Methods/design
Design and procedures
The present protocol design followed the recommended procedures and items to address in a trial protocol, namely SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) 2013 checklist. As aforementioned, this is an RCT (REGISTRATION NUMBER: NCT04099498) with three groups: a control group, not enrolled in any program; an intervention group, enrolled in the online program; and an enhanced-intervention group, enrolled in the online program with gamification strategies. The sample will be recruited by first contacting public elementary and middle schools from Portugal and inviting them to participate. Orientation sessions will be held at schools for all parents of children in the target grades to explain the project and its rationale and to invite them to enroll their child. The study will be introduced as a program to promote healthy lifestyles among late elementary school-age children, which aims to monitor relevant engagement variables and examine children’s, parents’, and teachers’ perceptions of the utility and feasibility of these interventions.
Retention and incentives for the participants
The following strategies will be undertaken to minimize attrition. First, prior to the enrollment of children in the intervention, the researchers will organize orientation sessions with parents to present the project and explain its pertinence and rationale. Second, parents who agree to enroll their child in the program will sign a participation agreement form. Third, parents will also be registered into the platform; moreover, reminder e-mails will be sent whenever there is an activity or task available in the platform, as well as the summary of the week’s chapter. Fourth, check-up sessions with parents will be carried mid-through the program. Lastly, to promote completion of follow-up phases, workshops will be offered to participant children, caregivers, and respective teachers about different themes (e.g., homework, inclusion of minorities). Nevertheless, participation is voluntary and, therefore, parents and children will be informed that they can withdraw from the program at any time. For that, participants only have to report their willingness to drop out of the program.
During the orientation session, parents who allow their child to participate in the program will complete the first assessment protocol in loco, as well as complete the informed consent form. In the informed consent, the following aspects will be covered: (i) description of the program, accompanied by a timeline, (ii) duties of the parents as guarantors of their child’s involvement in the program, and (iii) what is expected from parents regarding their involvement with the program and its activities. Children will also complete an informed consent form. Additionally, participants will be asked if they agree to the use of their data in case they choose to withdraw from the trial. Participants will be asked for permission for the research team to share relevant data with people from the Universities taking part in the research or from regulatory authorities, where relevant (e.g., the dissemination plan will be presented).
Genuine randomization at the individual level will not be possible as participants originating from the same class and school will have to be allocated to the same treatment condition to prevent between-group contamination. To address this difficulty, each school that accepts to enroll in the study will be randomly attributed a number associated to one of the groups (i.e., control, intervention, and enhanced-intervention) in a 1:1:1 basis. Moreover, for the intervention groups, small groups of five participants each will be randomly created though computed-generated random numbers.
Participants
Children from the 5th and 6th grades will be recruited. Children, instead of adolescents, were selected because it is during childhood that the formation of habits and behaviors takes place [
47]. Moreover, unlike children, adolescents are on a developmental stage where a sense of autonomy arises, and many adolescents are resistant to interventions [
48,
49]. That is, among adolescents, external control (e.g., parents) of food consumption decreases and internal control becomes increasingly predominant in the face of food choices. Moreover, considering the required degree of autonomy and skills on ICT to partake in the study, the late childhood spectrum seemed more feasible to implement the program.
For the purpose of this study, children attending the regular curriculum will be included. We will seek to exclude children attending alternative curricula, i.e., who have been identified by the school office as having special needs, since these children will not have the necessary autonomy required to attend the program. Additionally, despite the preventive nature of the present program, children will not be included/excluded from the intervention based on their weight status, nor will they need an alteration of their usual care pathways (e.g., medication) to participate in the intervention. Specific inclusion criteria include the following:
1.
Access to a computer equipped with a camera and speakers at home;
3.
Parents have to provide a written consent for their children to participate;
4.
Children have to provide a written agreement of willingness to participate;
5.
Parents have to be willing to participate in parental involvement activities;
6.
Parents have to own an e-mail account or be willing to create one.
Proposed sample size
Considering this longitudinal design, in the calculation of the sample, we considered a 0.25 effect size, an alpha level of 0.05, a desired statistical power of 0.8, five measures, and three research groups (i.e., control, online intervention, enhanced online intervention). To run a three-level cluster-randomized trial, previous analysis suggested 40 groups per group condition with 15 participants in each, measured at five different time points (see section “
Definition of the hierarchical model”).
