Inclusion criteria

Studies were included if they:

1. investigated the effectiveness of a school-based intervention targeting physical activity behavior (PA), sedentary behavior (SB) and/or nutrition behavior (NB);
2. considered the effects on children’s Body Mass Index (BMI), the BMI z-score (BMI adjusted for age and gender) and/or PA, SB and NB;
3. targeted children aged 4 to 12 years attending primary school;
4. studied a school-based intervention which consisted of at least one of the following types of interventions: (a) changes to the school’s physical environment, e.g. providing fruit or vegetables at school, or creating an activity-friendly playground; (b) changes to the school’s social environment, e.g. training school staff about health promotion, or the implementation of activity breaks by teachers; (c) changes to the school’s policies, e.g. rules about fruit and vegetable consumption at school or active transportation to school; or (d) economic support for the school, e.g. a budget for implementing activities promoting physical activity or providing fruit and vegetables;
5. studied a school-based intervention that directly involved parents. The definition of ‘direct parental involvement’ by Hingle et al. [14] was used: requesting parents to attend energy balance-related education sessions, e.g. workshops or lessons promoting physical activity, improving children’s diet or reducing sedentary behavior; or asking parents to attend or participate in family behavior counseling or parent training sessions (14). These sessions could have been conducted at home (in group sessions or one-on-one) or at a different location (at school, for example);
6. were written in English.

Exclusion criteria

A study was excluded if:

1. the intervention only indirectly involved parents, as defined by Hingle et al. [14]: (1) sending newsletters or information sheets to parents; (2) inviting parents to attend a health-related information evening; or (3) giving children homework that should be made with the help of their parents;
2. the intervention did not target the school environment (i.e. change the normal school’s routine);
3. the intervention was exclusively aimed at a particular subpopulation, e.g. overweight primary school children;
4. it was defined as a pilot study by the study authors. This was done because the aim of pilot studies is to test an intervention’s feasibility instead of its effectiveness.

Study selection

After removing the duplicates, the retrieved articles were screened independently on title/abstract by the two researchers (SV and IvdK). The remaining articles were screened as full text to assess the eligibility of the studies, based on the inclusion and exclusion criteria, determined a priori. Again this was done, independently, by two researchers (SV and IvdK). An overall agreement between the researchers of 74.6% existed. Discrepancies were discussed until consensus was reached. If no consensus was reached, a third researcher (SG) independently assessed the eligibility of the studies. The third researcher was consulted for four articles.

Data extraction

The PRISMA statement was used in writing this systematic review [19] (S1 File). One researcher (SV) conducted the data extraction regarding the study characteristics, intervention characteristics, and study effectiveness using predefined forms. The information on the following study characteristics was extracted: study design, setting, number of schools participating in the study, the timeframe in which follow-up measurements were made, number of participants, drop-out rates, characteristics of the participants and outcome measures.

The methodological quality of the studies was assessed using the quality assessment instrument of the Effective Public Health Practice Project (EPHPP) [20]. This instrument can be used to assess the quality of quantitative studies with a variety of study designs [20]. The studies were rated on six key quality components: selection bias, study design, confounders, blinding, data collection methods, withdrawals and drop-outs. Each of these quality components was rated ‘weak’, ‘moderate’ or ‘strong’. For example, a strong score was given when the study design was described as an RCT or a controlled clinical trial. An overall score was given based on the scores of the six quality components. The overall quality was rated: ‘strong’ when there were no weak and at least four strong ratings for the six quality components; ‘moderate’ when only one quality component was rated as weak; and ‘weak’ in case there were two or more weak ratings [20]. Two researchers (SV and IvdK) independently rated the quality of the articles. The interrater reliability was 72.1%. Differences were the result of different interpretations of the studies. The researchers compared their quality scoring results and reached consensus by discussion.

The effects on BMI, BMI z-score, physical activity behavior, sedentary behavior, and nutrition behavior were described. Study effectiveness was regarded as positive when all results regarding a particular outcome (e.g. BMI and BMI z-score) showed a statistically significant improvement for the intervention group. Study results were considered mixed when at least one finding of a particular outcome was statistically significant in favor of the intervention group but the other findings were not (e.g. a statistically significant improvement in fruit intake and ineffective or negative results regarding vegetable intake). The results were considered negative, when the results were statistically significant in favor of the control group. An intervention was considered ineffective when there were no statistically significant results for either the intervention or the control group.

Where possible, Cohen’s d effect sizes were calculated to indicate the standardized difference between the means of the intervention and control groups for the different outcome measures [21]. In case of multiple intervention arms, only the effects of the intervention arm that was school-based and included parental involvement were recorded. For studies without a control group, Cohen’s d was calculated by dividing the mean change in the outcome measure by the standard deviation of the baseline value. Lipsey’s cut-off points [22] were used to classify the magnitude of the effect: an effect size below 0.32 was considered ‘small’, between 0.33 and 0.55 ‘moderate’ and 0.56 or more ‘large’ [22]. When information required for calculating an effect size was missing (as was the case in five studies), a request was sent to the authors to provide the missing information. One author responded to this request.

