The WHO Health Promoting School framework for improving the health and well‐being of students and their academic achievement
Abstract
Background
The World Health Organization's (WHO’s) Health Promoting Schools (HPS) framework is an holistic, settings‐based approach to promoting health and educational attainment in school. The effectiveness of this approach has not been previously rigorously reviewed.
Objectives
To assess the effectiveness of the Health Promoting Schools (HPS) framework in improving the health and well‐being of students and their academic achievement.
Search methods
We searched the following electronic databases in January 2011 and again in March and April 2013: Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, PsycINFO, CINAHL, Campbell Library, ASSIA, BiblioMap, CAB Abstracts, IBSS, Social Science Citation Index, Sociological Abstracts, TRoPHI, Global Health Database, SIGLE, Australian Education Index, British Education Index, Education Resources Information Centre, Database of Education Research, Dissertation Express, Index to Theses in Great Britain and Ireland, ClinicalTrials.gov, Current controlled trials, and WHO International Clinical Trials Registry Platform. We also searched relevant websites, handsearched reference lists, and used citation tracking to identify other relevant articles.
Selection criteria
We included cluster‐randomised controlled trials where randomisation took place at the level of school, district or other geographical area. Participants were children and young people aged four to 18 years, attending schools or colleges. In this review, we define HPS interventions as comprising the following three elements: input to the curriculum; changes to the school’s ethos or environment or both; and engagement with families or communities, or both. We compared this intervention against schools that implemented either no intervention or continued with their usual practice, or any programme that included just one or two of the above mentioned HPS elements.
Data collection and analysis
At least two review authors identified relevant trials, extracted data, and assessed risk of bias in the trials. We grouped different types of interventions according to the health topic targeted or the approach used, or both. Where data permitted, we performed random‐effects meta‐analyses to provide a summary of results across studies.
Main results
We included 67 eligible cluster trials, randomising 1443 schools or districts. This is made up of 1345 schools and 98 districts. The studies tackled a range of health issues: physical activity (4), nutrition (12), physical activity and nutrition combined (18), bullying (7), tobacco (5), alcohol (2), sexual health (2), violence (2), mental health (2), hand‐washing (2), multiple risk behaviours (7), cycle‐helmet use (1), eating disorders (1), sun protection (1), and oral health (1). The quality of evidence overall was low to moderate as determined by the GRADE approach. 'Risk of bias' assessments identified methodological limitations, including heavy reliance on self‐reported data and high attrition rates for some studies. In addition, there was a lack of long‐term follow‐up data for most studies.
We found positive effects for some interventions for: body mass index (BMI), physical activity, physical fitness, fruit and vegetable intake, tobacco use, and being bullied. Intervention effects were generally small but have the potential to produce public health benefits at the population level. We found little evidence of effectiveness for standardised body mass index (zBMI) and no evidence of effectiveness for fat intake, alcohol use, drug use, mental health, violence and bullying others; however, only a small number of studies focused on these latter outcomes. It was not possible to meta‐analyse data on other health outcomes due to lack of data. Few studies provided details on adverse events or outcomes related to the interventions. In addition, few studies included any academic, attendance or school‐related outcomes. We therefore cannot draw any clear conclusions as to the effectiveness of this approach for improving academic achievement.
Authors' conclusions
The results of this review provide evidence for the effectiveness of some interventions based on the HPS framework for improving certain health outcomes but not others. More well‐designed research is required to establish the effectiveness of this approach for other health topics and academic achievement.
PICOs
Plain language summary
The WHO Health Promoting School framework for improving the health and well‐being of students and their academic achievement
Background
Health and education are strongly connected: healthy children achieve better results at school, which in turn are associated with improved health later in life. This relationship between health and education forms the basis of the World Health Organization's (WHO’s) Health Promoting Schools (HPS) framework, an approach to promoting health in schools that addresses the whole school environment. Although the HPS framework is used in many schools, we currently do not know if it is effective. This review aimed to assess whether the HPS framework can improve students’ health and well‐being and their performance at school.
Study characteristics
We searched 20 health, education, and social science databases, as well as trials registries and relevant websites, for cluster‐randomised controlled trials of school‐based interventions aiming to improve the health of young people aged four to 18 years. We only included trials of programmes that addressed all three points in the HPS framework: including health education in the curriculum; changing the school’s social or physical environment, or both; and involving students’ families or the local community, or both.
Key results
We found 67 trials, comprising 1345 schools and 98 districts, that fulfilled our criteria. These focused on a wide range of health topics, including physical activity, nutrition, substance use (tobacco, alcohol, and drugs), bullying, violence, mental health, sexual health, hand‐washing, cycle‐helmet use, sun protection, eating disorders, and oral health. For each study, two review authors independently extracted relevant data and assessed the risk of the study being biased. We grouped together studies according to the health topic(s) they focused on.
We found that interventions using the HPS approach were able to reduce students’ body mass index (BMI), increase physical activity and fitness levels, improve fruit and vegetable consumption, decrease cigarette use, and reduce reports of being bullied. However, we found little evidence of an effect on BMI when age and gender were taken into account (zBMI), and no evidence of effectiveness on fat intake, alcohol and drug use, mental health, violence, and bullying others. We did not have enough data to draw conclusions about the effectiveness of the HPS approach for sexual health, hand‐washing, cycle‐helmet use, eating disorders, sun protection, oral health or academic outcomes. Few studies discussed whether the health promotion activities, or the collection of data relating to these, could have caused any harm to the students involved.
Quality of the evidence
Overall, the quality of evidence was low to moderate. We identified some problems with the way studies were conducted, which may have introduced bias, including many studies relying on students’ accounts of their own behaviours (rather than these being measured objectively) and high numbers of students dropping out of studies. These problems, and the small number of studies included in our analysis, limit our ability to draw clear conclusions about the effectiveness of the HPS framework in general.
Conclusions
Overall, we found some evidence to suggest the HPS approach can produce improvements in certain areas of health, but there are not enough data to draw conclusions about its effectiveness for others. We need more studies to find out if this approach can improve other aspects of health and how students perform at school.
Authors' conclusions
Background
Promoting health in schools
The influence of childhood experiences on health status later in life is well documented (Felitti 1998; Galobardes 2006; Kessler 2010; Poulton 2002; Wadsworth 1997; Wright 2001). There is evidence to suggest that attitudes, beliefs, and behaviours learned during these early years ‐ for example, those relating to smoking, physical activity, and food choices ‐ show strong ‘tracking’ into adulthood (Kelder 1994; Singh 2008; Whitaker 1997). Promoting healthy habits during these early formative years is therefore of key importance.
Recognition of this has led to an interest in using schools as a means of promoting healthy behaviours in children and young people. Children spend a large proportion of their time at school and thus schools have the potential to be a powerful domain of influence on children's health. Additionally, there is a strong link between children’s health status and their capacity to learn (Powney 2000; Singh 2008). Creating positive and healthy school environments, therefore, can have numerous benefits in improving health, well‐being, and academic achievement, and reducing inequities.
Promoting health has long been an important role of schools, but traditionally activities have focused on health education, whereby information about health topics is imparted to students via the formal school curriculum, or on the development of specific skills such as communication skills or refusal techniques (Lynagh 1997). While a few programmes appear to have had some short‐term impact, there is little evidence to demonstrate that such approaches can effect sustainable behavioural change in the long term (Brown 2009; Faggiano 2005; Foxcroft 2011; Waters 2011).
The WHO Health Promoting Schools Framework
In recognition of the limited success of these interventions, a new holistic approach to school health promotion was developed in the late 1980s, influenced and underpinned by the values set out in the World Health Organization's Ottawa Charter (WHO 1986). This charter marked a significant shift in WHO public health policy, from a focus on individual behaviour to recognition of the wider social, political, and environmental influences on health.
The application of these principles to the educational setting led to the idea of the ‘Health Promoting School’ (HPS) whereby health is promoted through the whole school environment and not just through ‘health education’ in the curriculum. Thus, a Health Promoting School aims to:
-
Promote the adoption of lifestyles conducive to good health
-
Provide an environment that supports and encourages healthy lifestyles
-
Enable students and staff to take action for a healthier community and healthier living conditions (Health Education Boards 1996).
No strict definition of a Health Promoting School exists and it has been described in various ways in different documents (Denman 1999; IUHPE 2008; Lister‐Sharp 1999; Lynagh 1997; Nutbeam 1992; Parsons 1996; St Leger 1998; WHO 1997; Young 1989). The International Union for Health Promotion and Education, for example, provide a six‐point definition of Health Promoting Schools (school health policies; physical environment; social environment; individual health skills and action competencies; community links; and health services) (IUHPE 2008). Elsewhere in the literature a simpler, three‐point definition is employed, which subsumes the six points above (Denman 1999; Deschesnes 2003; Lister‐Sharp 1999; Marshall 2000; Mũkoma 2004; Nutbeam 1992; Parsons 1996; Rogers 1998; Young 1989). Additionally, whilst some interventions are explicitly labelled as adopting a HPS approach, others do not use this name but nonetheless are implicitly based upon HPS principles. In the United States, for example, this type of approach is commonly known as 'Comprehensive School Health Education'.
For the purposes of this review, we use the broad, three‐point definition of the HPS model in our selection criteria to ensure the review is inclusive of the somewhat varied and earlier approaches to HPS. According to this model, Health Promoting Schools require change in three areas of school life:
1. Formal health curriculum
Health education topics are given specific time allocation within the formal school curriculum in order to help students develop the knowledge, attitudes, and skills needed for healthy choices;
2. Ethos and environment of the school
Health and well‐being of students and staff are promoted through the ‘hidden’ or ‘informal’ curriculum, which encompasses the values and attitudes promoted within the school, and the physical environment and setting of the school; and
3. Engagement with families or communities or both
Schools seek to engage with families, outside agencies, and the wider community in recognition of the importance of these other spheres of influence on children’s attitudes and behaviours.
How Health Promoting Schools might influence health
We developed a logic model to capture the ways in which the Health Promoting Schools framework might influence health and educational outcomes (Figure 1). We identified important policy documents relevant to the intervention (HPS framework, Ottawa Charter) to inform the logic model, outlining key inputs and mechanisms of action, and providing examples of hypothesised changes in health behaviours or outcomes or both. The review authors refined and agreed the logic model.
Logic model
The Health Promoting Schools framework is based on an eco‐holistic model, recognising the physical, social, mental, emotional, and environmental dimensions of health and well‐being (Parsons 1996). The three domains described above recognise different levels of influence upon health ‐ moving from the individual, to the school environment, to the wider community context ‐ and emphasise the need to act upon all three levels in order to successfully influence health.
At the individual level, health education, through the formal curriculum, remains an important part of the HPS approach. Recognising that "to lead a healthy life is, to some degree, a matter of making the right choices" (Young 1989), students need accurate information about health issues in order to make informed choices. Thus, health education can increase knowledge and help establish positive attitudes and health behaviours. Developing the necessary skills in order to be able to act upon such information is also key; programmes may therefore emphasise communication skills, refusal techniques, and ways to promote self confidence and self efficacy. Ultimately improvements in knowledge, attitudes, and skills can enhance psychosocial health and help establish new positive social norms within the student population regarding health behaviours.
What children learn about health within the formal curriculum must be endorsed and promoted within the wider school environment to have credibility. The ‘hidden’ or ‘informal’ curriculum promoted within the school can help create a safe and supportive atmosphere that is conducive to healthy behaviours. Schools might, for example, provide secure cycle racks to promote active transport to school; implement a ‘no smoking’ policy; increase provision of healthy foods through the school catering service; develop peer mentoring approaches to tackle bullying; or increase student participation and engagement within schools through school councils.
Finally, it is important to recognise that the school environment is only one of the many domains of influence on children’s health. Families and the wider community in which children live also have an enormous impact on children’s health. It is necessary, therefore, to engage with the community beyond the school. To achieve this, schools should take into account the views and opinions of the families and communities they serve, and encourage their support and participation in health‐promoting activities. Health messages promoted at school need to be reinforced within the family and wider community settings if they are to have a significant impact on physical and social exposures and children’s behaviours.
Why it is important to do this review
A systematic review conducted in 1999 examined the impact of the HPS approach on a variety of student health outcomes (Lister‐Sharp 1999). However, the conclusions of this review were limited by the small number of studies available and weaknesses in their study designs. Results from these studies varied, but improvements in dietary intake, measures of physical fitness, self esteem, and rates of bullying were observed, and the authors concluded that there was "limited but promising" data to suggest that the HPS approach could have a positive impact on health (Lister‐Sharp 1999).
In the years since the Lister‐Sharp 1999 review was completed, interest in the HPS framework has continued to grow, with this approach being used in many countries in the absence of clear evidence of its effectiveness or potential harm. Focusing on studies with rigorous evaluation designs, we sought to re‐assess the current evidence of effectiveness of the Health Promoting Schools framework in order to inform future policy and research recommendations.
Objectives
To assess the effectiveness of the Health Promoting Schools (HPS) framework in improving the health and well‐being of students and their academic achievement.
Methods
Criteria for considering studies for this review
Types of studies
Cluster‐randomised controlled trials (RCTs), where clusters were at the level of school, district or other geographical area. As the HPS framework is an holistic, whole‐school approach, we excluded any studies where clusters were at the classroom level. We also excluded feasibility and pilot RCTs and any trials where only one school was allocated to intervention and control groups.
Public health interventions are often highly complex and context‐dependent (Rychetnik 2002), and as such may require different types of evaluative approaches. Many evaluations of the HPS framework have not been conducted using RCT methodology and offer important insights into both process and implementation. While we acknowledge the value of this body of evidence, we focus this review on cluster‐randomised trials as the most reliable form of evidence for evaluating the relative effects of interventions (Green 2011). For an overview of other evidence on the HPS framework (including non‐randomised study designs), see IUHPE 2010, Stewart‐Brown 2006 and Lister‐Sharp 1999.
Types of participants
Children and young people aged four to 18 years attending schools or colleges (including special schools). We excluded studies which covered both pre‐school and school‐aged students.
We made a post hoc change to the types of participants focused on in this review. We had originally intended to examine the impact of the Health Promoting Schools framework on staff as well as student health (Langford 2011). However, the definition of HPS interventions (as described in the published literature, referenced above) requires there to be curricular input as an essential criterion. This therefore eliminated any studies that focus on staff health, as they would not contain any curricular element. Consequently, this review is focused exclusively on students’ health and well‐being.
Types of interventions
Interventions (of any duration) based upon the HPS framework that demonstrate active engagement of the school in health promotion activities ineach of the following areas.
-
School curriculum;
-
Ethos or environment of the school or both;
-
Engagement with families or communities or both.
We present more specific inclusion criteria for these three categories in Appendix 1. Interventions did not have to explicitly state that they were based upon the HPS framework to be eligible for inclusion. If they addressed the three domains of the intervention we included them. It was not an eligibility requirement that studies reported academic outcomes.
Control schools were schools that implemented either no intervention or continued with their usual practice, or schools that implemented an alternative intervention that included only one or two of the HPS criteria.
Types of outcome measures
The HPS framework is a highly complex, multi‐dimensional intervention, which presented particular methodological challenges for this systematic review. The intervention seeks to improve ‘health’ in general, and does not restrict itself to specific health issues; the focus of each intervention is determined by the schools and researchers according to need. Thus, while individual studies may focus on a specific health topic (for example, obesity or substance misuse), the range of topics included in the review is very broad. Consequently this review defined its primary outcome ‐ health ‐ to reflect the broad focus of the HPS framework (improving health in its widest sense) as well as educational outcomes.
Primary outcomes
Health
For each health topic, we identified both positive and potentially adverse outcomes (where reported). We categorised health outcomes into the following topic areas:
-
Obesity or overweight or body size: body mass index or standardised body mass index (BMI or zBMI), height‐for‐age, weight‐for‐age, and weight‐for‐height z‐scores, skin‐fold thickness measures, waist circumference
-
Physical activity or sedentary behaviours: accelerometry, multi‐stage fitness tests (for example, shuttle runs, step tests), self‐reported levels of physical activity or sedentary behaviours
-
Nutrition: self‐reported food intake (particularly focusing on consumption of fruits and vegetables, water, high fat or sugar foods), indicators of specific nutritional deficiencies (for example, iron, iodine, and vitamin A deficiencies)
-
Tobacco use: salivary cotinine, carbon monoxide levels, self‐reported use of cigarettes or other tobacco products
-
Alcohol use: self‐reported use of alcohol
-
Other drug use: self‐reported use of other drugs (legal or illegal)
-
Sexual health: incidence of sexually transmitted infections, pregnancy or abortion, self‐reported use of condoms or other contraception, abstinence or delaying of sexual intercourse
-
Mental health and emotional well‐being: validated scales of well‐being or quality of life or both, incidence of self harm or suicide, use of validated scales such as Rosenberg’s self esteem scale, Beck Depression Inventory, Strengths and Difficulties Questionnaire
-
Violence: self‐reported violence (for example, carried weapon, got into a fight)
-
Bullying: self‐reported incidence of being bullied or bullying others
-
Infectious diseases: incidence of diseases such as diarrhoea, cold or influenza, skin disease, worms, head lice; observation or self report of hand‐washing with soap after visiting toilet or before handling food
-
Safety and accident prevention: incidence of traffic accidents or other accidents or injuries in school or at home; observation or self report of cycle‐helmet use
-
Body image or eating disorders: student (or teacher or parent) reports of disordered eating habits, body size acceptance, self esteem
-
Skin or sun safety: observation or self report of sunscreen, behaviours to reduce exposure to the sun (for example, wearing hat, seeking shade, covering up)
-
Oral health: decayed, missing or filled teeth index; self‐reported dental hygiene behaviours such as regular tooth brushing, dental check‐ups; self‐reported consumption of sugary snacks or drinks
Within each health topic, we measured outcomes using:
a. Objective measures of health or health behaviours, for example, validated methods or techniques such as BMI, accelerometry.
b. Subjective measures of health or health behaviours, for example, observation or self reports of behaviour or subjective ratings of health.
c. Measures of knowledge or attitudes or self efficacy (for example, knowledge of causes or consequences of specific health issues; attitudes towards behaviours that are known risk or protective factors for health; perceptions of one's ability to perform a certain behaviour).
Where studies presented an outcome measured in more than one way (for example, smoking in last seven days and smoking in last 30 days), we chose the category that indicated the highest frequency of the (harmful) behaviour within each respective study, assuming that this would be of the greatest public health importance.
Academic outcomes
Academic outcomes focused on: student‐standardised academic test scores, IQ tests or other validated scales; school academic performance.
Secondary outcomes
Secondary outcomes focused on:
-
School attendance outcomes.
-
Non‐academic school outcomes: for example, ratings of school climate, attachment to school, satisfaction with school.
-
Process outcomes: fidelity, acceptability, reach, and intensity of the intervention delivery.
-
Curriculum outcomes: evidence of health education topics within the formal school curriculum.
-
School environment outcomes: evidence of changes to the school’s social or physical environment or both. Examples might include: implementing no‐smoking policies, improving school catering services, developing peer mentoring programmes to tackle bullying, playground redesign.
-
Engagement with families or communities or both: participation of parents or families in relevant school‐based activities; evidence of engagement with local community organisations.
Timing of outcome assessment
The primary end point for outcome data extraction was immediately postintervention (or the closest time point to this, up to a maximum of six months postintervention). We then categorised follow‐up data after the end of the intervention (if presented) as being either short‐ (12 months or less), medium‐ (12 to 24 months) or long‐term (24 months or more).
Economic data
Where provided, we extracted data on the costs and cost effectiveness of studies.
Search methods for identification of studies
Electronic searches
We searched the following databases in January 2011. We conducted updated searches in 2013, beginning on 15 March 2013 and completed on 22 April 2013. We did not apply any date or language restrictions to our searches. Studies were not excluded on the basis of publication status. Abstracts, conference proceedings, and other 'grey' literature were included if they met the inclusion criteria.
-
Cochrane Central Register of Controlled Trials (CENTRAL) 2013, Issue 3, part of The Cochrane Library.
-
Ovid MEDLINE, 1950 to 15 March 2013.
-
EMBASE,1980 to 2013 week 16.
-
ASSIA ‐ Applied Social Science Index and Abstracts, 1987 to 2011.
-
Australian Education Index, 1979 to current.
-
BEI – British Education Index, 1975 to current.
-
BiblioMap – Database of Health Promotion Research (eppi.ioe.ac.uk/cms/).
-
CAB Abstracts, 1973 to 2013 week 11.
-
Campbell Library of Systematic Reviews (campbellcollaboration.org/lib/).
-
CINAHL ‐ Cumulative Index to Nursing and Allied Health Literature, 1982 to current.
-
Clinical Trials.gov (clinicaltrials.gov/).
-
Current Controlled Trials (controlled‐trials.com/mrct/)
-
Database of Abstracts of Reviews of Effects 2013, Issue 1, part of The Cochrane Library.
-
Database of Education Research (eppi.ioe.ac.uk/cms/).
-
Dissertation Express (dissexpress.umi.com/dxweb/search.html).
-
ERIC – Education Resources Information Centre, 1966 to current.
-
Global Health Database.
-
IBSS – International Bibliography of Social Sciences, 1950 to current.
-
International Clinical Trials Registry Platform (ICTRP) (who.int/ictrp/en/).
-
Index to Theses in Great Britain and Ireland.
-
PsycINFO, 1806 to 2013 week 10.
-
SIGLE – System for Information on Grey Literature in Europe (now known as OpenGrey) (www.opengrey.eu/).
-
Social Science Citation Index, 1956 to current.
-
Sociological Abstracts, 1952 to current.
-
TRoPHI ‐ Trials Register of Promoting Health Interventions (eppi.ioe.ac.uk/cms/).
The search strategies and search dates for these databases are shown in Appendix 2.
Searching other resources
We handsearched the reference lists of relevant articles and used citation tracking to identify and obtain relevant articles. In addition, we searched the following websites for relevant publications, including grey literature:
-
Australian Health Promoting Schools Association (www.ahpsa.org.au).
-
Barnardo’s (www.ahpsa.org.au).
-
Center for Disease Control and Prevention (www.cdc.gov).
-
Communities and Schools Promoting Health (www.safehealthyschools.org).
-
International Union for Health Promotion and Education (www.iuhpe.org).
-
International School Health Network (www.internationalschoolhealth.org).
-
National Centre for Social Research (www.natcen.ac.uk/).
-
National Children’s Bureau (www.ncb.org.uk).
-
National College for School Leadership (www.nationalcollege.org.uk).
-
National Foundation for Education Research (www.nfer.ac.uk).
-
National Healthy Schools Programme (home.healthyschools.gov.uk).
-
National Youth Agency (www.nya.org.uk).
-
Schools for Health in Europe (www.schoolsforhealth.eu).
-
School Health Education Unit (sheu.org.uk).
-
UNAIDS (www.unaids.org/).
-
UNFPA (www.unfpa.org).
-
UNICEF (www.unfpa.org).
-
World Bank (www.worldbank.org).
-
World Health Organization (www.who.int).
Several of the databases and the majority of websites that we searched in January 2011 yielded no or very few studies eligible for inclusion. The few eligible studies identified via these databases or websites were also identified through searches of MEDLINE, EMBASE, and PsycINFO. We therefore chose to exclude the following from our updated search in 2013: Global Health Database, Index to Theses in Great Britain and Ireland, Dissertation Express, SIGLE, Database of Educational Research, Bibliomap, and all websites. In addition, we no longer had access to ASSIA and therefore could not update our search of this database.
Data collection and analysis
Selection of studies
The initial search strategy produced over 35,000 reports, after removing duplicate records. A further 12,750 were retrieved in March and April 2013 after deduplication. One review author (RL) conducted an initial title screen to remove those which were obviously not pertinent to the review. For quality assurance purposes, a second review author (RC) double‐screened a random selection of 10% of these titles, yielding a kappa score of 0.88, reflecting excellent agreement. Thereafter, two authors independently screened all abstracts and full‐texts to determine eligibility. We resolved any disagreements regarding eligibility through discussion and, when necessary, in consultation with a third review author (usually RC).
Data extraction and management
For each study, two review authors (RL, and shared between LG, CB, SM, DM, and KK) independently completed data extraction forms created for the purposes of this review.
We extracted data pertaining to: basic study details (participant characteristics, study location, sample size, rates of attrition); study design and duration; intervention characteristics (including health focus, theoretical framework, content and activities, and details of any intervention offered to the control group); process evaluation of the intervention (including fidelity, acceptability, reach, intensity, and context of intervention); outcome measures postintervention and subsequent follow‐up; and costs of intervention. We used the PROGRESS PLUS check list to collect data relevant for equity (Kavanagh 2008).
Assessment of risk of bias in included studies
We assessed risk of bias within each included study using the tool outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). For each study two review authors (RL and DP) independently judged the likelihood of bias in the following domains: selection (sequence generation and allocation concealment), blinding (performance and detection bias), attrition (incomplete outcome data), reporting (selective outcome reporting), and any other potential sources of bias. For each domain, we rated studies as being at ‘high’, ‘low’ or ‘unclear’ risk of bias. We resolved any disagreements on categorisation through discussion, referring to a third review author when necessary (HJ).
Selection bias included an assessment of both adequate sequence generation and allocation concealment. We assessed sequence generation to be at low risk of bias when studies clearly specified a method for generating a truly random sequence. As all studies included in this review were cluster‐RCTs, we assessed studies as being at low risk of bias for allocation concealment if allocation was performed for all clusters at the start of the study.
The blinding domain covers both performance and detection bias. It was rarely (if ever) possible to blind students or staff to the fact that they were taking part in an intervention; we therefore assessed studies as being at high risk of performance bias unless authors explicitly stated that students were blind to group allocation. We assessed studies as being at low risk of detection bias if they clearly described the blinding of outcome assessors. If outcomes were assessed by self report, we rated the studies as being at high risk of bias where students were unlikely to have been adequately blinded.
In order to assess attrition bias we considered rates of attrition both overall and between groups, and considered whether this was likely to be related to intervention outcomes.
We assessed studies as being at low risk of reporting bias when a published protocol or study design paper was available and all prespecified outcomes were presented in the report. Where no protocol was available, we assessed studies as being at unclear risk of bias. If an outcome was specified in the study protocol but was not reported in any subsequent outcome papers, we assessed the study as being at high risk of bias.
We used the ‘other bias’ domain to note any additional concerns relating to study quality that did not fit into any of the previous five domains. For example, in this domain we included concerns about recruitment bias, baseline imbalances between groups, or selective reporting of subgroup analyses.
We assessed the overall quality of the body of evidence for each outcome using the GRADE approach (Schünemann 2011). Using this method, randomised trial evidence can be downgraded from high to moderate, low or very low quality on the basis of five factors: limitations in design or implementation (often indicative of high risk of bias); indirectness of evidence; unexplained heterogeneity; imprecision of results; or high probability of publication bias.
Measures of treatment effect
For dichotomous (binary) data, we used odds ratios (ORs) with 95% confidence intervals (CIs) to summarise results within each study. We summarised continuous outcomes using a mean difference (MD) with standard error. We extracted mean differences (adjusted for baseline) from an analysis of covariance (ANCOVA) model when these were presented. When ANCOVA results were not available we instead extracted or calculated mean differences based on final value measurements. We calculated a pooled standard deviation (SD) from intervention and control SDs at follow‐up.
Where studies used different scales to measure what we considered to represent the same underlying outcome, we first standardised results to a uniform scale by calculating standardised mean differences (SMDs). This involves dividing the estimated mean difference by the standard deviation of outcome measurements. Regardless of the method used to estimate the mean difference (ANCOVA or final values), standardisation was always performed using the standard deviation of outcome measurements at follow‐up. This was to avoid the problem of computed SMDs not being combinable across studies using different approaches to estimate the mean difference.
Where some studies reported an outcome as dichotomous and others provided a continuous measure, we converted results to the most commonly reported scale, assuming the underlying continuous measurement had an approximate logistic distribution, using methods described in Borenstein 2009 (Chapter seven).
Where data were presented separately by gender or age group, we combined these data using methods described in Borenstein 2009 (Chapter 23).
Unit of analysis issues
Interventions employing a 'whole school' approach require randomisation at the group (rather than individual) level. Where analysis took place at the school level (for example, school academic performance) no special statistical analysis is required. However, where studies reported results at the individual level, we determined whether or not the authors had accounted for the effect of clustering using appropriate statistical techniques such as multi‐level modelling. Where this had not been done (or it was not clear if it had been done), we attempted to contact the study authors to ask for the intra‐cluster correlation coefficient (ICC) and mean cluster size. This information allowed us to make an adjustment for clustering to their results before inclusion in the meta‐analyses (Higgins 2011b). If these data were not available, we examined the ICCs in similar studies. To be conservative, we selected the largest of these to adjust results prior to inclusion in the meta‐analyses.
When performing a meta‐analysis of SMDs from cluster‐RCTs, we had to decide whether to use the standard deviation of outcome measurements within clusters or the overall (‘total’) standard deviation across all individuals in a study (Grieve 2012; White 2005). The latter will be larger, since it also incorporates between‐cluster variability (specifically, Variance [total] = Variance [within clusters] + Variance [between clusters], White 2005), although the difference between the two measures is lessened if ICCs are small. Since within‐cluster standard deviations are rarely reported, we used the total standard deviation.