Assessment
Table
3 illustrates the timeline for the different assessment measures of the children participating in this project.
Table 3
Timeline for the different assessment measures of the project
Eligibility screen | X | – | – | – | – | – |
Informed consent | X | – | – | – | – | – |
Anthropometric | X | – | X | X | X | – |
Sociodemographic | X | – | X | X | X | – |
Self-Regulation1 | – | X | X | X | X | – |
Self-efficacy2 | – | X | X | X | X | – |
Declarative knowledge3 | – | X | X | X | X | – |
Attitudes and perceptions4 | – | X | X | X | X | – |
Healthy eating and physical activity behaviors5 | – | X | X | X | X | – |
Satisfaction | – | – | Intervention groups | – | – | – |
Journal | – | Control group | Control group | – | – | Intervention groups |
Anthropometric measures
Children participating in the intervention and control groups will have their weight, height, and waist circumference measured by the research team at the baseline, post-intervention, and at the 3- and 6-month follow-ups. This trial does not involve collecting biological specimens for storage.
Self-report questionnaires
Sociodemographic measures
These measures will include questions about participants’ sex, age, grade, academic achievement, and socioeconomic level, as well as questions about other healthy habits and behaviors, such as their daily screen time and physical activity.
Psychological self-report measures
Primary and secondary outcome measures
The primary outcomes will be the analysis of differences in SR, self-efficacy, knowledge towards healthy eating, and, additionally, the amount of healthy and unhealthy food consumption at baseline (T0), final (T1), and follow-up (T3, T4) assessments. Secondary outcome measures will include the anthropometric (i.e., body mass index [BMI] Z-scores), sociodemographic variables (−T1), other healthy behaviors (e.g., screen time and physical activity), psychological (e.g., attitudes and perceptions about healthy eating), and engagement (e.g., weekly journal) measures. Differences within the group (i.e., at baseline, final, and follow-ups assessments), as well as between groups (i.e., control, online-intervention, and enhanced-online-intervention), will be tested over time. The repeated measures assessment allows testing relevant factors affecting SR and self-efficacy on healthy eating and food consumption. Additionally, weekly engagement indicators collected with the journals and with the actual interactions with the platform will be tested as predictors of primary outcome measures in final and follow-up assessments. Lastly, moderators and mediators that can affect the outcomes of the intervention but that are not predictor variables include sex, socioeconomic level, and academic achievement.
Statistical analysis plan
The effectiveness of the HEP-S intervention will be assessed using linear mixed-effect models. Specifically, we will assess the impact of interventions groups vs. control group on changes over time regarding the primary outcomes (e.g., healthy eating behavior, SR, self-efficacy towards healthy eating). For that, we will first examine within-intervention group differences (i.e., over time) followed by between-intervention group assessments at each time point. Differences between groups will be supported if the beta-weight parameter for the interaction between group and time are statistically significant.
Moreover, statistics descriptive analyses will be computed to describe all participants’ characteristics by using chi-square tests and independent tests to compare groups on the distributions according to sex, age group, BMI Z-score classification, and others. In addition, secondary outcomes will be analyzed by using similar models, comparing the three research groups. Also, these secondary repeated measures (e.g., satisfaction with the program) will be examined by conducting an analysis of covariance to explore interference with main outcomes. Finally, no interim analyses of the primary and secondary outcomes are planned to be included in this trial.
All analyses will be completed by using SPSS software (version 26), and p values below .05 will be considered statistically significant. The normality, homoscedasticity, and linearity of the residuals of each model will be examined to ensure that the assumptions of the models are met.
