Given the variation in the CAG repeat length of the androgen receptor (AR) gene and its inverse effect on the receptor activity, the alleles with short CAG repeat length are expected to result in amplified androgen receptor activity, resulting in a state of hyperandrogenism. Thus, we hypothesized that the shorter alleles would be more frequent and preferentially more active among the PCOS women than the controls. To the best of our knowledge, this is the first study of its kind on a large cohort of Indian women to examine the association of AR polymorphism with PCOS. Our study revealed a range of 8–31 CAG repeats in the AR gene of South Indian PCOS women and the controls. However, neither the distribution of biallelic mean values nor the mean repeat sizes were found significantly different between the PCOS cases and controls. These results were concurrent to the observations of some previous studies which did not find significant difference in the mean values of CAG repeat sizes between PCOS cases and controls 
. However, there have been other studies which yielded divergent and contradictory results; while Xita et al 
and Shah et al 
found shorter alleles to be more frequent among the PCOS cases than controls, Hickey et al 
found longer alleles to be significantly higher in frequency in the PCOS women than controls. This inconsistent nature of association could be in part accounted for by the variation in the CAG repeat number in different populations and ethnic groups as observed by Edwards et al. 
. Our study did not reveal any association of CAG alleles, shorter or longer, with PCOS; the distribution of the cases and controls appear to be highly homogeneous across the spectrum of CAG alleles, with odds ratio close to 1 and effect size zero. Based on expected effects as presented in a previous study and given our sample size the minimum effect size expected would 0.14 to be able to detect significant association with power of 90%.Therefore, our study can be regarded as sufficiently powered to replicate the previous association between CAG repeats and PCOS.
Since AR gene is located on X chromosome, the epigenetic phenomenon of X chromosome inactivation leading to preferential activation of either the shorter or the longer allele may explain the differential androgen receptor activity resulting in the phenotypic or clinical heterogeneity in PCOS, chiefly the androgenic features. XCI analysis has been previously done in both PCOS and non-PCOS cases with hyperandrogenic features. Calvo et al. 
found no significant difference in the pattern of X inactivation between women with hyperandrogenic hirsutism/idiopathic hirsutism and controls. Subsequently, Shah et al 
demonstrated preferential inactivation of the longer CAG repeat alleles resulting in expression of the smaller alleles that predictably increase the androgen sensitivity. However, XCI analysis revealed random pattern of X-inactivation of the relatively larger proportion of our Indian samples and sub-samples. Nevertheless, concurrent to the functional hypothesis, analysis of samples with non-random XCI revealed that shorter alleles were preferentially more activated among the PCOS cases than controls, although the relative proportion of the activated alleles were not significantly different between them (OR
0.33). This trend is consistent with Shah et al 
observation which suggests that increased intrinsic androgenic activity associated with short AR alleles may have a role in the pathogenesis of PCOS-related hyperandrogenism.
Hyperandrogenism, which is ascertained biochemically by assessing the testosterone levels, and clinically by hirsutism using the Ferriman-Gallwey score, was also taken as a parameter for analyzing its association with CAG repeat numbers. We did not observe any significant association between testosterone levels and CAG repeat length which is in contrast to some of the previous studies 
. Consistent with the functional hypothesis, although there is a trend of shorter CAG alleles showing lower levels of testosterone than those with longer CAG alleles in the PCOS women, it is not statistically significant unlike in case of the previous studies 
. Given the relatively large sample size, this trend cannot be therefore considered as suggesting a relatively greater role of excessive androgen action (implying receptor activity) in PCOS than expected from hyperandrogenemia alone. The obese PCOS cases of our study exhibited contrasting trend in comparison to the lean cases, in that the shorter CAG repeat length is associated with a high testosterone value, albeit statistically not significant, supporting the hypothesis that increased androgen activity in women has a stimulatory effect on ovarian androgen production 
The pattern of body fat distribution can regulate androgen production and metabolism to a significant extent 
. Both production rates and metabolic clearance rates of androgens are equally increased in obesity. A condition of relative hyperandrogenism in women are associated with abdominal obesity and all the features of the metabolic syndrome, including insulin resistance and compensatory hyperinsulinemia, low high-density lipoprotein(HDL) cholesterol, increased triglycerides, and elevated arterial blood pressure 
that are commonly present in a PCOS phenotype. Thus, we also examined the association of hyperandrogenic features in obese and lean PCOS cases separately vis-à-vis the CAG repeat numbers. When CAG repeat distribution analysis was carried out among the lean and obese PCOS cases, there was a significant difference between them () whereby the lean PCOS cases had significantly higher frequency of biallelic mean in the middle range of 18–20 than the obese cases. Subsequent logistic regression analysis yielded significant odds ratio with a high statistical power of 92.8% suggesting that the PCOS women with CAG biallelic mean range of <18 and >20 are at a much higher risk to develop obesity and associated metabolic complications. However, there seemed to be no epigenetic influence over this heterogeneity, since majority of the cases followed random XCI pattern. Apart from this, we also found a significant positive correlation (r
0.403, p<0.001) between hirsutism and BMI within the PCOS cases. Among the hirsute cases, the proportion of obese PCOS women is significantly (p
0.02) higher (77%) than the lean PCOS women (23%). This could be explained by the increase in the metabolic clearance rate of androgen, particularly in obese women which finally leads to augmented hyperandrogenic features like hirsutism.
To sum up, we found homogeneity in CAG allele distribution profile of the cases and controls, hence CAG repeat polymorphism, by itself, cannot be considered as a useful discriminator between the PCOS cases and controls. Further, in spite of the lack of evidence for a major epigenetic effect, we observed a trend of preferential activation of the shorter allele among the PCOS cases with non random XCI pattern. On the other hand, the significant heterogeneity in the CAG biallelic mean between the obese and lean PCOS cases can explain the differential receptor activity among them, which may also probably explain the manifestation of hyperandrogenic features in such cases. Thus, the lean PCOS women, who have a significantly higher frequency of the middle range CAG repeats, would possibly be conferred with a moderate receptor activity as compared to the obese PCOS cases. In the latter group, the lower or higher CAG repeat numbers of the polymorphic spectrum would therefore lead to an increased or diminished receptor activity, respectively, both of which are expected to result in a hyperandrogenic state. We may further add that in case of the lean PCOS group the androgen receptor CAG repeats may not play a very instrumental role in manifestation of the hyperandrogenic features, while in the obese PCOS group this microsatellite variation may account for the hyperandrogenicity to a larger extent. XCI analysis among both these groups also revealed the same pattern as observed in the entire cohort implying that the epigenetic effect may not have any major effect on the androgen metabolism and receptor activity among Indian PCOS women. Overall, our results rule out the possibility of suggesting CAG repeat polymorphism as a prognostic marker for PCOS. Especially given that ours is one of the largest samples so far studied and from a relatively more homogenous population and due to internal consistency observed in the sub samples within our sample, we may repose fair degree of confidence in the reliability of our results. Nonetheless, our study is the first of its kind from this region and therefore further replicative studies are required among other ethnic groups, encompassing the geographic heterogeneity of India before reaching any unequivocal conclusions on the precise role of androgen receptor CAG repeat polymorphism in the manifestation of this immensely heterogeneous PCOS phenotype in the populations of this region.