The study included 3,503 children diagnosed by the Chinese age- and sex-specific BMI cutoffs (supplementary Table 1) (13
), of whom 1,229 were classified as obese and 655 as overweight. The mean age of subjects was 12.4 ± 3.1 years and 50.8% were boys (see supplementary Table 2 for additional characteristics). We genotyped six SNPs (rs7138803 near FAIM2
, rs1805081 in exon 6 of NPC1
, rs6499640 in intron 1 of FTO
, rs17782313 near MC4R
, rs6265 near BDNF
, and rs10938397 near GNPDA2
) in the cohort. The genotypes of the SNPs were in Hardy-Weinberg equilibrium in each group (P
> 0.05) (supplementary Table 3).
After adjusting for multiple testing, statistically significant associations of the SNPs (rs7138803, rs17782313, rs6265, and rs10938397) with BMI (adjusted for age and sex, P = 0.0097, 1.0 × 10−5, 0.038, and 0.00093, respectively) were found under the additive model (). The data of three models of inheritance (dominant, recessive, and additive) are provided in supplementary Table 4.
BMI according to genotypes and per effect allele change in BMI after age and sex adjustment
We also tested the associations between the six SNPs and anthropometric indexes, including waist circumference, WHtR, FMP, and birth weight (supplementary Table 5). Our results showed that the SNPs rs17782313 and rs10938397 were significantly associated with waist circumference, WHtR, and FMP, and the SNP rs7138803 was significantly associated with waist circumference and WHtR, but none of the six SNPs were associated with birth weight.
All six SNPs, including rs1805081 (borderline significance, P
= 0.048), were significantly associated with obesity in the Chinese children after age and sex adjustment (). After adjusting for multiple testing, the significant associations of the SNPs rs17782313, rs6265, and rs10938397 with obesity remained (P
= 5.0 × 10−6
, 0.043, and 0.00085, respectively). The power for rs1805081 was 0.994, assuming the effect size of 1.33 for the effect allele (9
) with the frequency of 0.744. In the analysis of overweight status, the only statistically significant association observed was with rs17782313 after age and sex adjustment and correction for multiple testing (). The data of three models (dominant, recessive, and additive) with power calculations are provided in supplementary Table 6.
Associations of SNPs with overweight and obesity in multinomial logistic regression with age and sex adjustment
For the 126 (six SNPs, three genetic models, and seven phenotypes) resulting P values, associated false discovery rate levels were performed (0.05 as criteria). After multiple testing, the significant associations of the SNPs rs17782313, rs6265, and rs10938397 with BMI and obesity remained (data not shown).
shows the frequency distribution of the number of effect alleles of the six SNPs in normal-weight, overweight, and obese groups. The associations of the number of effect alleles with overweight and obesity are shown in supplementary Table 7. Most subjects of each group carry four to six effect alleles. In the obese group, compared with subjects that carry three or fewer effect alleles, the risk of obesity of the subjects that carry four to six and greater than seven effect alleles was 1.54 (one-sided 95% CI >1.31; one-sided P = 5.9 × 10−6) and 2.50 (one-sided 95% CI >1.97; one-sided P = 1.3 × 10−10), respectively. There were significant associations of the number of effect alleles with anthropometric indexes (except birth weight) including BMI, waist circumference, WHtR, and FMP (supplementary Table 8).
Frequency distribution of the number of effect alleles in normal-weight, overweight, and obese groups.