Definition of the hierarchical model
To evaluate the effectiveness of program, a three-level cluster-randomized trial is considered where classrooms are randomly assigned to treatment and control groups. Let
Ytij denote the response at time
t for
ith participant (
i = 1,…,
nj) in the
jth group (
j = 1, …,
J), e.g., the
ith student in the
jth classroom. The three-level model for longitudinal experiments that involve clustered data is as follows:
$$ \mathrm{Level}-1:\kern0.5em {Y}_{tij}={\uppi}_{0 ij}+{\uppi}_{1 ij}{T}_{tij}+{e}_{tij},\kern0.5em {e}_{tij}\sim N\left(0,{\sigma}_{et}^2\right) $$
(1)
$$ \mathrm{Level}-2:\kern0.5em {\displaystyle \begin{array}{l}{\pi}_{0ij}={\beta}_{00j}+{r}_{0 ij}\\ {}{\pi}_{1ij}={\beta}_{10j}+{r}_{1 ij},\end{array}} $$
(2)
$$ {\displaystyle \begin{array}{l}\mathrm{Level}-3:\kern0.5em \begin{array}{l}{\upbeta}_{00j}={\gamma}_{000}+{\gamma}_{001}{G}_j+{u}_{00j}\\ {}{\upbeta}_{10j}={\gamma}_{100}+{\gamma}_{101}{G}_j+{u}_{10j},\end{array}\\ {}\left[\begin{array}{l}{r}_{0 ij}\\ {}{r}_{1 ij}\end{array}\right]\sim N\left[\begin{array}{l}0\\ {}0\end{array}\right],\left[\begin{array}{ll}{\sigma}_{r0}^2& \\ {}0& {\sigma}_{r1}^2\end{array}\right],\left[\begin{array}{l}{u}_{00j}\\ {}{u}_{10j}\end{array}\right]\sim N\left[\begin{array}{l}0\\ {}0\end{array}\right],\left[\begin{array}{ll}{\sigma}_{u0}^2& \\ {}0& {\sigma}_{u1}^2\end{array}\right]\end{array}} $$
(3)
Substituting Eqs.
2 and
3 into Eq.
1, we have the mixed model of interest
$$ {Y}_{tij}={\gamma}_{000}+{\gamma}_{00j}{G}_j+{u}_{00j}+{r}_{0 ij}+{\gamma}_{100}{T}_{tij}+{\gamma}_{101}{G}_j{T}_{tij}+{u}_{10j}{T}_{tij}+{r}_{1 ij}+{e}_{tij} $$
(4)
The lowest level variance component is represented by etij, and the two level 2 variances are as follows: the variability in intercepts across subjects (r0ij) and the variability in slopes across students (r1ij) nested within classrooms. The level 3 variance components are as follows: the variability in intercepts across classrooms (u00j) and the variability in slopes across classrooms (u10j).
As Heo et al. [
54] and others (e.g., [
55]) have indicated, the most important goal of a longitudinal intervention is to test whether there are differences between intervention groups with respect to their average growth rates. Applying the approach of these authors, the power to detect a specified treatment difference between average rates of change for the groups is defined as the probability of rejecting the null hypothesis of no treatment-by-linear-trend interaction
H0: γ
101 = 0, given that it is fact false (γ
101 ≠ 0). This hypothesis can be tested with
$$ {F}_o=\frac{{\hat{\gamma}}_{101}^2}{Var\left({\gamma}_{101}\right)}=\frac{\Delta^2\left[{N}_1{N}_2{N}_3 Var(T)\right]}{2\left(1-{\rho}_1\right){\sigma}^2{N}_1 Var(T){\sigma}_{u1}^2}, $$
(5)
where Δ is the effect standardized effect size at the last time point, N1 is the number of level 1 units (repeated measures) per student, N2 the number of level 2 units (students) per classroom, N3 the number of level 3 units (classroom) per treatment condition, Var(T) is the population variance of the time variable T, \( {\upsigma}_{u1}^2 \) is the three-level random slope variance, σ2 is the sum of the other variances, and ρ1 is the correlation between outcomes measures at different time points on the same student nested within classroom.
The power of the test statistic
F0, denotes 1 −
β, can be written as follows:
$$ 1=\beta =\phi \left\{{\left\{\frac{\Delta }{\left({N}_1-1\right)}\left[\frac{\left[{N}_1{N}_2{N}_3 Var(T)\right]}{2\left(1-{\rho}_1\right){\sigma}^2{N}_1 Var(T){\tau}_s}\right]\right.}^{\frac{1}{2}}-{Z}_{1-\left(\alpha /2\right)}\right\}, $$
(6)
where τs is the ratio of the three-level random slope variance to the sum of the other variances, ϕ is the cumulative distribution function of a standard normal distribution, and Z1 − (α/2) is the 100 (1 − α/2) percentile of the standard normal distribution for a bilateral test.
