Study design

This case–control study—the Brazilian PCOS study—encompasses several Collaborating Centres across the country. The sample consists of women with PCOS and a control group of healthy participants matched by age, socioeconomic status and geographical region. The use of standardised procedures allows incorporation of the data collected in each centre into a unified database (figure 1). Participants will be recruited until 2021.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow chart of inclusion and study procedures.

Setting

The first eight Collaborating Centres are distributed in the five Brazilian regions. The Coordinating Centre is located in the south, based at Universidade Federal do Rio Grande do Sul (Porto Alegre, South Region). Collaborating Centres include the following: in the southeast, Universidade Federal de Minas Gerais, Faculdade de Medicina, Universidade de São Paulo (USP), Faculdade de Medicina de Ribeirão Preto, USP and Instituto Estadual de Diabetes e Endocrinologia in Rio de Janeiro; in the north, Universidade Federal de Rondônia; in the northeast, Universidade Federal do Rio Grande do Norte; and in the midwest, Universidade Federal do Mato Grosso. Additional centres may join the study if they obtain local institutional ethics committee approval.

Sample

The study sample consists of women with PCOS receiving healthcare at university hospital outpatient clinics within the public healthcare network (Unified Health System). PCOS is diagnosed at inclusion using the Rotterdam criteria, and participants are further classified according to PCOS phenotypes, as follows: (1) classic PCOS—women presenting oligo/amenorrhoea and/or anovulation, clinical and/or biochemical hyperandrogenism, with or without polycystic ovary morphology (PCOM) (phenotypes A and B); (2) ovulatory PCOS—having clinical and/or biochemical hyperandrogenism, PCOM and regular menstrual cycles (phenotype C) and (3) normoandrogenic PCOS—defined by oligo/amenorrhoea, PCOM, absence of clinical hyperandrogenism and normal androgen levels (phenotype D).12

Non-PCOS control participants are recruited in each Collaborating Centre through public advertisement or among women consulting for contraception, vaccination or reasons not related with reproductive diseases. This control group will allow detection of regional and ethnic characteristics requiring adjustment and characterisation of genotype distribution in the Brazilian population.

Eligibility criteria

Participants in the case group must meet Rotterdam PCOS criteria at enrolment. Women aged 18–39 years, recruited at least 2 years after menarche, with body mass index (BMI) varying from 18.5 to 40 kg/m², and not taking any medicine that could interfere with metabolic or hormonal assays are eligible for the study. Use of oral contraceptives must be stopped 3 months, and of metformin at least 2 months, before study enrolment. The same criteria apply to the control group.

Exclusion criteria

Type 1 diabetes, pregnancy, androgen-secreting tumours, Cushing syndrome, non-classical congenital adrenal hyperplasia, hyperprolactinaemia and thyroid illness without adequate control entail exclusion from the study. The same exclusion criteria apply to the control group.

Sample size estimation

The sample size was calculated assuming a prevalence of 60% of the classic PCOS phenotype in Latin American women (Spritzer PM, PCOS metabolic profile and phenotypes in Brazil and comparison with other Latin-American populations, 2019), with 95% CI, maximum absolute error of 5% and a design effect of 2. The case–control ratio was defined as 2:1. Based on these parameters, the sample size was estimated as 800 PCOS cases and 400 controls. In addition, to ensure that the geographical distribution of the sample reflects the population of the five Brazilian regions, the following stratification was used: south (13.7%)—110 PCOS and 55 control participants; southeast (41.5%)—332 PCOS and 167 control participants; northeast (28.4%)—227 PCOS and 113 control participants; north (8.6%)—69 PCOS and 34 control participants; and midwest (7.8%)—62 PCOS and 31 control participants.

Ethics and dissemination

During data collection, care is taken to ensure confidentiality of participants’ information. Interim analyses and the full final data set will be reported in scientific publications and at research conferences.

Patient and public involvement

There is no patient and public involvement in this study.

Prospective database

Standardised questionnaires are used to collect information on demographic characteristics (age and self-reported skin colour), education, medical, family, reproductive history, lifestyle/behavioural factors, quality of life, medication/supplement inventory, alcohol consumption, cigarette smoking, dietary pattern, presence of asthma or migraine, comorbidities, menstrual pattern, presence of acne and/or hirsutism and obesity or overweight. Participants are interviewed and examinations are performed at outpatient clinics in each centre (table 1). All participants undergo a physical examination performed by a trained physician. The data are recorded in web-based forms. Data storage and management are centralised at the Coordinating Centre using a cloud service. Importantly, all data are immediately available after collection.

