Participants were Belgian children recruited by random cluster design (all children from 12 schools of the selected commune named Aalter) for the longitudinal ChiBS (Children’s Body composition and Stress) study (2010-2012). The study aimed at examining the relation of stress with lifestyle and body composition . At the start of the ChiBS study (2010), a sample size calculation was performed and is described by Michels et al. . In total, 523 children participated in the baseline measurement of the study. At the final follow-up measurement of the ChiBS study in 2012, a DXA measurement was added as an extra module to the ChiBS study, leading to the dataset used in this paper. No new sample size calculation was done in the framework of this follow-up measurement.
For this final follow-up measurement (February - April 2012), the ChiBS children were telephonically invited to make an appointment for the follow-up measurements concerning body composition (fat, lean and bone mass) and lifestyle. An individual appointment was made in the local sports facilities of the commune of Aalter. In total, 330 children participated in the follow-up measurement (193 children were lost between 2010 and 2012) of which 272 children underwent a DXA measurement, as this was an optional module. Complete dairy consumption data of 264 children were available. Only 210 children had matching data from accelerometry, since a limited number of accelerometers were available during the survey.
The study was conducted according to the guidelines laid down in the Declaration of Helsinki 1964 (revision of Edinburgh 2000) and was approved by the Ethics Committee of the Ghent University Hospital. Parents signed an informed consent.
Bone, lean and fat mass measurements by DXA
Children’s height (cm) was measured on bare feet using a stadiometer (SECA 225). DXA, more specifically the Hologic (Discovery-W apparatus using software version 12.8.0), was used to measure osseous and soft tissue. The device was calibrated daily using a lumbar spine phantom as recommended by the manufacturer. Children having a plaster were excluded (this was the case in one child), as well as children having an internal defibrillator, a pacemaker or osteo-synthetic material (this situation was not encountered). BMC (g) and aBMD (g/cm2) measurements were obtained from whole body as well as whole body minus the head measurements. Only the BMC and aBMD of the whole body minus the head were used in this study as this is – beside the lumbar spine - the most accurate and reproducible skeletal site for DXA assessment in children . Additionally, for each individual a BMC and aBMD z-score was calculated according to gender, age and height based on a Belgian reference population (data not published yet). This reference population consisted of 556 healthy Caucasian children (of which 311 girls) aged 5 to 19 years, recruited in six different geographical areas in Belgium (Keerbergen, Namur, Jette, Alsemberg, Ghent and Aalter) between April 2010 and April 2012. In the different provincial areas of Belgium, several schools were invited to participate; those who replied within the study period and were able to accommodate the equipment, were chosen to be included. The participating schools were not only from different geographical areas, but recruited children from different socio-economic status, as based on the income statistics of their respective city. In six of these cities (Keerbergen, Namur, Jette, Alsemberg and Ghent), schools were selected on their willingness to participate. All children and parents in the participating schools received an information letter with detailed explanation on the study, an informed consent and a detailed questionnaire. Participation was on a voluntary basis. The children in Aalter were recruited in the framework of the ChiBS study.
Besides bone mass, total fat mass (kg) and lean mass (kg) were measured by DXA. Fat mass index (FMI, kg/m2) and fat-free mass index (FFMI, kg/m2) were calculated by dividing respectively the fat and lean mass by the height2.
Daily physical activity was measured by an ActiGraph™ (GT3x (triaxial), GT1M (uniaxial) or actitrainer (uniaxial)) accelerometer which was worn on an elastic belt on the level of the right hip during waking hours for five (week and weekend) consecutive days. As uniaxial accelerometers only register velocity in vertical direction and as it was not possible to provide all children with a triaxial accelerometer, only the vertical axis counts were used for this study. The children were allowed to take it off for showering, bathing, water sports or in sports with a high risk of damaging the monitor. Activity counts were stored at a time interval (i.e. epoch) of 15 s intervals.
Only the data from children with readings of at least eight hours per day, but not exceeding 18 h per day and this for at least three days were used for analysis . When periods of more than 20 min of consecutive zero counts were found, these periods were also classified as non-wearing time [20, 21]. The overall time (minutes) spent in sedentary behaviour, light, moderate and vigorous physical activity were calculated using the cut-off points of Evenson depending on the counts per minute (cpm): sedentary behaviour (0-100 cpm), light (100-2295 cpm), moderate (2296-4011 cpm) and vigorous (≥4012 cpm) [22, 23]. Time spent in moderate-to-vigorous physical activity (MVPA) was calculated as the sum of both the time in moderate and vigorous activity. Next, the amount of minutes spent in sedentary behaviour as well as the amount of minutes spent in each category of physical activity was expressed as a percentage by dividing the amount of minutes per category by the total recording time in order to standardise the data for total wearing time (which was different between the children). The software Meter plus 4.2 was used to screen, score and clean the accelerometer data files.