Consider the following hypothetical values for three-level cluster-randomized trial design parameters: \( {\upsigma}_e^2=0.5 \), \( {\upsigma}_{r0}^2=0.3 \), \( {\upsigma}_{r1}^2=0.0 \), \( {\upsigma}_{u0}^2=0.2 \), and \( {\upsigma}_{u1}^2=0.1 \). Under the fixed slopes model, the correlation between the outcomes from different student nested within the same classroom is \( {\uprho}_2={\upsigma}_{u0}^2/\left({\upsigma}_e^2+{\upsigma}_{r0}^2+{\upsigma}_{u0}^2\right)=0.2 \), whereas the correlation between the outcomes measured at different time points on the same students nested within classrooms is \( {\uprho}_1={\upsigma}_{r0}^2+{\upsigma}_{u0}^2/\left({\upsigma}_e^2+{\upsigma}_{r0}^2+{\upsigma}_{u0}^2\right)=0.5 \). Further assumptions are as follows: (a) number of classrooms per treatment condition: N3 = 40; (b) average number of students per classroom: N3 = 15; (c) number of repeated measures per student: N1 = 5; (d) alpha significance level: α = 0.05; and (e) statistical power: 1 − β = 0.80.
Given design parameters and further assumptions, the minimum detectable effect size is 0.25. In our opinion, an effect size of Δ = .25 is a good first estimate of the smallest effect size of interest in psychological research, since it is not irrelevant and requires a considerable sample size to detect a treatment effect predicted by the researcher.
Finally, we consider that a study with 80% power is a properly powered study, despite several researchers (see, e.g., [
56]) consider this a value rather low, as it entails a 20% chance of not finding a theoretically important finding.
Blinding and data access
The assessment protocol will be completed by children on-site and will not be available for parents to consult or to the educational psychologist conducting the sessions. Each child will be attributed a unique code. Moreover, data will be analyzed by an external statistician who will be blind to the research groups of each data set. Additionally, only the principal investigator and the statistician will have access to the full data set, which will be used only for investigation. Also, all questionnaires will be destroyed once all data are published. Lastly, the procedures for personal data storage, handling, and protection will comply with the newest GDPR (General Data Protection Regulation) policies. Particularly, to ensure privacy and anonymity, pseudonymization will be carried out. This is a process that transforms personal data into a data set that cannot be linked to a particular subject unless another piece of information is added (e.g., decryption key). Usually, these data are stored in different locations.
Data monitoring and trial management
The trial will be supervised by a Trial Steering Committee (TSC) directed by the principal investigator (PI) of the trial. The TSC will meet every 4 months to monitor the trial progress, as well as review and manage the data. Besides the PI, the TSC will also include two independent researchers, one educational psychologist and a methodology expert, who are not directly involved in the study, and the CoPI. An independent statistician with no involvement in the trial will conduct all the data analysis and will inform the results to the TSC at a joint meeting. The co-investigators and research assistants of the project will be responsible for all aspects of local organization of the trial (e.g., identify potential recruits, take consent). To ensure the protocol is implemented as planned, the PI will be responsible for managing and supervising the trial and providing direction and administrative support. Three general meetings with the entire group of researchers will be held throughout the project and smaller monthly meetings will be conducted between the PI, the CoPI, and the researchers to monitor the development of activities and ensure compliance with the goals. Lastly, regular meetings will be conducted between the PI, the science manager, and the financial manager of the research center.
This trial does not have a Data Monitoring Committee (DMC) as the current trial is a low-risk intervention, without safety issues or foreseen risks associated. Moreover, the trial does not have a Stakeholder and Public Involvement Group (SPIG) as the researchers that will carry the intervention are trained and certified educational psychologists.
Safety aspects and ethical considerations
The standard procedure to conduct studies and interventions in the Portuguese school context requires an a priori evaluation and validation of the project by the Ministry of Education before researchers can invite schools to participate. Additionally, approval of the Ethics Committee of the University of Minho was obtained and can be consulted on the trial registration. Participation in the program is entirely voluntary, and parents and children are informed that their involvement in the school’s activities will not be affected by their decision on whether or not to participate. Additionally, because there are no anticipated detrimental consequences or real risks for the participants involved in this intervention, we did not include in the current trial interim assessment, stopping rules, nor criteria for discontinuing or modifying allocated interventions. In fact, parents and children are informed they can withdraw from the program at any time. We fully recognize our responsibilities for child protection, so the project must ensure that the children who participate are safe from all forms of abuse, injury, neglect, maltreatment, and exploitation. The project will ensure an environment where children feel secure, feel free to either participate or leave the study at any time, are encouraged to talk, and are listened to. Child welfare is paramount. Hence, children are safeguarded by the adoption of child protection guidelines and by a code of conduct for all involved. In order to work with children, researchers must present their criminal records during the job application. Additionally, parental consent must be obtained before carrying any procedure with children. This protocol will be held to meet high ethical standards.