It is useful to have estimates of ICCs for different outcomes within different population groups to inform future research. Additional Table 1 presents the ICCs that were either reported in the included studies, or obtained via correspondence with study authors.
| Study | Country | Age | Variable | Reported intra‐cluster correlation coefficient (ICC) | Published or correspondence |
| Australia | Grade 8 | Various ‐ including substance use, depressive symptoms and school engagement. | Not specifically reported for each outcome: ranged from 0.01 ‐ 0.06 | Published | |
| Germany | Grade 2 | BMI | 0.028 (NB this is the ICC for classroom, rather than school, clustering) | Correspondence | |
| USA | K‐Grade 2 | BMI | Not specifically reported for each outcome: ranged from 0 ‐ 0.019 | Published | |
| Physical activity | |||||
| Australia | Grades 5 ‐ 6 | zBMI | 0.02 | Correspondence | |
| BMI | 0.02 | ||||
| USA | Grade 5 | Various substance use outcomes | Not specifically reported: ranged from 0 ‐ 0.01 | Published | |
| USA | K‐Grade 1 | Portions of fruit and vegetables | 0.32 | Published | |
| Finland | Grade 8 | Fat intake | 0.004 | Correspondence | |
| Fruit consumption | 0.012 | ||||
| Vegetable consumption | 0.006 | ||||
| Netherlands | Grade 3 ‐ 8 | BMI | < 0.01 | Published | |
| Waist circumference | 0.014 | ||||
| Shuttle run | 0.166 | ||||
| Switzerland | Grade 1 and 5 | BMI | 0.01 | Published | |
| MVPA (accelerometry) | 0.08 | ||||
| Shuttle run | 0.06 | ||||
| Spain | 5 ‐ 6 year‐olds | BMI | 0.094 | Correspondence | |
| USA | Grades 7 ‐ 8 | Servings of fruits and vegetables | 0.0007 | Published | |
| % energy as fat | 0.0217 | ||||
| % energy as saturated fat | 0.0134 | ||||
| Finland | Grades 4 ‐ 6 | Self‐reported victimisation | 0.02 | Published | |
| Self‐reported bullying | 0.02 | ||||
| Well‐being at school | 0.03 | ||||
| Finland | Grades 2 ‐ 3 and 8 ‐ 9 | Self‐reported victimisation | Grade 2 ‐ 3: 0.05 Grade 8 ‐ 9: 0.03 | Published | |
| Self‐reported bullying | Grade 2 ‐ 3: 0.03 Grade 8 ‐ 9: 0.02 | ||||
| USA | Grades 6 ‐ 8 | Various – unclear if just referring to alcohol use or includes other substance use outcomes | Not specifically reported: ranged from 0.002 ‐ 0.03, with a median value of .015 | Published | |
| USA | Grades 4 ‐ 5 | Fruit and vegetable consumption | 0.03 | Published | |
| Australia | Grade 8 | Depression (CES‐D scores) | 0.02 | Published | |
| USA | Grades 4 ‐ 6 | % body fat | Not specifically reported: ranged from 0.0005 ‐ 0.026 | Published | |
| zBMI | |||||
| Food intake | Not specifically reported: ranged from 0.15 ‐ 0.38 | ||||
| Physical activity | 0.05 | ||||
| Sedentary behaviour | 0.03 | ||||
| Canada | Grade 9 | Physical dating violence | 0.02 | Published |
Dealing with missing data
In the event of missing or unclear data within published studies, we attempted to contact the study authors. Where multi‐level model data were presented but authors did not provide standard errors or specific P values (and we were unable to obtain these from authors), we used final value outcome measurements and adjusted for clustering as described above (three cases). To calculate standardised mean differences, we needed to divide the effect estimate by the standard deviation of the sample. Where this was not available, we imputed the standard deviation from baseline or from another similar study (Higgins 2011b).
Assessment of heterogeneity
We assessed statistical heterogeneity among studies initially by visual inspection of forest plots. We performed Chi² tests to assess evidence of variation in effect estimates beyond that expected by chance. However, since this test has low power to detect heterogeneity when studies have small sample sizes or are few in number, we calculated I², which is an estimate of the percentage of variation due to heterogeneity rather than sampling error or chance, where a value greater than 50% indicates moderate to substantial heterogeneity (Deeks 2011). For meta‐analyses where I² was greater than 50%, we performed subgroup analyses to explore this heterogeneity.
Assessment of reporting biases
Where possible, we drew funnel plots to assess the presence of possible publication bias or small study effects (Sterne 2011).
Data synthesis
Quantitative data
The HPS framework is a flexible intervention, which can be used to target a wide range of health behaviours. We identified a number of different types of HPS interventions based broadly on the health topic(s) that the studies sought to tackle. However, we also differentiated between the different approaches that were taken to tackling specific health issues. For example, we distinguished between studies that sought to tackle overweight or obesity by targeting physical activity, those that targeted nutrition, and those that targeted both physical activity and nutrition. Similarly, we also identified what we have termed Multiple Risk Behaviour interventions (Hurrelman 2006), which sought to target multiple health outcomes with one intervention. We mapped the review outcomes to which these intervention types contributed data in Additional Table 2 and they are described in detail in Appendix 3.
| Study ID | Intervention Name | Intervention outcomes | |||||||||||||||
| Overweight/ obesity | Physical activity | Nutrition | Tobacco | Alcohol | Drugs | Sexual | Mental | Violence | Bullying | Infectious disease | Safety/ accidents | Body image | Sun safety | Oral health | Aacdemic/ attendance/ school | ||
| Nutrition interventions | |||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| Fruits and Vegetables Make the Mark | X (MA) | ||||||||||||||||
| Project Tomato | X (MA) | ||||||||||||||||
| School Nutrition Policy Initiative | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Athletes in Service, Fruit and Vegetable Promotion Program | X | ||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| TEENS | X (MA) | ||||||||||||||||
| Gimme 5 | X | ||||||||||||||||
| 5 A DAY Power Plus | X (MA) | ||||||||||||||||
| ‐ | X | ||||||||||||||||
| High 5 | X (MA) | ||||||||||||||||
| Pro Children Study | X (MA) | ||||||||||||||||
| Physical activity interventions | |||||||||||||||||
| Fit 4 Fun | X (MA) | X (MA) | |||||||||||||||
| KISS | X (MA) | X (MA) | |||||||||||||||
| ICAPS | X (MA) | X (MA) | |||||||||||||||
| ‐ | X | ||||||||||||||||
| Physical activity + nutrition interventions | |||||||||||||||||
| Heart Smart | X | ||||||||||||||||
| URMEL ICE | X (MA) | ||||||||||||||||
| Pathways | X (MA) | X (MA) | X (MA) | ||||||||||||||
| RESCATE | X | X (MA) | |||||||||||||||
| Aventuras para Niños | X (MA) | X (MA) | |||||||||||||||
| HEALTHY | X (MA) | X (MA) | |||||||||||||||
| Health in Adolescents (HEIA) | X (MA) | X (MA) | |||||||||||||||
| ‐ | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Lekker Fit | X (MA) | X (MA) | |||||||||||||||
| AVall | X (MA) | ||||||||||||||||
| CATCH | X (MA) | X | X (MA) | X | |||||||||||||
| Project Energize | X (MA) | ||||||||||||||||
| APPLES | X (MA) | X (MA) | X (MA) | X | |||||||||||||
| M‐SPAN | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Nutrición en Movimiento | X (MA) | X | X (MA) | ||||||||||||||
| Bienestar (1) | X (MA) | ||||||||||||||||
| Bienestar (2) | X | X (MA) | X (MA) | ||||||||||||||
| Louisiana (LA) HEALTH | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Tobacco interventions | |||||||||||||||||
| ESFA (Denmark) | X (MA) | ||||||||||||||||
| ESFA (Finland) | X (MA) | ||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| Project MYTRI | X | ||||||||||||||||
| ‐ | X | ||||||||||||||||
| Alcohol interventions | |||||||||||||||||
| Project Northland (Chicago) | X (MA) | X (MA) | |||||||||||||||
| Project Northland (Minnesota) | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Multiple risk behaviour interventions | |||||||||||||||||
| Positive Action (Hawai’i) | X (MA) | X (MA) | X (MA) | X | X (MA) | X | |||||||||||
| LIFT | X | X | X | X | |||||||||||||
| Aban Aya | X | X | X | ||||||||||||||
| Positive Action (Chicago) | X (MA) | X (MA) | X (MA) | X (MA) | X (MA) | X | |||||||||||
| DARE Plus | X (MA) | X (MA) | X (MA) | X (MA) | X (MA) | ||||||||||||
| Hunter Regions Health Promoting Schools Program | X (MA) | ||||||||||||||||
| Going Places | X (MA) | X (MA) | X | X | |||||||||||||
| Sexual health interventions | |||||||||||||||||
| Safer choices | X | ||||||||||||||||
| MEMA Kwa Vijana | X | ||||||||||||||||
| Mental health and emotional well‐being interventions | |||||||||||||||||
| Gatehouse | X (MA) | X (MA) | X (MA) | X (MA) | X (MA) | X | |||||||||||
| beyondblue | X (MA) | X | |||||||||||||||
| Violence interventions | |||||||||||||||||
| Students for Peace | X (MA) | ||||||||||||||||
| Fourth R | X | X | X | ||||||||||||||
| Ant‐bullying interventions | |||||||||||||||||
| Friendly Schools | X (MA) | ||||||||||||||||
| Friendly Schools, Friendly Families | X | ||||||||||||||||
| ‐ | X (MA) | X (MA) | X | ||||||||||||||
| Steps to Respect | X (MA) | ||||||||||||||||
| KiVa (1) | X (MA) | X | |||||||||||||||
| KiVa (2) | X (MA) | ||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| Hand‐washing interventions | |||||||||||||||||
| ‐ | X | X | |||||||||||||||
| ‐ | X | X | |||||||||||||||
| Miscellaneous interventions | |||||||||||||||||
| School Bicycle Safety Project / The Helmet Files | X | ||||||||||||||||
| Healthy Schools ‐ Healthy Kids | X | X | |||||||||||||||
| SunSafe | X | ||||||||||||||||
| ‐ | X | ||||||||||||||||
MA: included in meta analysis for this outcome
Our meta‐analyses present summaries of the results of these different intervention types in separate subgroups; we felt it was inappropriate to pool data overall, given the heterogeneity of approaches used to target different health outcomes. At times, subgroups may include only one study; we have, however, included these data in the forest plots so that comparisons may be made ‘by eye’ with the other intervention approaches taken.
As these complex interventions differed in terms of participants, focus, implementation, and setting, we expected the true effect of the interventions to vary between studies. We therefore performed a random‐effects meta‐analysis for each outcome on all studies reporting that outcome. As a sensitivity analysis, we also calculated fixed‐effect summary estimates. We compared the point estimates from fixed‐effect meta‐analysis to those from random‐effects meta‐analysis as a check for the influence of small study effects, as recommended in Higgins 2011b.
We present data not included in meta‐analyses in Additional Table 3. We were unable to synthesise these data in the meta‐analysis for one or more of the following reasons: we considered outcome data too different to be combined with other studies; the intervention was compared against an alternative intervention rather than standard practice or no intervention; or they were not one of the main outcomes on which this review focused.
| Study ID | Name | Type | Outcome(s) | Authors’ conclusions |
| 1. Obesity or overweight or body size | ||||
| URMEL‐ICE | Physical activity + nutrition | Skinfold thickness (tricep and subscapular), waist circumference | Intervention students had lower measures for waist circumference (‐0.64, 95% CI ‐1.25 to ‐0.02) and subscapular skinfold thickness (‐0.85, 95% CI ‐1.59 to ‐0.12). However, after adjusting for the time‐lag between baseline and follow‐up, this difference was no longer apparent. No effect was seen for tricep skinfold thickness. | |
| Aventuras para Niños | Physical activity + nutrition | zBMI | Postintervention follow‐up: Data at the end of the intervention and at 1 and 2‐years postintervention. No impact on zBMI at any time point.No difference between control and intervention groups for % body fat. Adjusted difference = 0.18; 95% CI ‐0.45 to 0.81, P value = 0.56. | |
| Health in Adolescents (HEIA) | Physical activity + nutrition | Waist circumference, waist‐to‐hip ratio | No effect seen for waist circumference or waist‐to‐hip ratio for the total sample. | |
| KISS | Physical activity | Skinfolds thickness, waist circumference | Children in intervention group showed smaller increases in the sum of 4 skinfold z‐score units (‐0.12, 95% CI ‐0.21 to ‐0.03, P value = 0.009). No effect was seen for waist circumference. | |
| CATCH | Physical activity + nutrition | Tricep and subscapular skinfold | No difference between intervention and control group for tricep skin folds (difference = 0.14 mm, 95% CI ‐0.24 to 0.52, P value = 0.47), or subscapular skinfolds (difference = 0.13 mm; 95% CI ‐0.29 to 0.54, P value = 0.553) | |
| ICAPS | Physical activity | % body fat, Fat mass index, Fat‐free mass index | Among students who were not overweight at baseline, intervention students had lower fat mass index (‐0.2, 95% CI ‐0.39 to ‐0.01, P < 0.05). There was no difference for % body fat or fat‐free mass index. No differences were seen for any of these outcomes between the 2 groups for students who were initially overweight at baseline. Postintervention follow‐up: 2 years postintervention ‐ intervention students maintained lower age ‐ and gender‐adjusted BMI (0.37 kg/m², P value = 0.02) and waist circumference (1.6 cm, P < 0.01) than control counterparts. | |
| Bienestar (2) | Physical activity + nutrition | % body fat | No difference between control and intervention groups for % body fat. Adjusted difference = 0.18 (95% CI ‐0.45 to 0.81, P value = 0.56). | |
| LA Health | Physical activity + nutrition | % body fat | No difference between control and intervention (PP + PS group). | |
| 2. Physical activity | ||||
| HEARTSMART | Physical activity + nutrition | 1 mile run or walk test | 5th grade boys’ 1 mile run or walk times decreased by 1.3 minutes in intervention group, but increased by 0.8 minutes in the control group (P < 0.01). | |
| RESCATE | Physical activity + nutrition | % children engaging in moderate and moderate‐to‐vigorous physical activity and TV or computer time. | A greater % of children in the intervention group reported being moderately physically active more than 3 days a week, compared to control children (40% I, 8% C, P value for difference between groups not given). No difference between groups for moderate‐to‐vigorous physical activity or TV or computer time. | |
| Fit 4 Fun | Physical activity | Muscular fitness and flexibility | Positive treatment effects observed in intervention children for flexibility (sit and reach, adjusted mean difference, 1.52 cm, P value = 0.003), physical activity (adjusted mean difference, 3253 steps/day, P < 0.001) and 1 measure of muscular fitness (7‐stage sit‐up, adjusted mean difference, 0.62 stages, P value = 0.003). No effect was seen for 3 other measures of muscular fitness (basketball throw, push‐ups and standing jump). | |
| Nutricion en Movimiento | Physical activity + nutrition | % children active | No difference between control and intervention group. | |
| Avall | Physical activity + nutrition | TV screen time (hours). Proportion of students taking exercise | No difference between control and intervention group for TV screen time. Intervention students were more likely to report exercising (15.7% versus 10.9%, P value = 0.036). | |
| CATCH | Physical activity + nutrition | PE lesson length. Energy expenditure and energy expenditure rate (during PE lesson) | No difference between intervention and control schools for PE lesson length. However, intervention students had greater rates of energy expenditure (0.20 kJ/kg, 95% CI 0.12 to 0.27) and a higher energy expenditure ratio (0.35 kJ/kg per hour, 95% CI 0.26 to 0.45) in PE lessons than controls. | |
| M‐SPAN | Physical activity + nutrition | Physical activity at school (observations) | There was a greater rate of increase in physical activity at school over time in intervention schools, compared to controls (d = 0.93). Subgroup analyses reveal the effect was significant only for boy (d = 1.1). | |
| ICAPS | Physical activity | TV or video time, active commuting to and from school | Children in intervention group watched less television (‐15.71 minutes per day, 95% CI ‐28.49 to ‐2.92, P value = 0.02). No difference between groups for active commuting to and from schools (1.03 mins/day, 95% CI ‐2.16 to 4.22, P value = 0.53). Postintervention follow‐up: 2 years postintervention‐ intervention students spent less time watching television (29 mins/day, P < 0.01) and had higher active transport levels (+5 mins/days, P < 0.01). | |
| ‐ | Physical activity | Self reports on travel to and from school | No difference between intervention and control groups in number of children walking to and from school. | |
| LA Health | Physical activity + nutrition | Sedentary behaviour | No difference between control and intervention (PP + PS group). | |
| 3. Nutrition | ||||
| Aventuras Para Niños | Physical activity + nutrition | Consumption of sugary drinks and snacks | No effect seen for consumption of sugary drinks. There was an initial reduction in the number of snacks consumed by intervention group (‐0.38, SE 0.17). Postintervention follow‐up: This effect on snack consumption was not sustained at follow‐up. | |
| Athletes in Service, Fruit and Vegetable Promotion Program | Nutrition | Fruit and vegetable intake | Children in intervention consumed a greater amount of fruit (34 g, 95% CI 30 to 39) than control students (23 g, 95% CI 18 to 28) (P < 0.001). | |
| AVall | Physical activity + nutrition | Consumption of fruit and vegetable, and sugary snacks or drinks | No difference between groups for proportion of children eating fruit or vegetables daily. However, there was an increase in the daily intake of > 1 piece of fruit per day (P value = 0.005). No difference between groups for consumption of sugary snacks/drinks. | |
| GIMME FIVE | Nutrition | Fruit and vegetable intake, knowledge and confidence to eat more fruit and vegetables | Intervention students had higher fruit and vegetable consumption than controls for the first 2 years of the intervention (P < 0.05), but this effect was lost by the final year of the study. Intervention students had higher knowledge scores than controls in the final 2 years of intervention (P < 0.05 for both). No group effect was seen for student confidence in eating more fruit and vegetables. | |
| ‐ | Nutrition | % skipping breakfast. Healthy breakfast choices | No difference between groups for % of children skipping breakfast. No difference between groups for reported intake of any energy‐dense, micronutrient‐poor (EDMP) food or beverage breakfast choice. | |
| High 5 | Nutrition | Fruit and vegetable intake | Postintervention follow‐up: The increased consumption of fruit and vegetables in intervention students observed at the end of the intervention was maintained 12 months later (3.2 versus 2.21 servings for intervention and control groups, respectively, P < 0.0001). | |
| M‐SPAN | Physical activity + nutrition | School‐level fat intake levels (observations) | No effect was seen on school levels of fat intake. | |
| 4. Tobacco use | ||||
| LIFT | Multiple risk behaviours | Tobacco initiation | Postintervention follow‐up: Intervention was associated with a reduced risk (10%, β = ‐0.1, P < 0.01) in tobacco use initiation. After controlling for hypothesized mediators, the intervention was associated with less likelihood of tobacco use initiation (LR Chi² = 6.69, P < 0.05). | |
| CATCH | Physical activity + nutrition | Current smoker | No difference between intervention and control students. | |
| Project Mytri | Tobacco | Smoking in last 30 days, use of chewing tobacco and bidi. | The rates of smoking cigarettes, bidi smoking and any tobacco use increased over time in the control group; the rate of any tobacco use and bidi smoking decreased in the intervention group. Overall, tobacco use increased by 68% in the control group and decreased by 17% in the intervention group. | |
| ‐ | Tobacco | Ever and regular smoking | No effect was seen for students ever trying smoking (OR 0.72, 95% CI 0.44 to 1.16, P value = 0.178) but intervention students were less likely than controls to become regular smoker (OR 0.38, 95% CI 0.16 to 0.93, P value = 0.035). | |
| 5. Alcohol use | ||||
| LIFT | Multiple risk behaviours | Alcohol use | Postintervention follow‐up: Intervention was associated with a reduced risk (7%, β = ‐0.07, P < 0.05) in alcohol use initiation. | |
| 6. Drug use | ||||
| LIFT | Multiple risk behaviours | Illicit drug use | Postintervention follow‐up: No difference between groups for illicit drug use. The intervention had a marginal effect on initiation (9%, β = ‐0.09, P < 0.10). | |
| Aban Aya | Multiple risk behaviours | Substance use | Boys in intervention group were less likely than controls to report substance use (effect size 0.45, P value = 0.05, CIs not given) but this effect was of borderline significance. No effect was seen for girls. | |
| Fourth R | Violence prevention | Problem substance use | No effect seen on problem substance use (Adj. OR 1.11, 95% CI 0.84 to 1.44 P value = 0.43). | |
| 7. Sexual health | ||||
| Safer Choices | Sexual health | Delayed sexual initiation, condom use, number of partners | No difference between groups for incidence of sexual initiation (OR 0.83, 95% CI 0.54 to 1.27, P value = 0.39). Intervention students were less likely to have sex without a condom (effect size 0.63, P value = 0.05, CIs not given) and fewer partners with whom they had sex without a condom (effect size 0.73, P value = 0.02, CIs not given). | |
| Positive Action (Hawai’i) | Multiple risk behaviours | Sexual activity | Intervention students were less likely to have had sex than control student (OR 0.18, 90% CI 0.09 to 0.36). | |
| Aban Aya | Multiple risk behaviours | Recent sexual intercourse, Condom use. | Boys in the intervention group were less likely than controls to have had recent sexual intercourse (effect size 0.65, P value = 0.2) and more likely to use a condom (effect size 0.66, P value = 0.045, CIs not given). No effect was seen for girls. | |
| MEMA Kwa Vijana | Sexual health | HIV incidence. Prevalence of other STIs. Incidence of pregnancy. Condom use. Number of partners | No difference between groups for HIV incidence or prevalence of syphilis, Chlamydia and Trichomonas. Prevalence of gonorrhoea was higher in intervention women than control (Adj. RR 1.93, 95% CI 1.01 to 3.71). There was no difference between groups in the number of pregnancies. Intervention men and women were more likely to have first used a condom during the follow‐up period than controls (men: Adj. RR 1.41, 95% CI 1.15 to 1.73; women: Adj. RR 1.30, 95% CI 1.03 to 1.63). Intervention men (but not women) were more likely than controls to have used a condom at last sex (Adj. RR 1.47, 95% CI 1.12 to 1.93) and less likely to have had >1 partner in past 12 months (Adj. RR 0.69, 95% CI 0.49 to 0.95). Postintervention follow‐up: ≈6 years postintervention ‐ no difference between groups for HIV prevalence or any other STIs, number of pregnancies and condom use. There was an increase in men reporting < 4 sexual partners (Adj. prevalence rate 0.87, 95% CI 0.78 to 0.97). | |
| Fourth R | Dating violence prevention | Condom use | No difference seen between groups for condom use (Adj. OR 1.04 95% CI 0.51 to 2.2, P value = 0.91). | |
| 8. Mental health or emotional well‐being | ||||
| ‐ | Anti‐bullying | Depression | No difference observed between groups for depression. Postintervention follow‐up: 1 year postintervention, no difference observed between groups for depression. | |
| beyondblue | Emotional well‐being | Depression | Postintervention follow‐up: No difference between groups for depression. | |
| 9. Violence | ||||
| LIFT | Multiple risk behaviours | Physical aggression in playground | Postintervention follow‐up: Intervention students showed significant reductions in physical aggression in the playground, compared to controls (‐0.11, P < 0.01). | |
| ABAN AYA | Multiple risk behaviours | Violence | Boys in intervention group were less likely than controls to report violent behaviour (effect size 0.41, P value = 0.02, CIs not given). No effect was seen for girls. | |
| Going Places | Multiple risk behaviours | Antisocial behaviour (including violence and other 'social' problems) | No effect seen for antisocial behaviour. | |
| Fourth R | Dating violence prevention | Physical dating violence, peer violence | Postintervention follow‐up: (2½ years after start of intervention) No difference was seen for physical dating violence using unadjusted ORs (1.42, 95% CI, 0.87 to 2.33, P value = 0.15). When analyses were adjusted for baseline behaviour, stratifying variables and gender, intervention students were less likely to report physical dating violence (Adj. OR 2.42, 95% CI 1.00 to 6.02, P value = 0.05) but this effect was of borderline significance. No effect was seen for physical peer violence (OR 1.09, 95% CI 0.83 to 1.59). | |
| 10. Bullying | ||||
| Friendly Schools, Friendly Families | Anti‐bullying | Being bullied, bullying others, told if saw bullying | At the end of intervention, Grade 4 students in the low‐intensity group (control) were more likely to report having been bullied than students in the high‐intensity group (OR 1.39, 95% CI 1.02 to 1.91) but no effect was seen for Grade 6 students. No effect was seen for ‘bullying others’ in either Grade cohort at the end of intervention. Grade 6 students were more likely to tell someone if they saw bullying (OR 1.78, 95% CI 1.21 to 2.62). Postintervention follow‐up: 1 year postintervention (collected for Grade 4 students only) low‐intensity group (control) students were more likely to report having been bullied (OR 1.64, 95% CI 1.06 to 2.53) or bullying others (OR 1.74, 95% CI 1.09 to 2.78). | |
| ‐ | Anti‐bullying | Being bullied, active bullying | Postintervention follow‐up: 1 year postintervention, there were no differences between intervention and control students for being bullied (rate ratio 1.14, 95% CI 0.81 to 1.59) or active bullying (rate ratio 0.7, 95% CI 0.43 to 1.29). | |
| 11. Infectious disease prevention: Hand‐washing | ||||
| ‐ | Hygiene | Illness incidence | No difference seen between groups for overall illness incidence. However, intervention schools reported a 42% decrease in student absences.Intervention students were less likely than controls to be absent due to headaches (0.54 versus 0.73 episodes per 100 student weeks, P value = 0.04) and stomach aches (0 versus 0.3 episodes per 100 student weeks, P value = 0.03). | |
| ‐ | Hygiene | Absence caused by illness (influenza‐like infections, diarrhoea, conjunctivitis) | Overall, absences caused by illness were reduced by 21% in intervention schools (5.7 versus 7.2 median episodes). Absences due to influence‐like illness were reduced by 40% (0.3 versus 0.5 median episodes), diarrhoea by 33% (0.2 versus 0.3 median episodes) and conjunctivitis by 67% (0.1 versus 0.3 median episodes). P < 0.0001 for all. | |
| 12. Safety or accident prevention | ||||
| School Bicycle Safety Project (Helmet Files) | Safety | Observed and self‐reported helmet use, helmet worn correctly | No effect seen on observed helmet use. Of those who reported not always wearing a helmet at baseline, intervention students were more likely to report always wearing a helmet at post‐test 1 (OR 1.76, 95% CI 1.09 to 2.85) but this effect disappeared at post‐test 2. | |
| 13. Body image or eating disorders | ||||
| Healthy School – Healthy Kids | Body image | Student and teachers' body satisfaction, internalisation of media ideals, body size acceptance, weight‐based teasing, disordered eating, weight loss, muscle gaining behaviours | The intervention reported a positive effect in the "internalization of media ideals" for intervention students (F [2, 596] = 3.30, P value = 0.03) and a decrease in disordered eating (only measured in girls; F [2, 276) = 2.73, P value = 0.04). No effect was seen on body satisfaction, body size acceptance or perceptions of weight‐based teasing. Compared to controls, fewer intervention students were trying to lose weight at the end of the intervention (Chi² = 4.29, P value = 0.03) but this effect was lost at 6‐month follow‐up. No effect was seen at any point for muscle‐gaining behaviour. No effect was seen for teachers on any outcome. | |
| 14. Sun safety | ||||
| Sunsafe in Middle Schools | Sun protection | % Body Surface Area covered up in sun, sunscreen application | No effect was seen on the % of body surface area covered up on observed adolescents or reported sunscreen use at first follow‐up. However, by the end of the 2nd year, students from intervention areas were likely to be more covered up than control participants (66.1% versus 56.8% body surface area covered, P < 0.01). They were also more likely to report using sunscreen at this time than control participants (47% versus 13.8%, P < 0.001). | |
| 15. Oral health | ||||
| ‐ | Oral health | Net caries increment; Restoration, sealant, and decay score; Oral health care habits reported by mothers. | No difference between groups for number of decayed, missing or filled teeth (DMFT), although there was a slight reduction in number of decayed, missing or filled surfaces (DMFS) in intervention children (0.22 versus 0.35, P value = 0.013). Intervention students had a greater mean decrease in plaque index (0.32 versus 0.21, P value = 0.013) and sulcus bleeding index (0.14 versus 0.08, P value = 0.005). Intervention children were more likely than controls to have received restorants (10.3% versus 6.2%, P value = 0.006), have sealants placed (17.5% versus 4.1%, P < 0.001) and less likely to have untreated decay (7.6% versus 20.5%, P < 0.001). Mothers of children in intervention group were more likely to report their children brushed her or his teeth, had had a dental visit within the past year and used fluoride toothpaste (P < 0.001 for all). | |
| 16. Academic, attendance, and school‐related outcomes | ||||
| Positive Action (Hawai'i ) | Multiple risk behaviours | Test scores for reading and maths, absenteeism, suspensions, retentions in grade, school climate variables | Intervention schools had higher maths and reading scores than control schools (Hawai'i Content and Performance Standards, P < 0.05 for both), lower absenteeism (P < 0.001) and fewer suspensions (P < 0.001). No effect seen for retentions in grade. The effects indicate a 2% advantage per year in the intervention group compared to the control group. Student, teacher and parent School Quality Composite scores were all higher in intervention schools compared to control (P value = 0.015, 0.006, 0.007, respectively). | |
| Gatehouse Project | Emotional well‐being | Low school attachment | Unadjusted ORs revealed no effect seen on low school attachment. However, at final follow‐up, adjusted ORs suggest an improvement in school attachment in intervention students (Adj. OR 1.33, 95% CI 1.02 to 1.75). | |
| ‐ | Hygiene | Attendance | Intervention schools (expanded group) experienced 42% fewer absence episodes (P value = 0.03) and 54% fewer days of absence (P value = 0.03) than control schools. | |
| ‐ | Anti‐bullying | School satisfaction variables | No effect seen for general satisfaction with school life; satisfaction with contact with other pupils; or satisfaction with contact with teachers. | |
| KIVA (1) | Anti‐bullying | Well‐being at school | Intervention students reported higher levels of well‐being at school (0.096, P value = 0.011) compared to the control students. | |
| Positive Action (Chicago) | Multiple risk behaviours | Standardised test scores. Student and teacher reports of academic performance, motivation and disaffection. Absenteeism. | There was a significant decrease in student disaffection with learning in the intervention group compared to those in the control schools. No effect seen on teachers' ratings of students' academic performance but a positive effect on their rating of academic motivation was found. Lower rates of absenteeism found in intervention than in control schools (β= ‐0.16, one‐tailed P value = 0.015). No evidence of a programme effect on standardised test scores for reading and maths. | |
| Healthy School, Healthy Kids | Body image | Teachers' perceptions of school's social, behavioural and nutrition or physical climate | No effect on teachers' perceptions of school climate. | |
| APPLES | Physical activity + nutrition | Self‐perceived scholastic competence | No effect on self‐perceived scholastic competence. | |
| beyondblue | Emotional well‐being | Student and teacher ratings of school climate | No effect found for student rating of school climate. Teacher ratings significantly differed between intervention and control schools over time (β = 0.60, SE = 0.29, P value < 0.05). On average, school climate in intervention schools improved over time, while in control schools it declined. | |
| Going Places | Multiple risk behaviours | Students’ perceptions of school climate | No effect seen on students’ perceptions of school climate. | |
| ‐ | Hygiene | Attendance | Overall, absences caused by illness were reduced by 21% in intervention schools (5.7 versus 7.2 median episodes). | |
CI: confidence interval; OR: odds ratio; RR: risk ratio; SE: standard error; STI: sexually transmitted infection
Qualitative data
Few qualitative data were reported for any of the included studies outside of process evaluations. The exceptions to this were qualitative data collected during formative development of interventions for the studies conducted by Perry 2009 and Te Velde 2008. Given the paucity of qualitative data, and the differing populations, contexts, and focus of the interventions, we were unable to synthesise these data.
