- Esther M F van Sluijs, investigator scientist,
- Alison M McMinn, PhD student,
- Simon J Griffin, group leader
- Medical Research Council Epidemiology Unit, Institute of Metabolic Sciences, Addenbrooke's Hospital, Cambridge CB2 0QQ
- Correspondence to: E M F van Sluijs
- Accepted 11 July 2007
Objective To review the published literature on the effectiveness of interventions to promote physical activity in children and adolescents.
Design Systematic review.
Data sources Literature search using PubMed, SCOPUS, Psychlit, Ovid Medline, Sportdiscus, and Embase up to December 2006.
Review methods Two independent reviewers assessed studies against the following inclusion criteria: controlled trial, comparison of intervention to promote physical activity with no intervention control condition, participants younger than 18 years, and reported statistical analyses of a physical activity outcome measure. Levels of evidence, accounting for methodological quality, were assessed for three types of intervention, five settings, and three target populations.
Results The literature search identified 57 studies: 33 aimed at children and 24 at adolescents. Twenty four studies were of high methodological quality, including 13 studies in children. Interventions that were found to be effective achieved increases ranging from an additional 2.6 minutes of physical education related physical activity to 283 minutes per week of overall physical activity. Among children, limited evidence for an effect was found for interventions targeting children from low socioeconomic populations, and environmental interventions. Strong evidence was found that school based interventions with involvement of the family or community and multicomponent interventions can increase physical activity in adolescents.
Conclusion Some evidence was found for potentially effective strategies to increase children's levels of physical activity. For adolescents, multicomponent interventions and interventions that included both school and family or community involvement have the potential to make important differences to levels of physical activity and should be promoted. A lack of high quality evaluations hampers conclusions concerning effectiveness, especially among children.
The prevalence of childhood obesity and related health problems is increasing in many Western countries and is anticipated to continue to increase.1 Evidence of an association between physical activity and weight gain remains sparse.2 Nevertheless, in an effort to halt or reverse trends in obesity, promotion of physical activity in children and adolescents has been identified as a key focus of efforts to promote health.3 4 5 Physical activity among children and adolescents is believed to be insufficient,6 7 8 and low levels of activity seem to persist into adulthood.9 10 This makes physical inactivity among young people a risk factor for cardiovascular disease, cancer, and osteoporosis in later life.11 The development and evaluation of interventions to promote physical activity in young people is therefore a priority.
It is unclear how successful efforts have been to increase the activity levels of young people. Recently published reviews have mostly dealt with the prevention of obesity2 12 13 14 or included only adult populations.15 16 17 Previous attempts to summarise the evidence in young people were mostly narrative,18 19 20 did not assess the effects on children and adolescents separately,19 20 21 and did not assess the methodological quality of the studies.18 19 20 21 22 In addition these reviews have included studies without a no intervention control group18 19 20 21 and studies in which the promotion of physical activity was only a small part of an overall health promotion programme.18 19 20 21 22 We systematically reviewed the evidence on promotion of physical activity in children and adolescents.
We carried out a literature search of papers on interventions to promote physical activity in young people using six electronic databases (Pubmed, Psychlit, SCOPUS, Ovid Medline, Sportdiscus, and Embase) from the year of their inception up to and including December 2006. The search strategy focused on four key elements: population (for example, youth, children), study design (for example, controlled trial, random), behaviour (for example, physical activity, walking, exercise), and intervention (for example, health education, behaviour change). (The full list of terms is available at www.mrc-epid.cam.ac.uk/Publications/Supplementary_Material/VanSluijsBMJ2007/.) We also carried out a citation search of included papers and published relevant reviews.2 12 13 14 18 19 21 22 23 24 25 26 27 28 29 30 Native speakers translated potentially relevant foreign language papers.
We restricted the review to published trials, applying the following inclusion criteria: children and adolescents (≤18 years), not selected on the basis of having a specific disease or health problem; interventions in which the main component or one of the components was aimed at promotion of physical activity through behaviour change in any setting (we excluded interventions to reduce sedentary behaviour, or structured exercise programmes to prevent obesity); inclusion of a non-physical activity intervention for the control group; and reported statistical analyses of an outcome measure related to physical activity (self reported or objectively measured).
Two reviewers (EMFvS, AMMcM) independently reviewed the results from the initial search of the title then the abstract and finally the full paper. When opinions differed consensus was reached through discussion.