Regarding the interventions, information was extracted on the intervention duration and the behavior targeted by the intervention. In addition, the different types of socio-cognitive determinants of the children (knowledge, awareness, attitude, subjective norms, self-efficacy, intrinsic motivation) and the different environmental types (political, economic, physical and sociocultural) affecting the child were distinguished, according to the EnRG framework by Kremers et al. [17]. Examples of the environmental types are the parental rules regarding the child’s dietary behavior (political), the costs of healthy foods in the school canteen (economic), the availability of play equipment during school breaks (physical) and the stimulation of physical activity behavior at school by the teachers (sociocultural).

General study characteristics

The 25 included studies described the effectiveness of 24 school-based interventions with direct parental involvement. Of the 25 included studies, ten were randomized controlled trials [23–32], eleven were quasi-experimental studies [33–43], three had a pretest-posttest design [44–46] and one study had a repeated cross-sectional design [32] (Table 2). Most interventions (n = 9) were conducted in the USA [26, 33, 35–37, 40, 42, 43, 45]. The interventions were conducted between 1992 and 2015. The number of schools participating in the studies ranged from 1 [37] to 38 [47]. All studies performed the follow-up measurements immediately at the end of the intervention. One study conducted longer-term follow-up measurements (one year follow-up) [24]. The number of children participating in the study at baseline ranged from 97 [37] to 9867 [32]. Drop-out rates varied from 2% [25] to 48% [45]. Information on drop-out rates was missing in six studies [23, 35–37, 40, 43]. The average age of children participating in the included studies ranged from 5 [46] to 12 years [24]. Eighteen studies measured intervention effects on BMI or BMI z-score [23, 25, 26, 28–34, 36, 38, 39, 41, 43–45, 47]. Eleven studies measured PA [23, 26, 27, 29–32, 34, 39, 44, 46] and nine studies measured SB [26, 27, 30–32, 34, 35, 44, 46]. Twelve studies measured NB [23, 24, 26, 29, 31, 32, 36, 37, 40, 42, 44, 46].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. General characteristics of the included studies.

https://doi.org/10.1371/journal.pone.0204560.t002

Quality of the included studies

The methodological quality of eight studies (32%) was rated as weak [26, 32, 33, 36, 39–41, 46] (Table 3). Twelve studies (48%) were rated as being of moderate methodological quality [23, 24, 28, 30, 32, 35, 37, 38, 42–45] and five studies (20%) as being of high methodological quality [25, 27, 29, 31, 34]. Weak ratings were mainly due to information not being reported. For example, only four studies reported whether investigators were blinded to the intervention status of the participants [27, 32, 39, 42]. In addition, information on validity and reliability of data collection instruments or drop-out rates was missing in many studies [23, 26, 28, 30, 33, 35, 40, 41, 43, 46, 48].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Quality rating of included studies (based on Thomas et al. (20)).

https://doi.org/10.1371/journal.pone.0204560.t003

Intervention results regarding BMI or BMI z-score

Eleven of the eighteen studies measuring intervention effects on BMI or BMI z-score found favorable results [23, 25, 28, 29, 33, 34, 38, 39, 41, 45, 47]. Of these, seven studies were positively effective on BMI and BMI z-score [28, 29, 34, 39, 41, 45, 47] and four studies found mixed results [23, 25, 33, 38]. Of the studies with mixed results, one study found the intervention to be positively effective regarding BMI, but ineffective regarding the BMI z-score [23], the remaining three studies found effective results for particular subgroups: normal and overweight subgroups [33]; children who were overweight and obese at baseline [25]; and boys [38] (S2 Table). Of the studies with favorable results, effect sizes for BMI or BMI z-score were mainly small (ES -0.04 to -0.27) [25, 34, 38, 39, 41, 45]. Two studies had a moderate effect size (ES -0.34 and -0.48) [23, 29] and one study found a large effect on BMI (ES -0.79) [28] (S2 Table). The study quality of these studies was strong for three studies [25, 29, 34], moderate for five studies [23, 28, 38, 45, 47] and weak for three studies [33, 39, 41]. Six studies reported their intervention to be ineffective regarding BMI and BMI z-score [26, 30–32, 36, 43], although three of these showed a positive trend (ES -0.10 and -0.01, respectively) [31, 32, 36]. One study found negative results [44]. This was a pretest-posttest study without a comparison group.