Subgroup analysis and investigation of heterogeneity
We conducted prespecified subgroup analyses concerning intervention duration and participants’ age to explore heterogeneity between studies where I² was greater than 50%. We formally tested for differences between subgroups using meta‐regression. We classified studies as either of short (12 months or less) or long (greater than 12 months) duration. We also broadly categorised studies into those that target ‘younger’ students (12 years of age and under) and those that target ‘older’ students (over 12 years of age). Where overlap between these groupings occurred, we grouped studies according to the predominant age group. For example, a study targeting grades five to seven (10 to 13 years) would be categorised in the ‘younger’ age group.
Sensitivity analysis
Where data permitted, we undertook sensitivity analyses to explore the robustness of our findings. We assessed the impact of risk of bias in studies by restricting analyses to: (a) studies deemed to be at low risk of selection bias (associated with sequence generation or allocation concealment); (b) studies deemed to be at low risk of performance bias (associated with issues of blinding); and (c) studies deemed to be at low risk of attrition bias (associated with completeness of data). We performed additional sensitivity analyses to examine the impact of methodological choices, including: the use of standard deviations imputed from another study where original standard deviations were not available; combining accelerometry and self‐reported physical activity levels; and the choice of ‘fruit’ versus ‘vegetable’ intake where these data were presented separately.
Results
Description of studies
Figure 2 shows how references identified through searches were processed for this review. Our searches yielded 48,551 records after removal of duplicates. Of these, 46,324 were excluded on title, with a further 1097 excluded on abstract screening. We reviewed 1130 full‐text articles for eligibility. Sixty‐seven studies (from 293 reports) met the eligibility criteria for inclusion in the review.
Study flow diagram
Excluded studies
We identified 43 studies that initially appeared to be of relevance to this review but that we subsequently excluded for a variety of reasons, as documented in the Characteristics of excluded studies table. These were studies that: were not randomised or were randomised at classroom level; were pilot or feasibility studies; did not fulfil the criteria for a HPS intervention; included the wrong age‐group; targeted specific ‘at risk’ groups; or involved only two schools (one intervention, one control).
Ongoing studies
We found 11 ongoing studies that are potentially eligible for this review. These are detailed in the Characteristics of ongoing studies. Nine of these studies focus on physical activity or nutrition or both. The remaining two studies are Multiple Risk Behaviour interventions focusing on tobacco, alcohol, and drug use. In future updates of this review, we will contact authors of these studies to confirm eligibility and obtain data for inclusion in the review.
Included studies
Detailed information for each study can be found in the Characteristics of included studies tables. Below, we describe key elements of the 67 included studies. A summary of characteristics of the studies, organised by intervention type, can be found in the Study Design Table (Additional Table 4). This allows readers to assess the similarities and differences between studies in each intervention type. The outcomes to which each study contributes are mapped in Additional Table 2.
| Authors | Name | Review outcomes | Country | Target group | Duration | Theory |
| Nutrition interventions | ||||||
| ‐ | Nutrition | UK | 6 ‐ 7 and 10 ‐ 11 year‐ olds | 8 months | Health Promoting Schools framework | |
| Fruits and Vegetables Make the Mark | Nutrition | Norway | Grade 6 | 6 months. | Social cognitive theory | |
| Project Tomato | Nutrition | UK | Year 2 | 10 months | Framework for health maintenance behaviour | |
| School Nutrition Policy Initiative | Obesity or overweight. Nutrition | USA | Grades 4 ‐ 6 | 2 years | None stated | |
| Athletes in Service, Fruit and Vegetable Promotion Program | Nutrition | USA | Kindergarten and Grade 1 | 2½ years | Social learning theory | |
| ‐ | Nutrition | Finland | Grade 8 | 8 months | Social cognitive theory | |
| TEENS | Nutrition | USA | Grades 7 ‐ 8 | 2 years | Social cognitive theory | |
| Gimme 5 | Nutrition | USA | Grade 9 | 3 years | PRECEDE model of health education | |
| 5 A DAY Power Plus | Nutrition | USA | Grades 4 ‐ 5 | 6 months | Social learning theory | |
| ‐ | Nutrition | Australia | Grade 7 | 11 months | Health Promoting Schools framework | |
| High 5 | Nutrition | USA | Grade 4 | 1 year | Social cognitive theory | |
| Pro Children Study | Nutrition | Netherlands, Norway, Spain | Grades 5 ‐ 6 | 2 years | Social cognitive theory, Ecological model | |
| Physical activity interventions | ||||||
| Fit‐4‐Fun | Obesity or overweight. Physical activity | Australia | Grades 5 ‐ 6 | 8 weeks | Health Promoting Schools framework, Social cognitive theory, Harter's competence motivation theory. | |
| KISS | Obesity or overweight. Physical activity | Switzerland | Grades 1 ‐ 5 | 11 months | None stated | |
| ICAPS | Obesity or overweight. Physical activity | France | Grade 6 | 4 years | Says it is theory‐based but no details of a named theory given | |
| ‐ | Physical activity | Australia | Years 4 ‐ 5 | 2 years | Health Promoting Schools framework | |
| Physical activity + nutrition interventions | ||||||
| Heart Smart | Obesity or overweight, physical activity, nutrition | USA | Grades 4 ‐ 5 | 2½ years | Social cognitive theory | |
| URMEL ICE | Obesity or overweight, physical activity, nutrition | Germany | Grade 2 | 9 months | Social cognitive theory | |
| Pathways | Physical activity, nutrition | USA | Grade 3 | 3 years | Social learning theory | |
| RESCATE | Obesity or overweight, physical activity, nutrition | Mexico | Grades 4 ‐ 5 | 1 year | None stated | |
| Aventuras para Niños | Obesity or overweight, physical activity, nutrition | USA | K‐Grade 2 | 5 semesters | Social ecological theory, Social cognitive theory, Health belief model, Structural model of health behavior | |
| HEALTHY | Obesity or overweight | USA | Grades 6 ‐ 8 | 3 years | None stated | |
| Health in Adolescents (HEIA) | Obesity or overweight, physical activity, nutrition | Norway | Grade 6 | 20 months | Socioecological framework | |
| ‐ | Obesity or overweight, physical activity | Belgium | Grades 7 ‐ 8 | 2 years | Theory of planned behaviour, Transtheoretical model, Social cognitive theory, Attitude, Social influence and self‐Efficacy (ASE) Model | |
| Lekker Fit | Obesity or overweight, physical activity | Netherlands | Grades 3 ‐ 8 | 8 months | Theory of planned behaviour ecological model (Egger and Swinburn) | |
| Nutrición en Movimiento | Obesity or overweight, nutrition | Mexico | Grade 5 | 6 months | Not explicitly theory‐based, but does mention use of theory of peer learning for 1 element of the intervention (puppet theatre) | |
| AVall | Obesity or overweight, physical activity, nutrition | Spain | 5 ‐ 6 year‐olds | 2 years | Educational methodology 'IVAC'. | |
| CATCH | Physical activity, nutrition | USA | Grade 3 | 3 years | Social cognitive theory, Social learning theory | |
| Project Energize | Obesity or overweight | New Zealand | 5 and 10 year‐olds | 2 years | Health Promoting Schools framework | |
| APPLES | Obesity or overweight, physical activity, nutrition | UK | Years 4 ‐ 5 | 10 months | Health Promoting Schools framework | |
| M‐SPAN | Physical activity, nutrition | USA | Grades 6 ‐ 8 | 2 years | Ecological model | |
| Bienestar (1) | Physical activity, nutrition | USA | Grade 4 | 5 months | Social cognitive theory, Social ecological theory | |
| Bienestar (2) | Obesity or overweight, physical activity | USA | Grade 4 | 8 months | Social cognitive theory | |
| Louisiana (LA) HEALTH | Obesity or overweight, physical activity, nutrition | USA | Grades 4 ‐ 6 | 2½ years | Social learning theory | |
| Tobacco interventions | ||||||
| ESFA (Denmark) | Tobacco | Denmark | Grade 7 | 3 years | Attitude‐Social influence‐self‐Efficacy (ASE) model | |
| ESFA (Finland) | Tobacco | Finland | Grade 7 | 3 years | Attitude‐Social influence‐self‐Efficacy (ASE) model | |
| ‐ | Tobacco | Australia | Grade 9 students | 2 school years | Health Promoting Schools framework | |
| Project MYTRI | Tobacco | India | Grades 6 ‐ 8 | 2 years | Social cognitive theory, Social influences model | |
| ‐ | Tobacco | China | Grades 7 ‐ 8 | 2 years | Socioecological framework, PRECEDE‐PROCEED model | |
| Alcohol interventions | ||||||
| Project Northland (Chicago) | Alcohol, tobacco, drugs | USA | Grade 6 ‐ 8 | 3 years | Theory of triadic influence | |
| Project Northland (Minnesota) | Alcohol, tobacco, drugs | USA | Grades 6 ‐ 8 | 3 years. | Social learning theory | |
| Multiple risk behaviour interventions | ||||||
| Positive Action (Hawai’i) | Tobacco, alcohol, drugs, violence, sexual health, academic, and school‐related outcomes | USA | Grades 2 ‐ 3 | 3 years | Theory of self‐concept, Theory of triadic influence | |
| LIFT | Tobacco, alcohol, drugs | USA | Grades 1 and 5 | 10 weeks | Coercion theory | |
| Aban Aya | Violence, drugs, sexual health | USA | Grade 5 | 4 years | Theory of triadic influence | |
| Positive Action (Chicago) | Tobacco, alcohol, drugs, violence, academic, and school‐related outcomes | USA | Grade 3 | 6 years | Theory of self‐concept, Theory of triadic influence | |
| DARE Plus | Tobacco, alcohol, drugs, violence | USA | Grade 7 | 2 years | Theory of triadic influence | |
| Hunter Regions Health Promoting Schools Program | Tobacco | Australia | Years 7 ‐ 8 | 2 years | Health Promoting Schools framework, Community organisation theory | |
| Going Places | Tobacco, alcohol | USA | Grades 6 ‐ 8 | 3 years | Social cognitive theory | |
| Sexual health interventions | ||||||
| Safer Choices | Sexual health | USA | Grade 9 | 2 years | Social Cognitive Theory, Social Influence Theory and Models of School Change | |
| MEMA Kwa Vijana | Sexual health | Tanzania | Students aged 14+ years | 3 years | Social Learning Theory | |
| Mental health and emotional well‐being interventions | ||||||
| Gatehouse Project | Mental health and emotional well‐being, tobacco, drugs, bullying | Australia | Grade 8 | 3 years | Health Promoting Schools Framework, Attachment Theory | |
| beyondblue | Mental health and emotional well‐being | Australia | Year 8 | 3 years | Health Promoting Schools Framework | |
| Violence prevention interventions | ||||||
| Students for Peace | Violence | USA | Grades 6 ‐ 8 | 3 semesters. | Social cognitive theory | |
| Fourth R | Violence, sexual health | Canada | Grade 9 | 15 weeks | None stated | |
| Anti‐bullying interventions | ||||||
| Friendly Schools | Bullying | Australia | Grade 4 | 2 years | Health Promoting Schools framework, Social cognitive theory, Ecological theory, Social control theory, Health belief model, Problem behaviour theory | |
| Friendly Schools, Friendly Families | Bullying | Australia | Grades 2, 4, and 6 | 2 years | Health Promoting Schools framework | |
| ‐ | Bullying | Netherlands | 9 ‐ 12 year‐olds | 2 years | No specific theory but based on Olweus bullying programme | |
| Steps to Respect | Bullying | USA | Grades 3 ‐ 6 | 1 year | None stated | |
| KiVa (1) | Bullying | Finland | Grade 4 ‐ 6 | 9 months | Social cognitive theory | |
| KiVa (2) | Bullying | Finland | Grade 1 ‐ 3 and 7 ‐ 9 | 9 months | Social cognitive theory | |
| ‐ | Bullying | Belgium | 10 ‐ 16 year‐olds | Not clear | Social learning theory | |
| Hand‐washing interventions | ||||||
| ‐ | Illness from infectious diseases, attendance outcomes | China | Grade 1 | 5 months | None stated | |
| ‐ | Illness from infectious diseases | Egypt | Grades 1 ‐ 3 (for data collection, but all children in school targeted) | 12 weeks | None stated | |
| Miscellaneous interventions | ||||||
| School Bicycle Safety Project / The Helmet Files | Safety or accidents | Australia | Grade 5 | 2 years | Health Promoting Schools framework | |
| Healthy Schools‐ Healthy Kids | Body image | Canada | Grade 6 ‐ 7 | 8 months | Health Promoting Schools framework, Ecological approach | |
| SunSafe | Sun safety | USA | Grades 6 ‐ 8 | 3 years | Social cognitive theory, Socio‐ecological theory, Protection motivation theory | |
| ‐ | Oral health | China | Grade 1 | 3 years | Health Promoting Schools framework | |
Countries
Fifty‐nine of the 67 included studies were set in high‐income countries, as determined by the World Bank’s economic classification. Of these, 29 were conducted in North America (27 in USA, two in Canada), 19 in Europe (four in Finland; three in the UK; two each in Belgium, The Netherlands, and Norway; one each in Switzerland, France, Germany, Spain, and Denmark; and one multi‐country study conducted in Norway, The Netherlands, and Spain), and 11 in Australasia (10 in Australia and one in New Zealand). Of the remaining eight studies, five were conducted in upper‐middle income countries (three in China and two in Mexico), two in lower‐middle income countries (India and Egypt), and one in a low income country (Tanzania).
School types
Different countries have different schooling structures, which makes direct groupings and comparisons difficult. We describe the studies on the basis of the school ‘type’ indicated by authors. Thirty‐eight studies were conducted in primary or elementary schools (20 in primary schools and 18 in elementary schools; usually five to 11 year‐olds). Ten studies were conducted in middle schools (usually 11 to 14 year‐olds). One study from China was conducted in a junior high school (students in this study were 12 to 13 years of age). Seven studies were conducted in secondary schools (usually 11 to 16 year‐olds), and a further four were conducted in high schools (usually 14 to 18 year‐olds). A number of studies were conducted in more than one type of school: four studies were conducted in both elementary and middle or lower‐secondary schools; one study was conducted in middle and junior high schools; and one study was conducted in primary and secondary schools. A further study from Tanzania was conducted in primary schools but, as explained below, this study only included students aged 14 years and over.
Participants
All interventions took place in co‐education schools. Thus, the proportion of girls to boys was roughly 50% in all studies. Participants' ages ranged from five to 15 years (grades one to nine). Thirty‐eight studies focused on predominantly younger children (12 years of age and under), while 27 studies focused on older children (over 12 years of age). Two studies looked at both younger and older students and presented data for these separately. The majority of studies focusing on older students targeted those in grades six to eight; only four studies were conducted with students in grade nine (14 to 15 years of age).
The ethnic background of participants varied across trials. Studies conducted in the USA were the most ethnically diverse, including African American, Hispanic or Latino, Native American, Asian, and white participants. Some studies focused specifically on schools with a high proportion of a particular ethnic group. For example, the Pathways trial specifically targeted Native American students (Caballero 2003), while to be eligible for inclusion in the Aban Aya trial, (Flay 2004), schools had to have a student intake of more than 80% African American. In the studies from Europe, Australia, and New Zealand, the majority of studies did not specifically report participant ethnicity. Where it was reported, participants were predominantly white. No details of ethnicity were given for the trials conducted in China (Bowen 2007; Tai 2009; Wen 2010) , Egypt (Talaat 2011), India (Perry 2009) or Mexico (Colín‐Ramírez 2010; Levy 2012). The study conducted inTanzania (Ross 2007) provides the proportion of participants from the Sukuma tribe, as well as participants' religion.
About half of the studies did not report any measures of participants' socioeconomic status. Within the American studies that did report these data, over half targeted low‐income populations (usually indicated by percentage of students eligible for free school meals). In the remaining studies, the reported socioeconomic data appeared to broadly reflect the make‐up of the general population, with no specific emphasis on poorer populations.
Intervention duration
Twenty‐five of the studies reported on interventions that ran for less than one year (the shortest being eight weeks). The remaining studies included 41 with interventions that ran for more than one year (the longest being six years), and one study (Stevens 2000) where it was not possible to determine the length of the study. Broadly speaking, shorter interventions (12 months or less) were more likely to target physical activity or nutrition outcomes or both, while studies that focused on outcomes such as substance use, violence, sexual health or mental health tended to be of longer duration.
Postintervention follow‐up
Few studies examined the long‐term impact on outcomes once the intervention had finished. In 55 studies, the final data collection point was conducted immediately postintervention. Only 12 studies included any longer‐term data collection points after the intervention had finished. Five studies provided short‐term follow‐up (up to 12 months postintervention: Beets 2009; Cross 2012; Fekkes 2006; McVey 2004; Reynolds 2000), three provided medium‐term follow‐up data (between 12 and 24 months: Crespo 2012; Sawyer 2010; Simon 2006), and four provided long‐term follow‐up data (24 months and over: Eddy 2003; Luepker 1998; Ross 2007; Wolfe 2009).
Theoretical framework
Only 15 of the 67 studies were explicitly labelled as using the Health Promoting Schools framework to inform their intervention. Of these, 10 studies were from Australia (Bond 2004; Cross 2011; Cross 2012; Eather 2013; Hall 2004; Hamilton 2005; Radcliffe 2005; Sawyer 2010; Schofield 2003; Wen 2008), two from the UK (Anderson 2005; Sahota 2001), and one each from Canada (McVey 2004), New Zealand (Rush 2012), and China (Tai 2009).
All but 10 of the included studies stated that their intervention was informed by a named theory. A total of 22 different theoretical models were identified, although many studies were informed by more than one theoretical model. The most commonly cited theory was the social cognitive theory (20 studies), followed by ecological or socioecological models (11 studies), social learning theory (eight studies), and the theory of triadic influence (five studies).
Intervention focus
Half of the studies (34) focused on physical activity or nutrition or both, with the aim of decreasing overweight, obesity or associated risks for cardiovascular disease and Type II Diabetes. Of these 34 studies, four focused on physical activity, 12 focused on nutrition, and the remaining 18 studies targeted both of these areas.
Seven studies focused on bullying, five studies focused specifically on tobacco use, and we identified two studies for each of the following individual outcomes: alcohol use, sexual health, violence, mental health or emotional well‐being, and hand‐hygiene. Seven studies evaluated Multiple Risk Behaviour interventions that focused on a number of health behaviours in one programme. Different groups of topics were targeted in each intervention but included: alcohol, tobacco, drug use, sexual health, violence, and bullying. In addition, there were four studies that focused on ‘unique’ health topics. We identified only one study for each of the following health topics: accident prevention (cycle‐helmet use), eating disorders, sun protection, and oral health. The different intervention types and the outcomes on which they report are mapped in Additional Table 2.
Academic, attendance and school‐related outcomes
Few studies attempted to measure any form of academic attendance or school‐related outcomes. Just two studies presented any type of academic‐related outcomes (including student test scores, suspensions, and retentions in grade: Beets 2009; Li 2011) and only three presented any attendance data (Beets 2009; Bowen 2007; Talaat 2011). A further seven studies presented other school‐related outcomes: low school attachment (Bond 2004), school satisfaction (Fekkes 2006), school climate (McVey 2004; Sawyer 2010; Simons‐Morton 2005), well‐being at school (Kärnä 2011), and self perception of scholastic competence (Sahota 2001).
Process data
Some form of process data were presented in 54 of the 67 studies included in this review, although not all of these studies explicitly stated that they carried out a specific process evaluation of the intervention delivery. Thirteen studies did not provide any process data (Anderson 2005; Arbeit 1992; Colín‐Ramírez 2010; Kärnä 2011; Kärnä 2013; Llargues 2011; Perry 2003; Rush 2012; Levy 2012; Stevens 2000; Tai 2009; Trevino 2005; Wolfe 2009). Of the 54 studies presenting process data, the majority used quantitative methods only (41 studies), nine studies used both qualitative and quantitative methods, one study presented qualitative data only (in‐depth interviews, Wen 2008), and in three studies it was not possible to determine the methods used to collect the data (Hall 2004; McVey 2004; Sallis 2003).
In total, 48 studies provided data on how the intervention was implemented in schools (fidelity or intensity). This included documentation of the number of activities provided (for example, number of classroom sessions, newsletters sent out), assessment of how much of the intended intervention was implemented, reasons why full implementation was not achieved, and assessment of the quality of implementation (for example, lesson quality). In addition, 27 studies provided some data on the acceptability of the intervention to students, staff, and sometimes families. Very little information was provided across studies about the context in which interventions were implemented or discussion of causal pathways linking interventions with outcomes.
Economic data
Eight studies provided some indication of the costs involved in implementing their interventions, but only two studies (Basen‐Engquist 2001; Brandstetter 2012) provided comprehensive cost‐effectiveness analyses. Because these eight interventions varied in terms of outcomes, settings, and duration, it is not possible to draw any conclusions on the costs or cost effectiveness of these interventions. Details of the costs are summarised in Additional Table 5.
| Name | Approach | Country | Duration | Costs | Cost effectiveness |
| Nutrition | UK | 8 months | Costs estimated to be GP 378 for capital and development costs plus GBP 13.50 consumables per school | ‐ | |
| Sexual health | USA | 2 years | The total cost of the intervention was USD 105,243. | For every dollar invested in the program, USD 2.65 in total medical and social costs were saved. | |
| Physical activity and nutrition | Germany | 9 months | Intervention costs were EUR 24.09 per child. | The incremental cost‐effectiveness relation was EUR 11.11 (95% CI, 8.78 to 15.02) per cm waist circumference growth prevented and EUR 18.55 (95% CI, 14.04 to 26.86) per unit of waist‐to‐height ratio gain prevented. The authors conclude that based on a ‘maximum willingness to pay’ of EUR 35, the intervention can be considered cost‐effective. | |
| Tobacco | Finland | 3 years | Estimated costs per school each year were EUR 2500. | ‐ | |
| Nutrition | USA | 2½ years | No costs associated with the school‐wide loud‐speaker announcements or the CD‐ROM element which was available to schools free of charge. Costs associated with the lunchtime component were USD 0.04/sticker and a one‐time cost of approximately USD 100 to print the posters. Each family book cost USD 3.38. | ‐ | |
| Sexual health | Tanzania | 3 years | The 3‐year costs of trial implementation were USD 879,032. Initial start‐up costs were high but annual costs dropped from USD 16 per student in 1999 to USD 10 per student in 2001. Authors estimate that when scaled up, only an additional USD 1.54 is needed per pupil per year to continue the intervention. | ‐ | |
| Physical activity and nutrition | New Zealand | 2 years | Average cost estimated to be less than NZD 40. | ‐ | |
| Dating violence prevention | Canada | 15 weeks | Estimated costs of CAD 16 per student in initial year. Includes teacher release time for training (CAD 200 x 40 teachers = CAD 8000) and reusable curriculum materials (mean, CAD 700 per school or CAD 175 per teacher). | ‐ |
CI: confidence interval
Equity
We sought to identify studies which reported on characteristics known to be important from an equity perspective. The most commonly reported characteristics at baseline were participants’ gender (52 studies) and age (40 studies). About half of the studies (34 studies) also reported some indicator of socioeconomic status, for example: household income; eligibility for free or reduced‐price school meals; parental occupation or education levels; or area indices of deprivation. An indication of participants’ ethnicity was provided in 36 studies. Sixteen studies presented data on participants’ household structure, usually expressed as the proportion of students living in two‐parent households.
When analysing data on outcomes, 21 studies reported the effect of their intervention by gender, 10 reported effects by age or grade, six reported effects by ethnicity, and two studies reported effects by level of parental education.
Adverse events and outcomes
The majority of studies (57 studies) did not report any details on whether they had recorded any adverse events or outcomes as a result of the intervention. Of those studies that did record these data, seven studies reported no adverse events (Caballero 2003; Eather 2013; Eddy 2003; Foster 2008; Hamilton 2005; Tai 2009; Wolfe 2009), while three reported adverse events described below.
Foster 2010 and Grydeland 2013 reported adverse events related to data collection methods only and not to the intervention itself. The HEALTHY study (Foster 2010) reported that 2.4% of students experienced an adverse event at baseline and 1.7% of students at follow‐up; the most common event was dizziness during blood tests. The HEIA trial (Grydeland 2013) reported that approximately 2% of students had experienced an adverse event during health screening, again most commonly reported as dizziness.
The MEMA Kwa Vijana (Ross 2007) sexual health intervention implemented in Tanzania reported more serious adverse outcomes potentially associated with the trial, for a small minority of participants. These included reports of pregnant school girls being punished and expelled from school; rumours within the community that the curriculum materials were promoting immoral behaviour; and reports of sexual relationships and abuse between male teachers and students (although some of these instances preceded the trial).
Interpretation and implementation of the HPS framework
We aimed to describe how the HPS framework had been interpreted and implemented by documenting changes within the three HPS domains (curriculum, ethos or environment or both, and family or community or both). The majority of studies provided a brief description of the intervention and rarely gave details on exactly how the intervention had been implemented within the schools. We provide details on the intervention components (as described by study authors) for each individual study in the Characteristics of included studies tables. The following provides a brief summary of the types of activities undertaken within the three HPS domains, although obviously the specific content and activities of interventions varied according to the health topic(s) targeted. A more comprehensive description of these activities by intervention type is provided in Appendix 3.
Input into the curriculum
Intervention curricula focused primarily on providing information about particular health topics (for example, importance of physical activity or the health consequences of substance use), practising skills (for example, problem‐solving, refusal techniques, resisting peer pressure or general social or behaviour skills), and increasing students' self confidence and self efficacy.
Changes to ethos or environment or both
A common method used in a number of different types of interventions was to set up a school working group or committee, often composed of staff, students, and parents or community members or both. The aim of these committees was usually to assess current school practices, to develop or revise relevant health policies, and to implement a school‐wide plan to improve health outcomes. Social marketing campaigns were another commonly‐used method by which schools promoted health messages beyond the classroom in the wider school environment. These included posters, information displays, public service announcements, school assemblies, ‘health weeks’, competitions, and theatre productions. Staff training sessions were often implemented and some interventions used peer‐led activism or support groups. Interventions targeting physical activity or nutrition or both often made direct changes to the variety and quality of food served in school canteens, as well as making changes to the structure of the school day to provide greater opportunities for physical activity throughout the day (for example, during lessons, before or after school or during break times). In some studies, changes to the physical environment of the school were implemented, for example: increased provision of soap to facilitate hand‐washing, provision of games equipment to encourage physical activity, or changes to school boundaries to increase access to shaded areas.
Engagement with families or communities or both
Activity within this domain appeared to be the least intensive of the three HPS areas. The majority of studies only attempted to engage with families (rather than the community), most commonly by sending out newsletters to parents. Other activities included: family homework assignments, parent information evenings or training workshops, family events, or inviting parents to become members of the school health committee. The aim of this work was to provide parents with information about the curricular content of the intervention and to provide advice on how to support these messages within the home environment.
Fewer studies actively sought to engage with the local community. Examples of activities in this area include: inviting members from local organisations to join the school health committee or to give guest lectures to students; asking local policy makers to assess the ‘walkability’ of the local area or provide low‐cost access to sports facilities; improving local parks; asking local restaurants to provide healthy children’s menus; reminding local shops not to sell cigarettes to students; displaying intervention posters in local community settings; and conducting field trips to relevant organisations or institutions.
Risk of bias in included studies
We summarise the risks of bias across all domains for all studies included in the review in Figure 3.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Allocation
Overall, only 15 studies were assessed as being at low risk of bias for sequence generation. All remaining studies were assessed as being at unclear risk because authors simply stated that clusters had been ‘randomised’ without providing any further details on how this had been done.
We deemed 60 studies to be at low risk of bias for allocation concealment because allocation was performed for all clusters at the start of the study. In four studies, not enough detail was provided to assess how clusters had been allocated. We rated three studies at high risk of bias (Bowen 2007; Wen 2008; Wolfe 2009) because it was potentially possible to predict in advance to which group a school would be allocated.
Blinding
Because of the nature of these interventions, we deemed the majority of studies to be at high risk of bias because it was unlikely that participants could be adequately blinded to the fact they were taking part in an intervention. Three studies explicitly reported that students were blind to group allocation. Of these, one study stated that there was no evidence that students were aware of their group allocation and was rated at low risk of performance bias (Trevino 2004). We rated the remaining two studies as ‘unclear’ because it was not possible to determine how successful this blinding process was (Wen 2010; Wolfe 2009).
