variables

Internal consistency statistics (Cronbach’s α) are reported in Appendix 9. With the exception of the peer PA norm (acceptance) variable, which exhibited low internal consistency at all time points, all psychosocial composite variables had good internal consistency at all time points. Table 19 shows the adjusted multivariable regression results for psychosocial variables. There was little evidence of any differences between the PLAN-A and control arms on any of the variables at T1 or T2. There was no evidence of harm or detriment to these variables because of the intervention.

TABLE 19

Evidence of promise for intervention effects on psychosocial secondary outcomes

Physical activity social support

Figure 10 shows that in the PLAN-A schools, immediately after the intervention period, there was an increase (9%) in the number of Year 8 girls believing that someone in their year group had spoken to them recently about PA, compared with a decrease (12%) at this time point in the control arm schools. At T2, scores in both schools had returned to approximately baseline values, albeit slightly higher relative to baseline in PLAN-A schools. Figure 11 shows that, at baseline, a similar proportion of girls in control and PLAN-A schools thought that talking to someone in their year had helped them to be more active (41–42%). However, at T1, although this level of belief was maintained in PLAN-A schools, in control arm schools there was a 5% decrease. At T2, 34.7% and 27.8% of girls agreed with this statement in PLAN-A and control arm schools, respectively.

FIGURE 10

Responses (% yes, no, not sure) to peer support question 1 ‘has anyone in your year group talked with you recently about PA?’ at each time point.

FIGURE 11

Responses (% yes, no, not sure, NA) to peer support question 1 ’did talking to anyone in your year help you to be more active?’ at each time point. NA, not applicable.

Sample size for a definitive trial

The ICC on the primary outcome for a definitive trial (i.e. weekday minutes of MVPA) from the PLAN-A arm at each time point is shown in Table 20 alongside ICC estimates of accelerometer-assessed MVPA from other studies involving adolescent girls. Based on these findings, sample size calculations for a definitive trial were based on an ICC of 0.01, but scenarios were also modelled in which the ICC was 0.02 and 0.03.

TABLE 20

Intraclass correlations on accelerometer-assessed MVPA in PLAN-A and other studies including adolescent girls

The results of the sample size calculations are shown in Table 21. To detect a 10-minute between-arm difference in weekday MVPA with 90% power, at the 5% significance level, and with a design effect of 1.7 (based on 70 students per school), 560 students are required (i.e. 280 per arm), inflated to 800 students in 12 schools to account for a 30% loss to follow-up. Keeping all other parameters the same and using a smaller between-arm difference in MVPA of 6 minutes (i.e. the point estimate from this feasibility study), 980 students are required (i.e. 490 per arm), inflated to 1400 students (700 per arm) in 20 schools to account for a 30% loss to follow-up.

TABLE 21

Definitive trial sample size calculation results

Should the ICC be 0.02 or 0.03, 800 girls in 12 schools or 1000 girls in 16 schools would be required, respectively, (after inflation for loss to follow-up) to detect a 10-minute between-arm difference in weekday MVPA with 90% power. As such, conducting a definitive trial involving 20 schools and 1400 girls would provide acceptable statistical power to detect a 6-minute difference in weekday MVPA or a 10-minute difference in MVPA should the ICC be larger than anticipated.

Adverse events

No adverse events or instances of bullying, among the girls of PLAN-A, were reported in any school.