What is the prevalence of insulin resistance (IR) and the contributions of intrinsic and extrinsic IR in women diagnosed with polycystic ovary syndrome (PCOS) according to the Rotterdam criteria? We report novel clamp data in Rotterdam diagnosed PCOS women, using World Health Organization criteria for IR showing that women with PCOS have a high prevalence of IR, strengthening the evidence for an aetiological role of IR in both National Institutes of Health (NIH) and Rotterdam diagnosed PCOS in lean and overweight women. PCOS is a complex endocrine condition with a significant increased risk of gestational diabetes and type 2 diabetes. Using a cross-sectional study design, 20 overweight and 20 lean PCOS (Rotterdam criteria), 14 overweight and 19 lean body mass index (BMI)-matched control non-PCOS women underwent clinical measures of IR after a 3-month withdrawal of insulin sensitizers and the oral contraceptive pill. In an academic clinic setting, glucose infusion rate (GIR) on euglycaemic–hyperinsulinaemic clamp was investigated as a marker of insulin sensitivity. PCOS women were more IR than BMI-matched controls (main effect for BMI and PCOS; P < 0.001). IR was present in 75% of lean PCOS, 62% of overweight controls and 95% of overweight PCOS. Lean controls (mean ± SD; GIR 339 ± 76 mg min⁻¹ m^–2) were less IR than lean PCOS (270 ± 66 mg min⁻¹ m⁻²), overweight controls (264 ± 66 mg min⁻¹ m⁻²) and overweight PCOS (175 ± 96 mg min⁻¹ m⁻²). The negative relationship between BMI and IR reflected by GIR was more marked in PCOS (y = 445.1 – 7.7x, R² = 0.42 (P < 0.0001) than controls (y = 435.5 – 4.6x, R² = 0.04 (P < 0.01)). The study did not use glucose tracer techniques to completely characterize the IR, as well as the lack of matching for body composition and age. IR is exacerbated by increased BMI, supporting intrinsic IR in PCOS. BMI impact on IR is greater in PCOS, than in controls, irrespective of visceral fat, prioritizing lifestyle intervention and the need for effective therapeutic interventions to address intrinsic IR and prevent diabetes in this high-risk population. This investigator-initiated trial was supported by grants from the National Health & Medical Research Council (NHMRC) Grant number 606553 (H.J.T., N.K.S. and S.K.H.) as well as Monash University and The Jean Hailes Foundation. H.J.T. is an NHMRC Research Fellow. N.K.S. is supported through the Australian Government's Collaborative Research Networks (CRN) programme. A.E.J. is a Jean Hailes and NHMRC scholarship holder. The authors declare that there is no conflict of interest associated with this manuscript. ISRCTN84763265.

STUDY QUESTION What is the degree of intrinsic insulin resistance (IR) in women with polycystic ovary syndrome (PCOS) and the relative contribution of BMI to overall IR based on meta-analysis of gold standard insulin clamp studies? SUMMARY ANSWER We report an inherent reduction (−27%) of insulin sensitivity (IS) in PCOS patients, which was independent of BMI. WHAT IS ALREADY KNOWN PCOS is prevalent, complex and underpinned by IR but controversies surround the degree of intrinsic IR in PCOS, the effect of BMI and the impact of the different diagnostic criteria (NIH versus Rotterdam) in PCOS. STUDY DESIGN, SIZE, DURATION A systematic review and meta-analysis of Medline and All EBM databases was undertaken of studies published up to 30 May 2015. Studies were included if premenopausal women diagnosed with PCOS were compared with a control group for IS, measured by the gold standard euglycaemic–hyperinsulinaemic clamp. The systematic review adheres to the principles of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Meta-analyses were performed using mixed modelling and magnitude-based inferences expressed as mean effect ±99% CI. We inferred the effect was small, moderate or large relative to a smallest important change of −3.7% or 3.8% derived by standardisation. Effects were deemed unclear when the CI overlapped smallest important positive and negative values. Effects were qualified with probabilities reflecting uncertainty in the magnitude of the true value (likely, 75–95%; very likely, 95–99.5%; most likely, >99.5%). PARTICIPANTS/MATERIALS, SETTING, METHOD A total of 4881 articles were returned from the search. Of these, 28 articles were included in the meta-analysis. MAIN RESULTS AND THE ROLE OF CHANCE Overall IS was lower in women with PCOS compared with controls (mean effect −27%, 