Education

Educational attainment is assessed through the number of years of successful formal education, described as years at school.

Socioeconomic status

Socioeconomic status is assessed using an instrument developed by the Brazilian Association of Market Research Institutes (Associação Brasileira de Institutos de Pesquisa de Mercado). Participants are classified into five socioeconomic groups: A, B, C, D and E, with A representing the highest socioeconomic level. This classification is based on the responses to 15 questions that refer to variables such as educational attainment and items of home comfort.13

Alcohol intake

Alcohol consumption is self-reported: non-drinker (no drinking in the past year) or drinker. Participants are asked what type of alcohol (wine, beer, spirits, vodka, etc) they consume, the amount, and the frequency of consumption. After this detailed survey, the mean amount of alcohol (expressed as grams of alcohol consumed in a day) is estimated. Participants are then categorised as follows: low drinking (less than one drink per day), moderate drinking (less than two drinks per day), heavy drinking (more than seven drinks per week or three drinks per occasion) and binge drinking (four or more drinks on one drinking occasion).14

Smoking

For smokers, the number of cigarettes/day is registered. Former smokers are defined as individuals who report having quit smoking; they are also asked how old they were when they quit smoking. No minimum time since smoking cessation is employed in the definition of former smokers.15 16

Physical activity

Participants are asked about the type and frequency of the activity performed. The questions inquire about walking or riding a bicycle for commuting and about aerobic or anaerobic activities done for fitness. Specifically, participants are asked about physical training (walking, jogging, cycling, playing volleyball, basketball, handball, soccer, swimming, dancing and gym training). Besides their current habits, participants are surveyed about their physical activity in two past periods: from 12 to 18 years of age, and from 19 years until 1 year before study enrolment. For a given activity to be considered relevant, it must be performed for at least 10 uninterrupted minutes. Physical activity is classified according to its intensity, frequency and duration using the International Physical Activity Questionnaire.17 18

Household activities such as cooking, dishwashing and house cleaning are also covered in this form, since they may represent a major physical activity for some women.

Dietary pattern

The dietary pattern refers to the amount of fruits and vegetables participants eat on a normal day. Participants are first asked if they eat fruits and vegetables. If an affirmative answer is provided, the amount of fruits and vegetables consumed on a day is estimated. In addition, information is recorded on the daily and weekly frequency of fruit and vegetable consumption. Eating five or more portions of fruits and vegetables per day, 5–7 days in a week, is defined as a marker of healthy eating.19 20

Psychological assessment

Because patients with PCOS may have an increased propensity to psychological disorders,21 the modified Self-Reporting Questionnaire (SRQ-20) is used to determine the probability of psychological distress, mainly anxiety and depression. The SRQ-20 is recommended by the WHO for this purpose, and a validated Brazilian Portuguese version is available.22 Although the ideal cut-off is largely under debate, for the present sample a score >7 points in the SRQ-20 was established as the best estimate for the presence of a non-psychotic disorder. This cut-off had 86% sensitivity and 89% specificity in a Brazilian study including a population of which 54% were women. In that study, the discriminant power of SRQ-20 for psychiatric screening was 0.9, and Cronbach's alpha was 0.86.23 24 Also, quality of life is assessed using the WHO Quality of Life-BREF instrument.25 The same instruments will be administered to the control group.

Reproductive history and comorbidities

Women are asked about their age at menarche, the frequency and regularity of menses, use of medication to regulate menses, and the date of the last three menstrual periods. Menstrual pattern is then classified as regular, frequent, infrequent or amenorrhoea, according to the International Federation of Gynaecology and Obstetrics system.26 The use of hormonal contraception or any type of hormone therapy is verified by asking participants to show the medication box or the medical prescription. The number of years of contraceptive use is also recorded. Information is collected on the use of non-hormonal contraceptive methods and the age at which the contraceptive method was started.

Obstetric history is assessed by asking about the number of pregnancies, miscarriages, term and preterm births, mode of delivery, and diagnosis of gestational diabetes and preeclampsia in any pregnancy. Information on other former medical intercurrences in pregnancy is collected using an open question. Information is also collected about offspring gestational age, weight, and length at birth and about offspring with birth weight >4 kg.

The participants are then asked whether they take medication or have a known diagnosis of hypertension, diabetes and dyslipidaemia.

Family history

Participants are asked about diseases affecting family members. All diseases mentioned are recorded, but direct questions are asked about the following: hypertension, cardiovascular disease (defined as an acute myocardial infarction or angina pectoris), diabetes, thrombosis or pulmonary thromboembolism, infertility, hirsutism or thinning hair in women, and obesity.

