The CHAMPS study-DK can be described as a quasi experimental study evaluating a natural experiment  including 10 public schools – 6 intervention and 4 control schools - in the Municipality of Svendborg (explained in detail elsewhere) . The present study includes baseline and two years follow up data of body composition measures of the pupils attending 2nd-4th grade. All children and parents from the 10 participating schools received information about the study through school meetings and written information. Parents signed informed consent forms for joining the project and an additional one for participating in the DXA scans. Participation was at any time voluntary. Permission to conduct The CHAMPS study–DK was granted by the regional scientific ethical committee of Southern Denmark (ID S-20080047).
Collaboration with the municipality
Initially all 19 primary schools in the municipality of Svendborg, Denmark, were invited to participate in the project as sports (intervention) schools. Ten of the 19 schools agreed to be sports schools, but only six schools were willing to finance the additional lessons. The decision of additional research was made after the 9 schools had resigned from being a sports school. The municipality was asked to provide six matched control schools but only four schools agreed to become a control school. The six intervention schools and the four control schools were matched based on school size, urban-suburban/rural area and socio-economic position.
Though it is the capital of the municipality, Svendborg is a small town with surrounding rural districts. The 10 participating schools represent half of the public schools in the municipality. Four schools were urban/suburban (two intervention/two control) and six were rural (four intervention/two control). Of the non-participating nine schools, six were urban/suburban and three were rural schools.
Parents and children were unaware of the initiation of this project until two months before the following school year avoiding parents making an influenced school choice .
The school- based PE program
The school leaders and PE teachers of the intervention schools were invited to design the set-up for an optimal PE intervention. The number of children per PE teacher was on average 20, and girls and boys had PE together. The six intervention schools chose to implement four additional PE lessons per week to their usual PE program (resulting in a minimum of 4.5 hours PE per week divided over at least 3 sessions of at least 60 minutes) and to educate the specialized PE-teachers in specific age-related training principles. The four control schools continued their regular PE curriculum (i.e. 2 PE lessons/week resulting in 1.5 hours/week) .
Participants and measurements
All children attending 2nd to 4th grade in 2008 were invited for a DXA scan. DXA scans, height, weight and pubertal stage were assessed according to a standardized procedure at the same day and location. Only children with complete data at both time points were included in the analysis.
Body Mass Index (BMI)
Weight was measured to the nearest 0.1 kg on an electronic scale, (Tanita BWB-800S, Tanita Corporation, Tokyo, Japan) wearing light clothes. Height was measured to the nearest 0.5 cm using a portable stadiometer, (SECA 214, Seca Corporation, Hanover, MD). Both anthropometrics were conducted barefoot. BMI was calculated as [weight (kg)/height2 (m)].
BMI classifications for normal weight (NW), OW, and OB were defined using age- and sex specific cut-offs as recommended by the International Obesity Taskforce recommendations . Dichotomized categories were made for weight classes NW as one and OW/OB in another category to easier compare with the dichotomous variable of normal fat /adipose as described beneath according to Williams .
Total Body Fat Percentage (TBF%)
Total body fat mass was measured by Dual Energy X ray Absorptiometry (DXA), (GE Lunar Prodigy, GE Medical Systems, Madison, WI), ENCORE software (version 12.3, Prodigy; Lunar Corp, Madison, WI). The procedure took place at Hans Christian Andersen Children’s Hospital, Odense University Hospital, Denmark. The child was instructed to lie still in a supine position wearing underwear, a thin T-shirt, stockings and a blanket for the duration of the x-ray. All scans were performed by two different operators and analyzed by one on them. The DXA machine was reset every day, following standardized procedures. TBF% was calculated for each participant from the equation: [(TBF (g) x 100)/ weight (g)].
Cut-offs to classify children as normal-fat or adipose were defined using the cardiovascular health- and gender-related TBF% standards according to Williams et al . These standards were derived from a cross sectional study on 3320 children and adolescents aged 5 to 18 years. Equations developed specifically for children using the sum of subscapular and triceps skinfolds were used to estimate percentage fat. Body density was estimated from age and the sum of triceps and subscapular skinfolds and was subsequently used to derive total percentage body fat. Their analysis resulted in recommended health related cut-offs for adiposity for boys at ≥ 25% TBF and ≥ 30% TBF for girls .
Puberty was defined by self-assessment. The Tanner pubertal stages self-assessment questionnaire (SAQ) used in this study consists of drawings of the 5 Tanner stages . Boys were presented with pictures and text of Tanner staging for pubic hair development, whereas girls were presented with pictures and text representing breast development and pubic hair . Explanatory text in Danish supported the self-assessment. The children were asked to indicate which stage best referred to their own pubertal stage. The procedure took place in a private space with sufficient time to self assess the pubertal stage.
Summary statistics were calculated (means and SD) for the descriptive part on anthropometrics. Differences in OW/OB and adiposity prevalence were tested using Chi square tests. Fisher exact was used for testing differences in pubertal status between schools. Significance level was set at p ≤ 0.05.
To estimate the effect of school type multivariate multilevel mixed effect regression analysis using hierarchical models were used based on the intention to treat principle. Individual, class and school were considered random effects. Analyses were adjusted for age, gender and puberty (and height when TBF% was the outcome variable). Effect modification by gender, age and baseline OW/OB and adiposity category was explored by adding an interaction term between the moderator and school type (intervention versus control). If the interaction term was significant (p < 0.10), subgroup analyses were performed.
In a sensitivity analyses we compared the effect of the intervention based on the non-imputed sample with a sample with imputed data. We imputed missing information on covariates and outcomes (n = 22 to n = 84) using chained equations ("mi impute chained" in STATA) . All covariates, the respective outcomes, and the cluster variables school and class were included in the imputation approach. Beta coefficients and SEs were based on 20 imputed datasets.