Main meta-analysis results of the <i>INSIG2</i> rs7566605 association with obesity.

Pooled association for all Caucasian adult studies combined (All-CA) as well as stratified by study type (GP = general population, HP = healthy population, OB = obesity study), for all non-Caucasian studies (All-NC), or the children studies (All-CH) indicating differential results among GP or HP studies (Hypothesis 1). Numbers stated are the number of obese (BMI≥30 kg/m2, ‘cases’) and non-obese (BMI<30 kg/m2, ‘controls’) subjects with the number of studies, the recessive model ORs (p-value) comparing the odds for obesity among the subjects with the CC genotype versus subjects with the CG or GG genotype using fixed or random effects models, the I2 measure (p-value of the Q-statistics), and the p-value testing for pair-wise difference between GP, HP, and OB studies. aExcluding studies published before the response letter in Science by Herbert et al., December 2006 [1], in which the hypothesis of potential heterogeneity due to study design and a first call for this meta-analysis were stated (i.e., excluding American_Polish, NHS, KORA_S4, Essen_trios, EPIC_Norfolk, MRC_Ely, DESIR, SHIP, OB_adult). bP-values corrected for pair-wise comparison of three subgroups need to be multiplied by three. OR = Odds Ratios.