Design:
This pragmatic trial will be conducted in settings similar to the "real world" which will
include use of a school setting, teacher-led delivery during classroom time, no strict
exclusion criteria for students and comparison to routine lessons-as-usual. Data collection
time points will be selected to align with school academic calendars and will be conducted at
pre, post and two month follow up. Schools will be randomly allocated to an intervention
group (who receive the BodyKind programme) or a waitlist control. Participants in
intervention and waitlist groups complete measures at the same timepoints, and those in the
waitlist control will receive BodyKind after all assessments are completed. A mixed methods
evaluation of implementation quality will also be conducted to understand 'why' programme
outcomes were observed and inform efforts to optimise programme delivery in schools in the
future.
Randomization:
Prior to randomization, schools will be stratified based on school categorization (e.g.,
co-educational/ single sex girls/ single sex boys), size (i.e., estimated number of students
in Transition Year available to participate) and location (i.e., urban/rural categorisation)
to ensure the gender and geographical location of participants in each treatment group will
be closely balanced. Shortlisted schools will be then randomly allocated to condition by the
primary researcher using a computer-generated minimisation procedure. Minimisation is a
method of randomisation that allocates participants to the treatment group that best
maintains balance in prognostic factors (e.g., gender balance across condition). The first
participant is allocated a treatment at random. Each subsequent participants' allocation is
determined by what treatment allocation would lead to better balance between the groups in
the variables of interest. Minimisation ensures excellent balance across treatment groups and
is desirable when there are strong prognostic factors (e.g., gender) and modest treatment
effects. Shortlisted* stratified schools will be allocated using covariate-constrained
randomisation, a type of minimisation randomisation, for a two arm cluster randomized control
trial, with the R statistical package cvcrand
(https://cran.r-project.org/web/packages/cvcrand/cvcrand.pdf). Allocation of schools to
experimental/waitlist treatment groups will be constrained by a.) the number of boys in TY in
each school, b.) the number of girls in TY, and c.) the school's location categorisation
(urban/rural). Schools that drop out will be replaced by schools on the waitlist based on
minimisation or replacement by same category to same allocation (e.g., if an urban school
drops out/fails to sign principal consent form they will be replaced by an urban school on
the waitlist and to the allocation assigned to that urban school).
Data collection procedures:
Recruitment will occur at the school level. Schools will be recruited via emails (Educator
School Centre's Ireland's mailing list and cold call invitations), teacher training events
and advertisement on social media & website pages. Schools will be selected on the basis of
(1) their expressed interest and (2) their fit with recruitment criteria.
Shortlisted schools registering interest and meeting these criteria that decline to
participate after initial contact will be replaced by the next school from the waitlist of
eligible schools who registered interest with the researcher. Schools will receive a €200
financial honorarium for participating.
Sample size:
The aim is to recruit 600 students aged 15-17 years in 4th year who attend second-level
schools in Ireland. The aim is also to recruit 1-2 second-level teachers per school, from
12-28 schools, to deliver the BodyKind programme to adolescents.
Sample size rationale An RMASS power analysis (http://www.rmass.org/RMASS_Manual.pdf) based
on parameters; small Cohen's d effect size of .2, power .80, alpha .05, indicates that 504
students are required. To account for natural attrition and unanticipated events that reduce
participant numbers (e.g., student absence due to illness) in RCTs, approximately 20% (n=100)
more participants will be recruited giving a total N=600. A systematic review of consent
procedures and participation rates reported that the average consent rate for active parental
consent was 65.5% (range: 11-100%). The average number of students in TY in schools in
Ireland is 140; therefore, allowing for a 65% response rate would indicate that a minimum of
8 schools would be required to obtain a sample size of 600. However, a recent pilot trial of
a self-compassion body image intervention in Ireland recruited an average of 25 TY students
per school using opt in parental consent (4 schools, N=100), indicating that n=24 schools
would be required to attain a sample size of 600. Given the variability in class size in
Transition Year and range in consent rates for active parental consent the investigators will
seek to recruit within this range until sample size of 600 is met. To guard against drop out
at a cluster level the investigators will aim to oversample and recruit at least 12 schools
and a maximum of 28 schools.
Hypotheses:
The hypotheses are
Compared to participants in the waitlist control, participants who receive the BodyKind
programme will experience statistically significant increases in a.) body appreciation,
psychological wellbeing, self-compassion and compassion for others and b.) significant
decreases in body dissatisfaction and body ideal internalisation from pre- to
post-intervention.
Changes will be maintained in experimental groups at a 2-month follow up.
Changes in body appreciation and body dissatisfaction will be mediated by
self-compassion, compassion for others and body ideal internalisation. No a-priori
assumptions about the directionality of effects will be made.
Changes in body appreciation, body dissatisfaction and psychological wellbeing will be
moderated by appearance related social media use, appearance teasing, gender, baseline
body appreciation, body dissatisfaction, psychological wellbeing and implementation
quality. There are no a-priori assumptions about the directionality of effects.
The programme will be considered acceptable and feasible by adolescents and teachers
The programme can be implemented with high fidelity by teachers
Data analysis plan:
Prior to conducting inferential statistics, data will be screened for missingness, outliers,
normality and model assumptions. Data will be summarised graphically and by using descriptive
statistics.
H1. Generalised linear mixed effects models (GLMM) will be used to determine change on
outcome measures. Primary analysis will compare measures at time one (baseline) and two
(post-intervention), to identify fixed effects of the intervention. Potential clustering
effects will be examined and accounted for at the school level. The model will include 2 main
effects (1. Repeated measures effect of time and 2. Between subjects effect of condition) and
one two way interaction (time*condition). Gender effects will also be considered. With gender
in the model will have three main effects and corresponding two-way and three-way
interactions.
H2. Secondary analyses will examine data from time 3 (follow-up) to see if the intervention
condition has maintained hypothesised changes in dependent variables. Investigators will also
assess the proportion of within-group and between-group variance by calculating the ICC. As
complex models can fail to converge, if investigators encounter convergence issues in our
models, investigators may consider dropping the random effect accounting for the smallest
proportion of between-group variance. *Note: If the recruitment target (n=600) is not met, a
different statistical analytic technique, 2 (condition; experimental, control) x 3 (time;
pre, post, follow up) mixed factorial ANOVA will be used. This can detect small effects, but
not clustering effects, in a sample size of 238 participants, according to a MorePower
analysis. Post-hoc planned comparisons will be conducted using t-tests to examine changes
within groups (intervention vs control) at different time points.
H3. Mediation analyses will investigate whether changes in body ideal internalisation,
self-compassion and compassion for others mediate change in body dissatisfaction and positive
body image from T1 to T2.
H4. Moderation analyses will investigate whether appearance related social media use,
appearance teasing, gender, school type, baseline body appreciation, body dissatisfaction,
psychological wellbeing and implementation quality moderate changes from T1 to T2.
H5. Descriptive statistics will be used to assess teacher fidelity ratings