STUDY QUESTION: What are the best practices for undertaking epidemiologic and phenotypic studies in polycystic ovary syndrome (PCOS)? SUMMARY ANSWER: Best practices for the undertaking of epidemiologic and phenotypic studies in PCOS are outlined. WHAT IS KNOWN ALREADY: Currently methodologies used for studies of PCOS epidemiology and phenotypes vary widely, and the comparability of studies is low, reducing the ability to harmonize studies. STUDY DESIGN, SIZE, DURATION: The Androgen Excess and PCOS (AE-PCOS) Society established a Task Force to draft a research resource for epidemiologic and phenotypic studies in PCOS, with the aim of providing guidelines on study design and execution, insights into the limitations and alternatives and protocols to be used, taking into consideration a global perspective. PARTICIPANTS/MATERIALS, SETTING, METHODS: A targeted review of the literature was carried out as necessary. MAIN RESULTS AND THE ROLE OF CHANCE: High level recommendations include the following: (i) Before initiating the study, a number of critical factors should be addressed including selecting the population and diagnostic criteria (which should ideally align with the recommendations of the International Guidelines), the type of observational study to be undertaken and the primary and secondary endpoint(s) of the study.(ii) To assess the 'natural' or true phenotype and epidemiology of PCOS, the least medically biased, broadest and most generalizable population, and the broadest definition of PCOS, should be used.(iii) Four PCOS phenotypes (Phenotypes A through D), based on the presence or absence of three general features (oligo-anovulation, hyperandrogenism and polycystic ovarian morphology), should be ascertained.(iv) In epidemiologic and phenotypic studies, the detection of PCOS rests on the accuracy and sensitivity of the methods used for assessing the individual features of the disorder, and how 'normal' is defined.(v) Although an assessment algorithm that minimizes the use of certain measures (e.g. androgen levels and/or ovarian ultrasonography) can be devised, when possible it is preferable to uniformly assess all subjects for all parameters of interest. (vi) The inclusion of subjects in epidemiologic studies who do not appear to have PCOS (i.e. 'non-PCOS') will provide the necessary cohort to establish population-specific normative ranges for the various features of PCOS. (vii) Epidemiologic studies of PCOS in unselected populations will yield relatively limited numbers of PCOS subjects available for genetic study; alternatively, large population-based epidemiologic studies of PCOS will potentially generate large numbers of unaffected individuals that may serve as genetic controls. (viii) Epidemiologic studies of PCOS will benefit from a clear governance structure and should begin by informing, educating and engaging both the formal and informal leaders of the populations targeted for study. (ix) In designing their study investigators should, in advance, establish statistical power and recognize, manage and account for inherent biases. (x) Subjects suspected of having PCOS but who do not/cannot complete their evaluation (i.e. have 'possible PCOS') can be included by imputation, assigning them a 'diagnostic weight' based on those subjects of similar clinical phenotype that have completed the study. (xi) In obtaining, storing and retrieving subject data, subjects should be assessed consecutively using a uniform data collection form; providing as complete and in depth data as possible. (xii) Maintenance of both paper and electronic medical records should focus on ensuring data quality, accuracy and institutional ethical compliance, and familiarity with country-dependent laws, including biobanking-specific laws, tissue laws and research laws. (xiii) In obtaining and biobanking study samples, these should be ideally collected at the time of the first assessment. (xiv) Access to stored data sets should ideally be granted to other bona fide researchers conducting research in the public interest. (xv) SOPs detailing the exact method of each of the activities for handling the data and the samples are necessary to ensure that all methods are performed uniformly. (xvi) Epidemiologic studies of PCOS must be resourced adequately. LIMITATIONS, REASONS FOR CAUTION: As with all reports involving expert interpretation of experiential and published data, inherent individual biases are possible. This risk is minimized in the present study by including experts from varying fields of study, aligning with recent international evidence-based guidelines and obtaining consensus approval of the recommendations from the Task Force and the board of the AE-PCOS. WIDER IMPLICATIONS OF THE FINDINGS: These guidelines should encourage investigators worldwide to undertake much needed epidemiologic studies of PCOS, increasing the validity, integrity and comparability of the data. STUDY FUNDING/COMPETING INTEREST(S): The study received no funding. R.A. serves as consultant for Medtronic, Spruce Biosciences and Ansh Labs; has received research funding from Ferring Pharmaceuticals; and is on the advisory board of Martin Imaging; R.L. has received research funding from MSD Pharmaceuticals; J.L. has received fees and/or grant support from the Dutch Heart Association, The Netherlands Organisation for Health Research and Development (ZonMw), Ferring Pharmaceuticals, Danone, Euroscreen/Ogeda and Titus Health Care; H.T. receives grant funding from the National Health and Medical Research Council; K.K., L.M.-P., S.S.M. and B.O.Y. have no potential conflicts of interest. TRIAL REGISTRATION NUMBER: N/A.