Assessment of methodological quality
We assessed methodological quality using a 10 item quality assessment scale derived from previously used quality criteria13 16 29 31 and we focused on internal validity and analyses (table 1⇓). Two reviewers (EMFvS, AMMcM) independently assessed for each study whether its score on an item was “positive,” “negative,” or “not, or insufficiently, described.” In cases of disagreement, consensus was reached by discussion. We accumulated the positive scores and defined quality as high when a randomised controlled trial scored six or more or a controlled trial scored five or more. We analysed the level of agreement between the two reviewers using Cohen's κ, with agreement assessed on a dichotomous scale (negative and not described versus positive).
Data extraction on to standardised forms was undertaken separately for studies including children (<12 years, AMMcM) and adolescents (≥12 years, EMFvS). We included interventions in high schools and American middle schools (6th to 8th grade, ages 11 to 14) in the adolescent category. Data extracted included project title, country, study design, inclusion criteria, baseline descriptive data, randomisation procedure, descriptions of intervention and control conditions, length of follow-up, losses to follow-up, selective drop out (observed differences between drop outs and study completers), physical activity measures used, secondary outcome measures, and results. In addition both reviewers (EMFvS, AMMcM) extracted information on the specifics of the intervention (setting, target population, and intervention type), size of the baseline sample, and the overall effectiveness of the study for the main physical activity outcome measure (a measure of individual physical activity was used when available). We scored the size of the study as positive if there were more than 250 participants or if a power calculation was provided justifying the sample size (large) and as negative if there were 250 or fewer participants (small).
Studies used a wide range of methods to assess effectiveness and reported a variety of different outcome measures. We considered devising a common outcome metric for interpretational purposes, similar to a previous review focusing on interventions promoting one particular behaviour, walking.32 Although the studies included in this review assessed walking in different ways, it is a relatively homogeneous behaviour. In contrast, physical activity is more complex and consists of various domains, making it difficult to compare the results of the various measures used to assess this behaviour or particular sub-domains (for example, during breaks, out of school). Consequently because of the heterogeneity of the behaviour of interest and the outcome measures used we decided that calculating one common measure of outcome would not be valid or informative. Alternatively we used scores to indicate effectiveness—that is, whether there was no difference in effect between control and intervention group (0 score), a positive or negative trend (+ or −), or a statistically significant difference (P<0.05) in favour of the intervention or control group (++ or --, respectively). In cases of disagreement, consensus was reached by discussion.
Strength of the evidence
We thought a formal meta-analysis inappropriate owing to the heterogeneity of the interventions, settings, participants, and outcome measures. Instead we used a rating system of levels of evidence to draw conclusions on effectiveness, based on previously used best evidence syntheses.16 31 We defined five levels on the basis of study design, methodological quality, and sample size: strong, moderate, limited, inconclusive, or no evidence for effect (see www.mrc-epid.cam.ac.uk/Publications/Supplementary_Material/VanSluijsBMJ2007/), and conclusions were drawn on the basis of the consistency of results of studies with the highest available level of quality. If at least two thirds (66.6%) of the relevant studies were reported to have significant results in the same direction then we considered the overall results to be consistent. In a stratified analysis we assessed levels of evidence for studies according to setting, target population, and type of intervention (educational only, environmental or policy based only, or a combination of both, “multicomponent”).
Overall, 3045 references were retrieved from the database search (PubMed, n=2000; Psychinfo, n=340; Scopus, n=692; Ovid Medline, n=591; Sportdiscus, n=472; Embase, n=400). Fifty one studies met the inclusion criteria, a further six were added after searching the citations (figure⇓). Thirty three included childrenw1-w47 and 24 included adolescents.w48-w77 Tables 2⇓ and 3⇓ show the characteristics of the studies.
Agreement was 85.9% on the 570 items scored during the quality assessment (κ=0.72, substantial agreement); full consensus was reached after discussion. Overall 24 studies (42%) exhibited high methodological quality, of which 15 (25%) had more than 250 participants (see www.mrc-epid.cam.ac.uk/Publications/Supplementary_Material/VanSluijsBMJ2007/). Most studies applied intention to treat analyses and measured all study groups at similar times, but only 10 studies (18%) had a follow-up of six months or longer. Insufficient information was provided to score the adequacy of the randomisation procedure for 34 studies (60%), and 33 studies (58%) lacked information on allocation concealment at outcome assessment.
