Adriëtte Oostvogels

202 Chapter 8 SES SES was based on maternal education in years after primary school, reported in the pregnancy questionnaire and divided in three categories: low (0-5 years), middle (6- 10 years) and high (>10 years). Characteristics of the study population The following variables were used for demographic information: maternal age (years) and maternal smoking during pregnancy (yes/no) which were self-reported in the pregnancy questionnaire, sex (boy/girl), gestational age (days), parity (nulliparous/ multiparous), and birth weight of the child (gram) were retrieved from the Youth Health Care registration. Analyses Characteristics of the study sample were described according to weight status at age 5-6 years for boys and girls separately. Differences between the normal weight and overweight groups and the response and non-response group (live-born singletons with permission to follow-up meeting the inclusion criteria, but without growth data and/or BMI status at 5 years of age) were tested with chi-square tests for categorical variables and two-sample t-tests for continuous variables. To model the overall longitudinal BMI growth patterns of normal weight and overweight boys and girls, a skew-normal linear mixed model was fitted to the data. 34 We assumed that the measurement error of BMI was normally distributed. The random effects followed a multivariate skew-normal distribution to allow the child-specific growth offset to be non-symmetrically distributed around the average BMI growth pattern. To capture the non-linearity of the BMI growth pattern, both the fixed and random effects describing the relation between BMI and age were modeled with natural splines with five degrees of freedom. First, differences in BMI growth patterns between normal weight and overweight children were described. Second, the population was restricted to only children with overweight (N=487). The growth curves were compared for children who had overweight at age 5-6 years of European origin and of other origin. Third, to adjust for the interrelation of SES and ethnicity, the population was further restricted to children with overweight from only European origin (N=206). Growth curves were compared of overweight children from high SES, middle SES and low SES groups. We did not correct for any covariates as potentially important variables as birth weight