The fact that participants were unable to be adequately blinded had an impact on our assessment of blinding of outcome assessors. The majority of outcomes presented in studies were subjective, self‐reported measures; thus the outcome assessors (usually the participants themselves) were not blind. While alternatives to self reports may not be available or feasible for some health outcomes, the reliance on self reports in such studies does cause concern over the reliability of the data collected, especially when students know they are taking part in an intervention study. Although studies may have promised confidentiality to participants, because most data collection took place in school, it is possible that concerns over teachers or parents or both accessing student data may have increased reporting bias.
For the 25 studies that included some objective measures of health outcomes (for example, body mass index or standardised body mass index (BMI or zBMI), physical fitness tests), we assessed whether or not the outcome assessors were blind to group allocation. Eight studies reported that outcome assessors had been adequately blinded and thus were rated as being at low risk of bias (Caballero 2003; Crespo 2012; Eather 2013; Foster 2010; Kriemler 2010; Rush 2012; Tai 2009; Trevino 2004). Thirteen studies failed to provide any details on blinding of outcome assessors and were classified as being at unclear risk of bias. Four studies were assessed as being at high risk of bias because outcome assessors were not blind to group allocation (Foster 2008; Grydeland 2013; Sallis 2003; Simon 2006).
Incomplete outcome data
We assessed 18 studies as being at low risk of attrition bias. These studies had low overall levels of attrition, with missing data relatively balanced between study groups and judged unlikely to be related to study outcomes. For a further 15 studies it was not possible to determine the likelihood of attrition bias, due to a lack of clear information in study reports. We assessed 34 studies as being at high risk of bias due to the following reasons: high overall levels of attrition; significant differential attrition between study groups; loss of clusters; and significant differences between those who dropped out and those who completed the trial, which may have been related to outcomes measures.
Selective reporting
For the majority of studies (46) the risk of reporting bias was unclear; no protocol was available and therefore it was not possible to assess whether authors reported all relevant outcomes as intended. We assessed six studies as being at low risk of bias; a protocol (or study design paper) was available and all outcomes were reported (Caballero 2003; Eather 2013; Grydeland 2013; Luepker 1998; Ross 2007; Simon 2006). Fifteen studies were deemed to be at high risk of bias (Bond 2004; Colín‐Ramírez 2010; Cross 2012; Evans 2013; Foster 2010; Hoffman 2010; Jansen 2011; Kriemler 2010; Olson 2007; Rush 2012; Schofield 2003; Stevens 2000; Te Velde 2008; Trevino 2005; Williamson 2012). In these studies, either a protocol was available but not all outcomes had been reported, or a protocol was not available but there was reason to suspect that selective reporting had occurred (see Characteristics of included studies for more details).
Other potential sources of bias
We assessed 35 studies as being at low risk of other sources of bias. A further three studies provided insufficient data to be able to adequately assess their risk of other sources of bias. We rated 29 studies as being at high risk of bias (Anderson 2005; Arbeit 1992; Bond 2004; Bowen 2007; Brandstetter 2012; Cross 2012; De Vries (Denmark) 2003; Grydeland 2013; Hamilton 2005; Hoffman 2010; Hoppu 2010; Jansen 2011; Kriemler 2010; McVey 2004; Orpinas 2000; Perry 1996; Radcliffe 2005; Rush 2012; Sahota 2001; Sallis 2003; Sawyer 2010; Schofield 2003; Levy 2012; Stevens 2000; Te Velde 2008; Trevino 2004; Wen 2010; Williamson 2012; Wolfe 2009). Reasons for this assessment largely related to the external validity of the trials, such as low participation rates or important baseline imbalances or both between groups.
Missing data
We contacted authors from 29 studies to obtain missing data; 16 studies provided some or all of the data we required; data were not available for four studies; and we received no response from a further nine studies.
Assessment of quality of evidence
GRADE assessments for the quality of evidence for each outcome are summarised in Additional Table 6. In most cases, we assessed the quality of evidence to be low to moderate. While all included studies were cluster‐RCTs, evidence from these was often downgraded on the basis of risk of bias assessment (particularly concerning blinding and attrition) and unexplained heterogeneity.
| Review outcome | GRADE assessment | Justification |
| Obesity or overweight or body size | Moderate | RCT evidence downgraded on basis of high levels of unexplained heterogeneity |
| Physical activity | Low/moderate | RCT evidence downgraded on basis of high levels of unexplained heterogeneity and risk of bias (blinding of participants) for physical activity, but not physical fitness measures |
| Nutrition | Low | RCT evidence downgraded on basis of high levels of unexplained heterogeneity and lack of blinding of outcome measures |
| Tobacco | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants and attrition) |
| Alcohol | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Substance use | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Sexual health | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Mental health | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Violence | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Bullying | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Infectious disease | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Accident prevention | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants and attrition) |
| Body image or eating disorders | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants and attrition) |
| Skin or sun safety | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Oral health | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Academic or attendance outcomes | Moderate | RCT evidence downgraded in basis of risk of bias (attrition) |
RCT: randomised controlled trial.
The quality of the body of evidence from randomised trials is usually assessed as 'high' within the GRADE system. However, randomised trial evidence can be downgraded to moderate, low or very low quality on the basis of five factors: limitations in the design and implementation (often indicative of high bias risk); indirectness of evidence; unexplained heterogeneity or inconsistency of results; imprecision of results; and high probability of publication bias. For further description of GRADE levels of quality of a body of evidence see section 12.2 in Higgins 2011a.
Effects of interventions
Obesity or overweight or body size outcomes
Nineteen studies reported obesity or overweight or body size outcomes, of which 13 were included in the meta‐analysis for BMI and nine for the meta‐analysis of zBMI. No study focused on under‐nutrition or growth faltering. Other outcomes related to obesity or overweight or body size that were not synthesised in a meta‐analysis are presented in section 1 of Additional Table 3. These include measures of percentage body fat, skin fold thickness, waist circumference, and waist‐to‐hip ratio.
Of the 19 studies, 15 focused on both physical activity and nutrition (Brandstetter 2012; Caballero 2003; Crespo 2012; Foster 2010; Grydeland 2013; Haerens 2006; Jansen 2011; Levy 2012; Llargues 2011; Luepker 1998; Rush 2012; Sahota 2001; Sallis 2003; Trevino 2004; Williamson 2012), three focused on physical activity alone (Eather 2013; Kriemler 2010; Simon 2006), and one focused on nutrition (Foster 2008). Only two studies were specifically informed by the HPS framework (Rush 2012; Sahota 2001). Eight studies were conducted in the USA, eight in Europe (one each in UK, Belgium, Switzerland, France, Norway, Germany, Spain, and The Netherlands) and one study each was conducted in Australia, New Zealand, and Mexico. Thirteen studies focused on younger‐aged children (12 years of age and under). One study focused on grades three to eight (eight to 14 years of age), while the remaining five studies targeted older children (grades six to eight). Seven were implemented for less than one year (ranging from eight weeks to 11 months). One study ran for 20 months, seven studies ran for two to two and a half years, three studies ran for three years, and one study ran for four years.
Measures
Ten studies presented data on students’ BMI, six studies presented sex‐ and age‐adjusted zBMI scores, and two studies presented both BMI and zBMI. (Sallis 2003 calculated BMI based on student‐reported height and weight data). As zBMI is the preferred measure, because it allows for more meaningful comparisons of BMI levels between children of different ages, we present meta‐analyses for BMI and zBMI separately. Where studies presented both BMI and zBMI we have included both these data in the separate meta‐analyses.
Effectiveness
Analysis 1.1 and Analysis 1.2 present the results for the meta‐analyses for BMI and zBMI by intervention type. There is evidence that physical activity interventions were able to reduce BMI in students. These studies showed an average reduction in BMI of 0.38 kg/m² (95% confidence interval (CI) 0.73 to 0.03; 3 trials, 1430 participants) relative to control schools. Although there was a large amount of heterogeneity (I² = 86%), all three studies gave evidence in favour of the intervention. Nine studies targeted physical activity + nutrition and showed an average reduction in BMI of 0.11 kg/m² in intervention schools relative to controls, but with a wide confidence interval that crossed the null value (95% CI ‐0.24 to 0.02; 9 trials, 13,628 participants). The single nutrition intervention (Foster 2008), which measured BMI as an outcome, did not show any impact (mean difference (MD) ‐0.04, 95% CI ‐0.28 to 0.20; 843 participants).
When zBMI was used (which accounts for age and gender), only the single physical activity intervention (Eather 2013) showed a significant effect (MD ‐0.47, 95% CI ‐0.69 to ‐0.25; 196 participants). There was no evidence of effect for the nutrition only or the physical activity + nutrition interventions.
Follow‐up data
Only two studies presented any follow‐up data on overweight or obesity‐related outcomes (Crespo 2012; Simon 2006). These results are presented in section 1 of Additional Table 3.
Physical activity or sedentary behaviours
Eighteen studies reported outcomes related to physical activity or sedentary behaviours or both, of which nine contributed data to the meta‐analysis for physical activity, and five to the meta‐analysis for physical fitness. Three studies presented other physical activity data that could not be combined in the meta‐analysis (Colín‐Ramírez 2010; Crespo 2012; Wen 2008); results for these outcomes are described in section 2 of Table 3.
Four of these 18 studies focused only on promoting physical activity (Eather 2013; Kriemler 2010; Simon 2006; Wen 2008), 13 studies focused on both physical activity and nutrition (Arbeit 1992; Caballero 2003; Colín‐Ramírez 2010; Crespo 2012; Grydeland 2013; Haerens 2006; Jansen 2011; Llargues 2011; Sahota 2001; Sallis 2003; Trevino 2004; Trevino 2005; Williamson 2012), and one study focused on nutrition only, despite presenting outcome data for physical activity (Foster 2008). Three studies were specifically informed by the HPS framework (Eather 2013; Sahota 2001; Wen 2008). Eight studies were conducted in the USA, seven in Europe (Belgium, France, Switzerland, The Netherlands, United Kingdom, Norway, and Spain), two in Australia, and one in Mexico. Fourteen studies focused on younger‐aged children (12 years of age and under). One study focused on Grades three to eight (eight to 14 years of age), while three studies targeted older students (over 12 years of age). Seven studies reported on interventions that were implemented for up to one year (ranging from eight weeks to 12 months). One study ran for just under two years, seven studies ran for two to two and a half years, one study ran for three years, and one study ran for four years.
Measures
Physical fitness was measured in three studies using 20 metre shuttle runs (Eather 2013; Jansen 2011; Kriemler 2010) and in two studies using a modified version of the Harvard step test (Trevino 2004; Trevino 2005). For assessments of physical activity, four studies used student self reports (Haerens 2006; Sahota 2001; Simon 2006; Williamson 2012), one used observations (Sallis 2003), and four studies objectively measured physical activity using accelerometry (Caballero 2003; Grydeland 2013; Haerens 2006; Kriemler 2010). Two studies provided self‐reported data for all children with a subset of participants also providing accelerometry data (Caballero 2003; Haerens 2006). In this case, we chose to include the more objective measure of accelerometry in the meta‐analysis. Because physical activity and physical fitness outcomes were reported on different measurement scales, we converted results to standardised mean differences (SMDs) before pooling across studies.
Effectiveness
Analysis 2.1 and Analysis 2.2 present the results for the meta‐analyses for physical activity and physical fitness by intervention type. On average, across six studies, there was evidence that physical activity + nutrition interventions produced a small increase in physical activity in intervention students relative to control schools (SMD 0.14, 95% CI 0.03 to 0.26; 6 trials, 6190 participants) but there was a large amount of heterogeneity (I² = 66%). When analysis was restricted to just those studies using accelerometry data, heterogeneity was reduced (to I² = 0%) and the size of the effect increased slightly (SMD 0.18, 95% CI 0.10 to 0.26) (see Additional Table 7). The two physical activity interventions showed inconsistent results with one (using self reports) favouring the intervention (Simon 2006) and the other (using accelerometry) showing no effect (Kriemler 2010) (I² = 93%). There was no evidence of an effect for the single nutrition only intervention (Foster 2008).
| Accelerometry vs. self reported physical activity | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Physical activity | Physical activity only | accelerometry | 1 | 297 | 205 | 0.01 [‐0.01 to 0.03] | n/a |
| self report | 1 | 374 | 358 | 0.35 [0.17 to 0.53] | n/a | ||
| Physical activity + nutrition | accelerometry | 3 | 1475 | 1341 | 0.18 [0.10 to 0.26] | 0% | |
| self report | 3 | 1769 | 1605 | 0.12 [‐0.15 to 0.38] | 85% | ||
| Using 'vegetable intake' instead of 'fruit intake' where these were reported separately | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Fruit and vegetable intake | Nutrition only | fruit intake | 10 studies, 3 substitutions | 3293 | 2917 | 0.15 [0.02 to 0.29] | 83% |
| vegetable intake | 10 studies, 3 substitutions | 3293 | 2917 | 0.14 [0.01 to 0.27] | 83% | ||
| Physical activity + nutrition | fruit intake | 6 studies, 3 substitutions | 3507 | 3105 | 0.04 [‐0.18 to 0.26] | 79% | |
| vegetable intake | 6 studies, 3 substitutions | 3507 | 3105 | ‐0.07 [‐0.19 to0.04] | 26% | ||
| Excluding studies with borrowed standard deviations (SDs) | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| zBMI | Physical activity + nutrition | with borrowed SDs | 7 | 5672 | 5512 | ‐0.00 [‐0.04 to 0.03] | 41% |
| without borrowed SDs | 6 | 4980 | 4852 | ‐0.01 [‐0.05 to 0.03] | 39% | ||
| Fat intake | Nutrition only | with borrowed SDs | 7 | 2205 | 2011 | ‐0.08 [‐0.21 to 0.05] | 68% |
| without borrowed SDs | 4 | 1183 | 986 | 0.00 [‐0.08 to 0.08] | 27% | ||
| Physical activity + nutrition | with borrowed SDs | 10 | 6498 | 5962 | ‐0.04 [‐0.20 to0.12] | 95% | |
| without borrowed SDs | 9 | 6197 | 5643 | ‐0.00 [‐0.17 to 0.17] | 95% | ||
| Fruit and vegetable intake | Nutrition only | with borrowed SDs | 9 | 3293 | 2917 | 0.15 [0.02 to 0.29] | 83% |
| without borrowed SDs | 6 | 2188 | 1865 | 0.05 [‐0.06 to 0.16] | 67% | ||
| Physical activity | Physical activity + nutrition | with borrowed SDs | 6 | 3244 | 2946 | 0.14 [0.03 to 0.26] | 66% |
| without borrowed SDs | 5 | 3108 | 2804 | 0.14 [0.01 to 0.27] | 72% | ||
| Alcohol use | Alcohol intervention | with borrowed SDs | 2 | 3477 | 3817 | 0.72 [0.34 to1.52] | 82% |
| without borrowed SDs | 1 | 2501 | 3079 | 0.99 [0.97 to 1.01] | n/a | ||
| Random‐ versus fixed‐effect meta‐analyses | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Fruit and vegetable intake | Nutrition only | random | 9 | 2205 | 2011 | ‐0.08 [‐0.21 to 0.05] | 68% |
| fixed | 9 | 2205 | 2011 | ‐0.05 [‐0.10 to 0.00] | 68% | ||
| Alcohol use | Multiple risk behaviours | random | 4 | 4496 | 3644 | 0.75 [0.55 to 1.02] | 78% |
| fixed | 4 | 4496 | 3644 | 0.88 [0.78 to 1.00] | 78% | ||
| Substance use | Multiple risk behaviours | random | 3 | 3804 | 3016 | 0.57 [0.29 to1.14] | 71% |
| fixed | 3 | 3804 | 3016 | 0.76 [0.60 to 0.96] | 71% | ||
| Violence | Multiple risk behaviours | random | 3 | 3806 | 3014 | 0.50 [0.23 to 1.09] | 93% |
| fixed | 3 | 3806 | 3014 | 0.89 [0.82 to 0.96] | 93% | ||
| Bullying others | Anti‐bullying | random | 6 | 13949 | 12227 | 0.90 [0.78 to 1.04] | 67% |
| fixed | 6 | 13949 | 12227 | 0.81 [0.77 to 0.87] | 67% | ||
| ALLOCATION CONCEALMENT | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Being bullied | Anti‐bullying | All studies | 6 | 13993 | 12263 | 0.83 [0.72 to 0.96] | 61% |
| Low risk only | 4 | 12438 | 10694 | 0.85 [0.71 to 1.03] | 76% | ||
| BLINDING OF OUTCOME ASSESSORS FOR OBJECTIVE MEASURES | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| BMI | Physical activity + nutrition | All studies | 9 | 6520 | 7108 | ‐0.11 [‐0.24 to 0.02] | 84% |
| Low risk only | 1 | 727 | 682 | ‐0.20 [‐0.53 to 0.13] | n/a | ||
| zBMI | Physical activity + nutrition | All studies | 7 | 4980 | 4852 | ‐0.01 [‐0.05 to 0.03] | 39% |
| Low risk only | 3 | 3184 | 3172 | ‐0.01 [‐0.08 to 0.05] | 52% | ||
| Physical activity | Physical activity + nutrition | All studies | 6 | 3244 | 2946 | 0.14 [0.03 to 0.26] | 66% |
| Low risk only | 3 | 1475 | 1341 | 0.18 [0.10 to 0.26] | 0% | ||
| Physical fitness | Physical activity + nutrition | All studies | 3 | 2059 | 2171 | 0.12 [0.04 to 0.20] | 0% |
| Low risk only | 1 | 619 | 602 | 0.13 [0.01 to 0.25] | n/a | ||
| LOW ATTRITION RATES | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| BMI | Physical activity + nutrition | All studies | 9 | 6520 | 7108 | ‐0.11 [‐0.24 to 0.02] | 84% |
| Low risk only | 5 | 4095 | 4705 | ‐0.11 [‐0.29 to 0.07] | 76% | ||
| zBMI | Physical activity + nutrition | All studies | 7 | 4980 | 4852 | ‐0.01 [‐0.05 to 0.03] | 39% |
| Low risk only | 3 | 3544 | 3402 | ‐0.02 [‐0.05 to 0.02] | 0% | ||
| Physical activity | Physical activity + nutrition | All studies | 6 | 3244 | 2946 | 0.14 [0.03 to 0.26] | 66% |
| Low risk only | 2 | 428 | 443 | ‐0.03 [‐0.31 to 0.26] | 68% | ||
For physical fitness, there was evidence that physical activity + nutrition interventions were effective at increasing fitness levels in students (SMD 0.12, 95% CI 0.04 to 0.20; 3 trials, 4230 participants). Heterogeneity was large (I² = 82%) but the estimated effect in all three studies was in the direction of a benefit of the intervention. In addition, the two physical activity only interventions both showed a positive effect, but in one study the estimated effect was marginal (Kriemler 2010), while in the other (Eather 2013) it was moderate. Therefore, the resulting summary effect from a random‐effects meta‐analysis was positive, but with a wide confidence interval that crossed the null value (SMD 0.35. 95% CI ‐0.20 to 0.90, I² = 95%; 2 trials, 694 participants).
Follow‐up data
Only Simon 2006 presented any follow‐up data (two years postintervention); these results are presented in section 2 of Additional Table 3.
Nutrition
Twenty‐three studies reported on nutrition or diet‐related outcomes, of which 17 contributed data to the meta‐analysis for fat intake and 13 to the meta‐analysis for fruit and vegetable intake. Of these, 12 focused on nutrition alone (Anderson 2005; Bere 2006; Evans 2013; Foster 2008; Hoffman 2010; Hoppu 2010; Lytle 2004; Nicklas 1998; Perry 1998; Radcliffe 2005; Reynolds 2000; Te Velde 2008) and 11 focused on physical activity and nutrition (Caballero 2003; Colín‐Ramírez 2010; Crespo 2012; Foster 2010; Haerens 2006; Luepker 1998; Sahota 2001; Sallis 2003; Levy 2012; Trevino 2004; Williamson 2012). Two studies were specifically informed by the HPS framework (Anderson 2005; Sahota 2001). Thirteen were conducted in the USA, seven in Europe (three in the United Kingdom, one each in Norway, Finland, Belgium, and one multi‐country study), one in Australia, and two in Mexico. Sixteen studies focused on younger‐aged children (12 years of age and under) while seven studies targeted older students in grades six to nine (over 12 years of age). Eleven studies were implemented for less than one year, six studies were implemented for two years, two studies were implemented for two and a half years, and four studies were implemented for three years.
Measures
Nutrition intake was most commonly measured through student‐reported 24‐hour recalls. Hoppu 2010 used a 48‐hour recall period and Anderson 2005 used a three‐day food diary to assess food intake. Foster 2008, Haerens 2006, Levy 2012, and Sallis 2003 used food frequency questionnaires to assess nutritional intake. Williamson 2012 used digital photography to measure food selection and intake. Because outcomes were reported on different measurement scales, we converted results to SMDs before pooling across studies.
Three studies presented consumption of fruit and vegetables as two separate outcomes (Foster 2010; Hoppu 2010; Sahota 2001). In this case we used data for ‘fruit consumption’ in the meta‐analysis. A sensitivity analysis confirmed that using ‘vegetable consumption’ instead made no difference to our conclusions (Additional Table 7).
Other outcomes related to nutrition that were not synthesised in a meta‐analysis are presented in section 3 of Additional Table 3. These outcomes include measures of children’s consumption of sugary drinks or foods or breakfast intake. The intervention reported by Hoffman 2010 provided outcome data on intake of fruits and vegetables; however, because it was compared against an alternative intervention rather than standard practice, we did not include it in the meta‐analyses.
Analysis 3.1 and Analysis 3.2 present the results for the meta‐analyses for fat intake, and fruit and vegetable intake by intervention type. These analyses demonstrate that there was a large degree of heterogeneity in these outcomes across studies. On average across seven studies assessing the impact of nutrition only interventions on reducing self‐reported fat intake, the effect was in the direction of a slight benefit of the interventions (SMD ‐0.08) but the 95% CI was also consistent with the null hypothesis of no effect (‐0.21 to 0.05; 7 trials, 4216 participants). These nutrition only interventions, however, were effective on average at increasing reported fruit and vegetable intake among students (SMD 0.15, 95% CI 0.02 to 0.29, I² = 83%; 9 trials, 6210 participants). No overall effect was seen for physical activity + nutrition interventions on either fat intake or fruit and vegetable intake, although there was a very large degree of heterogeneity (I² = 95% and 79%, respectively), with some individual studies showing statistically significant effects in opposite directions.
Long‐term follow‐up
One study (Reynolds 2000) presented long‐term follow‐up data (12‐months postintervention); these results are presented in section 3 of Additional Table 3.
Tobacco
Fourteen studies provided data on tobacco use, of which 10 contributed data to the meta‐analysis (Beets 2009; Bond 2004; De Vries (Denmark) 2003; De Vries (Finland) 2003; Hamilton 2005; Li 2011; Perry 1996; Perry 2003; Schofield 2003; Simons‐Morton 2005). We did not include data in the meta‐analysis. We considered the studies conducted in India (Perry 2009) and China (Wen 2010) to be too dissimilar in context to be combined with data from high‐income countries such as the USA. The study by Eddy 2003 had no data available immediately postintervention. We did not include data from Luepker 1998 in the meta‐analysis since this intervention was primarily aimed at physical activity and nutrition outcomes. The results from these studies are summarised in section 4 of Additional Table 3.
Tobacco interventions
Five studies focused specifically on preventing or reducing tobacco use among students. Two of these studies came from the European Smoking Prevention Framework Approach (ESFA); this was a six‐country study conducted in Denmark, Finland, The Netherlands, Spain, Portugal, and the UK. Implementation of the intervention elements varied between countries and only two countries (Finland and Denmark) implemented a programme that met the HPS criteria, and were truly randomised. We have therefore included data from these two countries only, treating them as two separate studies (De Vries (Denmark) 2003; De Vries (Finland) 2003). These studies targeted students 12 to 13 years of age and the programme was implemented for three years.
Hamilton 2005 was conducted in Australia and was explicitly designed around the HPS framework. It targeted students in grade nine (14 to 15 year‐olds) and focused largely on harm minimisation (rather than prevention). It was implemented for two years. A study by Wen 2010 was conducted in Chinese schools in grades seven to eight (12 to 14 year‐olds) and was implemented for two years. Another study conducted by Perry 2009 was implemented in India, focusing on tobacco use among students in grades six to eight (11 to 14 year‐olds). This intervention also ran for two years.
Multiple risk behaviour interventions
Six multiple risk behaviour interventions reported tobacco use outcomes (Beets 2009; Eddy 2003; Li 2011; Perry 2003; Schofield 2003; Simons‐Morton 2005). All were conducted within the USA, with the exception of Schofield 2003, which was conducted in Australia and was specifically informed by the HPS framework. One of these studies was implemented for just 10 weeks (Eddy 2003). The remaining studies were implemented for two to six years. Three studies focused on younger children (12 years of age and under) (Beets 2009; Eddy 2003; Li 2011), and three on older children (Perry 2003; Schofield 2003; Simons‐Morton 2005).
Other interventions
A further three studies reported tobacco use outcomes but were not exclusively focused on this topic. Perry 1996 focused primarily on reducing alcohol use, but also measured impact on smoking outcomes. It was conducted in the USA with students in grades six to eight and was implemented for two years. The Gatehouse Project focused on emotional well‐being but measured impact on students' substance use. It was conducted in Australia with students in grade eight and was implemented for three years. Finally, the CATCH study conducted by Luepker 1998 was an intervention focused primarily on physical activity and nutrition, but included a very small element of smoking prevention in the fifth grade.
Measures
All studies used self reports of students’ behaviours to assess tobacco use.
Effectiveness
Analysis 4.1 presents the results for the meta‐analyses for tobacco use by intervention type. There is good evidence that both tobacco only (odds ratio (OR) 0.77, 95% CI 0.64 to 0.93, I² = 16%; 3 trials, 4747 participants) and multiple risk behaviour (OR 0.84, 95% CI 0.76 to 0.93, I² = 0%; 5 trials, 9992 participants) interventions are effective in reducing smoking in school‐aged children, with the estimated effect for the former being slightly larger. The alcohol intervention (Perry 1996), which also looked at the impact on tobacco use, also showed a positive intervention effect (OR 0.74, 95% CI 0.61 to 0.90; 1901 participants). The single emotional well‐being intervention gave an estimated effect in favour of the intervention (OR = 0.79) but with a wide confidence interval (95% CI 0.59 to 1.06; 630 participants).
Follow‐up data
Eddy 2003 presented follow‐up data over seven years (grades six to 12); results from this study are summarised in section 4 of Additional Table 3.
Alcohol
Eight studies provided data on alcohol use and all but one were included in the meta‐analysis. Eddy 2003 did not provide outcome data immediately postintervention and so could not be combined with data from other studies. The results from this study are described in section 5 of Additional Table 3.
Alcohol interventions
Only two studies targeted alcohol use as the main focus of their intervention. Project Northland was implemented in Minnesota, USA in 1991 (Perry 1996). It aimed to prevent alcohol use among adolescents (students in grades six to eight), although it also collected outcome data on tobacco and marijuana use. The intervention was conducted in three stages over seven years, but only the first phase met the HPS criteria. (Phase II did not include a curricular element through the intervention period, see Perry 2002). We therefore restrict analyses to the first three years of the intervention.
An adapted version of Project Northland was implemented in a separate trial in Chicago in 2002 (Komro 2008). Again, this intervention primarily targeted alcohol use, but also included tobacco and drug use as secondary outcomes. It targeted the same age group (grades six to eight) and was implemented for three years.
Multiple risk behaviour interventions
Five multiple risk behaviour interventions reported alcohol use outcomes (Beets 2009; Eddy 2003; Li 2011; Perry 2003; Simons‐Morton 2005). All of these studies were conducted in the USA. Three studies focused on younger children (12 years of age and under) (Beets 2009; Eddy 2003; Li 2011) and two studies targeted students in grades six and eight (Perry 2003; Simons‐Morton 2005). One of these studies was implemented for just 10 weeks (Eddy 2003), one study was implemented for two years (Perry 2003), two studies were implemented for three years (Beets 2009; Simons‐Morton 2005), and one study was implemented for six years (Li 2011).
Other interventions
Bond 2004 used an emotional well‐being intervention, which presented data on student alcohol use. It was conducted in Australia with students in grade eight and was specifically informed by the HPS framework. It was implemented for three years.
One final study (Schofield 2003) stated that it implemented an intervention to target alcohol, smoking, and sun safety. This study was informed by the HPS framework. However, data from this study were only presented for smoking outcomes and therefore we cannot provide any data from this study for this outcome.
Measures
All studies used self reports of students’ behaviours to assess alcohol use.
Effectiveness
Analysis 5.1 presents the results for the meta‐analyses for alcohol use by intervention type. Overall, there was no evidence that any of the different intervention approaches were effective in reducing alcohol intake.
The two alcohol only interventions produced conflicting results, with confidence intervals that do not overlap, one suggesting a positive effect of the intervention on alcohol intake (Perry 1996; OR 0.45, 95% CI 0.24 to 0.87; 1714 participants) and the other suggesting no effect (Komro 2008; OR 0.99, 95% CI 0.97 to 1.01; 5580 participants).
The multiple risk behaviour interventions similarly produced conflicting results. The two Positive Action trials both indicated a positive effect of the intervention, but with very wide confidence intervals (Beets 2009 OR 0.48, 95% CI 0.32 to 0.73; 1714 participants; Li 2011 OR 0.44, 95% CI 0.21 to 0.92; 363 participants). In contrast, the remaining two studies found no effect (Perry 2003 OR 0.95, 95% CI 0.80 to 1.13; 4743 participants; Simons‐Morton 2005 OR 0.97, 95% CI 0.80 to 1.18; 1320 participants).