Eighteen of the 33 studies in children were carried out in the United States, seven in the United Kingdom, and the remainder in other countries. Most of the child studies (82%) evaluated school based interventions, 14 of which included a community or family component. Around half of the child interventions were educational whereas a third used a multicomponent approach. Just over half of the studies relied on child reported or parent reported questionnaires or diaries as the main measure of physical activity (see www.mrc-epid.cam.ac.uk/Publications/Supplementary_Material/VanSluijsBMJ2007/). Twelve studies used an objective measure of physical activity whereas three used observation methods. Only five of these assessed overall physical activity; the remainder mostly assessed activity during physical education or playtime. Nineteen of the studies assessed overall physical activity, eight measured school based physical activity only, and six assessed out of school physical activity.
Eighteen of the 24 studies in adolescents were carried out in the United States and six in other countries. Almost all evaluated school based interventions, six of which included involvement of the family or community. Five studies included a follow-up measurement of six months or more (see www.mrc-epid.cam.ac.uk/Publications/Supplementary_Material/VanSluijsBMJ2007/). Measurement of physical activity was mostly focused on non-school related activities and carried out with self reported questionnaire or recall instruments. Four studies used an objective measure as the main measure of physical activity, all assessing total physical activity, and one study used direct observation to assess physical education related physical activity.
Evidence of effect on physical activity
Thirty eight studies reported a positive intervention effect (67%), achieving statistical significance in 27 (47%). This included 14 studies in children (42%) and 13 in adolescents (54%). Significant results ranged from an increase of 2.6 minutes during physical education classes to a 42% increase in participation in regular physical activity and an increase of 83 minutes per week in moderate to vigorous physical activity.
Table 4⇓ summarises the stratified levels of evidence for the effectiveness of interventions to promote physical activity in children and adolescents.
Intervention types in children
Nineteen studies evaluated education only interventions, including one large high quality randomised controlled trial,w11 two large high quality controlled trials,w3w4 four small high quality randomised controlled trials,w15-w17w23 and seven low quality randomised controlled trials.w1w2w8w10w19-w21 Four of these reported a statistically significant intervention effect, therefore no overall evidence of an effect of education only interventions was identified. Four studies evaluated changes in the school environment, including two low quality randomised controlled trials.w24w26 Both of these reported a significant intervention effect, providing limited evidence of an effect of environmental interventions.
Ten studies evaluated multicomponent interventions in children, including three large high quality randomised controlled trials.w29w39w44 Only one of these high quality trials reported a significant positive effect, equating to inconclusive evidence of effectiveness.
Intervention types in adolescents
Seventeen studies evaluated education only interventions in adolescents, including four large high quality randomised controlled trials.w48w58w60w63 No evidence of an effect was found, with only one of the large high quality trials reporting statistically significant positive results. Only one study, a low quality randomised controlled trial,w67 evaluated the effect of an environmental intervention, providing inconclusive evidence of an effect.
Six studies evaluated multicomponent interventions, all in the school setting. Three were large high quality randomised controlled trials,w69w72w76 which all showed significant positive results, providing strong evidence of an effect of multicomponent interventions.
Settings in studies of children
Twenty seven studies evaluated school based interventions. Thirteen of these were restricted to the school setting only, including five randomised controlled trials, one of high qualityw39 and four of lower quality.w8w21w24w26 Three of these randomised controlled trials reported significant positive intervention effects, resulting in the classification of inconclusive evidence of an effect of school only interventions. The other 14 school based interventions also included family or community components, such as homework assignments to do with parents or incorporation of physical activity into existing community events. Two large high quality randomised controlled trials were identified,w29w44 one of which showed a significant positive intervention effect, suggesting inconclusive evidence of an effect.
Four studies evaluated family based interventions, including three small high quality randomised controlled trials.w16w17w23 Only one reported a significant effect, which favoured the control group. One large high quality randomised controlled trialw11 and one small high quality randomised controlled trialw15 evaluated the effect of a community based intervention, and neither reported a positive intervention effect. This review therefore provides no evidence of an effect of either family based or community based interventions among children.
Settings in studies of adolescents
Of the 20 studies that evaluated school based interventions, 14 were restricted to the school setting, including two large high quality randomised controlled trials,w58w60 one of which reported a statistically significant intervention effect.w58 This represents inconclusive evidence of an effect. Six studies evaluated school based interventions also including family or community involvement, three of which were large high quality randomised controlled trials.w69w72w76 Two of these large high quality trials showed statistically significant positive results suggesting strong evidence of an effect of school based interventions including family or community involvement.