Lastly, for ethical reasons, once all the assessments have been conducted, the control group will be given the opportunity to receive the intervention. This will guarantee that all children share the same benefits of the program. In this sense, all participants will have the opportunity to develop not only knowledge but also a set of transversal skills and strategies on the healthy habits’ domain. The participation in this intervention will bring no harm, and no risks were identified. Therefore, there will be no compensation for trial participation, and no post-trial care for participants is anticipated.
Dissemination
The plan for disseminating the research outcomes of the HEP-S program include the following: (a) write and publish papers in peer-reviewed journals, including the efficacy of the intervention program; (b) write a document systematizing guidelines for online interventions focused on the promotion of healthy eating through SRL; (c) conduct workshops about implications for practice targeting participants (both children and caregivers), educators, health professionals, and relevant stakeholders; and (d) organize advocacy groups comprised of children, parents, teachers, and the health community to advocate for the importance of launching policies on healthy eating habits. Advocacy groups for healthy eating habits could consider working with governmental agencies and food companies to help set environmental and structural changes likely to facilitate individual healthy choices. The work done by these advocacy groups is expected to generate public debate on this issue and, hopefully, help flourish innovative strategies to prevent obesity.
Discussion
Prevention of overweightness and obesity remains poorly effective, representing a great challenge and a major concern for governments and relevant stakeholders worldwide. Despite the numerous efforts carried to mitigate this major health issue, overweightness and obesity do not seem to be diminishing [
1]. There is, therefore, a pressing need to provide evidence-based options to improve healthful eating among individuals of all ages, particularly among children.
This study will develop and evaluate a preventive online-based intervention to promote the development of self-regulation skills for healthy eating among elementary school children. Although many online preventive interventions regarding healthful lifestyles have been developed, seldom the approaches have focused on children or skill development. Moreover, although the program has children as the target population, parents and families will be deeply involved throughout the intervention. All these aspects confer to this intervention a unique and innovative character.
An online-based intervention will, expectedly, help overcome some challenges associated with the implementation of school-based programs. For instance, educational psychologists are still a scarce resource at the school setting, being often difficult to develop programs of preventive nature instead of remedial ones. Thus, it is aimed that HEP-S helps narrow the gap between children and their access to preventive, skill promotion, psychoeducational programs, in a cost-effective format by providing at-distance support to several groups of children. This strategy allows allocating human resources more effectively as, often, an education psychologist is responsible for several schools within the same district. Furthermore, it is expected that by actively involving parents in the intervention, adherence and engagement raises. Thus, we believe the HEP-S program may play an important role in preventing risky and unhealthy eating behaviors and contribute to more healthful lifestyles within families.
This study has some strengths worth highlighting. First, it presents an online-based intervention focused on promoting SR skills and strategies to promote healthy eating behaviors among elementary school-aged children. Second, it is conceived as a preventive approach, focused on promoting health-related behaviors, not on treatment. Third, gamification strategies will be implemented to reduce high rates of attrition and low feelings of engagement with the tool. Lastly, the research design foresees an implementation period longer than most of the intervention studies in this domain, with long follow-up. This will allow searching for sustained behavioral changes as a result of the intervention.
There are potential limitations to the present study proposal. First, although we include a weekly journal for participants to complete, during the study it will be difficult to monitor the actual implementation of the strategies trained in the program. Second, full adherence to the program will be a challenge as it implies some level of parental involvement and it lasts for 20 weeks. Third, although the study design is optimized by the selection of relevant psychological variables and the introduction of gamification strategies, the small sample size of this study confers a limited scope to examine full effects. In fact, there is the possibility that smaller effects than anticipated are found. This particular aspect may pose limitations when reflecting about the implications and conclusions of the study. Nevertheless, the study attempts to minimize this aspect by having conducted a priori power analyses. If the present trial shows positive effects at the primary outcomes level, future trials should try to use larger samples, from distinct backgrounds and cultures and include longer follow-up measures (> 6 months).
Altogether, we expect that the results of this study will provide evidence on whether training and fostering SR strategies among elementary school-aged children helps in the promotion of healthy eating habits.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.