The emotional well‐being intervention similarly found no effect (Bond 2004 OR 1.13, 95% CI 0.76 to 1.67; 1619 participants).
Follow‐up data
Eddy 2003 presented follow‐up data over seven years (Grades six to 12); we summarise results from this study in section 5 of Table 3.
Substance use
Nine studies provided data on substance use and six of these were included in the meta‐analysis (Beets 2009; Bond 2004; Komro 2008; Li 2011; Perry 1996; Perry 2003). We could not include three studies in the meta‐analysis either because they did not provide outcome data immediately postintervention (Eddy 2003; Wolfe 2009) or because the intervention was compared against an alternative intervention rather than standard practice (Flay 2004). The results from these studies are described in section 6 of Additional Table 3.
Multiple risk behaviour interventions
Five multiple risk behaviour interventions reported substance use outcomes (Beets 2009; Eddy 2003; Flay 2004; Li 2011; Perry 2003). All of these studies were conducted in the USA. Beets 2009, Li 2011, and Flay 2004 focused on younger children (12 years of age and under), while Eddy 2003 and Perry 2003 targeted older students. One study was implemented for 10 weeks (Eddy 2003). The remaining studies were implemented between two and six years.
Alcohol interventions
Two studies were primarily focused on alcohol use but also included data on other student substance use. Komro 2008 and Perry 1996 were implemented in the USA with students in Grades six to eight for three years.
Other interventions
A further two studies presented substance use outcomes. Wolfe 2009 was a Canadian intervention that sought to reduce dating violence. It targeted students in grade nine and was implemented for 15 weeks. The emotional well‐being intervention by Bond 2004 also reported on substance use. It was conducted in Australia with grade eight students for three years and was specifically informed by the HPS framework.
Measures
All studies used self reports of students’ behaviours to assess substance use. In most cases, studies looked at cannabis use or did not specify which drugs the intervention sought to target.
Effectiveness
Analysis 6.1 presents the results for the meta‐analyses for substance use by intervention type. Overall, there was no evidence that any of the three intervention approaches were effective in reducing substance use.
One multiple risk behaviour intervention (Beets 2009) found a positive effect on substance use (OR 0.28, 95% CI 0.13 to 0.63; 1714 participants). The two other multiple risk behaviour interventions also showed effects in favour of the intervention, but in both cases their confidence intervals overlapped the null value (Li 2011; Perry 2003).
Neither the alcohol only interventions (Komro 2008; Perry 1996) nor the emotional well‐being intervention (Bond 2004) showed evidence of effectiveness.
Follow‐up data
Eddy 2003 presented follow‐up data over seven years (Grades six to 12); we summarise results from this study in section 6 of Additional Table 3.
Sexual health
Five studies reported on student sexual health outcomes. We considered the results of the interventions reporting on sexual health outcomes too heterogeneous in terms of approach, setting, and outcomes to combine them in a meta‐analysis.
Sexual health interventions
Only two studies focused specifically and exclusively on sexual health. The Safer Choices intervention was conducted in the USA and focused on students in grade nine (Basen‐Engquist 2001). This intervention lasted two years. The second study was from Tanzania (Mema Kwa Vijana, Ross 2007). This study was implemented with students aged 14 years and over. The one‐year intervention was conducted for three consecutive years. A long‐term follow‐up was conducted six years after the end of the original intervention.
Other interventions
A further three studies reported sexual health outcomes but were not exclusively focused on this topic. Two of these (Beets 2009; Flay 2004) were multiple risk behaviour interventions targeting sexual health among a suite of other health behaviours. These studies were conducted in the USA, were implemented for three or four years, and targeted younger (12 years of age and under) children. The study by Flay 2004 compared the intervention against an alternative ‘Health Enhancement Curriculum’ rather than usual practice. The Fourth R intervention (Wolfe 2009) was conducted in Canada and targeted grade nine students (14 to 15 year‐olds). It primarily focused on preventing dating violence but also reported on condom usage. It lasted for one semester (15 weeks) and collected outcome measures two and a half years later.
Measures
All studies used student self reports of sexual behaviours, including having had sexual intercourse and use of condoms. However, Mema Kwa Vijana (Ross 2007) also included laboratory testing to determine HIV incidence and prevalence of other sexually‐transmitted infections (STIs).
Effectiveness
As it was not possible to meta‐analyse data from these studies, we summarise the results of the individual studies in section 7 of Additional Table 3.
Follow‐up data
Mema Kwa Vijana (Ross 2007) presented long‐term follow‐up data six years postintervention: we present results from this study in section 7 of Additional Table 3.
Mental health and emotional well‐being
Three studies presented data on student mental health (depression) and we include all three in the meta‐analysis.
Emotional well‐being interventions
Two studies focused on mental health and emotional well‐being. The beyondblue project (Sawyer 2010) was a three‐year intervention programme aimed at students in grade eight (13 to 14 year olds), which sought to reduce depressive symptoms and increase individual‐level protective factors (such as social skills and coping skills). TheGatehouse Project (Bond 2004) was similarly targeted at students in grade eight and was implemented for three years. It sought to increase emotional well‐being and reduce rates of substance use, known to be related to emotional well‐being. Both of these interventions were implemented in Australia and were explicitly designed around the Health Promoting Schools framework.
Other interventions
Only one other study reported any mental health or emotional well‐being outcomes (Fekkes 2006). This anti‐bullying intervention was implemented for two years and targeted children aged nine to 12 years.
Measures
All three studies used validated but different scales to assess levels of student depression. Sawyer 2010 used the Center for Epidemiologic Studies Depression scale, a 20‐item scale describing a wide range of depressive symptomatology (Radloff 1977). Bond 2004 used a computerised version of the revised clinical interview schedule (CIS‐R), a structured psychiatric interview for non‐clinical populations (Angold 1995). Fekkes 2006 used the Short Depression Inventory for Children (Kroesbergen 1996). In all three cases, higher scores indicated greater risk of depression. Because these three studies reported outcomes using different measurement scales, we converted results to SMDs before pooling across studies.
Effectiveness
Analysis 7.1 presents the results for the meta‐analyses for depression by intervention type. Overall, there was no evidence that these interventions were effective at reducing rates of depression in students. Indeed, for the two studies focused specifically on mental health and emotional well‐being, there appears to be a trend in the opposite direction with intervention students reporting poorer mental health (OR 0.06, 95% CI ‐0.00 to 0.13; 2 trials, 6099 participants). The anti‐bullying intervention by Fekkes 2006 found no effect on levels of depression in students.
Long‐term follow‐up
Follow‐up data are presented for Sawyer 2010 (one and two years postintervention) and Fekkes 2006 (one year postintervention); these results are presented in section 8 of Additional Table 3.
Violence or aggressive behaviours
Eight studies presented data on violent or aggressive behaviours in students, of which we include four in the meta analysis (Beets 2009; Li 2011; Orpinas 2000; Perry 2003). The remaining four studies were not included in the meta‐analysis for the following reasons. Eddy 2003 and Wolfe 2009 did not provide data immediately postintervention. The intervention implemented by Flay 2004 was compared against an alternative intervention rather than usual practice. Simons‐Morton 2005 reported on ‘anti‐social behaviour’, which aggregated both violence and other ‘social’ problems in one score. The results of these studies are presented in section 9 of Additional Table 3.
Violence prevention interventions
Two studies focused specifically on preventing violence and aggressive behaviours. Students for Peace (Orpinas 2000) was an American programme implemented over three semesters with sixth to eighth grade students (11 to 14 year‐olds) in middle schools. It aimed to reduce aggressive behaviours between students. The Fourth R intervention (Wolfe 2009) was implemented in Canada with grade nine students (14 to 15 year‐olds) over one semester. However, this intervention focused specifically on preventing dating violence. In this intervention, dating violence prevention was integrated with lessons on healthy relationships, sexual health, and substance use.
Multiple risk behaviour interventions
A further six studies focused on violence as an outcome within a multiple risk behaviour intervention (Beets 2009; Eddy 2003; Flay 2004; Li 2011; Perry 2003; Simons‐Morton 2005). All of these studies were conducted in the USA. One of these studies was implemented for 10 weeks (Eddy 2003). All of the remaining studies were long‐term interventions implemented for between two and six years. Four of these studies focused on younger children (12 years of age and under) and two focused on students in grades six to eight (12 to 14 year‐olds).
Measures
All studies used self reports of students’ behaviours to assess violent behaviours.
Effectiveness
Analysis 8.1 presents the results for the meta‐analyses for violence by intervention type. Overall, there was no evidence that violence prevention or multiple risk behaviour interventions were effective in reducing violent behaviour in students.
The violence prevention intervention by Orpinas 2000 found no effect on rates of student violence. The multiple risk behaviour interventions produced conflicting results. The two Positive Action trials both found evidence of a reduction in violent behaviours (Beets 2009 OR 0.32, 95% CI 0.16 to 0.62; 1714 participants; Li 2011 OR 0.38, 95% CI 0.25 to 0.56; 363 participants). However, the large study by Perry 2003 found no evidence of effect (OR 0.93, 95% CI 0.86 to 1.01; 4743 participants).
Follow‐up data
Follow‐up data are presented for Wolfe 2009 (two and a half years after the start of the intervention) and Eddy 2003 (over seven years); these results are presented in section 9 of Additional Table 3.
Bullying
Ten studies reported on bullying outcomes (being bullied or bullying others), with eight contributing data to the meta‐analysis for being bullied (Bond 2004; Cross 2011; Fekkes 2006; Frey 2005; Kärnä 2011; Kärnä 2013; Perry 1996; Stevens 2000) and seven contributing data to the meta‐analysis on bullying others (Cross 2011; Fekkes 2006; Frey 2005; Kärnä 2011; Kärnä 2013; Li 2011; Stevens 2000). All interventions were compared against usual practice, with the exception of Friendly Schools, Friendly Families (Cross 2012), where all control schools received the Friendly Schools anti‐bullying manual but had no further input. We therefore excluded this study from the two bullying meta‐analyses; the results from this study are reported in section 10 of Additional Table 3.
Anti‐bullying interventions
Seven studies focused specifically on reducing or preventing incidence of bullying in schools. These studies were conducted in Belgium (Stevens 2000), The Netherlands (Fekkes 2006), Finland (Kärnä 2011; Kärnä 2013), Australia (Cross 2011; Cross 2012), and the United States (Frey 2005). The two studies implemented in Australia were Friendly Schools and a follow‐up intervention called Friendly Schools, Friendly Families, which extended the Friendly Schools programme to include greater family input. Both of these studies were specifically informed by the Health Promoting Schools framework. The two studies conducted in Finland evaluated the effectiveness of the KiVa programme in different grades of children (grades four to six in the first study and grades one to three and seven to nine in the second study). Although these studies were evaluating the same intervention, they randomised different schools for each evaluation, and we therefore treat them as two separate studies.
Three studies were implemented for one school year and two for two years. Frey 2005 was implemented for two years, but control schools received the intervention after the first year of implementation. For the purposes of this review, therefore, we have only included data from the first year of this programme. It was unclear in the case of one study exactly how long the intervention had been implemented (Stevens 2000).
Five studies focused on younger students (usually 12 years of age and under). Kärnä 2013 included both younger (grades one to three) and older (grades seven to nine) students. Stevens 2000 focused on students aged 10 to 16 years.
Other interventions
A further three studies presented bullying outcomes but were not exclusively focused on this topic. Two of these were multiple risk behaviour interventions conducted in the USA (Li 2011; Perry 2003). Li 2011 focused on younger children (12 years of age and under) and was implemented for six years. Perry 2003 targeted older students in grade seven and was implemented for two years. The final study focused on promoting emotional well‐being in students and was conducted in Australia (Bond 2004). It targeted students in grade eight and ran for three years.
Measures
All studies used self reports of students’ behaviours to assess bullying behaviour.
Effectiveness
Analysis 9.1 and Analysis 9.2 present the results for the meta‐analyses for being bullied and bullying others by intervention type. Anti‐bullying interventions showed an average reduction of 17% for reports of being bullied (OR 0.83, 95% CI 0.72 to 0.96, I² = 61%; 6 trials, 26,256 participants), relative to control schools, although there was a considerable amount of heterogeneity. For bullying others, the confidence interval for the pooled effect crossed the null (OR 0·90, 95% CI 0·78 to 1·04, I² = 67%; 6 trials, 26,176 participants), but the two largest studies (Kärnä 2011; Kärnä 2013) investigating the same intervention showed strong evidence of an effect. The emotional well‐being intervention (Bond 2004 ) failed to show any impact on both being bullied and bullying others. No effect was seen for being bullied for the single multiple risk behaviour intervention reporting this outcome ( Perry 2003). However, another multiple risk behaviour intervention (Li 2011) reported the effect of their intervention on bullying others and found evidence of a large reduction in this behaviour (OR 0.49, 95% CI 0.34 to 0.71; 363 participants).
Follow‐up data
Two studies presented follow‐up data after one year (Cross 2012; Fekkes 2006); we present the results from these studies in section 10 of Additional Table 3.
Infectious disease prevention
Two studies focused on preventing infectious disease by promoting hand‐hygiene among primary or elementary school students. Both studies were implemented in middle‐income countries (China and Egypt).
The study by Bowen 2007 focused on promoting hand‐washing in schools to reduce illness and absences from school. This intervention was conducted in rural primary schools in China’s Fujian province and targeted first grade students (seven to eight year‐olds). The exact length of the intervention is unclear but data were collected over a five‐month period. The study by Talaat 2011 similarly focused on promoting hand‐washing to reduce infectious disease and absenteeism. It was conducted in elementary schools in Cairo, Egypt over 12 weeks. The intervention targeted all school students but outcome data were collected for children in grades one to three.
Measures
In Bowen 2007, teachers were asked to record student absences each day. They were trained by a paediatrician to identify 10 symptoms of illness using standardised case definitions and were asked to record these in association with student absences. In the study conducted by Talaat 2011, school administrators collected absenteeism data. The hand hygiene teams within the school telephoned parents to collect information on symptoms. Laboratory testing of nasal swabs was also conducted on children with influenza‐like symptoms.
Effectiveness
It was not possible to combine data from these two studies in a meta‐analysis; results are therefore presented in section 11 of Additional Table 3.
Safety and accident prevention
One study focused on safety and accident prevention by encouraging students to wear helmets while cycling. TheSchool Bicycle Safety Project (Hall 2004) was conducted over two years and targeted students in grade five (10 to 11 year‐olds). It was conducted in Australia and was explicitly informed by the Health Promoting Schools framework.
Measures
Measures include self‐reported use of a helmet while cycling and observations of correct helmet usage in schools by trained staff.
Effectiveness
See section 12 of Additional Table 3 for a summary of the results of this study.
Body image or eating disorders
One study (McVey 2004) focused on body image and eating disorders. The Healthy School, Healthy Kids study was implemented in Canada over an eight‐month period with children in grades six to seven (11 to 13 year‐olds). The impact of the intervention was also measured in a subset of teachers within the schools. The intervention was specifically informed by the Health Promoting Schools framework.
Measures
Body image outcomes were assessed by student or teacher self reports.
Effectiveness
See section 13 of Additional Table 3 for a summary of the results of this study.
Skin or sun safety behaviours
Olson 2007 reported on an intervention to promote skin or sun safety. This intervention was implemented in schools and local communities in New Hampshire and Vermont, USA, and promoted covering up in the sun to prevent the harmful effects of sun exposure. It was a three‐year intervention that targeted students in grades six to eight (11 to 14 year‐olds).
One other study (Schofield 2003) implemented an intervention to target sun safety behaviours as part of a multiple risk behaviour intervention that also focused on smoking and alcohol. However, data from this study were only presented for smoking outcomes and therefore we cannot provide any data from this study for this outcome.
Measures
Coverage of body surface area was assessed by direct observation. Use of sunscreen was assessed through self reports.
Effectiveness
The results of this intervention are reported in section 14 of Additional Table 3.
Oral health
One study, conducted in China, focused on oral health (Tai 2009). A three‐year oral and dental health project was conducted in primary schools in Yichang city with first‐grade students (six to seven year‐olds). This intervention was explicitly informed by the HPS framework.
Measures
Student caries and decay were assessed by dentists. Oral health care habits were reported by students’ mothers.
Effectiveness
The results of this intervention are reported in section 15 of Additional Table 3.
Academic, attendance, and school‐related outcomes
Very few studies reported on any academic or attendance outcomes. Only the two Positive Action trials (Beets 2009; Li 2011) specifically measured the impact of their intervention on academic achievement, and only four presented any attendance data (Beets 2009; Bowen 2007; Li 2011; Talaat 2011). In both cases, the authors collected data on standardised test scores for reading and maths. Beets 2009 also presented data on suspensions and retentions in grade, and Li 2011 reported absenteeism data as well as student disaffection with learning and teachers’ perceptions of student motivation and performance. Sahota 2001 included data on students’ self perception of academic competence. Bowen 2007 and Talaat 2011 presented data on attendance outcomes.
Some studies collected data on measures of school climate and satisfaction with school. Beets 2009 reported composite scores on school climate or quality, while Li 2011 presented data on student‐reported levels of disaffection. Fekkes 2006 reported outcome data on school satisfaction, covering the following areas: contact with other pupils, contact with teachers, and satisfaction with school life. Kärnä 2011 reported on student well‐being at school. The Gatehouse Project (Bond 2004) reported the number of students with low school attachment. Sawyer 2010 included measures of student‐ and teacher‐ratings of school climate. Going Places (Simons‐Morton 2005) assessed students’ perceptions of teacher supportiveness, and clarity and fairness of school rules. Healthy School Healthy Kids (McVey 2004) reported on teachers’ perceptions of school climate, including the school’s social, behavioural, and nutrition or physical activity environments. The HEALTHY study (Foster 2010) reported in their protocol that they would assess impact on academic outcomes, but these have not been reported in subsequent trial papers.
Effectiveness
The results of the academic, attendance, and school‐related outcomes are reported in section 16 of Additional Table 3.
Subgroup analyses
We performed subgroup analyses by age (12 years of age and under) and intervention duration (12 months or less), and formally tested for a difference between subgroups using meta‐regression (Additional Table 8). Due to the paucity of data, we were unable to perform subgroup analyses by gender and socioeconomic status (SES). Only six studies presented outcome data by gender across a range of outcomes, and we report these in Additional Table 8. No study presented outcome data by SES.
| Age group (< 12 years>) subgroup analyses | ||||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | MD or SMD [95% CI] | I² | Meta‐regression |
| BMI | Physical activity only | younger (≤ 12 years) | 1 | 297 | 205 | ‐0.12 [‐0.20 to ‐0.04] | n/a | n/a |
| older (> 12 years) | 1 | 374 | 358 | ‐0.28 [‐0.52 to ‐0.04] | n/a | |||
| Physical activity + nutrition | younger (≤ 12 years) | 8 | 4350 | 5242 | ‐0.28 [‐0.47 to ‐0.10] | 86% | 0.47 [‐0.11 to 1.05] | |
| older (> 12 years) | 3 | 2271 | 1961 | 0.08 [‐0.08 to 0.24] | 68% | |||
| zBMI | Physical activity + nutrition | younger (≤ 12 years) | 6 | 2507 | 2708 | ‐0.05 [‐0.12 to 0.02] | 78% | 0.12 [‐0.12 to 0.43] |
| older (> 12 years) | 2 | 3267 | 2898 | 0.04 [‐0.08 to 0.17] | 73% | |||
| Physical activity | Physical activity only | younger (≤ 12 years) | 1 | 297 | 205 | 0.01 [‐0.01 to 0.03] | n/a | n/a |
| older (> 12 years) | 1 | 374 | 358 | 0.35 [0.17 to 0.53] | n/a | |||
| Physical activity + nutrition | younger (≤ 12 years) | 4 | 1403 | 1515 | 0.06 [‐0.10 to 0.23] | 54% | 0.18 [‐0.10 to 0.46] | |
| older (> 12 years) | 2 | 1841 | 1431 | 0.24 [0.17 to 0.31] | 0% | |||
| Fat intake | Nutrition only | younger (≤ 12 years) | 5 | 1770 | 1704 | ‐0.17 [‐0.35 to 0.00] | 73% | 0.28 [‐0.17 to 0.73] |
| older (> 12 years) | 2 | 435 | 307 | 0.10 [‐0.05 to 0.25] | 0% | |||
| Physical activity + nutrition | younger (≤ 12 years) | 7 | 2762 | 2646 | 0.00 [‐0.32 to 0.33] | 94% | ‐0.18 [‐ 0.78 to 0.42] | |
| older (> 12 years) | 3 | 3736 | 3316 | ‐0.17 [‐0.41 to 0.07] | 97% | |||
| Fruit and vegetable intake | Nutrition only | younger (≤ 12 years) | 7 | 2858 | 2610 | 0.20 [0.05 to 0.35] | 85% | ‐0.24 [‐0.65 to 0.16] |
| older (> 12 years) | 2 | 435 | 307 | ‐0.04 [‐0.36 to 0.28] | 76% | |||
| Physical activity + nutrition | younger (≤ 12 years) | 2 | 488 | 506 | ‐0.06 [‐0.22 to 0.11] | 0% | 0.18 [‐1.11 to 1.49] | |
| older (> 12 years) | 2 | 3019 | 2599 | 0.16 [‐0.42 to 0.74] | 93% | |||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | OR [95% CI] | I² | Meta‐regression OR [95% CI] |
| Tobacco use | Multiple risk behaviours | younger (≤ 12 years) | 2 | 1169 | 908 | 0.68 [0.35 to 1.31] | 32% | 1.31 [0.55, 3.11] |
| older (> 12 years) | 3 | 4334 | 3581 | 0.85 [0.77 to 0.94] | 0% | |||
| Alcohol use | Multiple risk behaviours | younger (≤ 12 years) | 2 | 1169 | 908 | 0.47 [0.33 to 0.67] | 0% | 2.04 [0.88, 4.73] |
| older (> 12 years) | 2 | 3327 | 2736 | 0.96 [0.84 to 1.09] | 0% | |||
| Substance use | Multiple risk behaviours | younger (≤ 12 years) | 2 | 1169 | 908 | 0.41 [0.18 to 0.93] | 44% | 2.07 [0.00, 33.42] |
| older (> 12 years) | 1 | 2635 | 2108 | 0.85 [0.66 to 1.10] | n/a | |||
| Violence | Multiple risk behaviours | younger (≤12 years) | 2 | 1171 | 906 | 0.36 [0.26 to 0.50] | 0% | 2.60 [0.27, 24.59] |
| older (> 12 years) | 1 | 2635 | 2108 | 0.93 [0.86 to 1.01] | n/a | |||
| Being bullied | Anti‐bullying | younger (≤12 years) | 6 | 8556 | 8301 | 0.84 [0.70 to 1.01] | 71% | 1.15 [0.70, 1.89] |
| older (> 12 years) | 2 | 5437 | 3962 | 1.01 [0.86 to 1.19] | 0% | |||
| Bullying others | Anti‐bullying | younger (≤12 years) | 6 | 8550 | 8292 | 0.84 [0.70 to 1.02] | 70% | 1.05 [0.57, 1.95] |
| older (> 12 years) | 2 | 5399 | 3935 | 0.92 [0.77 to 1.09] | 0% | |||
| Duration (< 12 months>) subgroup analyses | ||||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | MD or SMD [95% CI] | I² | Meta‐regression MD or SMD [95% CI] |
| BMI | Physical activity only | shorter (≤ 12 months) | 1 | 297 | 205 | ‐0.12 [‐0.20 to ‐0.04] | n/a | n/a |
| longer (> 12 months) | 1 | 374 | 358 | ‐0.28 [‐0.52 to ‐0.04] | n/a | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 4 | 2289 | 2471 | ‐0.37 [‐0.70 to ‐0.03] | 88% | 0.29 [‐0.39 to 0.97] | |
| longer (> 12 months) | 6 | 4332 | 4732 | ‐0.08 [‐0.26 to 0.10] | 87% | |||
| zBMI | Physical activity + nutrition | shorter (≤ 12 months) | 2 | 394 | 397 | ‐0.22 [‐0.68 to 0.24] | 93% | 0.18 [‐0.12 to 0.48] |
| longer (> 12 months) | 6 | 5380 | 5209 | ‐0.00 [‐0.04 to 0.04] | 50% | |||
| Physical activity | Physical activity only | shorter (≤ 12 months) | 1 | 297 | 205 | 0.01 [‐0.01 to 0.03] | n/a | n/a |
| longer (> 12 months) | 1 | 374 | 358 | 0.35 [0.17 to 0.53] | n/a | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 1 | 292 | 301 | ‐0.17 [‐0.39 to 0.05] | n/a | 0.39 [0.07 to 0.71] | |
| longer (> 12 months) | 5 | 2952 | 2645 | 0.22 [0.16 to 0.28] | 93% | |||
| Fat intake | Nutrition only | shorter (≤ 12 months) | 5 | 1480 | 1512 | ‐0.17 [‐0.42 to 0.07] | 76% | 0.18 [‐0.34 to 0.69] |
| longer (> 12 months) | 2 | 725 | 499 | ‐0.02 [‐0.13 to 0.09] | 36% | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 4 | 1616 | 1622 | 0.20 [‐0.23 to 0.62] | 96% | ‐0.42 [‐0.90 to 0.07] | |
| longer (> 12 months) | 6 | 4882 | 4340 | ‐0.21 [‐0.39 to ‐0.02] | 94% | |||
| Fruit and vegetable intake | Nutrition only | shorter (≤ 12 months) | 6 | 1766 | 1743 | 0.24 [0.07 to 0.41] | 78% | ‐0.22 [‐0.55 to 0.11] |
| longer (>12 months) | 3 | 1527 | 1174 | 0.02 [‐0.18 to 0.21] | 84% | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 1 | 292 | 301 | 0.14 [‐0.15 to 0.43] | n/a | 0.07 [‐1.59 to 1.73] | |
| longer (> 12 months) | 3 | 3215 | 2804 | 0.06 [‐0.22 to 0.34] | 86% | |||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | OR [95% CI] | I² | Meta‐regression OR [95% CI] |
| Being bullied | Anti‐bullying | shorter (≤ 12 months) | 3 | 12209 | 10472 | 0.74 [0.69 to 0.80] | 0% | 1.49 [0.97 to 2.27] |
| longer (> 12 months) | 2 | 1784 | 1791 | 1.08 [0.76 to 1.53] | 46% | |||
| Bullying others | Anti‐bullying | shorter (≤ 12 months) | 3 | 11887 | 10256 | 0.77 [0.72 to 0.82] | 0% | 1.28 [0.81 to 2.02] |
| longer (> 12 months) | 2 | 1777 | 1786 | 0.99 [0.75 to 1.30] | 0% | |||
| Gender subgroup analyses (as presented by authors) | ||||||||
| Outcome | Intervention type | Study name | Authors' results | |||||
| BMI | Physical activity + nutrition | Effect found for girls (increase in BMI: 1.11 kg/m² versus 1.66 kg/m² for intervention and control groups, respectively, P < 0.05) but not for boys | ||||||
| Effect found for boys (BMI: ‐0.28 kg/m² versus 0.36 kg/m² for intervention and control groups, respectively, P value = 0.04) but not for girls | ||||||||
| zBMI | Physical activity + nutrition | Effect found for girls (increase in zBMI: 0 versus 0.17 for intervention and control groups, respectively, P < 0.05) but not for boys | ||||||
| No effect found in either boys or girls | ||||||||
| Physical activity | Physical activity + nutrition | No difference between girls and boys in terms of self‐reported physical activity | ||||||
| No difference between girls and boys in terms of self‐reported physical activity | ||||||||
| Fat intake | Physical activity + nutrition | Significant reductions in intervention compared to controls for fat intake and % energy from fat in girls (P < 0.001 for both). No effect was seen for boys | ||||||
| No difference between girls and boys in terms of fat intake | ||||||||
| Tobacco | Multiple risk behaviours | Positive effect in boys (0.18 versus 0.31 for intervention and control groups, respectively, P value = 0.02) but not in girls | ||||||
| Alcohol | Multiple risk behaviours | Positive effect in boys but not in girls (1.19 versus 1.64, for intervention and control groups, respectively, P value = 0.04) but not in girls | ||||||
| Substance use | Multiple risk behaviours | No effect found in either boys or girls | ||||||
| Violence | Multiple risk behaviours | No effect found in either boys or girls | ||||||
| Violence prevention | No effect found in either boys or girls | |||||||
| Bullying | Multiple risk behaviours | Reduction in physical victimisation effect in boys (‐0.10 versus 0.03, for intervention and control groups, respectively, P value = 0.02) but not in girls | ||||||
The only statistically significant difference between groups was for physical activity in the physical activity and nutrition interventions, where interventions of a longer duration were, on average, more likely to increase physical activity levels than interventions of shorter duration (meta‐regression SMD 0.39, 95% CI 0.07 to 0.71). However, there was only one intervention of shorter duration included in this analysis (versus five of longer duration) and these comparisons are subject to high levels of confounding.
The confidence intervals for all other comparisons were very wide, which is to be expected given that meta‐regression is low powered, and we had so few studies contributing data to each of these comparisons (range: three to 11 studies). It is possible that there may be true differences in intervention effectiveness by age and duration, but at present we do not have enough data to be able to detect these.
Sensitivity analyses
In addition to those already mentioned above (regarding use of accelerometry and differences in fruit or vegetable intake), we performed further sensitivity analyses to check the robustness of a number of methodological decisions. These include: comparing point estimates for random‐ or fixed‐effect analyses; restricting analyses to studies with low risk for selection, performance, and attrition bias (where possible); assessing the impact of using imputed standard deviations where original data were not available; using 'vegetable' rather than 'fruit' intake, where both were reported separately; and comparing the effects of self‐reported and objectively measured (accelerometry) levels of physical activity.
