Van de Gaar 2014.

Methods	Design: Cluster‐RCT (included as a NRCT in our GRADE assessment due to the small number of randomised units) Timing: Prospective Allocation to group: Quote: "A total of four schools, located in four different non adjacent neighbourhoods in Rotterdam, the Netherlands, were matched in pairs of two. (...) One of the schools in the school pair was then allocated to either the intervention or control condition by the flip of a coin" Number of clusters or sites: 4 schools Number of individuals: 1009 Length of intervention: 13 months
Participants	General description of participants: Children attending elementary school in multi‐ethnic, socially‐deprived neighbourhoods in Rotterdam, the Netherlands Age: Children (age range 6 ‐ 12 years) Inclusion criteria: School‐level (Quote: "The matched schools had a roughly equal number of pupils, had pupils of similar socio‐economic status, and had a similar prevalence of overweight. On the basis of these criteria, we were able to select only six school pairs from the 80 schools that were eligible for the study. These pairs were then approached based on convenience, in the knowledge that a school pair could only be included in the study if both schools in the pair provided consent and a maximum of two school pairs in total could be included in the study." Student‐level (quote): "At the four participating schools, all children in grades 2 to 7 (aged 6‐12 years) were invited to participate, as were their parents") Exclusion criteria: N/R Recruitment: Schools were "approached on convenience" (quote); how students were approaches is not reported Weight status at baseline: The combined prevalence of overweight and obesity at baseline was 25.1% in the CG and 21.6% in the IG (based on the observation report, see table 2 of the study's primary report) SSB consumption at baseline: Baseline SSB consumption based on a survey among the parents of participating children was 760 ml/day in the IG and 850 ml/day in the CG. Based on a survey among the participants themselves it was 1310 ml/day in the IG and 1500 ml/day in the CG Equity considerations: The intervention was performed in neighbourhoods described as socially deprived and ethnically diverse. Approximately 25% of participants had a Dutch ethnic background, 30% Surinamese/Antillean, 27% Moroccan/Turkish, and 20% Other/Missing. Quotes: "The water campaign is an intervention tailored to children (...) and their families who live in multi‐ethnic, socially more deprived neighbourhoods; populations who remain disproportionately affected by childhood overweight. (...) [T]he local government intervention‐development team decided to focus the water campaign on Turkish and Moroccan families. These families form a large group of non‐Western immigrants in the study area, a group disproportionately affected by childhood overweight. (...) [W]hen we explored whether such tailoring of the water campaign specifically to these minorities improved the effects seen among these children, we were unable to detect significant differences in intervention effect between children of Turkish and Moroccan background and children from other ethnic backgrounds (p > 0.05 in all three data sets; data not shown)"
Interventions	Intervention: Improved access to drinking water at school. (Intervention components included "[p]rovision of free water at school throughout the day; Taking a water break during physical education lessons; Provision of free water bottles by community organizations during summer activities") Behavioural co‐intervention: Educational and promotional activities based on social marketing techniques. (Behavioural intervention components included the following: "Special event: water campaign kick‐off ‘Drinking water is fun!’; Use of promotional material: posters ‘Water is the best thing I can give to my child!’ Activity for children: Pimp up your water bottle; Activity for parents: Pimp up your water jug; Water theme week, including [the following] activities for children: special educational water lessons, fun games such as happy families, board and card games involving water consumption, and a special water show provided by children’s role models. [Activities for] parents: storytelling about promoting water consumption, different fun games involving water consumption and other aspects of water, including a water magazine for mothers; and promotion by water ambassadors") Control: No intervention. (Quote: "The intervention and control schools continued with their regular health promotion programme, the effective school‐based curriculum ‘Enjoy Being Fit’")
Outcomes	Measures of SSB intake: Average SSB consumption in l/day and servings/day, and percentage of participants consuming SSB every day, assessed through a questionnaire completed by participants and their parents at baseline and 13 months; and percentage of participants bringing SSB to school, assessed by trained observers on 1 randomly‐chosen day at baseline and 15 months Measures of intake of alternatives to SSB: Average water consumption in l/day and servings/day, and percentage of participants consuming water every day, assessed through a questionnaire completed by participants and their parents at baseline and 13 months Anthropometric measures: Mean BMI, and combined prevalence of overweight and obesity, assessed by trained personnel with a standardised protocol at baseline and 15 months Adverse outcomes: The study does not report how data on adverse outcomes were collected, and if adverse outcomes were observed or not, but notes an increase in the BMI in the IG compared to the CG (see above) Other outcomes: None included in this review
Context and implementation	Setting: Schools Sector: Education Country: The Netherlands Year(s) when implemented: 2011 ‐ 2012 Mode of implementation: Pilot trial by researchers in co‐operation with local government Level of implementation: Setting‐based intervention
Declarations	COI: "All authors (VM van de Gaar, W Jansen, A van Grieken, GJJM Borsboom, S Kremers and H Raat) declare that they have no competing interests" Funding: "This study is funded by a grant from the major funding body ZonMw, the Netherlands Organization for Health Research and Development, project no. 50‐50102‐96‐015. This study is part of the Dutch project CIAO, which stands for Consortium Integrated Approach Overweight. Within CIAO, several studies are being conducted that investigate the different components of the EPODE approach. This study reflects on the component ‘social marketing’. The publication of this study was supported by a grant of the Netherlands Organization for Scientific Research (NWO)" Trial registration: Registered at the Nederlands Trial Register at trialregister.nl/trialreg/admin/rctview.asp?TC=3400 Protocol availability: Protocol mentioned in the report but not publicly available