Intervention results regarding physical activity behavior

Except for two studies, all eleven studies on PA found favorable results on at least one PA outcome measure [23, 26, 27, 29, 31, 34, 39, 44, 46]: four studies found significant positive results for all PA outcome measures [29, 34, 39, 46] and five studies found significant positive results for at least one PA outcome but were ineffective regarding other PA outcomes [26, 27, 31, 34, 44]. Small to large significant positive effects were found on total daily moderate-to-vigorous PA (MVPA) (ES 0.41 and 0.48) [23, 39], MVPA during school break time (ES 0.19) [27], leisure-time MVPA (ES 0.98) [29], time spent in outdoor activities (ES 0.49) [34], daily physical activities [46], frequency of jogging/running [31], number of sports participated in during the past year (ES 0.10) [26], and supervised sports or dancing per week (ES 0.34) [44]. PA was mostly self-reported by the children [23, 31, 39, 44, 46] or the parents [26, 29, 34]. One study measured PA by the use of accelerometers [27]. The methodological study quality of the studies with favorable results was strong for four studies [27, 29, 31, 34], moderate for two studies [23, 44] and weak for three studies [26, 39, 46]. Two studies found the intervention to be ineffective on PA [30, 32], however one study found a positive trend (ES 0.20) [30].

Intervention results regarding sedentary behavior

Six of the eight studies measuring SB found favorable results [27, 31, 34, 35, 44, 46]. Of these, four studies reported merely significant positive results [31, 34, 35, 46] and two reported significant positive result for at least one SB outcome but were ineffective regarding other SB outcomes [27, 44]. Small to moderate significant positive effects were found on TV watching per week (ES -0.38) [34] and TV watching per day (ES -0.15) [44, 46], screen-time behavior [35], TV viewing and computer use [31], sitting per day (ES -0.20) [44] and SB during school break time (defined by the use of accelerometer-specific cut-off points) (ES -0.02) [27]. SB was mainly measured via child-questionnaire [31, 44, 46] or parent-questionnaire [34, 35]. One study used accelerometers to measure SB [27]. Three studies were of strong methodological quality [27, 31, 34], two of moderate [35, 44] and one of weak quality [46]. Two interventions were found to be ineffective regarding frequency TV viewing. Although not significant, the results were in favor of the intervention group (ES -0.41) [26, 32].

Intervention results regarding nutrition behavior

Twelve studies measured intervention effects on NB. Five studies found favorable results regarding NB outcomes measured [31, 36, 37, 42, 46], of which two studies reported merely statistically significant positive results for all NB outcomes [36, 46] and three studies reported significant positive results, but were ineffective regarding other NB outcomes [31, 37, 41]. Small to moderate significant positive effects were seen on daily sugar-sweetened beverages consumption (ES -0.42) [36], fruit and vegetable consumption (ES 0.17, 0.21, 0.35) [37, 42, 46], added sugar intake (ES -0.21) [42], and red meat consumption [31]. The methodological study quality was mainly weak [36, 41, 46]; two studies were of moderate [37] and strong [31] methodological quality. Two studies reported mixed results: with statistically significant positive results for some NB outcomes and statistically negative results for other outcomes [23, 44]. Five studies found the intervention to be ineffective on NB [24, 26, 29, 32, 40].

Intervention characteristics

The intervention durations varied from ten weeks [37] to four years [46] (Table 2). It seems that interventions of longer duration (at least one year) were more likely to lead to favorable results regarding weight status [23, 25, 28, 29, 41, 45, 47], but not for PA, SB or NB. Two interventions targeted PA only [27, 30] and three interventions targeted NB only [24, 40, 42]. These interventions were mainly ineffective. The remaining interventions targeted children’s PA as well as their NB. Eight studies reported to target children’s SB (e.g. reduce TV viewing) [26, 30, 35, 36, 39, 41, 44–46].

The child’s socio-cognitive determinants targeted.

Five socio-cognitive determinants of the children targeted by the interventions could be distinguished: knowledge, awareness, attitude, self-efficacy and intrinsic motivation (Table 4). Except for three studies [27, 32, 34], all interventions educated the children on nutrition, physical activity, or health with the aim of increasing knowledge about EBRBs. Active ways of nutrition education (cooking classes) and physical education (in the form of extra and better-quality physical education) were implemented by 14 interventions, aiming to increase energy balance-related skills and self-efficacy [23, 24, 26, 29–37, 43–45]. Seven interventions aimed to increase the children’s awareness on their own PA or NB, by asking them to monitor their behavior [23, 24, 30, 31, 34, 39, 44]. Six studies emphasized that their intervention was fun/enjoyable for the children, in order to increase intrinsic motivational regulation [23, 27, 29–31, 34]. Four interventions aimed to change children’s attitudes toward PA and/or nutrition [23, 24, 34, 40]. No pattern was found between the child’s socio-cognitive determinants targeted and intervention effectiveness.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Children’s socio-cognitive determinants, the community, school and family environmental types targeted, and effectiveness of the interventions.

https://doi.org/10.1371/journal.pone.0204560.t004

Parental involvement components.