Anthropometric measurements

Participants undergo a physical examination performed by a trained physician. Weight and height measurements are performed without shoes or heavy clothing. Weight is measured using a mechanical scale and height is measured using a stadiometer. Waist circumference is measured at the midpoint between the iliac crest and the last rib. Hip circumference is measured at the widest point of the relaxed abdomen. For both measures, participants are asked to stand with feet together and arms hanging at the sides. All these measures are taken twice, and a mean value is computed.

Blood pressure measurements

Standardised blood pressure measurements are performed twice, using the same calibrated mercury manometer. The mean of both assessments is used in the analysis. Hypertension is confirmed by asking about a previous diagnosis of hypertension or use of antihypertensive medication, or else identified during the physical examination in the presence of systolic blood pressure ≥130 mm Hg or diastolic blood pressure ≥80 mm Hg.27 28

Hirsutism and acne

Hirsutism is assessed with the modified Ferriman-Gallwey scale. The score is registered as a continuous variable. A score of 8 points or more indicates hirsutism.29 Hirsutism score is initially assigned by a trained investigator. After that, the participant provides her own self-reported score using a printed guide showing body areas. Shaved areas are tagged as ‘shaved’ and are taken into account in the analysis.

In the next step of the physical examination, the physician records the presence of acanthosis nigricans, galactorrhoea, reddish-purple streaks, acne and androgenetic alopecia. Acne is classified as follows: grade I (presence of non-inflammatory lesions, comedones only), grade II (presence of inflammatory lesions such as pustules and skin redness plus comedones), grade III (presence of widely distributed larger pustules) and grade IV (presence of cysts, nodular pustules and abscesses, with moderate to severe scarring).30 Alopecia is classified according to severity as type I, II or III.31

Blood and saliva sampling

Blood samples are collected from all participants between 08:00 and 10:00 after an overnight fast of 10–12 hours, during the early follicular phase of the menstrual cycle or at a random day in case of oligo/amenorrhoea. Blood is quickly centrifuged at 1036g for 8 min at room temperature and serum is separated and stored in aliquots at −80°C for future hormonal and epigenetic analyses. Additional drop samples are collected in FTA gene cards (Whatman FTA) and stored at room temperature for genetic analysis. These samples are regularly transferred to the Coordinating Centre in the city of Porto Alegre for processing, and will constitute a bio-repository.

Salivary samples are obtained at the mid-luteal phase of the menstrual cycle using a chewing roll and a collector tube especially designed for salivary hormonal tests (Salivette, Sarstedt, Nümbrecht, Germany). The tubes are centrifuged at 1000g for 2 min and stored at −20°C for subsequent progesterone assays.

Dyslipidaemia

History of dyslipidaemia is defined by self-reported hypercholesterolaemia and/or use of anti-cholesterol drugs. Blood tests are also performed to measure total cholesterol (TC) and fractions and triglyceride levels.

Diabetes and metabolic syndrome

Diabetes is determined by self-report, use of antidiabetic drugs or a fasting blood glucose level of 126 mg/dL or higher. Metabolic syndrome is defined as the presence of at least three of the following components: waist circumference >88 cm, high-density lipoprotein cholesterol (HDL-C) level <50 mg/dL, triglyceride (TG) level of 150 mg/dL or higher, blood pressure ≥130/85 mm Hg and glucose level of 100 mg/dL or higher.32

Pelvic ultrasound

A pelvic ultrasound image is obtained from all participants at each centre. The preferred method is the transvaginal approach, considered to have the best sensitivity and specificity for evaluation of the uterus and ovaries. The transabdominal procedure is used for participants with no previous vaginal intercourse. Ultrasound examination is performed during the follicular phase in participants with regular menses or on any day in those with oligo/amenorrhoea. Ovarian and uterine volumes and the number of antral follicles in each ovary are recorded. Ovarian volume ≥10 cm³ in at least one ovary, in the absence of corpus luteum, cyst or dominant follicle, will be considered as PCOM.33

