Coleman 2005.

Study characteristics
Methods	Study design: cluster‐RCT Intervention period: 4 years Follow‐up period (post‐intervention): nil Differences in baseline characteristics: reported Reliable outcomes: reported Protection against contamination: NR Unit of allocation: school Unit of analysis: school All analyses were performed according to ITT principles
Participants	N (controls baseline) = 473 N (interventions baseline) = 423 N (interventions follow‐up) = 744 Setting (and number by study group): 8 schools (N = 4 intervention; N = 4 control) Recruitment: intervention schools chosen randomly from schools that had applied to participate in the programme in 1999. Control schools matched by district and geographic location. All children in 3rd grade invited to participate Geographic region: El Paso, Texas ‐ along US‐Mexico border region Percentage of eligible population enrolled: 94% Mean age: intervention: 8.3 ± 0.5 years (boys), 8.2 ± 0.45 years (girls); control: 8.3 ± 0.5 years (boys), 8.3 ± 0.5 years (girls) Sex: intervention: 47% female; control: 47% female
Interventions	Intervention schools: received money (USD 3500 in 1st year, USD 2500 in 2nd year, USD 1500 for 3rd year and USD 1000 for 4th year) for purchasing equipment and paying substitutes so that PE teachers and food service staff could attend training, and for promotion of CATCH programme at each school. Classroom materials were also subsidised (CATCH PE guidebook, PE activity box for grades 3 through 5, curriculum material for grades 3 through 5 and the EATSMART manual). Control schools: did not receive any of the El Paso CATCH programme materials and did not attend any training for the programme. Received USD 1000 at the start of each school year to encourage participation Also received some data i.e. at start of 4th grade, the 3rd grade summary results were provided to both intervention and control schools Combined effects of dietary interventions and PA interventions vs control
Outcomes	Risk of overweight or overweight Anthropometry (height, weight, waist to hip ratio, BMI) Aerobic fitness PE outcomes (time spent in moderate PA (goal ≥ 50%), time spent in vigorous PA (goal ≥ 20%)) Cafeteria outcomes (fat in school lunches ≥ 30%), sodium in school lunches (goal = 600‐1000 mg)) Process evaluation: reported
Implementation‐related factors	Theoretical basis: NR Resources for intervention implementation (e.g. funding needed or staff hours required): reported Who delivered the intervention: reported PROGRESS categories assessed at baseline: reported (race, gender, SES) PROGRESS categories analysed at outcome: reported (gender) Outcomes relating to harms/unintended effects: NR Intervention included strategies to address diversity or disadvantage: reported Economic evaluation: NR
Notes	Funding: this work was funded by the Patient Care and Outcomes Research Grant program from the American Heart Association, Dallas, Tex (9970182N)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	"Participant schools were chosen randomly from those schools that had completed an application to participate" in CATCH programme. Not clear how this was done. Control schools matched and assigned, probably not using randomly generated sequence. Study authors describe design as quasi‐experimental
Allocation concealment (selection bias)	Unclear risk	Allocation may have been concealed but it is not clear. There was cluster allocation. Control schools were first matched to these schools primarily by district and geographic location, and then 4 were randomly selected to participate.
Blinding (performance bias and detection bias) All outcomes	High risk	Blinding probably not carried out for participants or outcome assessors
Incomplete outcome data (attrition bias) All outcomes	Low risk	ITT analysis conducted
Selective reporting (reporting bias)	High risk	Protocol/trial registration documents were unavailable. Incomplete reporting of outcome data. No anthropometry data at endpoint (study authors state no effect but no data provided)
Other bias	Low risk	No other threats to validity
Other bias‐ timing of recruitment of clusters	Low risk	Clusters were recruited before randomisation.