Herbert et al. (1) tested 86,604 single-nucleotide polymorphisms (SNPs) for an association with obesity in 694 individuals of the National Heart, Lung, and Blood Institute–Framingham Heart Study. They found that minor-allele homozygotes (CC) of the rs7566605 variant, located 10 kb upstream of INSIG2, had an increased risk of obesity (P = 0.0026). This finding was replicated in four separate samples of different ethnicity. However, we found no evidence of association between this variant and obesity risk in two large ethnically homogeneous population-based cohorts (n = 6599) aged 35 to 79 years.

We genotyped the rs7566605 variant in 4916 individuals (BMI 26.0 kg/m² ± 3.7, 12% obese) of the European Prospective Investigation of Cancer (EPIC) Norfolk study (EPIC5000) (2) and in 1683 individuals (BMI 27.2 kg/m² ± 4.8, 22% obese) of the Medical Research Council (MRC) Ely study (3). All individuals were of Caucasian origin and unrelated. The association between BMI and the rs7566605 SNP genotype was tested using generalized linear models adjusted for age and sex. We first tested for association under an additive model, which assumes an additive effect for each additional minor allele. In post hoc analyses, we applied (i) the recessive model, which compares the CC homozygotes with G-allele carriers and (ii) the dominant model, which compares GG homozygotes with C-allele carriers. Analyses were performed for each cohort separately, as well as for the combined cohort (P value for heterogeneity between cohorts = 0.42). In the combined analyses, we adjusted for cohort in addition to BMI and sex and tested for genotype-by-cohort interaction. Sex-specific effects were tested by including a genotype-by-sex interaction term in the models.

The minor-allele frequency (C allele) is 32% in both cohorts. Genotype frequencies are in Hardy-Weinberg equilibrium (P > 0.25) and similar to those reported for Caucasians in Herbert et al. (1). In contrast to Herbert et al., however, we found a tendency (P = 0.06) for the C allele of the rs7566605 variant to be associated with a lower BMI in the EPIC5000 cohort (Table 1). The association was significant (P = 0.02) under the dominant model [GG: 26.1 kg/m² ± 0.08 (mean ± SEM); C-allele carriers: 25.8 kg/m² ± 0.07]. No significant association (P = 0.61) was observed in the MRC Ely cohort (Table 2) or when both cohorts were combined (GG: 26.7 kg/m² ± 0.08; CG: 26.5 kg/m² ± 0.08; CC: 26.5 kg/m² ± 0.15; P = 0.09). Also under a recessive model, as reported by Herbert et al. (1), no significant associations were observed (EPIC5000 P = 0.87, MRC Ely P = 0.48, combined P = 0.57). Furthermore, there was no evidence of sex-specific effects (genotype-by-sex interaction: EPIC5000 P = 0.30, MRC Ely P = 0.28, combined P = 0.14) or cohort-specific effects (additive model P = 0.69). Each cohort had sufficient power (>90%) to detect the genetic effects reported by Herbert et al. (1) under a recessive model and at a significance of 1%.

Despite replicated findings reported by Herbert et al. (1), we found no evidence of an increased obesity risk in minor-allele homozygotes in two large cohorts of Caucasian origin. On the contrary, we found a tendency toward a lower BMI in carriers of minor alleles, although the effect was small. We conclude that the rs7566605 variant is most likely not a major player in the genetic causes of obesity, at least not in Caucasian populations.