This work received approval from the Ethical Review Committee, Faculty of Medicine, University of Colombo, Sri Lanka. Consecutive women conforming to inclusion criteria were recruited from the Endocrine Clinic of the University Unit, Colombo, Sri Lanka. Diagnosis of PCOS was based on the Rotterdam criteria . Details of sample size calculation , inclusion and exclusion criteria of study subjects and control group are described in previous reports [27-29] were given below;
Sample size calculation
The Schlesselman case control study formula was used for sample size calculation ;
The exposure rate of allele frequency of SNPs among controls (30%–44%) was based on literature from other countries as no studies have been done in Sri Lanka [10, 30-32].
The estimated proportion of 38% was used in sample size calculation.
R = odds ratio associated with exposure
The OR used in this study was 2.7; because Sri Lankans with anovulatory PCOS manifest severe symptoms at a younger age, with greater IR and a higher prevalence of metabolic syndrome than white Europeans .
P1 = Proportion of exposure among cases was calculated using the following formula;
Recruitment of Subjects
Inclusion criteria: Inclusion criteria were adolescent women whose symptoms manifested between 11–19 years of age (WHO), and all 3 diagnostic criteria should be present between 16 –19 years of age . The lower limit of age selection in this study was 15 years, based on the previous study in Sri Lanka and mean age of menarche in Sri Lanka being 13 years and leaving an allowance of two additional years for regularization of menstruation .
Anovular PCOS or amenorrhoea/ oligomenorrhea: Anovular cycles — length of the menstrual cycle is more than 35 days . Amenorrhoea — absence of menstrual periods for six months or more. Oligomenorrhea — menstrual periods occurs at intervals of greater than 35 days, with only four to nine periods in a year.
Polycystic ovaries on ultrasound: transvaginal or transabdominal ultrasound scan of ovaries, performed within the first 5 days from the onset of menstruation, defined by 24 or more follicles, measuring between 2 and 9 mm and/or an ovarian volume >10 cm3 .
Hyperandrogenism: Clinical evidence of hirsutism was determined by modified Ferriman-Gallwey score (mFG) with the cut off value of ≥ 8 along with serum testosterone levels (T) > 3.5nmol/L .
Exclusion criteria: Exclusion criteria included inherited disorders of IR such as Rabson–Mendenhall syndrome, Cushing syndrome, hyperprolactinaemia, untreated primary hypothyroidism, congenital adrenal hyperplasia or an androgen secreting ovarian/adrenal tumor; those taking corticosteroid, antiepileptic or antipsychotic drugs, history of hormonal contraception within the previous 6 months, pregnancy and the first postpartum year.
Control sample: Working women of similar ethnic and social background as the affected subjects were recruited for the study. Consenting, asymptomatic, normo-androgenic, normal cycling since adolescence and women of reproductive age in whom PCOS was objectively excluded by clinical, biochemical and ultrasound assessment, were recruited as controls.
Clinical and Biochemical Evaluation
Socio demographic, reproductive information (menstruation, fertility), age and degree of severity of clinical feature of PCOS, drug history, family history of diabetes, anthropometry (BMI and central obesity), resting blood pressure, hyperandrogensim assessed by a single clinical observer that includes hirsutism (modified FG score), temporal hair loss, acne, acanthosis nigricans, ovarian ultrasound, serum kisspeptin and testosterone concentrations of all subjects were utilized from our previously reported studies [27-29].
Statistical analysis was carried out with previously reported SNPs of the obesity gene (FTO), selected candidate genes of HPG axis (Kiss1, GPR54, GnRH, FSHB, FSHR, LHB, LHCGR) and insulin receptor gene (INSR) [28,29] and serum kisspeptin and testosterone concentration .
The Kolmogorov-Smirnov test was used to test the normality of distribution. Values with a biological distribution are presented as mean ± standard error for mean. Comparison of means between cases (those with PCOS) and controls (those without PCOS) was performed with independent sample t test. Chi-square test was used for comparison of genotype frequency between groups. It was also used for describing the correlation of genetic alleles with other numeric variables. To assess the magnitude of the risk factors in the development of PCOS, first all cases and controls were compared using crude odds ratio (OR) and 95% confidence interval (CI) by binary logistic regression (forward LR) method. Factors assessed were: BMI, serum kisspeptin and testosterone concentration with the obesity gene - FTO (rs9939609), HPG axis genes – Kiss1 SNPs (rs5780218, rs4889), GPR54 SNPs (rs10407968, rs12507294, rs350131, chr19:918686, chr19:918735), GnRH (rs6185), FSHB (rs6169), FSHR (rs6165/rs6166) and LHCGR (rs2293275) and insulin receptor gene (rs1799817), all of which were included to the logistic regression analysis in order identify risk factors for PCOS. To assess the risk for PCOS associated with the FTO rs9939609 polymorphism after controlling for confounders, logistic regression analysis was performed using forward LR method to obtain adjusted OR. In the regression model, case control status was included as the dependent variable; and the main predictor (FTO rs9939609 polymorphism was categorized as AA [homozygous], AT [heterozygous] and TT [wild type]) along with confounders (all significant factors in the univariate analysis) as independent variables. To determine the risk for FTO rs9939609 polymorphism to be associated with BMI, mFG scale, serum testosterone and kisspeptin levels, forward LR was performed. This analysis categorized the FTO rs9939609 into 2 groups - mutant allele (AA and AT) and normal allele (TT) and included them as the dependent variable. Any interaction of FTO rs9939609 polymorphism with each variable was analyzed individually to test for significance. Variables found to have a significant interaction were then included to the model. Several regression models were developed for both analysis, and the best model was selected based on the goodness of fit. All analyses were performed by SPSS software (v.18.0 SPSS, Inc., Chicago, IL). The level of significance was set as 5%.
Deviations from the Hardy–Weinberg equilibrium were tested by comparison of observed and expected genotype frequencies with χ2 test. Calculation of genotype and haplotype associations for all the SNPs was carried out using SNPSTATS program (http://bioinfo.iconcologia.net/index.php?module=Snpstats). Five inheritance models (co-dominant, dominant, recessive, over-dominant and additive) were applied for statistical analysis. The best inheritance model was assessed using the Akaike information criteria (AIC) and the Bayesian information criteria (BIC) with the model with the lowest values being the best fit.