A parentally reported, semi-quantitative food frequency questionnaire (FFQ) was used for dietary analysis. This FFQ was based on the validated FFQ of Huybrechts et al. . The central question was ‘How often did your child eat or drink the following food item during the last month and which quantity of the food items is on average consumed on a regular day?’. To report the consumption frequency, 6 possible answers could be chosen ranging from ‘never’ to ‘daily’ (never, less than once a week, one day a week, 2 to 4 days a week, 5 to 6 days a week, daily). To report the consumed quantity, 5 or 6 possible answers could be chosen, expressed in g or ml per day, depending on the food item. Examples of portion sizes were given to facilitate answering the questions. In total, 67 food items were considered in the FFQ, grouped in 16 food groups. The last food group focused on food supplements. The consumption of milk, soy milk, quark, cheese and yoghurt was summed up and is refered to as total dairy consumption (g/week).
Stage of pubertal development was assessed by a paediatrician using the Tanner score of pubic hair distribution and genital development for boys and pubic hair distribution and breast development for girls . In the absence of the paediatrician on the examination day, a parentally reported questionnaire with images was used to determine Tanner stage (9 % of the study population). For the analysis, the different Tanner stages were recorded in a dichotomous variable (no signs of puberty, signs of puberty) because of the low percentage of children who had started pubertal development.
To describe socio-economic status, parental education level (PEL) according to the International Standard Classification of Education classification (level 0 ‘pre-primary education’ , 1 ‘primary education’ , 2 ‘lower secondary education’ , 3 ‘upper secondary education’ , 4 ‘post-secondary non-tertiary education’ , 5 ‘first stage of tertiary education’ , 6 ‘second stage of tertiary education’) was used. The highest classification among the two parents was used in the analysis. Because none of the parents were classified in group 1, and only a few in group 2, the PEL levels were recoded into three groups: ‘equal or lower than level 3’ , ‘level 4’ and ‘level 5 or 6'.
IBM SPSS Statistics 22.0 for Windows (SPSS Inc, Chicago, IL) was used to perform the statistical analyses. The level of statistical significance was set at 5 %, also for the interaction terms. Independent samples t-test or Mann-Whitney U test (for continuous variables) and chi square test (for categorical variables) were used to explore possible differences in bone parameters, body composition, accelerometer data and dairy consumption between boys and girls. BMC or aBMD differences between PEL groups were tested using ANOVA.
Multiple linear regression analyses were used to examine the associations of accelerometer data and dairy consumption with aBMD and BMC as outcome, using both the absolute values of the bone parameters as well as the z-scores. All residuals showed a satisfactory pattern. When absolute values of aBMD and BMC were included in the regression analyses as dependent variable, we adjusted for age, gender, Tanner stage, height and body composition (FMI and FFMI). When z-scores of aBMD and BMC (standardised for age, gender and height) were included in the regression analyses as dependent variable, we adjusted only for Tanner stage and body composition (FMI and FFMI). In these regression models, FMI and FFMI were included after ln-transformation due to their skewed distribution. Age was included in the models as a dichotomous variable (below or above the median age) as age was too highly correlated with height. In each multiple regression model, the collinearity between the different independent variables was checked using the variance inflation factor.
In a first set of regression analyses, we investigated the associations of sedentary behaviour, the time spent in different categories of physical activity (light, moderate, vigorous and moderate to vigorous) and of dairy consumption on bone parameters separately. Standardised regression coefficients, p-values and eta2 for the independent variables are reported for each model. The eta2 describes the proportion out of the total variation in the outcome that can be attributed to a specific predictor. Eta2 levels around 0.01 were considered to be small, around 0.10 as medium and around 0.25 as large . In these models, the interaction effect of gender in the relation between sedentary behaviour, physical activity as well as dairy consumption and BMC/aBMD was tested by including an interaction factor. To calculate this interaction factor, gender was coded ‘+1’ for girls and ‘-1’ for boys and multiplied with the dairy consumption, sedentary behaviour or one of the different physical activity parameters after standardisation. If interaction was found in a model, this model was re-run after stratification for gender.
In a second set of regression analyses, we included both a parameter of sedentary behaviour or physical activity as well as the dairy consumption and looked at the associations with BMC and aBMD. First, these models were run including interaction factors to test the interaction effect between dairy consumption and sedentary behaviour as well as physical activity on BMC and aBMD. To calculate these interaction factors, the predictors were first standardised and then multiplied with each other. When an interaction factor showed a significant association with the dependent variable, plots of interactions were created where each factor variable was split into two groups (below and above the median value) to reflect low and high levels of the factors and to ensure an even distribution of the data. The interaction plots were created using the statistical software R3.1.0 (http://www.R-project.org).