The only study evaluating a family based intervention, a high quality randomised controlled trial,w63 did not report a positive effect, as did the high quality randomised controlled trialw48 evaluating a community based intervention. One of the two low quality randomised controlled trials evaluating primary care based interventionsw49w57 showed a significant positive effect. Consequently evidence of an effect of either family based, community based, or primary care based interventions in adolescents is inconclusive.
Target populations in children
Five studies evaluated interventions targeted specifically at girls: one large high quality randomised controlled trial,w11 two small high quality randomised controlled trials,w15w16 and two low quality randomised controlled trials.w8w19 Four reported positive effects but only one was significant. Ten studies evaluated interventions specifically aimed at ethnic minority groups, including four small high quality randomised controlled trialsw15-w17w23 and four low quality randomised controlled trials.w2w10w19w33 Only one low quality randomised controlled trial reported a significant positive effect. Therefore no overall evidence of an effect for interventions targeting girls or ethnic minority groups was found. Three controlled trials, including two of high quality,w3w28 assessed the effect of interventions targeting children from low socioeconomic backgrounds. All reported a significant positive effect, resulting in a classification of limited evidence of an effect for these interventions.
Target populations in adolescents
Eight studies evaluated interventions specifically aimed at adolescent girls and one aimed at boys. Three were large high quality randomised controlled trials,w48w69w72 of which one showed significant positive results. In two small high quality studies—one randomised controlled trialw50 and one controlled trialw54—interventions were aimed at low socioeconomic groups, with only one reporting a significant intervention effect. Consequently evidence of an effect of interventions targeting adolescents of one sex or from low socioeconomic groups is inconclusive.
We found that in children there is limited evidence of an effect of interventions targeting low socioeconomic populations and environmental interventions and the evidence of an effect for multicomponent interventions and the two types of school based interventions is inconclusive. More adequately powered high quality research is needed to strengthen and confirm these results. In general, interventions achieved important changes, such as a 13% increase in play time spent in moderate to vigorous physical activity. No evidence of effectiveness was observed in six of the intervention categories. This review especially raises questions about the usefulness of targeting interventions at children from ethnic minority populations or carrying out family based or community based interventions, as most of the studies identified did not report positive results. Before pursuing these strategies further, it is necessary to identify and learn from the limitations of these interventions and their evaluations. In addition, no evidence of an effect was found for educational interventions and the interventions targeting females, despite more than 67% of studies evaluating these interventions reporting positive effects. Most of these were low quality studies and did not always achieve statistical significance.
Overall there was more evidence for an effect of interventions among adolescents than among children. However, more studies in adolescents compared with studies in children were of high quality and included a large sample size (33% v 21%). Adolescents are also known to be less active than children33 34 so may exhibit greater potential for change. Effects ranged from increases of three minutes during physical education to a 50% increase in the number of participants being regularly active. Strong evidence was found for the effectiveness of school based interventions including family or community involvement and multicomponent interventions. No evidence of an effect was observed for educational interventions although an overall positive trend was observed. This trend is, however, mainly due to the results of studies with lower methodological quality and should therefore be interpreted with caution. This review also shows inconclusive evidence of an effect in adolescents in other categories, warranting further investigation.
Almost a third of the included studies in children were targeted at minority ethnic groups, although the evidence on the association between ethnicity and physical activity in children is fairly inconsistent.35 36 In contrast an association has often been reported in adolescents, with levels of physical activity tending to be lower in non-white ethnic groups,35 37 38 39 yet no interventions targeting adolescents from minority ethnic groups were identified. Low socioeconomic status has been identified as a possible determinant of physical inactivity.40 41 42 43 Recently, trials of interventions targeting socially disadvantaged people provided some evidence of the potential of this strategy. A higher level of activity in males compared with females is consistently observed throughout childhood and adolescence and evidence also suggests that both sexes tend to become less active with increasing age.33 35 This review raises doubts about whether targeting females and males separately is an effective approach, although a positive trend was observed among children. Most of the studies investigating differential response by sex did not find one, casting more doubt on the need for separate approaches for the sexes.
Parental factors and the home environment are believed to influence physical activity,44 45 yet few interventions were specifically aimed at the home and those that have been carried out did not show significant positive results. An increasing number of school based interventions do, however, include some parental involvement, although usually limited to newsletters and homework assignments. Evidence of effectiveness of these interventions in adolescents was strong, although in children the evidence is still inconclusive. Whether the strategy of involving parents in interventions will be as effective for children should be the focus of future research.
