Risk factors for PCOS development
Table 1 shows the association of cases of PCOS and controls based on the univariate
and logistic regression analysis. PCOS is associated with BMI, serum testosterone
and kisspeptin concentrations and the FTO rs9939609 polymorphism by univariate analysis.
Our model showed that BMI, serum kisspeptin levels and FTO rs9939609 (AA) polymorphism
remained significant after adjusting for potential confounders. The model was statistically
significant with a Hosmer and Lemeshow test value of 0.34 (p > 0.1 taken as significant).
The model also explained 24% (Nagelkerke R2) of the variance among those with PCOS
and correctly classified 72% of the cases, which confirms to a good fit. Meanwhile,
this model demonstrated there was no significant association between genes of the
HPG axis (Kiss1, GPR54, GnRH, FSHB, FSHR, LHCGR) and insulin receptor gene (INSR)
with PCOS.
Hardy–Weinberg equilibrium and mode of inheritance analysis
The genotype distributions of the FTO SNP (rs9939609) was not in the Hardy–Weinberg
equilibrium in both patients and controls (p<0.05).
Association between the FTO SNP (rs9939609) and PCOS risk was analyzed under five
gene models (co-dominant, dominant, recessive, over-dominant and log additive).
The FTO gene rs9939609 polymorphism - under the co-dominant model, the genotypes “AA”
(OR = 5.49; 95% CI -2.34-12.88; p < 0.05); under the dominant model genotype “A/T-A/A”
(OR = 3.21 95% CI -1.62-6.37; p <0.05); under recessive model genotype “A/A” (OR=
4.45, 95% CI -2.01-9.90; p<0.05) and the log additive model (OR = 2.30; 95%CI -1.51-3.51;
p<0.05) were associated with increased risk for PCOS (Table 2).
The model with the lowest AIC and BIC values for a given polymorphism was considered
the best-fit model. The AIC and BIC values indicated that the log additive model may
serve as the best-fit model of rs9939609 polymorphism of FTO gene (Table 2).
Association of FTO rs9939609 polymorphism with clinical and hormonal characteristics
of PCOS
The best predictors of FTO rs9939609 polymorphism, with a statistically significant
association were the mFG scale and serum testosterone (Table 3). Furthermore, interaction
between FTO rs9939609 polymorphism with the variables (BMI, mFG, serum testosterone
and kisspeptin levels) showed significant interaction between FTO AA genotype and
mFG (p=0.042); whereas the AT genotype of FTO gene showed a marginally significant
interaction (p=0.054) with mFG.
Association of FTO genotype with serum kisspeptin and testosterone levels
Although kisspeptin levels were significantly associated with FTO genotype in the
univariate analysis; the association was insignificant when adjusted for confounders
(Table 3). Nevertheless, when the subjects were divided based on the FTO genotypes
and compared with serum kisspeptin and testosterone levels, subjects with AA genotype
had higher mean serum kisspeptin levels when compared to AT and TT genotypes (Fig
1) while the mean testosterone concentrations were greater in subjects with the AT
genotype (p>0.05) (Fig 1). In addition, serum testosterone levels were significantly
higher in subjects with mutant alleles (AA + AT) when compared to those with normal
allele (TT) (p<0.05).
Association of FTO genotypes with body mass index (BMI)
When the subjects were subdivided based on their BMI using the Asian cut off (BMI
< 25 kg/m2) and compared with the FTO genotype (Fig 2) we found a significant correlation
between FTO gene polymorphism and BMI (chi square value =17.05, p<0.05). The frequency
of AA genotype was greater among obese PCOS subjects (BMI ≥ 25 kg/m2) while the TT
allele was seen in a greater proportion of controls.
FTO with cutaneous markers of insulin resistance (Acanthosis nigricans)
Acanthosis nigricans was used as a clinical marker of insulin resistance. PCOS subjects
with the AA genotype had a significantly higher frequency of acanthosis nigricans
when compared to those with the TT genotype (p<0.05) (Fig.3). Acanthosis was seen
most commonly among women with PCOS who were obese or overweight.