Adriëtte Oostvogels

3 81 pBMI, lipids and offspring’s body composition combining recordings from the Dutch perinatal registration, medication use and self- reported hypertension. 23 Covariables from the child were sex of the child, duration of exclusive breastfeeding (no breastfeeding, <1 month, 1–3 months, >3 months) screen (computer and television) time hours/day at age 5–6 years (hours, continuous) and saturated fat intake (grams/day, continuous) at age 5–6 years. Duration of breastfeeding was available from the infancy questionnaire received when the child was aged 3 months and from the Youth Health Care Registration. This prospectively collected information was combined with retrospective information of the 5-year questionnaire to complete the data (19.9% was from 5-year questionnaire). Number of screen-time hours was obtained from the 5-year questionnaire as well as the child’s saturated fat intake. Which was assessed by the food frequency questionnaire, which is validated in children. 24 As both the number of missings in screen time- hours and satured fat intake were relatively high, 7.8% and 15.0% respectively (in the current study population), these values were imputed in the ABCD biomarker study (n = 4389) by a random imputation procedure using linear regression analysis and all variables used in the current study. Because we considered birth weight and gestational age as intermediates between pBMI, maternal lipid profile and child’s adiposity, no adjustment was made for birth weight. 5,6,11 Statistical analysis Analyses were performed with R version 2.14.1 and SPSS version 16.0 (SPSS Inc, Chicago, Ill, USA). No departure from linearity was observed. Associations were explored using linear (where applicable on continuous) and logistic (where applicable on dichotomous outcomes) regression models. First, the association between maternal pBMI (continuous measure) and lipids was explored (a), adjusting for gestational age at blood sampling. Next, we explored the relation between maternal lipids and the adiposity of the child (b), adjusting for gestational age at blood sampling, sex and age of the child. Then we explored the association between maternal pBMI and the adiposity of the child, adjusting for sex and age of the child (c). Finally, all analyses (a–c) were repeated (model 2) with adjustment for potential confounders. The confounders were based on previous literature, and were defined a priori. The following potential confounders were identified: age mother, parity, ethnicity, height, years of education, alcohol, smoking, and hypertension and from the child, duration of exclusive breastfeeding, screen time hours/day and saturated fat intake. Subsequently, logistic regression analyses were conducted following the same procedure as the linear regression models, to explore the association between pBMI, lipids and the risk of offspring’s overweight.