Adriëtte Oostvogels

92 Chapter 3 Strengths and limitations Strengths of this study are the large study population, the number of available covariables and the long follow-up period. Also, several childhood adiposity measures were available, i.e. besides BMI we also assessed the percentage of body fat and WHtR. This study also has some limitations. First, blood collection for maternal plasma lipids was accomplished randomly i.e. on only one occasion and in a non-fasting state. The non-fasting state might have slightly diluted our results. However, in a cohort of initially healthy women, Bansal et al. found that non-fasting TG levels were associated with the prevalence of cardiovascular events; in that study fasting TG levels showed little independent relationship. 41 Also, another prospective cohort study found that elevated non-fasting TG levels were associated with increased risk of myocardial infarction, ischemic heart disease, and death. 42 As most of the day is spent in a non-fasting state, these non-fasting lipid levels may even more accurately estimate health risks than fasting lipid levels. Second, selection bias may have occurred as this is a cohort study and selective loss to follow-up may have taken place. The women that participated in the ABCD biomarker study, but who were lost to follow-up during the health check measurement of their children at age 5 years(n = 1378), were significantly younger, less often of Dutch origin, had fewer years of education, and were less likely to consume alcohol (see Table S1). Both groups were comparable for maternal pBMI, standardized birth weight and gestational age. Due to this selective loss to follow-up and our exclusion criteria, the present results apply to a relatively healthy sample. Therefore, associations found in this study might be an underestimation of the effects of the whole community. Third, although we adjusted for several factors we cannot exclude some residual confounding. Maternal pBMI was based on self-reported questionnaires. Self- reported height tends to be slighted overestimated and weight underestimated, resulting in a (possible) underestimation of BMI. Nevertheless, if present this underestimation will be apparent across almost the whole study population. 43,44 We deliberately excluded birth weight as (according to our hypothesis) it lies within the causal pathway; maternal lipid status may influence offspring somatic growth 13 and thereby influence birth weight and finally influence childhood adiposity. Our explorative analyses showed that the results remained the same after the inclusion of birth weight (data not shown). Furthermore, gestational weight gain, specifically in combination with maternal pBMI, is associated with childhood adiposity. 45 Unfortunately information on gestational weight gain was not available, but we did