The conclusion of strong evidence of effect of multicomponent interventions in adolescents and the limited evidence of effect of environmental interventions in children is in keeping with the ecological approach to behaviour change as advocated in recent years.46 47 48 49 The environmental or policy element of multicomponent interventions mostly consisted of alterations to the physical education programme, such as additional classes, physical education teacher training, or the availability of additional equipment. Observations of physical activity during physical education classes mostly showed some increases, but few effects were observed in overall physical activity. This raises the question of whether children might compensate during the rest of the day. It is therefore important to use objective measures to assess the overall effect of the intervention on total activity levels.
Young children's activity is typically intermittent34 in contrast to that of adolescents, which is more structured and planned. Although traditional cognitive approaches, potentially combined with environmental approaches, may increase activity among adolescents and older children (≥10 years), more structural environmental or policy changes might be needed to change younger children's physical activity behaviour.50 This is supported by the evidence of an effect of environmental interventions, which tended to be evaluated in children in the lower grades (grades 1-4) at primary school. Few studies, however, included preschool aged children.
Implementation of interventions
Factors that may have limited effectiveness are the levels of exposure to the intervention and adherence. Several papers reported problems in these areas. For example, Pate et alw37 described that only 5% of participants attended at least half of the sessions offered. Most papers, however, did not describe attendance, implementation, or quality assurance of the intervention, making it impossible to assess the impact these factors may have had on the overall findings.
Most of the studies included in this review were carried out in the United States, raising questions about the generalisability of these results to other countries. Feasibility and effectiveness of cross national implementation is potentially limited owing to known differences in infrastructure, school systems, environments, and social norms. To assess the usefulness of these strategies across different cultures we would advocate replicating evaluations of previously successful interventions adapted to a specific country.
Various limitations in study design and subsequent reporting were identified. In particular, information was lacking on the randomisation procedure and blinding at outcome assessment, limiting the interpretation of the methodological quality. Brief descriptions of interventions hampered stratification of the studies and analyses of potential effective components. Overall, methodological limitations across the studies included short duration of follow-up, inadequate adjustment for potential confounders, and a lack of adjustment for clustering when randomisation was carried out at group level. Another limitation was the lack of precision of the physical activity outcome measures. Eighteen of the studies in children used self reported or parent reported measures (55%), just over half of which were not reported to be previously validated, possibly limiting responsiveness. Studies using observation or objective measures of physical activity were in fact more likely to report significant positive results than studies with a self reported measure, both in children and in adolescents.
Various policy documents have called for the development of effective strategies to increase physical activity in children and adolescents to help halt or reverse the increase in obesity and to improve other aspects of health.4 11 51 52 Based on the published evidence to date it seems that a multilevel approach to promoting physical activity, combining school based interventions with family or community involvement and educational interventions with policy and environmental changes, is likely to be effective among adolescents and should be promoted. For children, there is limited evidence of an effect for environmental interventions and interventions targeting those from low socioeconomic groups. Research should focus on filling the gaps identified in this review, such as the lack of studies among adolescent ethnic minority populations and preschool children, and of interventions outside the school setting. Furthermore, future studies should aim to strengthen the evidence with rigorous design, appropriate sample size, follow-up beyond post intervention to assess maintenance, use of objective measures of overall activity, and assessment of factors along the causal pathway.53 54 Moreover, studies should include assessment of implementation issues and carry out cost effectiveness analyses to further inform future public health strategies in this topic.
What is already known on this topic
Children and adolescents are believed to have low levels of physical activity, which is associated with obesity and other health problems
It is unclear what strategies might be effective to promote physical activity
What this study adds
Multicomponent interventions and interventions including school and family or community involvement may make important differences in physical activity levels in adolescents
For children some evidence of effect was shown for environmental interventions and those targeted at children from low socioeconomic backgrounds
We thank Stephen Sharp for his help in interpreting the statistical methods applied and results presented in the included papers.
Contributors: EMFvS led the review, identified the research question, and designed the search strategy. She is guarantor for the paper. EMFvS and AMMcM carried out the literature searches and screened the initial results, assessed methodological quality, extracted data, analysed the findings, and drafted the tables. EMFvS drafted the manuscript. All authors contributed to synthesising the results and critical revision of the manuscript, and all approved the final version.
Funding: This work was supported by programme grant funding from the Medical Research Council.
Competing interests: None declared.
Ethical approval: Not required.