Assays

TC, HDL-C, TGs and fasting glucose levels will be determined by colorimetric-enzymatic methods in each centre. Low-density lipoprotein cholesterol (LDL-C) is estimated indirectly with the Friedewald formula:(LDL-C=TC − (HDL-C +TG/5)). Hormone determinations will be performed at the Coordinating Centre. Steroid hormone binding globulin (SHBG), insulin and 25‐hydroxyvitamin D (25(OH)D) levels will be measured by chemiluminescence (Abbott Architect, IL, USA), with sensitivity of 0.02 nmol/L, 1.0 μU/mL and 1.6 ng/mL, respectively. The intra-assay and interassay coefficients of variation (CVs) for SHBG, insulin and 25(OH)D are ≤5.2% and≤9.5%, ≤4.2% and ≤5.2%, and ≤5.1% and≤7.1%, respectively. Total testosterone (TT) is measured by high-performance liquid chromatography-tandem mass spectrometry, with a sensitivity of 7.0 ng/dL and intra-assay and interassay CVs of 4.5% and 4.7%, respectively. Androstenedione, dehydroepiandrosterone sulfate (DHEA-S) and 17 alpha hydroxyprogesterone (17α-OHP) are determined by radioimmunoassay (Beckman Coulter, Prague, Czech Republic) with sensitivity of 0.05 ng/mL, 12.33 ng/mL and 0.046 ng/mL, respectively. The intra-assay and interassay CVs for androstenedione, DHEA-S and 17α-OHP are ≤7.5% and≤11.3%,≤5.2% and≤5.3%,≤7.8% and≤15.7%, respectively. Free androgen index is estimated by dividing TT (nmol/L) by SHBG (nmol/L) × 100. Homeostasis model assessment index is calculated by multiplying insulin (μIU/mL) by glucose (mmol/L) and dividing this product by 22.5. The lipid accumulation product index is calculated using the formula (waist (cm) − 58) × triglyceride concentration (mmol/L). Salivary progesterone and serum anti-Müllerian hormone (AMH) will be assayed after enrolment is completed.

Molecular analyses

For later epigenetic/microRNA analysis, serum samples are stored at −80°C. MicroRNA extraction will be made with a commercial kit (miRNeasy Serum/Plasma Advanced Kit, Qiagen, Venlo, The Netherlands). Real-time quantitative PCR will be performed to determine relative expression levels of a panel of microRNAs that have been previously screened and expressed differentially in PCOS women.34

DNA will be extracted from blood samples in FTA gene cards according to the manufacturer's instructions. Single nucleotide polymorphisms (SNP) will be assessed by real-time PCR using allelic discrimination assays with TaqMan MGB primers and probes (Applied Biosystems, ThermoFisher Scientific, Waltham, MA, USA).

Genotypic analyses will target genes related to endocrine and metabolic dysfunctions of PCOS, including androgen excess and metabolism, insulin resistance, metabolic syndrome and vitamin D metabolism, adipokines, inflammatory markers, hormone receptors, obesity and diabetes mellitus, hormone-binding proteins, lipoproteins, angiotensin, interleukins, β-adrenergic, Peroxisome proliferator-activated receptor (PPAR) family, cytochrome p450, irisin, kisspeptin, uncoupling protein family, and immunological factors.35 36

Statistics

The Shapiro-Wilk normality test and descriptive statistics will be used to evaluate the distribution of data. Parametric variables will be described as mean and SD. Variables with non-parametric distribution will be log-transformed before statistical analysis and reported as geometric mean of their original units of measure. Non-parametric variables will be presented as median and IQR. Categorical data will be expressed as percentage and 95% CIs. Statistical tests will be applied according to data distribution in the different groups under comparison (eg, control vs PCOS or across PCOS categories). The preferred statistical tests will be χ² test for heterogeneity to compare proportions; one-way analysis of variance (ANOVA), followed by Bonferroni post-hoc tests for parametric variables or their equivalent of non-parametric statistics (ie, Kruskal-Wallis test); co-variance analysis will also be considered (eg, analysis of covariance) to examine the influence of variables such as age and BMI.

The association of demographic, anthropometric and metabolic variables with lifestyle habits and epigenetic markers will be investigated by unadjusted analyses, multivariate models as well as stratified by groups. An important focus of these analyses is the potential heterogeneity of risk relationships in different regions and PCOS categories. Therefore, interactions will be tested when appropriate. Logistic regression will be preferred to estimate the OR for the associations under the investigation.

Regarding molecular analyses, the agreement of genotype frequencies with Hardy-Weinberg equilibrium for each SNP will be analysed using the Pearson χ² test. Lewontin’s D’ statistic for linkage disequilibrium will be calculated for each pair of polymorphisms, and haplotypes will be inferred using the Phase 2.1 software, which uses Bayesian statistics. Linear trends will be tested to check for codominant effects of genotypes on scale data. The relationship between the outcome of interest and genotypes will be evaluated using ORs. All analyses will be performed in PASW Statistics for Windows, V.21.0 (SPSS, Chicago, IL, USA). All tests will be two-tailed and significance will be accepted at a p value<0.05.