Overall, there was little difference between point estimates for analyses using random‐ or fixed‐effect models, with no impact on the overall conclusions of effectiveness, with the exception of the following outcomes: using fixed‐effect meta‐analysis, we found positive intervention effects for multiple risk behaviour interventions for substance use and violence, and a marginal effect for alcohol use. We found anti‐bullying interventions to be effective for reducing reports of bullying others using fixed‐effect analyses. We found a marginal intervention effect for nutrition only interventions for fat intake. (See Additional Table 7 for details).
Due to the high numbers of studies identified as being at high or unclear risk of selection, performance, and attrition bias, we were only able to perform a small number of sensitivity analyses on specific outcomes (see Additional Table 7). Where analyses were possible, restricting analyses to studies marked as being at low risk of bias tended to reduce intervention effectiveness. However, it should be noted that in most cases only a very small number of studies could be included in each analysis and that these data should therefore be treated cautiously.
We excluded studies where we had to impute a standard deviation from another similar study to create a standardised mean difference (eight cases). This made no difference to the overall conclusions with the exception of one case: we found nutrition only interventions no longer effective at increasing fruit and vegetable intake. (See Additional Table 7 for details).
Funnel plots
For the majority of outcomes, there were too few studies (fewer than 10) to be able to create funnel plots to explore the possibility of publication bias. We generated funnel plots for BMI and the two nutrition outcomes (Figure 4; Figure 5; Figure 6). For BMI and fat intake, studies were unevenly distributed indicating that there may be small study bias. This could potentially lead to an inflated estimate of intervention effectiveness as small negative studies appear to be under‐represented.
Funnel plot of comparison: 1 Overweight/Obesity, outcome: 1.1 BMI.
Funnel plot of comparison: 3 Nutrition, outcome: 3.1 Fat intake.
Funnel plot of comparison: 3 Nutrition, outcome: 3.2 Fruit and vegetable intake.
Discussion
Summary of main results
This is the first systematic review of cluster‐randomised controlled trials (C‐RCTs) to assess the effectiveness of the World Health Organization's (WHO’s) Health Promoting Schools (HPS) framework in improving the health and well‐being of students and their academic achievement. We identified 67 eligible studies, although only 10 of these were explicitly based upon the HPS framework.
The 67 included studies focused on a wide range of health outcomes and we were able to meta‐analyse data for 13 outcomes (body mass index (BMI), standardised body mass index (zBMI), physical activity, physical fitness, fat intake, fruit and vegetable intake, tobacco use, alcohol use, drug use, violence, depression, being bullied, and bullying others).
The results of meta‐analyses demonstrated evidence of effectiveness for HPS interventions seeking to reduce BMI and increase physical activity or fitness and fruit and vegetable intake. We also found positive intervention effects for HPS interventions seeking to reduce tobacco use and incidence of being bullied. For the HPS interventions that addressed alcohol and substance use, violence, mental health or bullying others, there was no evidence of effect.
It was not possible to meta‐analyse data from a number of studies with HPS interventions relating to sexual health, hand‐washing, accident prevention, body image, sun safety, and oral health. Few studies examined the impact of their intervention on academic achievement or other school‐related outcomes.
BMI or zBMI
The findings suggest that physical activity interventions reduce BMI (3 trials, 1430 participants). This represents a small but important shift in BMI at the school population level and is comparable with results from another recent review focusing on the prevention of obesity in childhood (Waters 2011; 34 school‐based interventions, including four from this review). The only physical activity intervention reporting an alternative measure of adiposity in children (zBMI) also reported a positive effect. However, no evidence of effect for zBMI was found for physical activity + nutrition interventions. It is important that future research in this area includes both BMI and zBMI as measures of childhood adiposity.
Physical activity and fitness
Physical activity + nutrition interventions also appear to be effective at increasing physical activity and fitness levels in students, an effect which remains when analyses are restricted to objective (accelerometry) measures (physical activity; 6 trials, 4230 participants). The effect sizes are equivalent to an increase of approximately three minutes of moderate‐to‐vigorous activity per day or a 0.25 level increase in the shuttle run test. Importantly, small increases that are successfully sustained have the potential to produce public health benefits at the population level (Rose 1985). Our results for physical activity are within the range reported by a recent Cochrane review by Dobbins 2013, which focused on all types of school‐based interventions to increase physical activity. This review of 26 studies (of which six were also included in this review) reported an increase of five to 45 minutes of moderate‐to‐vigorous physical activity per week.
Nutrition
The evidence of effect on nutrition outcomes was less promising. No evidence of effect was seen for either nutrition only or physical activity + nutrition interventions for fat intake; the latter intervention type also failed to increase fruit and vegetable intake. However, nutrition only interventions produced a small increase in fruit and vegetable consumption (9 trials, 6210 participants). This equates to an additional 30g of fruit and vegetables per day, roughly equivalent to half a portion. This finding is comparable with another review of school‐based nutrition programmes (including both RCTs and uncontrolled studies), which reported a 0.38 increase in servings of fruits and vegetables across seven studies (of which three were included in this review, Howerton 2007). A Cochrane review of community‐based interventions (including school settings) to promote consumption of fruits and vegetables in children (five to 18 year‐olds) is currently underway (Ganann 2010).
Tobacco use
Reductions in smoking behaviour were also apparent from our analyses. Among the studies that focused on tobacco use alone, intervention students were 23% less likely to smoke at follow‐up than their control counterparts (3 trials, 4747participants). Tackling tobacco use alongside other health outcomes in a multiple risk behaviour intervention was also effective (5 trials, 9992 participants). These effects are smaller in comparison to those found for social competence curricula (OR 0.52, 95% CI 0.3 to 0.88), and combined social competence and social influences programmes (OR 0.50, 95% CI 0.28 to 0.87) at longest follow‐up in a recent review of school‐based programmes for the prevention of smoking (Thomas 2013). Interestingly, the seven multimodal programmes included in Thomas's review that most closely resemble HPS interventions (and involved four studies also included in this review) were not found to be effective.
Bullying
We also found some evidence to suggest that HPS interventions may reduce bullying in schools, with reductions in reports of being bullied of 17% (6 trials, 26,256 participants), although no evidence of effect was found for reports of bullying others. A Campbell Collaboration review by Farrington 2009 reviewed 89 school‐based anti‐bullying interventions, including both randomised and non‐randomised study designs (four of which were also included in this review). They reported an overall reduction in being bullied of a similar magnitude to that reported here (17% to 20%). However, they also found substantial reductions in bullying others (20% to 23%).
Other substance use and violence
We found no evidence of effect for alcohol use, drug use or violent behaviours (4 trials, 8140 participants). It is important to note, however, that these meta‐analyses contained a small number of studies and more evidence is required in order to be able to determine whether the HPS framework is effective for these outcomes. Recent Cochrane reviews on school‐based interventions for alcohol use, drug use, and violence have produced mixed evidence for the effectiveness of these interventions. Faggiano 2005 found some evidence that skills‐based programmes can reduce drug use (risk ratio (RR) 0.81, 95% CI 0.64 to 1.02; 2 studies) and marijuana use (RR 0.82, 95% CI 0.73 to 0.92; 4 studies), but no effect was seen on drug use for knowledge‐based or affect interventions. Foxcroft 2011 provided a narrative review of 53 alcohol interventions (involving two studies also included in this review), and identified both studies that showed no preventive effect, as well as those that demonstrated statistically significant effects. Mytton 2006 reported significant reductions in student aggressive behaviour in 34 trials focusing on improving social skills or non‐response or both (SMD ‐0.41, 95% CI ‐0.56 to ‐0.26). A recent synthesis of multi‐level studies focusing specifically on the school environment found that schools with higher attainment and lower truancy than might be expected from students’ socioeconomic profile had lower rates of substance use and aggressive behaviours, suggesting that institutional factors may be protective (Bonell 2013), but have not to date been adequately addressed in HPS interventions evaluated through trials.
Mental health
Similarly, we found no evidence of effect for depression in the three studies that focused on this outcome. Where HPS interventions focused specifically on mental health, we observed a small, non‐significant increase in depressive symptoms in intervention students (2 trials, 6099 participants). The authors of these studies suggested a number of potential explanations for this, including: insufficient intervention duration; difficulties in establishing whole‐school change; and inability to address risk factors occurring outside of school (for example, family problems) (Bond 2004; Sawyer 2010). As noted above, we need more research in this area to determine the effect of this approach on improving mental health; however, given the findings reported here, future interventions should pay attention to potential harms that might arise from such programmes. A recent review by Kidger 2012 of nine studies (including two studies from this review) found limited evidence to suggest that changes to the school environment had a major impact on student mental health and well‐being. The authors conclude that whole‐school change can be difficult to establish and sustain, and that interventions that focus on one or two ‘active ingredients’ may be more effective. Future HPS interventions into this area should include comprehensive process evaluations and factorial designs to help identify critical elements of intervention success.
Hetereogeneity
The majority of our analyses displayed high levels of heterogeneity. Unlike clinical trials where interventions are highly standardised, eligibility criteria for participants ensure a relatively homogeneous population, and outcome measures are standardised, public health interventions inevitably display much greater levels of heterogeneity. This is particularly the case for largely non‐prescriptive interventions, such as the HPS framework, which allows a great deal of flexibility in intervention components. We attempted to address some of this heterogeneity by identifying distinct intervention ‘types’ within the HPS framework; for example, differentiating between physical activity only, nutrition only, and physical activity + nutrition interventions. However, we recognise that even within these groupings, interventions will have included different elements and activities. As the number of studies using the HPS approach continues to grow, it may be possible to further differentiate between different types of interventions to help identify the key elements for successful HPS interventions, as well as exploring differences in effectiveness between different populations.
RCTs and complex interventions
The use of cluster RCTs to evaluate complex interventions, such as the HPS framework, is much debated. Some have argued that RCTs are too rigid and inflexible to be able to adequately evaluate complex public health programmes (Nutbeam 1998; Tones 2000; WHO 1999). This is based on the assumption that RCTs require highly standardised intervention components and methods of delivery, thus precluding the possibility of local adaptation, which many health promotion specialists see as critical to intervention success. As Hawe 2004 and Rychetnik 2002 point out, however, this assumption is unfounded. It is possible to implement well‐designed cluster RCTS that can capture complexity and allow for local adaptation. As Hawe 2004 points out, the critical issue is ‘what’ is standardised (the intervention components or the steps in the change process). This review identified 67 cluster RCTs that successfully implemented the HPS framework approach. As such, it represents an important contribution to the body of evidence on the effectiveness of the HPS approach. Focusing on the most robust evidence available and using a conservative approach to assess effectiveness, we have found evidence in favour of the HPS framework for a number of important outcomes. To contextualise these findings, it is important that this review be read alongside other evaluations of the HPS framework employing different evaluation study designs (for example, IUHPE 2008; IUHPE 2010) which offer insight into the process and practicalities of implementation.
Overall completeness and applicability of evidence
Our review identified a large number of eligible HPS interventions. However, because this framework can be used to focus on any health outcome, the actual numbers of studies reporting data on a particular outcome were often quite small. The greatest amount of evidence we have is on overweight or obesity, physical activity or fitness, and nutrition. Half of the studies included in the review (34 studies) focused on one or more of these outcomes and all but four of these contributed to one or more meta‐analyses. By contrast, relatively few studies focused on substance use, violence, sexual health or mental health. Where meta‐analyses for these outcomes were possible, few studies are included and we require more evidence in order to be able to determine whether the HPS framework is effective for these outcomes.
We identified a broad division between the types of health issues focused on at particular ages. With some exceptions, we found that physical activity or nutrition interventions or both tended to focus on younger children, while substance use, violence, sexual activity, and mental health tended to target older children. While this latter approach may seem intuitive given that adolescence is often when these behaviours begin and many mental health conditions first emerge, the two Positive Action trials (Beets 2009; Li 2011) were conducted in elementary school children and showed promising effects for a number of outcomes, suggesting that tackling these issues at a younger age may be beneficial. Equally, while establishing healthy eating and promoting physical activity in younger children is clearly of importance, we also need effective interventions of these types in older children too. Physical activity levels, particularly in young women, are known to decrease during adolescence (Allison 2007; Nader 2008), and this is also a period when young people potentially start to gain greater freedom over their food choices.
We note a similar division for intervention duration. With some exceptions, studies focusing on physical activity or nutrition or both tended be shorter in duration (12 months or less) while those focusing on substance use, violence, mental health or sexual health tended to be of longer duration. It was unclear why this was the case.
Few studies measured the impact of their intervention on academic, attendance or other school‐related outcomes (10 studies). Only two studies measured the impact of their interventions on both academic achievement and attendance; Beets 2009, but not Li 2011, reported positive impacts on test scores for maths and reading, and both studies found a reduction in student absenteeism. A further two studies (both conducted in middle‐income countries on hand‐washing: Bowen 2007; Talaat 2011) assessed the impact of their intervention on attendance rates; both found a substantial decrease in illness‐related absences in intervention students. One study measured the effect on self‐perceived scholastic competence (Sahota 2001). The remaining studies focused on outcomes relating to school climate or satisfaction with mixed effects. Given that the HPS framework is based upon a recognition of the intrinsic link between health and education, the paucity of data on academic attendance and school‐related outcomes is both surprising and disappointing. Admittedly, only 10 studies included in this review were explicitly based upon the HPS framework, but even among these 10 studies only the two emotional well‐being studies presented school‐related data (school attachment, Bond 2004; school climate, Sawyer 2010). Given this lack of data, it is not possible to draw any definitive conclusions on the effectiveness of the HPS framework in improving academic achievement in students.The WHO recently highlighted the lack of attention paid to the impact of child health on educational outcomes in high‐income countries (Suhrcke 2011). We acknowledge that education‐related data are usually collected within education administrative processes and may be more difficult to obtain within research studies. Nevertheless, future evaluations of the HPS framework should seek to address this gap, not least because evidence of educational improvements is likely to be an important factor in determining whether interventions are scaled up.
An important limitation to the conclusions of this review is the lack of postintervention follow‐up in the majority of studies. Only 10 studies provided any postintervention follow‐up measures (ranging from six months to six years). While interventions may be able to produce short‐term changes in behaviours or health outcomes, unless these prove sustainable they are likely to be of little public health importance. Research funding needs to be invested into implementation (Phase IV) studies in order to determine the longer‐term impact of interventions (MRC 2000). This might include the use of anonymised data linkage with routinely collected health, education, social security, and criminal justice data (Lyons 2009; Lyons 2012).
The evidence for the HPS approach to school health promotion is dominated by studies from North America (27 USA, two Canada), which constituted almost half of the included studies. It is also notable that the multiple risk behaviour approach, whereby several health behaviours are targeted simultaneously, is almost exclusively used in an American context. When looking in detail at the components of these American trials, there is little to suggest that these intervention elements could not be implemented in other country contexts, given appropriate local adaptation. However, it is disappointing to note how few studies addressed issues such as social, cultural or political context within their documentation or process evaluations; the majority of studies focused exclusively on fidelity or acceptability or both. While these elements are important, additional contextual details are needed to allow policy‐makers to determine how a programme should be adapted and if it could produce similar results in their local area.
It is also disappointing to note the small number of studies coming from low‐ and middle‐income countries. Only eight studies were conducted in these areas and only one of these (Ross 2007) was implemented in a low‐income country (Tanzania). Given the well‐established links between poor nutrition and infectious disease on children’s cognitive development (Berkman 2002; Grantham‐McGregor 1995), it would seem that the HPS approach potentially has much to offer in the poorest parts of the world. For example, the two hand‐washing trials included in this review both reported reductions in illness‐related absences from school. The potential of this approach has been explicitly recognised with the development of the FRESH framework (Focusing Resources on Effective School Health), which adapts the HPS framework for use in low‐income contexts (World Education Forum 2008). However, little of this work appears to be evaluated with high quality evaluation study designs. Well‐designed research is required using the HPS approach in countries or areas where resources are constrained if we are to establish the efficacy of this approach outside of well‐resourced contexts.
The majority of studies compared the HPS intervention against no intervention or usual practice. We are therefore not able to assess the effectiveness of the HPS approach against simpler, less holistic interventions except via comparisons between our own results and those of reviews specifically focusing on health education curricula. Factorial designs would be useful to identify the importance of the three different intervention levels (curriculum, ethos or environment or both, and family or community or both) and how they interact.
Many studies failed to report data on a number of pertinent issues. Few studies assessed whether or not their intervention caused harm to students, either through assessment procedures or, more importantly, as a result of the intervention itself. Given the sensitive nature of the health topics focused on by these studies, it is important that researchers fully explore the potential for unintended negative consequences on students’ health and well‐being. Disappointingly few studies examined the impact of interventions by relevant equity criteria such as socioeconomic status, gender, and ethnicity. It is well acknowledged that interventions can increase health inequities (MacIntyre 2003). Reporting intervention outcomes within prespecified subgroups will help identify for whom the intervention works, as well as highlighting potential impacts on health inequities. Qualitative data collected in process evaluations could also provide important insights into issues of equity.The majority of studies failed to provide any details of the costs of their intervention and only two studies included cost‐effectiveness evaluations. Finally, although the majority of studies stated their intervention was informed by theory, very few provided specific details on how these theories were expected to produce changes in health behaviours or outcomes in students.
We were unable to evaluate the impact of the HPS framework on staff health because of the way this intervention has been defined (requiring input into the formal school curriculum as a key criterion). It is ironic that an intervention which originally envisaged healthy school environments that benefited both students and staff (WHO 1998) precludes the latter by its very definition. Staff health is clearly important, both in its own right and in terms of the impact it can have on student health and educational attainment (Bowers 2004; Lang 2013). A slightly modified definition of the HPS framework would be required to identify holistic interventions that specifically seek to target staff health.
Quality of the evidence
The quality of evidence overall, as determined by the GRADE approach, was low to moderate. RCT evidence was often downgraded on the basis of high levels of unexplained heterogeneity or high risk of bias (particularly for blinding of participants and for attrition). However, as noted above, the presence of heterogeneity in public health interventions is often inevitable. In addition, blinding of participants in such interventions is generally not possible.
Poor quality of reporting and insufficient detail often hampered our ability to assess risks of bias in a number of domains, particularly with regard to random sequence generation, where the majority of studies were assessed as being at unclear risk of bias. Similarly, a lack of published protocols for many studies hampered our ability to assess risk of bias for selective reporting of outcome data.
We assessed the majority of studies as being at high risk of performance bias: to a certain extent this is an unavoidable feature of interventions of this type whereby blinding of participants is difficult, if not impossible. However, this limitation has important implications for the reliability of outcome data included in this review. The majority of studies relied on student self reports to assess impact on outcome measures and thus were assessed as being at high risk of bias due to their lack of blinding. While feasible alternatives to self report may not be available or appropriate for some outcomes, researchers should be encouraged to use validated, objective measures assessed by researchers blind to group allocation wherever possible to mitigate this problem.
Attrition was also noted to be a problem in a number of studies, with high numbers of students lost to follow‐up. Attrition was particularly problematic in the multiple risk behaviour interventions. These studies tended to be of longer duration (two years or more), which inevitably increases the possibility of attrition over time. However, these studies often targeted low‐income areas where student turnover in schools can be high and may often be related to the outcomes being measured (for example, expulsions due to substance use or violent behaviours). Loss of clusters was a problem in a number of studies which could similarly introduce bias if schools with more challenging student intakes were more likely to withdraw from the study.
Only 37 studies reported their sample size power calculation, and only 27 adjusted this calculation to take into account the impact of clustering. It is therefore possible that many included studies did not have enough power to detect true statistical differences between groups. More worryingly, nine studies failed to adjust their analyses for the impact of clustering, despite analysing data at the student level. This would result in an overestimation of the precision of the effect estimate. We used reported or imputed intra‐cluster correlation coefficients (ICCs) to correct for this where these data were included in meta‐analyses.
Conducting systematic reviews of complex interventions is challenging (Jackson 2005; Shepperd 2009). This is the first Cochrane Review of this intervention and the very large number of hits generated by our searches (78,651 before de‐duplication), the substantial number of review outcomes and the complexity of synthesising data on a complex, multi‐level intervention meant that this review has taken a long time to complete. We conducted our original search in January 2010 and updated this in March and April 2013. Consequently, the latest search upon which this review is based began just over 12 months from the date of publication.
Potential biases in the review process
One limitation of this review is the potential for publication bias. It is possible that eligible studies have been carried out but have not been submitted or accepted for publication because of their null findings. The likelihood of this is difficult to judge, as in many cases we did not have enough studies contributing data to the meta‐analyses to be able to draw funnel plots. The move towards the registration of trials and protocols should help to identify (if not alleviate) this problem in the future.
One further limitation concerns our decision to have only one author complete the initial title screen to exclude those papers which were obviously not relevant to the review. This was a pragmatic decision based on the extremely large number of hits our search strategy generated (≈ 50,000). A very broad search strategy was necessary because of the absence of consistent key words for these interventions. It is therefore possible that we may have missed some eligible studies during this initial single‐person screening. However, we feel this is unlikely for two reasons. First, we double‐screened a random 10% of titles to check accuracy, and consistency and agreement between the two authors was excellent (kappa = 0.88). Second, we handsearched references lists from eligible trials and relevant systematic reviews to identify any potentially relevant trials.
We did not search the ASSIA database or any websites during our updated searches in March and April 2013, and it is possible that we may have missed relevant studies as a result. However, during our original search these sources did not identify any relevant studies, which were not also identified in other database searches.
We acknowledge a number of methodological limitations with regard to our meta‐analyses. First, in a small minority of studies in which no adjustment for clustering had been made in the reported analysis, and for which ICCs were not available (either from study publications or from attempted contact with the authors), we used ICCs from similar studies in order to make an adjustment for clustering. To ensure that our analyses were conservative, if multiple ICCs were available we chose the largest. Second, where standard deviations for the study population were not reported, we imputed a standard deviation from another similar study in order to calculate a standardised mean difference (SMD). Unlike imputation of missing ICCs, this decision impacts upon the point estimate of intervention effectiveness from the specific study, rather than just its precision. We conducted sensitivity analyses to examine the impact of this decision on our analyses, as reported in Additional Table 7. Third, where studies provided model data but no standard errors or P values, we used the final values for outcome measurements and adjusted for clustering using methods described above. Finally, to calculate SMDs we used the overall (‘total’) standard deviation across all individuals in a study rather than the ‘within‐cluster’ standard deviation, as studies rarely reported the latter. However, because we found ICCs to be generally small in this review, this is unlikely to have substantially affected our results.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Funnel plot of comparison: 1 Overweight/Obesity, outcome: 1.1 BMI.
Funnel plot of comparison: 3 Nutrition, outcome: 3.1 Fat intake.
Funnel plot of comparison: 3 Nutrition, outcome: 3.2 Fruit and vegetable intake.
Comparison 1 Overweight or obesity, Outcome 1 BMI.
Comparison 1 Overweight or obesity, Outcome 2 zBMI.
Comparison 2 Physical activity, Outcome 1 Physical activity.
Comparison 2 Physical activity, Outcome 2 Physical fitness.
Comparison 3 Nutrition, Outcome 1 Fat intake.
Comparison 3 Nutrition, Outcome 2 Fruit and vegetable intake.
Comparison 4 Tobacco use, Outcome 1 Tobacco use.
Comparison 5 Alcohol use, Outcome 1 Alcohol use.
Comparison 6 Substance use, Outcome 1 Substance use.
Comparison 7 Mental health, Outcome 1 Depression.
Comparison 8 Violence, Outcome 1 Violence.
Comparison 9 Bullying, Outcome 1 Being bullied.
Comparison 9 Bullying, Outcome 2 Bullying others.
