Genotypes
A valid MR13 study is predicated on three assumptions: (i) each genetic variant is associated with a risk factor; (ii) it affect the outcome only through their effect on the risk factor; and (iii) each genetic variant is independent of potential confounders. We used the following criteria to identify the candidate single-nucleotide polymorphism (SNP) for obesity and the mediators: (i) minor allele frequency of ≥5%, (ii) genotyping call rate of >98% for all children, (iii) no linkage disequilibrium with candidate SNPs, and (iv) primers designed using the National Genotyping Center of the Academia Sinica platform (http://lims.ngc.sinica.edu.tw/service/).27Genotyping was performed using Sequenom iPLEX matrix-assisted laser desorption/ionization–time of flight mass spectrometry at the National Center for Genome Medicine, Taiwan.
Using genomic information from large genome-wide association studies, we chose well-known genetic variants associated with BMI or the mediators. To address the concern of potential pleiotropic effects of genetic variants, SNPs previously reported to be associated with asthma and confounders were excluded. The online supplementary file demonstrates the details of how candidate SNPs were selected. In summary, we selected 28, 16, 4, 13, 14, and 11 SNPs for BMI, FEV1/FVC, FeNO, physical fitness, SDB, and early puberty, respectively. Weighted genetic risk scores (GRSs) composed of SNPs of BMI and the mediators were used as genetic IVs. The dosage of effect allele was multiplied by the regression coefficients of each gene on associated BMI/mediators divided by the mean value of all regression coefficients to generate GRSs.28