Teun Remmers

Relationship between PA and the development of BMI in children | 21 Only measurements that contained at least two week days and one weekend day were considered to be valid and were used in the analyses (37). Each day should consist of at least 400 minutes of registration time. Periods of 60 consecutive minutes of no accelerometry counts were considered to be non-wearing time (37). We used the thresholds of Evenson (12) to distinguish between sedentary behavior (0-25 counts per epoch), and PA of light (26-573 counts per epoch), moderate (574-1002 counts per epoch), and vigorous intensity (≥ 1003 counts per epoch). The validation study by Evenson was performed in 5-8 year-old children, which is comparable to our study sample. Time spent in the moderate and vigorous categories were summed into moderate to vigorous PA (MVPA), as internationally recommended (24). As accelerometers tend to underestimate cycling and cannot measure swimming/water activities, parents were asked to directly report the minutes that their child spent on these activities in a diary. The daily average number of minutes of cycling and swimming were calculated from weekly totals. As these activities were non-normally distributed they were categorized into three categories (cycling: ≤ 10, 11-30, and ≥ 31 minutes per day; swimming: 0, 1-20, and ≥21 minutes per day). Body mass index Parental height and weight were assessed in an annual questionnaire at the child's age of 4-5 years old. The child's body height and weight were measured by trained research assistants with a portable stadiometer (Leicester height measure) and digital scale (CAS personal scale, HE-5) and recorded as millimeters and grams (rounded off to 100g), respectively. This was done at T 0 in all children, and at T 1 and T 2 in those children who wore the accelerometer at that time. At other times parents were asked to measure height and weight as part of annual questionnaires. The method of measurement of height and weight at follow-up (i.e. objectively measured or parent reported) was controlled for in multivariate analyses. Children’s BMI z-scores were calculated using reference values from the Fourth Dutch Growth Study to standardize for age and gender (13). In both periods, we combined objectively and parent-reported height and weight measurement methods. In the first period, all height and weight measurements were obtained by research assistants at baseline (i.e. T 0 ), and 24.6% measurements were obtained by research assistants at follow up (i.e. T 1 ). In the second period, 49.3% of measurements were obtained by research assistants at baseline (i.e. T 1 ), and 24.8% of measurements were obtained by research assistants at follow up (i.e. T 2 ). Data analyses Data from both time periods were combined using SPSS 19 for Windows. The distributions of the main continuous variables (BMI z-scores, percentage time spent in different PA intensity categories) were approximately normal, so no transformations were needed. The relationship between PA and the development of BMI z-scores was investigated, controlling for time period, recruitment group, gender, method of height and weight measurement, cycling, swimming, seasonality, maternal BMI and paternal BMI. In the first