All interventions involved the parents directly; 17 interventions additionally applied indirect involvement strategies by providing written information to parents, like newsletters, brochures, information sheets, recipe cards and lists of tips (n = 13) [24–26, 28–32, 35, 36, 41–43, 47], and/or by requesting parents to assist their child with the intervention-related homework (n = 8) [23, 25, 26, 29, 30, 37, 42, 43] (Table 5). Parents were predominantly directly involved in the intervention by attending educational sessions (n = 17) [23–25, 27–32, 34, 38–40, 42, 44, 46, 47]. These sessions were mainly organized as group sessions for the parents [24, 27, 28, 31, 34, 38–40, 44, 46]. Eight interventions implemented energy balance-related activities for parents and children (family activities), e.g. family cooking nights, fruit and vegetable bazaars, activities in the supermarket [23, 24, 33, 35, 36, 41–43, 45]. In seven interventions one-on-one counseling was provided, mostly by home visits [26, 33, 35, 36, 46], or telephone counseling [34, 37]. One study did not report the counseling method used [45]. Five interventions additionally targeted parents of overweight children [28, 33, 35, 36, 45, 46]; in these interventions parental counseling sessions were held [33, 35, 36, 45, 46] or information meetings were arranged [28]. These interventions were effective on children’s BMI [28, 33], BMI z-score [45], PA [46], SB and NB [35, 36, 46]. In the four interventions in which parents were provided a report on their child’s health status [23, 29, 44, 45] favorable results were found regarding the intervention effect on children’s BMI [23, 29] and BMI z-score [45], their PA [29, 44] and SB [44].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Parental involvement components of interventions and effectiveness of the interventions.

https://doi.org/10.1371/journal.pone.0204560.t005

Limitations of the studies

There are some limitations of the studies included in this review. Methodological study quality was difficult to rate in most papers because of a lack of detail. As a result, it might be underestimated. A second limitation was the great variation in outcome measures of EBRBs. We tried to overcome this problem by calculating effect sizes. Since only one author answered our request for additional information to allow the calculation of effect sizes, it was impossible to calculate the effect sizes for all studies. Another limitation was the incomplete description of most interventions. This limitation impeded comparison and extraction of information and may have biased the results as presented in this systematic review.

Strengths and limitations of the systematic review

To our knowledge this is the first systematic review aiming to disentangle the socio-cognitive determinants and different environmental types targeted in the school and home environment to explain the intervention effectiveness of school-based physical activity and nutrition interventions with direct parental involvement. We used the EnRG framework [17], which has been employed in other studies to analyze intervention content (e.g. [68]). Methodological strengths of this systematic review were the use of the EPHPP tool to assess the methodological quality of the studies (this quality assessment tool has proven content and construct validity) [20]; the use of the PRISMA statement [19] and the calculation of effect sizes (Cohen’s d).

The limitations of this systematic review should also be acknowledged. There is a risk of publication bias, as we used only four databases and included only articles written in English. Another limitation may be the inclusion of studies with a weak methodology. In most cases, the weak methodological ratings were due to missing information. We decided to include these studies anyway, as we did not know whether the components determining quality were indeed not implemented by the researchers or whether they merely failed to report the information. The inclusion of studies with any other design than a randomized controlled trial (RCT) can also be a limitation of this study, since RCTs are considered the gold standard [69]. However, the results of quasi-experimental study designs are valuable because of their external validity, and that study design is considered more appropriate for these types of interventions [70].

Recommendations

This systematic review demonstrates the potential of school-based interventions with direct parental involvement to improve BMI, BMI z, and physical activity and decrease sedentary behavior. We recommend that policymakers and practitioners develop and implement school-based interventions with direct parental involvement, focus on multiple EBRBs simultaneously to take into account the total energy balance, and target different environmental types, in particular the social and physical environments, both within the school and the home. We recommend that sustainability of interventions should be carefully considered as sustainable intervention (twelve months or longer) appear to be more effective compared to studies of shorter duration. This may require a shift in focus and budgeting, implementing less but more extensive intervention activities. To enable the implementation of these interventions, research should focus on effective strategies to engage parents and enhance parental involvement. This may need a shift in focus from primarily focusing on schools and children towards parents, making this the key element and taking into account their needs and perspectives. Further, we recommend an extended exploration of the role of behavioral change techniques alongside the types of environments and socio-cognitive determinants. This may add to the ability to explain intervention effectiveness, however fell out of scope for the current review. In addition, we recommend future studies to study the effectiveness on intermediate outcomes (e.g. socio-cognitive and environmental determinants) in order to explore the pathways of effectiveness of these types of interventions.