Murdoch Children's Research Institute, Parkville, VIC, Australia.
Department of Paediatrics, University of Melbourne, Parkville, VIC, Australia.
Int J Obes (Lond). 2022 Jul;46(7):1375-1383. doi: 10.1038/s41366-022-01130-2. Epub 2022 May 3.
BACKGROUND/OBJECTIVES: Modelling genetic pre-disposition may identify children at risk of obesity. However, most polygenic scores (PGSs) have been derived in adults, and lack validation during childhood. This study compared the utility of existing large-scale adult-derived PGSs to predict common anthropometric traits (body mass index (BMI), waist circumference, and body fat) in children and adults, and examined whether childhood BMI prediction could be improved by combining PGSs and non-genetic factors (maternal and earlier child BMI).
SUBJECTS/METHODS: Participants (n = 1365 children, and n = 2094 adults made up of their parents) were drawn from the Longitudinal Study of Australian Children. Children were weighed and measured every two years from 0-1 to 12-13 years, and adults were measured or self-reported measurements were obtained concurrently (average analysed). Participants were genotyped from blood or oral samples, and PGSs were derived based on published genome-wide association studies. We used linear regression to compare the relative utility of these PGSs to predict their respective traits at different ages.
BMI PGSs explained up to 12% of child BMI z-score variance in 10-13 year olds, compared with up to 15% in adults. PGSs for waist circumference and body fat explained less variance (up to 8%). An interaction between BMI PGSs and puberty (p = 0.001-0.002) suggests the effect of some variants may differ across the life course. Individual BMI measures across childhood predicted 10-60% of the variance in BMI at 12-13 years, and maternal BMI and BMI PGS each added 1-9% above this.
Adult-derived PGSs for BMI, particularly those derived by modelling between-variant interactions, may be useful for predicting BMI during adolescence with similar accuracy to that obtained in adulthood. The level of precision presented here to predict BMI during childhood may be relevant to public health, but is likely to be less useful for individual clinical purposes.
背景/目的：建模遗传易感性可能会识别出肥胖风险的儿童。然而，大多数多基因评分（PGS）都是在成年人中得出的，并且在儿童期缺乏验证。本研究比较了现有的大规模成人衍生 PGS 在预测儿童和成人常见人体测量特征（体重指数（BMI）、腰围和体脂肪）方面的效用，并研究了是否可以通过结合 PGS 和非遗传因素（母亲和儿童早期 BMI）来改善儿童 BMI 的预测。
受试者/方法：参与者（n=1365 名儿童和 n=2094 名由其父母组成的成年人）来自澳大利亚儿童纵向研究。儿童从 0-1 岁到 12-13 岁每两年测量一次体重和身高，成年人同时进行测量或自我报告测量（平均分析）。参与者从血液或口腔样本中进行基因分型，并根据已发表的全基因组关联研究得出 PGS。我们使用线性回归比较这些 PGS 预测不同年龄段各自特征的相对效用。
在 10-13 岁的儿童中，BMI PGS 可解释高达 12%的儿童 BMI z 分数变异，而在成年人中则高达 15%。腰围和体脂肪的 PGS 可解释的变异较少（高达 8%）。BMI PGS 与青春期之间的相互作用（p=0.001-0.002）表明，某些变体的作用可能在整个生命周期中有所不同。儿童期的个体 BMI 测量值可预测 12-13 岁时 BMI 的 10-60%，而母亲 BMI 和 BMI PGS 在此基础上分别增加了 1-9%。
BMI 的成人衍生 PGS，特别是通过建模变体之间相互作用得出的 PGS，可能有助于预测青春期 BMI，其准确性与成年人获得的准确性相当。这里提出的预测儿童 BMI 的精度水平可能与公共卫生相关，但可能对个体临床目的的用处不大。