| Study | Country | Age | Variable | Reported intra‐cluster correlation coefficient (ICC) | Published or correspondence |
| Australia | Grade 8 | Various ‐ including substance use, depressive symptoms and school engagement. | Not specifically reported for each outcome: ranged from 0.01 ‐ 0.06 | Published | |
| Germany | Grade 2 | BMI | 0.028 (NB this is the ICC for classroom, rather than school, clustering) | Correspondence | |
| USA | K‐Grade 2 | BMI | Not specifically reported for each outcome: ranged from 0 ‐ 0.019 | Published | |
| Physical activity | |||||
| Australia | Grades 5 ‐ 6 | zBMI | 0.02 | Correspondence | |
| BMI | 0.02 | ||||
| USA | Grade 5 | Various substance use outcomes | Not specifically reported: ranged from 0 ‐ 0.01 | Published | |
| USA | K‐Grade 1 | Portions of fruit and vegetables | 0.32 | Published | |
| Finland | Grade 8 | Fat intake | 0.004 | Correspondence | |
| Fruit consumption | 0.012 | ||||
| Vegetable consumption | 0.006 | ||||
| Netherlands | Grade 3 ‐ 8 | BMI | < 0.01 | Published | |
| Waist circumference | 0.014 | ||||
| Shuttle run | 0.166 | ||||
| Switzerland | Grade 1 and 5 | BMI | 0.01 | Published | |
| MVPA (accelerometry) | 0.08 | ||||
| Shuttle run | 0.06 | ||||
| Spain | 5 ‐ 6 year‐olds | BMI | 0.094 | Correspondence | |
| USA | Grades 7 ‐ 8 | Servings of fruits and vegetables | 0.0007 | Published | |
| % energy as fat | 0.0217 | ||||
| % energy as saturated fat | 0.0134 | ||||
| Finland | Grades 4 ‐ 6 | Self‐reported victimisation | 0.02 | Published | |
| Self‐reported bullying | 0.02 | ||||
| Well‐being at school | 0.03 | ||||
| Finland | Grades 2 ‐ 3 and 8 ‐ 9 | Self‐reported victimisation | Grade 2 ‐ 3: 0.05 Grade 8 ‐ 9: 0.03 | Published | |
| Self‐reported bullying | Grade 2 ‐ 3: 0.03 Grade 8 ‐ 9: 0.02 | ||||
| USA | Grades 6 ‐ 8 | Various – unclear if just referring to alcohol use or includes other substance use outcomes | Not specifically reported: ranged from 0.002 ‐ 0.03, with a median value of .015 | Published | |
| USA | Grades 4 ‐ 5 | Fruit and vegetable consumption | 0.03 | Published | |
| Australia | Grade 8 | Depression (CES‐D scores) | 0.02 | Published | |
| USA | Grades 4 ‐ 6 | % body fat | Not specifically reported: ranged from 0.0005 ‐ 0.026 | Published | |
| zBMI | |||||
| Food intake | Not specifically reported: ranged from 0.15 ‐ 0.38 | ||||
| Physical activity | 0.05 | ||||
| Sedentary behaviour | 0.03 | ||||
| Canada | Grade 9 | Physical dating violence | 0.02 | Published |
| Study ID | Intervention Name | Intervention outcomes | |||||||||||||||
| Overweight/ obesity | Physical activity | Nutrition | Tobacco | Alcohol | Drugs | Sexual | Mental | Violence | Bullying | Infectious disease | Safety/ accidents | Body image | Sun safety | Oral health | Aacdemic/ attendance/ school | ||
| Nutrition interventions | |||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| Fruits and Vegetables Make the Mark | X (MA) | ||||||||||||||||
| Project Tomato | X (MA) | ||||||||||||||||
| School Nutrition Policy Initiative | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Athletes in Service, Fruit and Vegetable Promotion Program | X | ||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| TEENS | X (MA) | ||||||||||||||||
| Gimme 5 | X | ||||||||||||||||
| 5 A DAY Power Plus | X (MA) | ||||||||||||||||
| ‐ | X | ||||||||||||||||
| High 5 | X (MA) | ||||||||||||||||
| Pro Children Study | X (MA) | ||||||||||||||||
| Physical activity interventions | |||||||||||||||||
| Fit 4 Fun | X (MA) | X (MA) | |||||||||||||||
| KISS | X (MA) | X (MA) | |||||||||||||||
| ICAPS | X (MA) | X (MA) | |||||||||||||||
| ‐ | X | ||||||||||||||||
| Physical activity + nutrition interventions | |||||||||||||||||
| Heart Smart | X | ||||||||||||||||
| URMEL ICE | X (MA) | ||||||||||||||||
| Pathways | X (MA) | X (MA) | X (MA) | ||||||||||||||
| RESCATE | X | X (MA) | |||||||||||||||
| Aventuras para Niños | X (MA) | X (MA) | |||||||||||||||
| HEALTHY | X (MA) | X (MA) | |||||||||||||||
| Health in Adolescents (HEIA) | X (MA) | X (MA) | |||||||||||||||
| ‐ | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Lekker Fit | X (MA) | X (MA) | |||||||||||||||
| AVall | X (MA) | ||||||||||||||||
| CATCH | X (MA) | X | X (MA) | X | |||||||||||||
| Project Energize | X (MA) | ||||||||||||||||
| APPLES | X (MA) | X (MA) | X (MA) | X | |||||||||||||
| M‐SPAN | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Nutrición en Movimiento | X (MA) | X | X (MA) | ||||||||||||||
| Bienestar (1) | X (MA) | ||||||||||||||||
| Bienestar (2) | X | X (MA) | X (MA) | ||||||||||||||
| Louisiana (LA) HEALTH | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Tobacco interventions | |||||||||||||||||
| ESFA (Denmark) | X (MA) | ||||||||||||||||
| ESFA (Finland) | X (MA) | ||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| Project MYTRI | X | ||||||||||||||||
| ‐ | X | ||||||||||||||||
| Alcohol interventions | |||||||||||||||||
| Project Northland (Chicago) | X (MA) | X (MA) | |||||||||||||||
| Project Northland (Minnesota) | X (MA) | X (MA) | X (MA) | ||||||||||||||
| Multiple risk behaviour interventions | |||||||||||||||||
| Positive Action (Hawai’i) | X (MA) | X (MA) | X (MA) | X | X (MA) | X | |||||||||||
| LIFT | X | X | X | X | |||||||||||||
| Aban Aya | X | X | X | ||||||||||||||
| Positive Action (Chicago) | X (MA) | X (MA) | X (MA) | X (MA) | X (MA) | X | |||||||||||
| DARE Plus | X (MA) | X (MA) | X (MA) | X (MA) | X (MA) | ||||||||||||
| Hunter Regions Health Promoting Schools Program | X (MA) | ||||||||||||||||
| Going Places | X (MA) | X (MA) | X | X | |||||||||||||
| Sexual health interventions | |||||||||||||||||
| Safer choices | X | ||||||||||||||||
| MEMA Kwa Vijana | X | ||||||||||||||||
| Mental health and emotional well‐being interventions | |||||||||||||||||
| Gatehouse | X (MA) | X (MA) | X (MA) | X (MA) | X (MA) | X | |||||||||||
| beyondblue | X (MA) | X | |||||||||||||||
| Violence interventions | |||||||||||||||||
| Students for Peace | X (MA) | ||||||||||||||||
| Fourth R | X | X | X | ||||||||||||||
| Ant‐bullying interventions | |||||||||||||||||
| Friendly Schools | X (MA) | ||||||||||||||||
| Friendly Schools, Friendly Families | X | ||||||||||||||||
| ‐ | X (MA) | X (MA) | X | ||||||||||||||
| Steps to Respect | X (MA) | ||||||||||||||||
| KiVa (1) | X (MA) | X | |||||||||||||||
| KiVa (2) | X (MA) | ||||||||||||||||
| ‐ | X (MA) | ||||||||||||||||
| Hand‐washing interventions | |||||||||||||||||
| ‐ | X | X | |||||||||||||||
| ‐ | X | X | |||||||||||||||
| Miscellaneous interventions | |||||||||||||||||
| School Bicycle Safety Project / The Helmet Files | X | ||||||||||||||||
| Healthy Schools ‐ Healthy Kids | X | X | |||||||||||||||
| SunSafe | X | ||||||||||||||||
| ‐ | X | ||||||||||||||||
| MA: included in meta analysis for this outcome | |||||||||||||||||
| Study ID | Name | Type | Outcome(s) | Authors’ conclusions |
| 1. Obesity or overweight or body size | ||||
| URMEL‐ICE | Physical activity + nutrition | Skinfold thickness (tricep and subscapular), waist circumference | Intervention students had lower measures for waist circumference (‐0.64, 95% CI ‐1.25 to ‐0.02) and subscapular skinfold thickness (‐0.85, 95% CI ‐1.59 to ‐0.12). However, after adjusting for the time‐lag between baseline and follow‐up, this difference was no longer apparent. No effect was seen for tricep skinfold thickness. | |
| Aventuras para Niños | Physical activity + nutrition | zBMI | Postintervention follow‐up: Data at the end of the intervention and at 1 and 2‐years postintervention. No impact on zBMI at any time point.No difference between control and intervention groups for % body fat. Adjusted difference = 0.18; 95% CI ‐0.45 to 0.81, P value = 0.56. | |
| Health in Adolescents (HEIA) | Physical activity + nutrition | Waist circumference, waist‐to‐hip ratio | No effect seen for waist circumference or waist‐to‐hip ratio for the total sample. | |
| KISS | Physical activity | Skinfolds thickness, waist circumference | Children in intervention group showed smaller increases in the sum of 4 skinfold z‐score units (‐0.12, 95% CI ‐0.21 to ‐0.03, P value = 0.009). No effect was seen for waist circumference. | |
| CATCH | Physical activity + nutrition | Tricep and subscapular skinfold | No difference between intervention and control group for tricep skin folds (difference = 0.14 mm, 95% CI ‐0.24 to 0.52, P value = 0.47), or subscapular skinfolds (difference = 0.13 mm; 95% CI ‐0.29 to 0.54, P value = 0.553) | |
| ICAPS | Physical activity | % body fat, Fat mass index, Fat‐free mass index | Among students who were not overweight at baseline, intervention students had lower fat mass index (‐0.2, 95% CI ‐0.39 to ‐0.01, P < 0.05). There was no difference for % body fat or fat‐free mass index. No differences were seen for any of these outcomes between the 2 groups for students who were initially overweight at baseline. Postintervention follow‐up: 2 years postintervention ‐ intervention students maintained lower age ‐ and gender‐adjusted BMI (0.37 kg/m², P value = 0.02) and waist circumference (1.6 cm, P < 0.01) than control counterparts. | |
| Bienestar (2) | Physical activity + nutrition | % body fat | No difference between control and intervention groups for % body fat. Adjusted difference = 0.18 (95% CI ‐0.45 to 0.81, P value = 0.56). | |
| LA Health | Physical activity + nutrition | % body fat | No difference between control and intervention (PP + PS group). | |
| 2. Physical activity | ||||
| HEARTSMART | Physical activity + nutrition | 1 mile run or walk test | 5th grade boys’ 1 mile run or walk times decreased by 1.3 minutes in intervention group, but increased by 0.8 minutes in the control group (P < 0.01). | |
| RESCATE | Physical activity + nutrition | % children engaging in moderate and moderate‐to‐vigorous physical activity and TV or computer time. | A greater % of children in the intervention group reported being moderately physically active more than 3 days a week, compared to control children (40% I, 8% C, P value for difference between groups not given). No difference between groups for moderate‐to‐vigorous physical activity or TV or computer time. | |
| Fit 4 Fun | Physical activity | Muscular fitness and flexibility | Positive treatment effects observed in intervention children for flexibility (sit and reach, adjusted mean difference, 1.52 cm, P value = 0.003), physical activity (adjusted mean difference, 3253 steps/day, P < 0.001) and 1 measure of muscular fitness (7‐stage sit‐up, adjusted mean difference, 0.62 stages, P value = 0.003). No effect was seen for 3 other measures of muscular fitness (basketball throw, push‐ups and standing jump). | |
| Nutricion en Movimiento | Physical activity + nutrition | % children active | No difference between control and intervention group. | |
| Avall | Physical activity + nutrition | TV screen time (hours). Proportion of students taking exercise | No difference between control and intervention group for TV screen time. Intervention students were more likely to report exercising (15.7% versus 10.9%, P value = 0.036). | |
| CATCH | Physical activity + nutrition | PE lesson length. Energy expenditure and energy expenditure rate (during PE lesson) | No difference between intervention and control schools for PE lesson length. However, intervention students had greater rates of energy expenditure (0.20 kJ/kg, 95% CI 0.12 to 0.27) and a higher energy expenditure ratio (0.35 kJ/kg per hour, 95% CI 0.26 to 0.45) in PE lessons than controls. | |
| M‐SPAN | Physical activity + nutrition | Physical activity at school (observations) | There was a greater rate of increase in physical activity at school over time in intervention schools, compared to controls (d = 0.93). Subgroup analyses reveal the effect was significant only for boy (d = 1.1). | |
| ICAPS | Physical activity | TV or video time, active commuting to and from school | Children in intervention group watched less television (‐15.71 minutes per day, 95% CI ‐28.49 to ‐2.92, P value = 0.02). No difference between groups for active commuting to and from schools (1.03 mins/day, 95% CI ‐2.16 to 4.22, P value = 0.53). Postintervention follow‐up: 2 years postintervention‐ intervention students spent less time watching television (29 mins/day, P < 0.01) and had higher active transport levels (+5 mins/days, P < 0.01). | |
| ‐ | Physical activity | Self reports on travel to and from school | No difference between intervention and control groups in number of children walking to and from school. | |
| LA Health | Physical activity + nutrition | Sedentary behaviour | No difference between control and intervention (PP + PS group). | |
| 3. Nutrition | ||||
| Aventuras Para Niños | Physical activity + nutrition | Consumption of sugary drinks and snacks | No effect seen for consumption of sugary drinks. There was an initial reduction in the number of snacks consumed by intervention group (‐0.38, SE 0.17). Postintervention follow‐up: This effect on snack consumption was not sustained at follow‐up. | |
| Athletes in Service, Fruit and Vegetable Promotion Program | Nutrition | Fruit and vegetable intake | Children in intervention consumed a greater amount of fruit (34 g, 95% CI 30 to 39) than control students (23 g, 95% CI 18 to 28) (P < 0.001). | |
| AVall | Physical activity + nutrition | Consumption of fruit and vegetable, and sugary snacks or drinks | No difference between groups for proportion of children eating fruit or vegetables daily. However, there was an increase in the daily intake of > 1 piece of fruit per day (P value = 0.005). No difference between groups for consumption of sugary snacks/drinks. | |
| GIMME FIVE | Nutrition | Fruit and vegetable intake, knowledge and confidence to eat more fruit and vegetables | Intervention students had higher fruit and vegetable consumption than controls for the first 2 years of the intervention (P < 0.05), but this effect was lost by the final year of the study. Intervention students had higher knowledge scores than controls in the final 2 years of intervention (P < 0.05 for both). No group effect was seen for student confidence in eating more fruit and vegetables. | |
| ‐ | Nutrition | % skipping breakfast. Healthy breakfast choices | No difference between groups for % of children skipping breakfast. No difference between groups for reported intake of any energy‐dense, micronutrient‐poor (EDMP) food or beverage breakfast choice. | |
| High 5 | Nutrition | Fruit and vegetable intake | Postintervention follow‐up: The increased consumption of fruit and vegetables in intervention students observed at the end of the intervention was maintained 12 months later (3.2 versus 2.21 servings for intervention and control groups, respectively, P < 0.0001). | |
| M‐SPAN | Physical activity + nutrition | School‐level fat intake levels (observations) | No effect was seen on school levels of fat intake. | |
| 4. Tobacco use | ||||
| LIFT | Multiple risk behaviours | Tobacco initiation | Postintervention follow‐up: Intervention was associated with a reduced risk (10%, β = ‐0.1, P < 0.01) in tobacco use initiation. After controlling for hypothesized mediators, the intervention was associated with less likelihood of tobacco use initiation (LR Chi² = 6.69, P < 0.05). | |
| CATCH | Physical activity + nutrition | Current smoker | No difference between intervention and control students. | |
| Project Mytri | Tobacco | Smoking in last 30 days, use of chewing tobacco and bidi. | The rates of smoking cigarettes, bidi smoking and any tobacco use increased over time in the control group; the rate of any tobacco use and bidi smoking decreased in the intervention group. Overall, tobacco use increased by 68% in the control group and decreased by 17% in the intervention group. | |
| ‐ | Tobacco | Ever and regular smoking | No effect was seen for students ever trying smoking (OR 0.72, 95% CI 0.44 to 1.16, P value = 0.178) but intervention students were less likely than controls to become regular smoker (OR 0.38, 95% CI 0.16 to 0.93, P value = 0.035). | |
| 5. Alcohol use | ||||
| LIFT | Multiple risk behaviours | Alcohol use | Postintervention follow‐up: Intervention was associated with a reduced risk (7%, β = ‐0.07, P < 0.05) in alcohol use initiation. | |
| 6. Drug use | ||||
| LIFT | Multiple risk behaviours | Illicit drug use | Postintervention follow‐up: No difference between groups for illicit drug use. The intervention had a marginal effect on initiation (9%, β = ‐0.09, P < 0.10). | |
| Aban Aya | Multiple risk behaviours | Substance use | Boys in intervention group were less likely than controls to report substance use (effect size 0.45, P value = 0.05, CIs not given) but this effect was of borderline significance. No effect was seen for girls. | |
| Fourth R | Violence prevention | Problem substance use | No effect seen on problem substance use (Adj. OR 1.11, 95% CI 0.84 to 1.44 P value = 0.43). | |
| 7. Sexual health | ||||
| Safer Choices | Sexual health | Delayed sexual initiation, condom use, number of partners | No difference between groups for incidence of sexual initiation (OR 0.83, 95% CI 0.54 to 1.27, P value = 0.39). Intervention students were less likely to have sex without a condom (effect size 0.63, P value = 0.05, CIs not given) and fewer partners with whom they had sex without a condom (effect size 0.73, P value = 0.02, CIs not given). | |
| Positive Action (Hawai’i) | Multiple risk behaviours | Sexual activity | Intervention students were less likely to have had sex than control student (OR 0.18, 90% CI 0.09 to 0.36). | |
| Aban Aya | Multiple risk behaviours | Recent sexual intercourse, Condom use. | Boys in the intervention group were less likely than controls to have had recent sexual intercourse (effect size 0.65, P value = 0.2) and more likely to use a condom (effect size 0.66, P value = 0.045, CIs not given). No effect was seen for girls. | |
| MEMA Kwa Vijana | Sexual health | HIV incidence. Prevalence of other STIs. Incidence of pregnancy. Condom use. Number of partners | No difference between groups for HIV incidence or prevalence of syphilis, Chlamydia and Trichomonas. Prevalence of gonorrhoea was higher in intervention women than control (Adj. RR 1.93, 95% CI 1.01 to 3.71). There was no difference between groups in the number of pregnancies. Intervention men and women were more likely to have first used a condom during the follow‐up period than controls (men: Adj. RR 1.41, 95% CI 1.15 to 1.73; women: Adj. RR 1.30, 95% CI 1.03 to 1.63). Intervention men (but not women) were more likely than controls to have used a condom at last sex (Adj. RR 1.47, 95% CI 1.12 to 1.93) and less likely to have had >1 partner in past 12 months (Adj. RR 0.69, 95% CI 0.49 to 0.95). Postintervention follow‐up: ≈6 years postintervention ‐ no difference between groups for HIV prevalence or any other STIs, number of pregnancies and condom use. There was an increase in men reporting < 4 sexual partners (Adj. prevalence rate 0.87, 95% CI 0.78 to 0.97). | |
| Fourth R | Dating violence prevention | Condom use | No difference seen between groups for condom use (Adj. OR 1.04 95% CI 0.51 to 2.2, P value = 0.91). | |
| 8. Mental health or emotional well‐being | ||||
| ‐ | Anti‐bullying | Depression | No difference observed between groups for depression. Postintervention follow‐up: 1 year postintervention, no difference observed between groups for depression. | |
| beyondblue | Emotional well‐being | Depression | Postintervention follow‐up: No difference between groups for depression. | |
| 9. Violence | ||||
| LIFT | Multiple risk behaviours | Physical aggression in playground | Postintervention follow‐up: Intervention students showed significant reductions in physical aggression in the playground, compared to controls (‐0.11, P < 0.01). | |
| ABAN AYA | Multiple risk behaviours | Violence | Boys in intervention group were less likely than controls to report violent behaviour (effect size 0.41, P value = 0.02, CIs not given). No effect was seen for girls. | |
| Going Places | Multiple risk behaviours | Antisocial behaviour (including violence and other 'social' problems) | No effect seen for antisocial behaviour. | |
| Fourth R | Dating violence prevention | Physical dating violence, peer violence | Postintervention follow‐up: (2½ years after start of intervention) No difference was seen for physical dating violence using unadjusted ORs (1.42, 95% CI, 0.87 to 2.33, P value = 0.15). When analyses were adjusted for baseline behaviour, stratifying variables and gender, intervention students were less likely to report physical dating violence (Adj. OR 2.42, 95% CI 1.00 to 6.02, P value = 0.05) but this effect was of borderline significance. No effect was seen for physical peer violence (OR 1.09, 95% CI 0.83 to 1.59). | |
| 10. Bullying | ||||
| Friendly Schools, Friendly Families | Anti‐bullying | Being bullied, bullying others, told if saw bullying | At the end of intervention, Grade 4 students in the low‐intensity group (control) were more likely to report having been bullied than students in the high‐intensity group (OR 1.39, 95% CI 1.02 to 1.91) but no effect was seen for Grade 6 students. No effect was seen for ‘bullying others’ in either Grade cohort at the end of intervention. Grade 6 students were more likely to tell someone if they saw bullying (OR 1.78, 95% CI 1.21 to 2.62). Postintervention follow‐up: 1 year postintervention (collected for Grade 4 students only) low‐intensity group (control) students were more likely to report having been bullied (OR 1.64, 95% CI 1.06 to 2.53) or bullying others (OR 1.74, 95% CI 1.09 to 2.78). | |
| ‐ | Anti‐bullying | Being bullied, active bullying | Postintervention follow‐up: 1 year postintervention, there were no differences between intervention and control students for being bullied (rate ratio 1.14, 95% CI 0.81 to 1.59) or active bullying (rate ratio 0.7, 95% CI 0.43 to 1.29). | |
| 11. Infectious disease prevention: Hand‐washing | ||||
| ‐ | Hygiene | Illness incidence | No difference seen between groups for overall illness incidence. However, intervention schools reported a 42% decrease in student absences.Intervention students were less likely than controls to be absent due to headaches (0.54 versus 0.73 episodes per 100 student weeks, P value = 0.04) and stomach aches (0 versus 0.3 episodes per 100 student weeks, P value = 0.03). | |
| ‐ | Hygiene | Absence caused by illness (influenza‐like infections, diarrhoea, conjunctivitis) | Overall, absences caused by illness were reduced by 21% in intervention schools (5.7 versus 7.2 median episodes). Absences due to influence‐like illness were reduced by 40% (0.3 versus 0.5 median episodes), diarrhoea by 33% (0.2 versus 0.3 median episodes) and conjunctivitis by 67% (0.1 versus 0.3 median episodes). P < 0.0001 for all. | |
| 12. Safety or accident prevention | ||||
| School Bicycle Safety Project (Helmet Files) | Safety | Observed and self‐reported helmet use, helmet worn correctly | No effect seen on observed helmet use. Of those who reported not always wearing a helmet at baseline, intervention students were more likely to report always wearing a helmet at post‐test 1 (OR 1.76, 95% CI 1.09 to 2.85) but this effect disappeared at post‐test 2. | |
| 13. Body image or eating disorders | ||||
| Healthy School – Healthy Kids | Body image | Student and teachers' body satisfaction, internalisation of media ideals, body size acceptance, weight‐based teasing, disordered eating, weight loss, muscle gaining behaviours | The intervention reported a positive effect in the "internalization of media ideals" for intervention students (F [2, 596] = 3.30, P value = 0.03) and a decrease in disordered eating (only measured in girls; F [2, 276) = 2.73, P value = 0.04). No effect was seen on body satisfaction, body size acceptance or perceptions of weight‐based teasing. Compared to controls, fewer intervention students were trying to lose weight at the end of the intervention (Chi² = 4.29, P value = 0.03) but this effect was lost at 6‐month follow‐up. No effect was seen at any point for muscle‐gaining behaviour. No effect was seen for teachers on any outcome. | |
| 14. Sun safety | ||||
| Sunsafe in Middle Schools | Sun protection | % Body Surface Area covered up in sun, sunscreen application | No effect was seen on the % of body surface area covered up on observed adolescents or reported sunscreen use at first follow‐up. However, by the end of the 2nd year, students from intervention areas were likely to be more covered up than control participants (66.1% versus 56.8% body surface area covered, P < 0.01). They were also more likely to report using sunscreen at this time than control participants (47% versus 13.8%, P < 0.001). | |
| 15. Oral health | ||||
| ‐ | Oral health | Net caries increment; Restoration, sealant, and decay score; Oral health care habits reported by mothers. | No difference between groups for number of decayed, missing or filled teeth (DMFT), although there was a slight reduction in number of decayed, missing or filled surfaces (DMFS) in intervention children (0.22 versus 0.35, P value = 0.013). Intervention students had a greater mean decrease in plaque index (0.32 versus 0.21, P value = 0.013) and sulcus bleeding index (0.14 versus 0.08, P value = 0.005). Intervention children were more likely than controls to have received restorants (10.3% versus 6.2%, P value = 0.006), have sealants placed (17.5% versus 4.1%, P < 0.001) and less likely to have untreated decay (7.6% versus 20.5%, P < 0.001). Mothers of children in intervention group were more likely to report their children brushed her or his teeth, had had a dental visit within the past year and used fluoride toothpaste (P < 0.001 for all). | |
| 16. Academic, attendance, and school‐related outcomes | ||||
| Positive Action (Hawai'i ) | Multiple risk behaviours | Test scores for reading and maths, absenteeism, suspensions, retentions in grade, school climate variables | Intervention schools had higher maths and reading scores than control schools (Hawai'i Content and Performance Standards, P < 0.05 for both), lower absenteeism (P < 0.001) and fewer suspensions (P < 0.001). No effect seen for retentions in grade. The effects indicate a 2% advantage per year in the intervention group compared to the control group. Student, teacher and parent School Quality Composite scores were all higher in intervention schools compared to control (P value = 0.015, 0.006, 0.007, respectively). | |
| Gatehouse Project | Emotional well‐being | Low school attachment | Unadjusted ORs revealed no effect seen on low school attachment. However, at final follow‐up, adjusted ORs suggest an improvement in school attachment in intervention students (Adj. OR 1.33, 95% CI 1.02 to 1.75). | |
| ‐ | Hygiene | Attendance | Intervention schools (expanded group) experienced 42% fewer absence episodes (P value = 0.03) and 54% fewer days of absence (P value = 0.03) than control schools. | |
| ‐ | Anti‐bullying | School satisfaction variables | No effect seen for general satisfaction with school life; satisfaction with contact with other pupils; or satisfaction with contact with teachers. | |
| KIVA (1) | Anti‐bullying | Well‐being at school | Intervention students reported higher levels of well‐being at school (0.096, P value = 0.011) compared to the control students. | |
| Positive Action (Chicago) | Multiple risk behaviours | Standardised test scores. Student and teacher reports of academic performance, motivation and disaffection. Absenteeism. | There was a significant decrease in student disaffection with learning in the intervention group compared to those in the control schools. No effect seen on teachers' ratings of students' academic performance but a positive effect on their rating of academic motivation was found. Lower rates of absenteeism found in intervention than in control schools (β= ‐0.16, one‐tailed P value = 0.015). No evidence of a programme effect on standardised test scores for reading and maths. | |
| Healthy School, Healthy Kids | Body image | Teachers' perceptions of school's social, behavioural and nutrition or physical climate | No effect on teachers' perceptions of school climate. | |
| APPLES | Physical activity + nutrition | Self‐perceived scholastic competence | No effect on self‐perceived scholastic competence. | |
| beyondblue | Emotional well‐being | Student and teacher ratings of school climate | No effect found for student rating of school climate. Teacher ratings significantly differed between intervention and control schools over time (β = 0.60, SE = 0.29, P value < 0.05). On average, school climate in intervention schools improved over time, while in control schools it declined. | |
| Going Places | Multiple risk behaviours | Students’ perceptions of school climate | No effect seen on students’ perceptions of school climate. | |
| ‐ | Hygiene | Attendance | Overall, absences caused by illness were reduced by 21% in intervention schools (5.7 versus 7.2 median episodes). | |
| CI: confidence interval; OR: odds ratio; RR: risk ratio; SE: standard error; STI: sexually transmitted infection | ||||
| Authors | Name | Review outcomes | Country | Target group | Duration | Theory |
| Nutrition interventions | ||||||
| ‐ | Nutrition | UK | 6 ‐ 7 and 10 ‐ 11 year‐ olds | 8 months | Health Promoting Schools framework | |
| Fruits and Vegetables Make the Mark | Nutrition | Norway | Grade 6 | 6 months. | Social cognitive theory | |
| Project Tomato | Nutrition | UK | Year 2 | 10 months | Framework for health maintenance behaviour | |
| School Nutrition Policy Initiative | Obesity or overweight. Nutrition | USA | Grades 4 ‐ 6 | 2 years | None stated | |
| Athletes in Service, Fruit and Vegetable Promotion Program | Nutrition | USA | Kindergarten and Grade 1 | 2½ years | Social learning theory | |
| ‐ | Nutrition | Finland | Grade 8 | 8 months | Social cognitive theory | |
| TEENS | Nutrition | USA | Grades 7 ‐ 8 | 2 years | Social cognitive theory | |
| Gimme 5 | Nutrition | USA | Grade 9 | 3 years | PRECEDE model of health education | |