99% CI ±6%; large, most likely lower). A higher BMI exacerbated the reduction in IS by −15% (±8%; moderate, most likely lower) in PCOS compared with control women. There was no clear difference in IS between women diagnosed by the original National Institutes of Health (NIH) criteria alone compared with those diagnosed by the Rotterdam criteria. Low levels of sex hormone-binding globulin (SHBG) were associated with reduced levels of IS (−10%, ±10%; small, very likely negative), which was not confounded by BMI. LIMITATIONS, REASONS FOR CAUTION This systematic review and meta-analysis inherited the confounding problems of small sample sizes, missing data (e.g. some hormones, waist and hip girths) and the lack of Rotterdam criteria phenotype reporting, limiting the evidence synthesis and meta-analysis. WIDER IMPLICATIONS OF THE FINDINGS BMI has a greater impact on IS in PCOS than in controls. SHBG appears a potentially valuable marker of IR in PCOS, whereas testosterone after adjustment for BMI demonstrated an unexpected interplay with IS which warrants further investigation. STUDY FUNDING/COMPETING INTERESTS This work was supported by grants from the National Health & Medical Research Council (NHMRC), grant number 606553 (H.J.T., N.K.S.), as well as Monash University. H.J.T. is an NHMRC Research Fellow. N.K.S. is supported through the Australian Government's Collaborative Research Networks (CRN) programme. The funding bodies played no role in the design, methods, data management or analysis or in the decision to publish. All authors declare no conflict of interests. REGISTRATION NUMBER N/A

Key points: The discovery of microRNAs (miRNAs) has established new mechanisms that control health, but little is known about the regulation of skeletal muscle miRNAs in response to exercise. This study investigated components of the miRNA biogenesis pathway (Drosha, Dicer and Exportin‐5), muscle enriched miRNAs, (miR‐1, ‐133a, ‐133b and 206), and several miRNAs dysregulated in muscle myopathies, and showed that 3 h following an acute exercise bout, Drosha, Dicer and Exportin‐5, as well as miR‐1, ‐133a, ‐133‐b and miR‐181a were all increased, while miR‐9, ‐23a, ‐23b and ‐31 were decreased. Short‐term training increased miR‐1 and miR‐29b, while miR‐31 remained decreased. Negative correlations were observed between miR‐9 and HDAC4 protein, miR‐31 and HDAC4 protein and between miR‐31 and NRF1 protein, 3 h after exercise. miR‐31 binding to the HDAC4 and NRF1 3′ untranslated region (UTR) reduced luciferase reporter activity. Exercise rapidly and transiently regulates several miRNA species potentially involved in the regulation of skeletal muscle regeneration, gene transcription and mitochondrial biogenesis. Abstract  The identification of microRNAs (miRNAs) has established new mechanisms that control skeletal muscle adaptation to exercise. The present study investigated the mRNA regulation of components of the miRNA biogenesis pathway (Drosha, Dicer and Exportin‐5), muscle enriched miRNAs, (miR‐1, ‐133a, ‐133b and ‐206), and several miRNAs dysregulated in muscle myopathies (miR‐9, ‐23, ‐29, ‐31 and ‐181). Measurements were made in muscle biopsies from nine healthy untrained males at rest, 3 h following an acute bout of moderate‐intensity endurance cycling and following 10 days of endurance training. Bioinformatics analysis was used to predict potential miRNA targets. In the 3 h period following the acute exercise bout, Drosha, Dicer and Exportin‐5, as well as miR‐1, ‐133a, ‐133‐b and ‐181a were all increased. In contrast miR‐9, ‐23a, ‐23b and ‐31 were decreased. Short‐term training increased miR‐1 and ‐29b, while miR‐31 remained decreased. Negative correlations were observed between miR‐9 and HDAC4 protein (r=−0.71; P= 0.04), miR‐31 and HDAC4 protein (r =−0.87; P= 0.026) and miR‐31 and NRF1 protein (r =−0.77; P= 0.01) 3 h following exercise. miR‐31 binding to the HDAC4 and NRF1 3′ untranslated region (UTR) reduced luciferase reporter activity. Exercise rapidly and transiently regulates several miRNA species in muscle. Several of these miRNAs may be involved in the regulation of skeletal muscle regeneration, gene transcription and mitochondrial biogenesis. Identifying endurance exercise‐mediated stress signals regulating skeletal muscle miRNAs, as well as validating their targets and regulatory pathways post exercise, will advance our understanding of their potential role/s in human health.