Notes	Given the small number of randomised units, randomisation may not have been sufficient to attain baseline comparability of the IG and CG. We therefore included Van de Gaar 2014 as a NRCT in our GRADE assessment
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "One of the schools in the school pair was then allocated to either the intervention or control condition by the flip of a coin"
Allocation concealment (selection bias)	Low risk	Allocation was by institution (school), and allocation was done at the start of the study
Similarity of baseline outcome measurements (selection bias)	Unclear risk	Important baseline outcome measurement imbalances were present. The combined prevalence of overweight and obesity based on the observation report was 25.1% in the CG and 21.6% in the IG, and the share of students bringing SSB to school was 69% in the IG and 80% in the CG. Based on the parent report, baseline SSB consumption was 769 ml/day in the IG and 850 ml/day in the CG. While baseline differences were taken into account in the analysis, adjustment may not have been complete. Quote: "In all analyses, outcome measures were adjusted for baseline SSB values, several socio‐demographic characteristics (grade, gender and ethnic background of the child and educational level of the caregiver) and child’s weight status at baseline. This was done by also entering them as independent variables"
Similarity of other baseline characteristics (selection and performance bias)	Unclear risk	Potentially relevant baseline differences between the IG and CG were observed. The share of participants with Dutch ethnic background based on the parent report was 29% in the CG and 41% in the IG, and the share of parents with higher education was 50% in the CG and 55% in the IG. While baseline differences were taken into account in the analysis, adjustment may not have been complete. These differences, and the large differences in baseline outcome measurements, may hint at relevant differences in unobserved baseline characteristics of the IG and CG
Incomplete outcome data (attrition bias) All outcomes	High risk	The analysis was per‐protocol. Participants lost to follow‐up differed significantly from those who stayed in the study, and it is unclear to what extent there was different attrition in the IG and CG. Quotes: "Only complete case analyses were performed, meaning we analysed data only from children whose data from both time points was complete. (...) The non‐response of parents to the parent questionnaire (complete case analyses only possible for 35%) is another limitation of this study. (...) We were able to conduct non‐response analyses for the variables gender, grade and ethnic background of the child. Parents of children who participated in the study were more often parents of children in the lower grades (p < 0.001) and of children with a Dutch ethnic background (p < 0.001) as compared to parents lost to follow‐up. Children who completed a questionnaire were more often children in the lower grades (p < 0.001) and children with a Dutch ethnic background (p = 0.007) as compared to children lost to follow‐up. Children who were observed were more often children with a non‐Dutch ethnic background (p < 0.001) as compared to children lost to follow‐up. Non‐response analyses were also conducted for the condition variable. Here we saw a difference between the intervention and control conditions in parents of children who participated in the study compared to parents lost to follow‐up (p = 0.006) and for children who underwent observation compared to children lost to follow‐up (p = 0.014)"
Blinding (performance and detection bias) Subjective outcomes	High risk	Participants and data collectors were not blinded, and the intensive behavioural co‐intervention may have introduced a relevant degree of social desirability bias. Quotes: "Blinding of participants and data collectors was not possible since the water campaign’s activities were visible at the intervention schools and throughout the neighbourhoods. (...) [R]esearch has shown that parents are more prone to reporting socially desirable answers compared to children. This could also partly explain the fact that SSB consumption reported by children was higher than that reported by parents"
Blinding (performance and detection bias) Objective outcomes	Low risk	Participants were not blinded, but body weight outcomes are objective, and were assessed by trained personnel with a standardised protocol
Contamination (performance bias)	Unclear risk	Contamination (in particular through the community component) is possible, and would have biased results towards null
Selective reporting (reporting bias)	High risk	The trial registry entry for this study (www.trialregister.nl/trialreg/admin/rctview.asp?TC=3400) mentions 3 potentially important outcomes which are not reported in the study: waist circumference, snack consumption and intake of fruit. Moreover, results on the prespecified primary outcome (BMI) and the first of the 4 prespecified secondary outcomes (water consumption) are not reported in the Results section of the paper, but only in the annexe, with a short reference in the Discussion section of the paper. For the first of the 4 secondary outcomes prespecified in the trial registry entry the study's primary report states explicitly that it was not an outcome Quote: "Although the intervention was aimed at reducing the intake of children’s SSB consumption by promoting the intake of water, water consumption was not an outcome measure of our study")
Other bias	Unclear risk	The unit of analysis was not the unit of control, and it is unclear to what extent this was taken into account in the analysis. Quote: "A further limitation is the small number of clusters (i.e. four), which inhibited multi‐level analyses but was countered by adding the ‘school pair’ variable in the analyses." For the beverage intake data it is not reported if the questionnaires which were used were validated (for the anthropometric measures standardised measurement protocols were used)