| 5 A DAY Power Plus | Nutrition | USA | Grades 4 ‐ 5 | 6 months | Social learning theory | |
| ‐ | Nutrition | Australia | Grade 7 | 11 months | Health Promoting Schools framework | |
| High 5 | Nutrition | USA | Grade 4 | 1 year | Social cognitive theory | |
| Pro Children Study | Nutrition | Netherlands, Norway, Spain | Grades 5 ‐ 6 | 2 years | Social cognitive theory, Ecological model | |
| Physical activity interventions | ||||||
| Fit‐4‐Fun | Obesity or overweight. Physical activity | Australia | Grades 5 ‐ 6 | 8 weeks | Health Promoting Schools framework, Social cognitive theory, Harter's competence motivation theory. | |
| KISS | Obesity or overweight. Physical activity | Switzerland | Grades 1 ‐ 5 | 11 months | None stated | |
| ICAPS | Obesity or overweight. Physical activity | France | Grade 6 | 4 years | Says it is theory‐based but no details of a named theory given | |
| ‐ | Physical activity | Australia | Years 4 ‐ 5 | 2 years | Health Promoting Schools framework | |
| Physical activity + nutrition interventions | ||||||
| Heart Smart | Obesity or overweight, physical activity, nutrition | USA | Grades 4 ‐ 5 | 2½ years | Social cognitive theory | |
| URMEL ICE | Obesity or overweight, physical activity, nutrition | Germany | Grade 2 | 9 months | Social cognitive theory | |
| Pathways | Physical activity, nutrition | USA | Grade 3 | 3 years | Social learning theory | |
| RESCATE | Obesity or overweight, physical activity, nutrition | Mexico | Grades 4 ‐ 5 | 1 year | None stated | |
| Aventuras para Niños | Obesity or overweight, physical activity, nutrition | USA | K‐Grade 2 | 5 semesters | Social ecological theory, Social cognitive theory, Health belief model, Structural model of health behavior | |
| HEALTHY | Obesity or overweight | USA | Grades 6 ‐ 8 | 3 years | None stated | |
| Health in Adolescents (HEIA) | Obesity or overweight, physical activity, nutrition | Norway | Grade 6 | 20 months | Socioecological framework | |
| ‐ | Obesity or overweight, physical activity | Belgium | Grades 7 ‐ 8 | 2 years | Theory of planned behaviour, Transtheoretical model, Social cognitive theory, Attitude, Social influence and self‐Efficacy (ASE) Model | |
| Lekker Fit | Obesity or overweight, physical activity | Netherlands | Grades 3 ‐ 8 | 8 months | Theory of planned behaviour ecological model (Egger and Swinburn) | |
| Nutrición en Movimiento | Obesity or overweight, nutrition | Mexico | Grade 5 | 6 months | Not explicitly theory‐based, but does mention use of theory of peer learning for 1 element of the intervention (puppet theatre) | |
| AVall | Obesity or overweight, physical activity, nutrition | Spain | 5 ‐ 6 year‐olds | 2 years | Educational methodology 'IVAC'. | |
| CATCH | Physical activity, nutrition | USA | Grade 3 | 3 years | Social cognitive theory, Social learning theory | |
| Project Energize | Obesity or overweight | New Zealand | 5 and 10 year‐olds | 2 years | Health Promoting Schools framework | |
| APPLES | Obesity or overweight, physical activity, nutrition | UK | Years 4 ‐ 5 | 10 months | Health Promoting Schools framework | |
| M‐SPAN | Physical activity, nutrition | USA | Grades 6 ‐ 8 | 2 years | Ecological model | |
| Bienestar (1) | Physical activity, nutrition | USA | Grade 4 | 5 months | Social cognitive theory, Social ecological theory | |
| Bienestar (2) | Obesity or overweight, physical activity | USA | Grade 4 | 8 months | Social cognitive theory | |
| Louisiana (LA) HEALTH | Obesity or overweight, physical activity, nutrition | USA | Grades 4 ‐ 6 | 2½ years | Social learning theory | |
| Tobacco interventions | ||||||
| ESFA (Denmark) | Tobacco | Denmark | Grade 7 | 3 years | Attitude‐Social influence‐self‐Efficacy (ASE) model | |
| ESFA (Finland) | Tobacco | Finland | Grade 7 | 3 years | Attitude‐Social influence‐self‐Efficacy (ASE) model | |
| ‐ | Tobacco | Australia | Grade 9 students | 2 school years | Health Promoting Schools framework | |
| Project MYTRI | Tobacco | India | Grades 6 ‐ 8 | 2 years | Social cognitive theory, Social influences model | |
| ‐ | Tobacco | China | Grades 7 ‐ 8 | 2 years | Socioecological framework, PRECEDE‐PROCEED model | |
| Alcohol interventions | ||||||
| Project Northland (Chicago) | Alcohol, tobacco, drugs | USA | Grade 6 ‐ 8 | 3 years | Theory of triadic influence | |
| Project Northland (Minnesota) | Alcohol, tobacco, drugs | USA | Grades 6 ‐ 8 | 3 years. | Social learning theory | |
| Multiple risk behaviour interventions | ||||||
| Positive Action (Hawai’i) | Tobacco, alcohol, drugs, violence, sexual health, academic, and school‐related outcomes | USA | Grades 2 ‐ 3 | 3 years | Theory of self‐concept, Theory of triadic influence | |
| LIFT | Tobacco, alcohol, drugs | USA | Grades 1 and 5 | 10 weeks | Coercion theory | |
| Aban Aya | Violence, drugs, sexual health | USA | Grade 5 | 4 years | Theory of triadic influence | |
| Positive Action (Chicago) | Tobacco, alcohol, drugs, violence, academic, and school‐related outcomes | USA | Grade 3 | 6 years | Theory of self‐concept, Theory of triadic influence | |
| DARE Plus | Tobacco, alcohol, drugs, violence | USA | Grade 7 | 2 years | Theory of triadic influence | |
| Hunter Regions Health Promoting Schools Program | Tobacco | Australia | Years 7 ‐ 8 | 2 years | Health Promoting Schools framework, Community organisation theory | |
| Going Places | Tobacco, alcohol | USA | Grades 6 ‐ 8 | 3 years | Social cognitive theory | |
| Sexual health interventions | ||||||
| Safer Choices | Sexual health | USA | Grade 9 | 2 years | Social Cognitive Theory, Social Influence Theory and Models of School Change | |
| MEMA Kwa Vijana | Sexual health | Tanzania | Students aged 14+ years | 3 years | Social Learning Theory | |
| Mental health and emotional well‐being interventions | ||||||
| Gatehouse Project | Mental health and emotional well‐being, tobacco, drugs, bullying | Australia | Grade 8 | 3 years | Health Promoting Schools Framework, Attachment Theory | |
| beyondblue | Mental health and emotional well‐being | Australia | Year 8 | 3 years | Health Promoting Schools Framework | |
| Violence prevention interventions | ||||||
| Students for Peace | Violence | USA | Grades 6 ‐ 8 | 3 semesters. | Social cognitive theory | |
| Fourth R | Violence, sexual health | Canada | Grade 9 | 15 weeks | None stated | |
| Anti‐bullying interventions | ||||||
| Friendly Schools | Bullying | Australia | Grade 4 | 2 years | Health Promoting Schools framework, Social cognitive theory, Ecological theory, Social control theory, Health belief model, Problem behaviour theory | |
| Friendly Schools, Friendly Families | Bullying | Australia | Grades 2, 4, and 6 | 2 years | Health Promoting Schools framework | |
| ‐ | Bullying | Netherlands | 9 ‐ 12 year‐olds | 2 years | No specific theory but based on Olweus bullying programme | |
| Steps to Respect | Bullying | USA | Grades 3 ‐ 6 | 1 year | None stated | |
| KiVa (1) | Bullying | Finland | Grade 4 ‐ 6 | 9 months | Social cognitive theory | |
| KiVa (2) | Bullying | Finland | Grade 1 ‐ 3 and 7 ‐ 9 | 9 months | Social cognitive theory | |
| ‐ | Bullying | Belgium | 10 ‐ 16 year‐olds | Not clear | Social learning theory | |
| Hand‐washing interventions | ||||||
| ‐ | Illness from infectious diseases, attendance outcomes | China | Grade 1 | 5 months | None stated | |
| ‐ | Illness from infectious diseases | Egypt | Grades 1 ‐ 3 (for data collection, but all children in school targeted) | 12 weeks | None stated | |
| Miscellaneous interventions | ||||||
| School Bicycle Safety Project / The Helmet Files | Safety or accidents | Australia | Grade 5 | 2 years | Health Promoting Schools framework | |
| Healthy Schools‐ Healthy Kids | Body image | Canada | Grade 6 ‐ 7 | 8 months | Health Promoting Schools framework, Ecological approach | |
| SunSafe | Sun safety | USA | Grades 6 ‐ 8 | 3 years | Social cognitive theory, Socio‐ecological theory, Protection motivation theory | |
| ‐ | Oral health | China | Grade 1 | 3 years | Health Promoting Schools framework | |
| Name | Approach | Country | Duration | Costs | Cost effectiveness |
| Nutrition | UK | 8 months | Costs estimated to be GP 378 for capital and development costs plus GBP 13.50 consumables per school | ‐ | |
| Sexual health | USA | 2 years | The total cost of the intervention was USD 105,243. | For every dollar invested in the program, USD 2.65 in total medical and social costs were saved. | |
| Physical activity and nutrition | Germany | 9 months | Intervention costs were EUR 24.09 per child. | The incremental cost‐effectiveness relation was EUR 11.11 (95% CI, 8.78 to 15.02) per cm waist circumference growth prevented and EUR 18.55 (95% CI, 14.04 to 26.86) per unit of waist‐to‐height ratio gain prevented. The authors conclude that based on a ‘maximum willingness to pay’ of EUR 35, the intervention can be considered cost‐effective. | |
| Tobacco | Finland | 3 years | Estimated costs per school each year were EUR 2500. | ‐ | |
| Nutrition | USA | 2½ years | No costs associated with the school‐wide loud‐speaker announcements or the CD‐ROM element which was available to schools free of charge. Costs associated with the lunchtime component were USD 0.04/sticker and a one‐time cost of approximately USD 100 to print the posters. Each family book cost USD 3.38. | ‐ | |
| Sexual health | Tanzania | 3 years | The 3‐year costs of trial implementation were USD 879,032. Initial start‐up costs were high but annual costs dropped from USD 16 per student in 1999 to USD 10 per student in 2001. Authors estimate that when scaled up, only an additional USD 1.54 is needed per pupil per year to continue the intervention. | ‐ | |
| Physical activity and nutrition | New Zealand | 2 years | Average cost estimated to be less than NZD 40. | ‐ | |
| Dating violence prevention | Canada | 15 weeks | Estimated costs of CAD 16 per student in initial year. Includes teacher release time for training (CAD 200 x 40 teachers = CAD 8000) and reusable curriculum materials (mean, CAD 700 per school or CAD 175 per teacher). | ‐ | |
| CI: confidence interval | |||||
| Review outcome | GRADE assessment | Justification |
| Obesity or overweight or body size | Moderate | RCT evidence downgraded on basis of high levels of unexplained heterogeneity |
| Physical activity | Low/moderate | RCT evidence downgraded on basis of high levels of unexplained heterogeneity and risk of bias (blinding of participants) for physical activity, but not physical fitness measures |
| Nutrition | Low | RCT evidence downgraded on basis of high levels of unexplained heterogeneity and lack of blinding of outcome measures |
| Tobacco | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants and attrition) |
| Alcohol | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Substance use | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Sexual health | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Mental health | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Violence | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Bullying | Low | RCT evidence downgraded on basis of high levels of unexpected heterogeneity and risk of bias (blinding of participants and attrition) |
| Infectious disease | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Accident prevention | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants and attrition) |
| Body image or eating disorders | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants and attrition) |
| Skin or sun safety | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Oral health | Moderate | RCT evidence downgraded on basis of risk of bias (blinding of participants) |
| Academic or attendance outcomes | Moderate | RCT evidence downgraded in basis of risk of bias (attrition) |
| RCT: randomised controlled trial. The quality of the body of evidence from randomised trials is usually assessed as 'high' within the GRADE system. However, randomised trial evidence can be downgraded to moderate, low or very low quality on the basis of five factors: limitations in the design and implementation (often indicative of high bias risk); indirectness of evidence; unexplained heterogeneity or inconsistency of results; imprecision of results; and high probability of publication bias. For further description of GRADE levels of quality of a body of evidence see section 12.2 in Higgins 2011a. | ||
| Accelerometry vs. self reported physical activity | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Physical activity | Physical activity only | accelerometry | 1 | 297 | 205 | 0.01 [‐0.01 to 0.03] | n/a |
| self report | 1 | 374 | 358 | 0.35 [0.17 to 0.53] | n/a | ||
| Physical activity + nutrition | accelerometry | 3 | 1475 | 1341 | 0.18 [0.10 to 0.26] | 0% | |
| self report | 3 | 1769 | 1605 | 0.12 [‐0.15 to 0.38] | 85% | ||
| Using 'vegetable intake' instead of 'fruit intake' where these were reported separately | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Fruit and vegetable intake | Nutrition only | fruit intake | 10 studies, 3 substitutions | 3293 | 2917 | 0.15 [0.02 to 0.29] | 83% |
| vegetable intake | 10 studies, 3 substitutions | 3293 | 2917 | 0.14 [0.01 to 0.27] | 83% | ||
| Physical activity + nutrition | fruit intake | 6 studies, 3 substitutions | 3507 | 3105 | 0.04 [‐0.18 to 0.26] | 79% | |
| vegetable intake | 6 studies, 3 substitutions | 3507 | 3105 | ‐0.07 [‐0.19 to0.04] | 26% | ||
| Excluding studies with borrowed standard deviations (SDs) | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| zBMI | Physical activity + nutrition | with borrowed SDs | 7 | 5672 | 5512 | ‐0.00 [‐0.04 to 0.03] | 41% |
| without borrowed SDs | 6 | 4980 | 4852 | ‐0.01 [‐0.05 to 0.03] | 39% | ||
| Fat intake | Nutrition only | with borrowed SDs | 7 | 2205 | 2011 | ‐0.08 [‐0.21 to 0.05] | 68% |
| without borrowed SDs | 4 | 1183 | 986 | 0.00 [‐0.08 to 0.08] | 27% | ||
| Physical activity + nutrition | with borrowed SDs | 10 | 6498 | 5962 | ‐0.04 [‐0.20 to0.12] | 95% | |
| without borrowed SDs | 9 | 6197 | 5643 | ‐0.00 [‐0.17 to 0.17] | 95% | ||
| Fruit and vegetable intake | Nutrition only | with borrowed SDs | 9 | 3293 | 2917 | 0.15 [0.02 to 0.29] | 83% |
| without borrowed SDs | 6 | 2188 | 1865 | 0.05 [‐0.06 to 0.16] | 67% | ||
| Physical activity | Physical activity + nutrition | with borrowed SDs | 6 | 3244 | 2946 | 0.14 [0.03 to 0.26] | 66% |
| without borrowed SDs | 5 | 3108 | 2804 | 0.14 [0.01 to 0.27] | 72% | ||
| Alcohol use | Alcohol intervention | with borrowed SDs | 2 | 3477 | 3817 | 0.72 [0.34 to1.52] | 82% |
| without borrowed SDs | 1 | 2501 | 3079 | 0.99 [0.97 to 1.01] | n/a | ||
| Random‐ versus fixed‐effect meta‐analyses | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Fruit and vegetable intake | Nutrition only | random | 9 | 2205 | 2011 | ‐0.08 [‐0.21 to 0.05] | 68% |
| fixed | 9 | 2205 | 2011 | ‐0.05 [‐0.10 to 0.00] | 68% | ||
| Alcohol use | Multiple risk behaviours | random | 4 | 4496 | 3644 | 0.75 [0.55 to 1.02] | 78% |
| fixed | 4 | 4496 | 3644 | 0.88 [0.78 to 1.00] | 78% | ||
| Substance use | Multiple risk behaviours | random | 3 | 3804 | 3016 | 0.57 [0.29 to1.14] | 71% |
| fixed | 3 | 3804 | 3016 | 0.76 [0.60 to 0.96] | 71% | ||
| Violence | Multiple risk behaviours | random | 3 | 3806 | 3014 | 0.50 [0.23 to 1.09] | 93% |
| fixed | 3 | 3806 | 3014 | 0.89 [0.82 to 0.96] | 93% | ||
| Bullying others | Anti‐bullying | random | 6 | 13949 | 12227 | 0.90 [0.78 to 1.04] | 67% |
| fixed | 6 | 13949 | 12227 | 0.81 [0.77 to 0.87] | 67% | ||
| ALLOCATION CONCEALMENT | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| Being bullied | Anti‐bullying | All studies | 6 | 13993 | 12263 | 0.83 [0.72 to 0.96] | 61% |
| Low risk only | 4 | 12438 | 10694 | 0.85 [0.71 to 1.03] | 76% | ||
| BLINDING OF OUTCOME ASSESSORS FOR OBJECTIVE MEASURES | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| BMI | Physical activity + nutrition | All studies | 9 | 6520 | 7108 | ‐0.11 [‐0.24 to 0.02] | 84% |
| Low risk only | 1 | 727 | 682 | ‐0.20 [‐0.53 to 0.13] | n/a | ||
| zBMI | Physical activity + nutrition | All studies | 7 | 4980 | 4852 | ‐0.01 [‐0.05 to 0.03] | 39% |
| Low risk only | 3 | 3184 | 3172 | ‐0.01 [‐0.08 to 0.05] | 52% | ||
| Physical activity | Physical activity + nutrition | All studies | 6 | 3244 | 2946 | 0.14 [0.03 to 0.26] | 66% |
| Low risk only | 3 | 1475 | 1341 | 0.18 [0.10 to 0.26] | 0% | ||
| Physical fitness | Physical activity + nutrition | All studies | 3 | 2059 | 2171 | 0.12 [0.04 to 0.20] | 0% |
| Low risk only | 1 | 619 | 602 | 0.13 [0.01 to 0.25] | n/a | ||
| LOW ATTRITION RATES | |||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | Estimate [95% CI] | I² |
| BMI | Physical activity + nutrition | All studies | 9 | 6520 | 7108 | ‐0.11 [‐0.24 to 0.02] | 84% |
| Low risk only | 5 | 4095 | 4705 | ‐0.11 [‐0.29 to 0.07] | 76% | ||
| zBMI | Physical activity + nutrition | All studies | 7 | 4980 | 4852 | ‐0.01 [‐0.05 to 0.03] | 39% |
| Low risk only | 3 | 3544 | 3402 | ‐0.02 [‐0.05 to 0.02] | 0% | ||
| Physical activity | Physical activity + nutrition | All studies | 6 | 3244 | 2946 | 0.14 [0.03 to 0.26] | 66% |
| Low risk only | 2 | 428 | 443 | ‐0.03 [‐0.31 to 0.26] | 68% | ||
| Age group (< 12 years>) subgroup analyses | ||||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | MD or SMD [95% CI] | I² | Meta‐regression |
| BMI | Physical activity only | younger (≤ 12 years) | 1 | 297 | 205 | ‐0.12 [‐0.20 to ‐0.04] | n/a | n/a |
| older (> 12 years) | 1 | 374 | 358 | ‐0.28 [‐0.52 to ‐0.04] | n/a | |||
| Physical activity + nutrition | younger (≤ 12 years) | 8 | 4350 | 5242 | ‐0.28 [‐0.47 to ‐0.10] | 86% | 0.47 [‐0.11 to 1.05] | |
| older (> 12 years) | 3 | 2271 | 1961 | 0.08 [‐0.08 to 0.24] | 68% | |||
| zBMI | Physical activity + nutrition | younger (≤ 12 years) | 6 | 2507 | 2708 | ‐0.05 [‐0.12 to 0.02] | 78% | 0.12 [‐0.12 to 0.43] |
| older (> 12 years) | 2 | 3267 | 2898 | 0.04 [‐0.08 to 0.17] | 73% | |||
| Physical activity | Physical activity only | younger (≤ 12 years) | 1 | 297 | 205 | 0.01 [‐0.01 to 0.03] | n/a | n/a |
| older (> 12 years) | 1 | 374 | 358 | 0.35 [0.17 to 0.53] | n/a | |||
| Physical activity + nutrition | younger (≤ 12 years) | 4 | 1403 | 1515 | 0.06 [‐0.10 to 0.23] | 54% | 0.18 [‐0.10 to 0.46] | |
| older (> 12 years) | 2 | 1841 | 1431 | 0.24 [0.17 to 0.31] | 0% | |||
| Fat intake | Nutrition only | younger (≤ 12 years) | 5 | 1770 | 1704 | ‐0.17 [‐0.35 to 0.00] | 73% | 0.28 [‐0.17 to 0.73] |
| older (> 12 years) | 2 | 435 | 307 | 0.10 [‐0.05 to 0.25] | 0% | |||
| Physical activity + nutrition | younger (≤ 12 years) | 7 | 2762 | 2646 | 0.00 [‐0.32 to 0.33] | 94% | ‐0.18 [‐ 0.78 to 0.42] | |
| older (> 12 years) | 3 | 3736 | 3316 | ‐0.17 [‐0.41 to 0.07] | 97% | |||
| Fruit and vegetable intake | Nutrition only | younger (≤ 12 years) | 7 | 2858 | 2610 | 0.20 [0.05 to 0.35] | 85% | ‐0.24 [‐0.65 to 0.16] |
| older (> 12 years) | 2 | 435 | 307 | ‐0.04 [‐0.36 to 0.28] | 76% | |||
| Physical activity + nutrition | younger (≤ 12 years) | 2 | 488 | 506 | ‐0.06 [‐0.22 to 0.11] | 0% | 0.18 [‐1.11 to 1.49] | |
| older (> 12 years) | 2 | 3019 | 2599 | 0.16 [‐0.42 to 0.74] | 93% | |||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | OR [95% CI] | I² | Meta‐regression OR [95% CI] |
| Tobacco use | Multiple risk behaviours | younger (≤ 12 years) | 2 | 1169 | 908 | 0.68 [0.35 to 1.31] | 32% | 1.31 [0.55, 3.11] |
| older (> 12 years) | 3 | 4334 | 3581 | 0.85 [0.77 to 0.94] | 0% | |||
| Alcohol use | Multiple risk behaviours | younger (≤ 12 years) | 2 | 1169 | 908 | 0.47 [0.33 to 0.67] | 0% | 2.04 [0.88, 4.73] |
| older (> 12 years) | 2 | 3327 | 2736 | 0.96 [0.84 to 1.09] | 0% | |||
| Substance use | Multiple risk behaviours | younger (≤ 12 years) | 2 | 1169 | 908 | 0.41 [0.18 to 0.93] | 44% | 2.07 [0.00, 33.42] |
| older (> 12 years) | 1 | 2635 | 2108 | 0.85 [0.66 to 1.10] | n/a | |||
| Violence | Multiple risk behaviours | younger (≤12 years) | 2 | 1171 | 906 | 0.36 [0.26 to 0.50] | 0% | 2.60 [0.27, 24.59] |
| older (> 12 years) | 1 | 2635 | 2108 | 0.93 [0.86 to 1.01] | n/a | |||
| Being bullied | Anti‐bullying | younger (≤12 years) | 6 | 8556 | 8301 | 0.84 [0.70 to 1.01] | 71% | 1.15 [0.70, 1.89] |
| older (> 12 years) | 2 | 5437 | 3962 | 1.01 [0.86 to 1.19] | 0% | |||
| Bullying others | Anti‐bullying | younger (≤12 years) | 6 | 8550 | 8292 | 0.84 [0.70 to 1.02] | 70% | 1.05 [0.57, 1.95] |
| older (> 12 years) | 2 | 5399 | 3935 | 0.92 [0.77 to 1.09] | 0% | |||
| Duration (< 12 months>) subgroup analyses | ||||||||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | MD or SMD [95% CI] | I² | Meta‐regression MD or SMD [95% CI] |
| BMI | Physical activity only | shorter (≤ 12 months) | 1 | 297 | 205 | ‐0.12 [‐0.20 to ‐0.04] | n/a | n/a |
| longer (> 12 months) | 1 | 374 | 358 | ‐0.28 [‐0.52 to ‐0.04] | n/a | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 4 | 2289 | 2471 | ‐0.37 [‐0.70 to ‐0.03] | 88% | 0.29 [‐0.39 to 0.97] | |
| longer (> 12 months) | 6 | 4332 | 4732 | ‐0.08 [‐0.26 to 0.10] | 87% | |||
| zBMI | Physical activity + nutrition | shorter (≤ 12 months) | 2 | 394 | 397 | ‐0.22 [‐0.68 to 0.24] | 93% | 0.18 [‐0.12 to 0.48] |
| longer (> 12 months) | 6 | 5380 | 5209 | ‐0.00 [‐0.04 to 0.04] | 50% | |||
| Physical activity | Physical activity only | shorter (≤ 12 months) | 1 | 297 | 205 | 0.01 [‐0.01 to 0.03] | n/a | n/a |
| longer (> 12 months) | 1 | 374 | 358 | 0.35 [0.17 to 0.53] | n/a | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 1 | 292 | 301 | ‐0.17 [‐0.39 to 0.05] | n/a | 0.39 [0.07 to 0.71] | |
| longer (> 12 months) | 5 | 2952 | 2645 | 0.22 [0.16 to 0.28] | 93% | |||
| Fat intake | Nutrition only | shorter (≤ 12 months) | 5 | 1480 | 1512 | ‐0.17 [‐0.42 to 0.07] | 76% | 0.18 [‐0.34 to 0.69] |
| longer (> 12 months) | 2 | 725 | 499 | ‐0.02 [‐0.13 to 0.09] | 36% | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 4 | 1616 | 1622 | 0.20 [‐0.23 to 0.62] | 96% | ‐0.42 [‐0.90 to 0.07] | |
| longer (> 12 months) | 6 | 4882 | 4340 | ‐0.21 [‐0.39 to ‐0.02] | 94% | |||
| Fruit and vegetable intake | Nutrition only | shorter (≤ 12 months) | 6 | 1766 | 1743 | 0.24 [0.07 to 0.41] | 78% | ‐0.22 [‐0.55 to 0.11] |
| longer (>12 months) | 3 | 1527 | 1174 | 0.02 [‐0.18 to 0.21] | 84% | |||
| Physical activity + nutrition | shorter (≤ 12 months) | 1 | 292 | 301 | 0.14 [‐0.15 to 0.43] | n/a | 0.07 [‐1.59 to 1.73] | |
| longer (> 12 months) | 3 | 3215 | 2804 | 0.06 [‐0.22 to 0.34] | 86% | |||
| Outcome | Intervention type | Subgroup | N Studies | N intervention | N control | OR [95% CI] | I² | Meta‐regression OR [95% CI] |
| Being bullied | Anti‐bullying | shorter (≤ 12 months) | 3 | 12209 | 10472 | 0.74 [0.69 to 0.80] | 0% | 1.49 [0.97 to 2.27] |
| longer (> 12 months) | 2 | 1784 | 1791 | 1.08 [0.76 to 1.53] | 46% | |||
| Bullying others | Anti‐bullying | shorter (≤ 12 months) | 3 | 11887 | 10256 | 0.77 [0.72 to 0.82] | 0% | 1.28 [0.81 to 2.02] |
| longer (> 12 months) | 2 | 1777 | 1786 | 0.99 [0.75 to 1.30] | 0% | |||
| Gender subgroup analyses (as presented by authors) | ||||||||
| Outcome | Intervention type | Study name | Authors' results | |||||
| BMI | Physical activity + nutrition | Effect found for girls (increase in BMI: 1.11 kg/m² versus 1.66 kg/m² for intervention and control groups, respectively, P < 0.05) but not for boys | ||||||
| Effect found for boys (BMI: ‐0.28 kg/m² versus 0.36 kg/m² for intervention and control groups, respectively, P value = 0.04) but not for girls | ||||||||
| zBMI | Physical activity + nutrition | Effect found for girls (increase in zBMI: 0 versus 0.17 for intervention and control groups, respectively, P < 0.05) but not for boys | ||||||
| No effect found in either boys or girls | ||||||||
| Physical activity | Physical activity + nutrition | No difference between girls and boys in terms of self‐reported physical activity | ||||||
| No difference between girls and boys in terms of self‐reported physical activity | ||||||||
| Fat intake | Physical activity + nutrition | Significant reductions in intervention compared to controls for fat intake and % energy from fat in girls (P < 0.001 for both). No effect was seen for boys | ||||||
| No difference between girls and boys in terms of fat intake | ||||||||
| Tobacco | Multiple risk behaviours | Positive effect in boys (0.18 versus 0.31 for intervention and control groups, respectively, P value = 0.02) but not in girls | ||||||
| Alcohol | Multiple risk behaviours | Positive effect in boys but not in girls (1.19 versus 1.64, for intervention and control groups, respectively, P value = 0.04) but not in girls | ||||||
| Substance use | Multiple risk behaviours | No effect found in either boys or girls | ||||||
| Violence | Multiple risk behaviours | No effect found in either boys or girls | ||||||
| Violence prevention | No effect found in either boys or girls | |||||||
| Bullying | Multiple risk behaviours | Reduction in physical victimisation effect in boys (‐0.10 versus 0.03, for intervention and control groups, respectively, P value = 0.02) but not in girls | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 BMI Show forest plot | 13 | Mean Difference (Random, 95% CI) | Subtotals only | |
| 1.1 Nutrition only | 1 | 843 | Mean Difference (Random, 95% CI) | ‐0.04 [‐0.28, 0.20] |
| 1.2 Physical activity only | 3 | 1430 | Mean Difference (Random, 95% CI) | ‐0.38 [‐0.73, ‐0.03] |
| 1.3 Physical activity + nutrition | 9 | 13628 | Mean Difference (Random, 95% CI) | ‐0.11 [‐0.24, 0.02] |
| 2 zBMI Show forest plot | 9 | Mean Difference (Random, 95% CI) | Subtotals only | |
| 2.1 Nutrition only | 1 | 843 | Mean Difference (Random, 95% CI) | ‐0.01 [‐0.09, 0.07] |
| 2.2 Physical activity only | 1 | 196 | Mean Difference (Random, 95% CI) | ‐0.47 [‐0.69, ‐0.25] |
| 2.3 Physical activity + nutrition | 7 | 11184 | Mean Difference (Random, 95% CI) | ‐0.00 [‐0.04, 0.03] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Physical activity Show forest plot | 9 | Std. Mean Difference (Random, 95% CI) | Subtotals only | |
| 1.1 Nutrition only | 1 | 751 | Std. Mean Difference (Random, 95% CI) | 0.02 [‐0.02, 0.06] |
| 1.2 Physical activity only | 2 | 1234 | Std. Mean Difference (Random, 95% CI) | 0.17 [‐0.16, 0.50] |
| 1.3 Physical activity + nutrition | 6 | 6190 | Std. Mean Difference (Random, 95% CI) | 0.14 [0.03, 0.26] |
| 2 Physical fitness Show forest plot | 5 | Std. Mean Difference (Random, 95% CI) | Subtotals only | |
| 2.1 Physical activity only | 2 | 694 | Std. Mean Difference (Random, 95% CI) | 0.35 [‐0.20, 0.90] |
| 2.2 Physical activity + nutrition | 3 | 4230 | Std. Mean Difference (Random, 95% CI) | 0.12 [0.04, 0.20] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Fat intake Show forest plot | 17 | Std. Mean Difference (Random, 95% CI) | Subtotals only | |
| 1.1 Nutrition only | 7 | 4216 | Std. Mean Difference (Random, 95% CI) | ‐0.08 [‐0.21, 0.05] |
| 1.2 Physical activity + nutrition | 10 | 12460 | Std. Mean Difference (Random, 95% CI) | ‐0.04 [‐0.20, 0.12] |
| 2 Fruit and vegetable intake Show forest plot | 13 | Std. Mean Difference (Random, 95% CI) | Subtotals only | |
| 2.1 Nutrition only | 9 | 6210 | Std. Mean Difference (Random, 95% CI) | 0.15 [0.02, 0.29] |
| 2.2 Physical activity + nutrition | 4 | 6612 | Std. Mean Difference (Random, 95% CI) | 0.04 [‐0.18, 0.26] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Tobacco use Show forest plot | 10 | Odds Ratio (Random, 95% CI) | Subtotals only | |
| 1.1 Tobacco interventions | 3 | 4747 | Odds Ratio (Random, 95% CI) | 0.77 [0.64, 0.93] |
| 1.2 Multiple risk behaviours interventions | 5 | 9992 | Odds Ratio (Random, 95% CI) | 0.84 [0.76, 0.93] |
| 1.3 Emotional well‐being interventions | 1 | 630 | Odds Ratio (Random, 95% CI) | 0.79 [0.59, 1.06] |
| 1.4 Alcohol interventions | 1 | 1901 | Odds Ratio (Random, 95% CI) | 0.74 [0.61, 0.90] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Alcohol use Show forest plot | 7 | Odds Ratio (Random, 95% CI) | Subtotals only | |
| 1.1 Alcohol interventions | 2 | 7481 | Odds Ratio (Random, 95% CI) | 0.72 [0.34, 1.52] |
| 1.2 Multiple risk behaviour interventions | 4 | 8140 | Odds Ratio (Random, 95% CI) | 0.75 [0.55, 1.02] |
| 1.3 Emotional well‐being interventions | 1 | 1619 | Odds Ratio (Random, 95% CI) | 1.13 [0.76, 1.67] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Substance use Show forest plot | 6 | Odds Ratio (Random, 95% CI) | Subtotals only | |
| 1.1 Multiple risk behaviour interventions | 3 | 6820 | Odds Ratio (Random, 95% CI) | 0.57 [0.29, 1.14] |
| 1.2 Alcohol interventions | 2 | 7481 | Odds Ratio (Random, 95% CI) | 0.94 [0.78, 1.12] |
| 1.3 Emotional well‐being interventions | 1 | 466 | Odds Ratio (Random, 95% CI) | 0.81 [0.57, 1.15] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Depression Show forest plot | 3 | Std. Mean Difference (Random, 95% CI) | Subtotals only | |
| 1.1 Emotional well‐being interventions | 2 | 6099 | Std. Mean Difference (Random, 95% CI) | 0.06 [‐0.00, 0.13] |
| 1.2 Anti‐bullying interventions | 1 | 2224 | Std. Mean Difference (Random, 95% CI) | 0.0 [‐0.08, 0.08] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Violence Show forest plot | 4 | Odds Ratio (Random, 95% CI) | Subtotals only | |
| 1.1 Violence prevention interventions | 1 | 2090 | Odds Ratio (Random, 95% CI) | 1.13 [0.61, 2.07] |
| 1.2 Multiple risk behaviour interventions | 3 | 6820 | Odds Ratio (Random, 95% CI) | 0.50 [0.23, 1.09] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Being bullied Show forest plot | 8 | Odds Ratio (Random, 95% CI) | Subtotals only | |
| 1.1 Anti‐bullying interventions | 6 | 26256 | Odds Ratio (Random, 95% CI) | 0.83 [0.72, 0.96] |
| 1.2 Multiple risk behaviour interventions | 1 | 4743 | Odds Ratio (Random, 95% CI) | 0.97 [0.90, 1.05] |
| 1.3 Emotional well‐being interventions | 1 | 963 | Odds Ratio (Random, 95% CI) | 0.88 [0.68, 1.13] |
| 2 Bullying others Show forest plot | 7 | Odds Ratio (Random, 95% CI) | Subtotals only | |
| 2.1 Anti‐bullying interventions | 6 | 26176 | Odds Ratio (Random, 95% CI) | 0.90 [0.78, 1.04] |
| 2.2 Multiple risk behaviours interventions | 1 | 363 | Odds Ratio (Random, 95% CI) | 0.49 [0.34, 0.71] |
