Polygenic Scores of Body Mass Index, Mediation Through Eating Behaviours, and Their Interaction with Dietary Intakes in Childhood.

Abstract

Previous studies found that the effect of polygenic scores (PGS) on body mass index (BMI) is at least partly mediated through eating behaviours, and that PGS can mitigate or elevate the association of dietary intakes on BMI. However, this evidence has several limitations. For instance, there are few investigations in child populations or using longitudinal data. Also, the majority of PGS for BMI in studies of children are constructed from adult genome-wide association studies (GWAS) summary statistics, even though the genetic architecture of obesity may change throughout life. To address those limitations, the thesis has three specific objectives: 1) to measure the genetic susceptibility to obesity in children using two PGS derived from adults and children GWAS summary statistics, respectively, and compare their association with BMI and discrimination of obesity, 2) to determine the extent to which the association between adult- and child-derived PGS and BMI is mediated through eating behavioural traits (over-eating, fussy eating) assessed in early childhood, and 3) to estimate the association between preschool dietary (food and macronutrient) intakes and BMI (mean and change with age) displayed in childhood and identify the ability of adult- and child-derived PGS to modify those associations. The objectives are developed in three manuscripts. In the first manuscript, we showed that a PGS derived from adult summary statistics has a stronger association with BMI z-scores from 4 to 13 years, compared to a PGS derived from child summary statistics. Using longitudinal growth curve mediation analysis in the second manuscript, we observed that over-eating expressed from 2 to 6 years of age mediated the association between both PGS and BMI z-scores from 6 to 13 years of age. Notably, the proportion mediated by over-eating decreased as children grew older. Leveraging linear mixed effects models in the third manuscript, we showed that the association between specific dietary intakes (protein, lipid, and total energy) and BMI increased with the value of the adult-derived PGS, but not the child-derived PGS. Overall, this thesis sheds light on the underlying role of dietary habits in genetic susceptibility to obesity across childhood and teenage years.