Impact of FTO genotypes on BMI and weight in polycystic ovary syndrome: a systematic review and meta-analysis

Abstract

Aims/hypothesis: FTO gene single nucleotide polymorphisms (SNPs) have been shown to be associated with obesity-related traits and type 2 diabetes. Several small studies have suggested a greater than expected effect of the FTO rs9939609 SNP on weight in polycystic ovary syndrome (PCOS). We therefore aimed to examine the impact of FTO genotype on BMI and weight in PCOS.

Methods: A systematic search of medical databases (PubMed, EMBASE and Cochrane CENTRAL) was conducted up to the end of April 2011. Seven studies describing eight distinct PCOS cohorts were retrieved; seven were genotyped for SNP rs9939609 and one for SNP rs1421085. The per allele effect on BMI and body weight increase was calculated and subjected to meta-analysis.

Results: A total of 2,548 women with PCOS were included in the study; 762 were TT homozygotes, 1,253 had an AT/CT genotype, and 533 were AA/CC homozygotes. Each additional copy of the effect allele (A/C) increased the BMI by a mean of 0.19 z score units (95% CI 0.13, 0.24; p = 2.26 × 10(-11)) and body weight by a mean of 0.20 z score units (95% CI 0.14, 0.26; p = 1.02 × 10(-10)). This translated into an approximately 3.3 kg/m(2) increase in BMI and an approximately 9.6 kg gain in body weight between TT and AA/CC homozygotes. The association between FTO genotypes and BMI was stronger in the cohorts with PCOS than in the general female populations from large genome-wide association studies. Deviation from an additive genetic model was observed in heavier populations.

Conclusions/interpretation: The effect of FTO SNPs on obesity-related traits in PCOS seems to be more than two times greater than the effect found in large population-based studies. This suggests an interaction between FTO and the metabolic context or polygenic background of PCOS.

Fig. 1

PRISMA diagram of the study selection process

Fig. 2

Inverse variance (IV) meta-analysis of per allele (A/C) effect increase in log-transformed BMI, expressed in z score units. Forest plot and corresponding values represent point estimates for per allele effect together with 95% CI. Pooled weighted mean difference effect size with 95% CI is also shown. Each cohort is represented by the name of the first author. Test for heterogeneity: Q = 2.57, df = 7, p = 0.9217, I ² = 0.00%. Test overall effect: Z = 6.69, p < 0.0001

Fig. 3

Inverse variance (IV) meta-analysis of per allele (A) effect increase in log-transformed body weight, expressed in z score units. Forest plot and corresponding values represent point estimates for per allele effect together with 95% CI. Pooled weighted effect size with 95% CI is also shown. Each cohort is represented by the name of the first author. Test for heterogeneity: Q = 1.78, df = 6, p = 0.9388, I ² = 0.00%. Test overall effect: z = 6.46, p < 0.0001

Fig. 4

Comparison of the effect of the FTO polymorphism on BMI between women with PCOS and the general female population from the GIANT Consortium (a) and the study of Frayling et al [3] (b). White columns represent the per allele effect increase in the PCOS population, whereas grey columns represent the corresponding effect calculated for the general population. Data are expressed as either rank-based inverse normally transformed BMI (a) or log-transformed BMI z scores (b). Error bars represent SE. Comparison between cohorts was performed with a two-sided t test. ^† p = 0.0002 for comparison between women with PCOS and the general female population from the GIANT Consortium. *p = 0.0146 for comparison between women with PCOS and the general female population described by Frayling et al [3]

Fig. 5

Mixed-effect meta-regression analysis correlating the mean difference in (a) log-transformed BMI z score units between AT/CT and TT (dependent variable) and mean BMI in the TT group as independent variable or (b) log-transformed body weight z score units between AT and TT (dependent variable) and mean body weight in the TT group (independent variable). The solid lines indicate the predicted effects with 95% CI (dashed lines). Circles represent point estimates for each cohort. Study weight is represented by the diameter of the corresponding circle. Error bars represent the SE of each point estimate. Dotted lines represent a lack of phenotypic difference between the TT and AT/CT polymorphism. There was a statistically significant inverse correlation between mean BMI in the TT group and the AT-associated increase in logBMI (β = −0.0381, p = 0.0033). Similarly, a significant inverse correlation was observed between mean body weight in the TT group and the AT-associated increase in log-body weight (β = −0.0161, p = 0.0041)