Polycystic ovary syndrome (PCOS) is the most common endocrine disorder, affecting up to 13% of reproductive-aged women (1). It is associated with a range of reproductive, metabolic and psychological sequelae (2, 3). Diagnosis of PCOS in adults is based on the presence of two of three features: (i) oligo/amenorrhea; (ii) clinical/biochemical hyperandrogenism and/or (iii) polycystic ovary morphology (4) or elevated serum anti-Müllerian hormone levels, according to the latest 2023 International Evidence-based PCOS Guideline (1). The key pathophysiological drivers of PCOS include hyperandrogenism and intrinsic insulin resistance that is mechanistically distinct from obesity-associated insulin resistance (5, 6).

Several of these pathophysiological risk factors in women with PCOS have been theoretically linked to obstructive sleep apnea (OSA) (7, 8). OSA is characterized by the repetitive collapse of the upper airway during sleep for at least 10 seconds, and associated with oxygen desaturation and/or arousal from sleep (9). Additionally, obstructive sleep apnea syndrome (OSAS) is defined as OSA (apnea-hypopnea index (AHI) ≥5 events per hour) together with features of sleep disturbances, such as excessive daytime sleepiness and/or cardiometabolic comorbidities (e.g., arterial hypertension) (10–12). OSAS and associated snoring is the most common sleep-disordered breathing abnormality (12). In the general population and in PCOS, OSAS has been associated with multiple comorbidities such as obesity, insulin resistance, gestational diabetes, type 2 diabetes, hypertension, and impaired quality of life (12–14). Furthermore, cardiometabolic risk factors such as chronic inflammation, oxidative stress, and impaired fibrinolysis are elevated in PCOS (1) that can lead to higher risk of OSAS.

Past cross-sectional studies have shown that OSAS is highly prevalent in women with PCOS (15–19) compared with the general population of reproductive-aged women (17-75% vs. 9-28%) (9, 10), with lower rates in adolescents with PCOS (0-57%) (20, 21). Previous systematic reviews have also reported higher risk of OSA in PCOS relative to those without PCOS, but this varies between the odds ratios (ORs) of 2.86 and 8.30 (22–26). The highly variable OSA prevalence rates in past reviews relate to the heterogenous measurements, including polysomnography (PSG), cardiorespiratory polygraph and screening questionnaires (e.g., Berlin Questionnaire) (22–26). One recent review also included studies that did not report OSA measurements (26). Indeed, formal sleep studies using PSG assessments are essential for accurate OSA prevalence estimates (11). The wide risk estimates reported in prior reviews are also attributed to the inconsistent study inclusion and exclusion criteria employed, with some including conference abstracts and grey literature (23, 26) and others limiting searches to only peer-reviewed and database-indexed articles (22, 24–26).

To date, there has been no quantitative synthesis of the available evidence linking OSAS and related symptoms with PCOS status. Further, important confounders which may influence this relationship, including body mass index (BMI), age, PCOS diagnostic criteria, and ethnicity, have not been thoroughly interrogated. Higher BMI in adults can lead to physiological changes that contribute to OSA development through accumulation of fat deposits in the upper respiratory tract, thereby narrowing the airway and decreasing muscle activity in this region (7). Some studies however, reported higher prevalence of OSA irrespective of BMI in PCOS, raising the possibility that body fat distribution rather than total body mass may be more useful in determining OSAS risk in PCOS or that other factors may contribute to this association (8, 27, 28). Racial and ethnic disparities in PCOS phenotypes and in OSA diagnosis have also been reported, and may be related to genetic inheritance or health inequities or both (29).

To address these important gaps, we conducted the first systematic review and meta-analysis to comprehensively synthesize the prevalence and related symptoms of OSAS severity in PCOS compared with non-PCOS populations, with inclusion of peer-reviewed published studies that utilized the PSG tool. The methodology applied for this evidence synthesis was aligned with the 2023 International Evidence-based PCOS Guideline (1), which directly informed current recommendations for assessing OSAS in PCOS.

This is the first comprehensive systematic review and meta-analysis to analyze the presence of OSAS and related symptoms in PCOS compared with non-PCOS populations using PSG/polygraph with or without validated sleep screening questionnaires. Our pooled analyses of eight studies totaling 942 participants showed that, overall, 37.0% of individuals with PCOS had OSA (29.0% adolescents; 40.0% adults), while only 6.0% without PCOS had OSA (7.0% adolescents; 7.0% adults). Moreover, those with PCOS were approximately 10 times more likely to have OSAS than those without PCOS. The risk of OSAS tended to be higher in those who had overweight or obesity, adult women and of white ethnicity compared with normal weight, adolescent and Asian populations respectively, but was largely independent of PCOS diagnostic criteria.

Compared to previous meta-analyses, the present study showed relatively higher pooled OSA prevalence (37.0% vs. 20.8-35.0%) and higher risk of OSAS (OR: 9.52 vs. ORs: 2.86-8.30) in women with PCOS relative to without PCOS (14, 22, 26). Differences in study methodology likely explain the varied estimates of OSAS risk compared with previous meta-analyses. For example, He et al. (26) included studies that screened for OSA using the Berlin questionnaire, studies that utilized PSG to diagnosed OSAS as well as studies that did not specify OSAS measurement. Inclusion of studies with objective, subjective and unknown measurements of OSAS could have underestimated the overall OSAS risk (OR: 2.86), as demonstrated in the present meta-analysis, where we report a markedly higher risk of OSAS (OR: 9.52) when including only those articles that employed formal sleep studies to assess OSAS.

The present meta-analysis reported, for the first time, a ~5-fold higher effect estimate for AHI≥5 with symptoms compared to AHI≥5 only outcomes among individuals with PCOS (OR: 17.95 vs. OR: 3.90) (15, 17–21, 40). This indicates that sleep disturbances and associated symptoms may drive the inflammation, oxidative stress and increased sympathetic excitability, plausibly alter the regulation of gonadotropins and gonadotropin-releasing hormones, thus contributing to the expression of pathophysiological features of PCOS such as hyperandrogenism and insulin resistance in a bidirectional association (41). Although treatment with continuous positive airway pressure (CPAP) therapy has shown to improve nocturnal apnea, daytime sleepiness, as well as metabolic outcomes for those with OSA, evidence regarding the impact of CPAP use in PCOS populations is lacking. The long-term cardiometabolic health benefits for treating OSA with CPAP are also not clearly established from RCTs, but research from observational studies seems to suggest a benefit on health outcomes from treatment in general populations, which may extend to those with PCOS (42). With a high proportion of the general population being minimally symptomatic, it is considered optimal to target and treat those with symptoms of OSAS such as non-restorative sleep, sleep-related complaints, daytime fatigue and sleepiness or hypertension. In light of this, targeted screening approaches for symptoms of OSAS, using questionnaires for example, have been recommended for high prevalence populations that include PCOS (43).

Similar to previous meta-analyses (22, 23, 26), our subgroup analysis demonstrated that the overall prevalence of OSA was higher in adults with PCOS, but not in adolescents with PCOS. This is expected, as PCOS is likely to precede the onset of OSA given that OSA may develop over time as features of PCOS like hyperandrogenism worsen, predisposing to OSA via effects on neural control of breathing and upper airway mechanics (44). Regarding differences by diagnostic criteria, we found significantly higher risks of OSAS in PCOS overall compared with those without PCOS, regardless of diagnostic criteria (45). Previous research has shown mixed results, with one meta-analysis reporting a greater proportion of OSAS in PCOS based on NIH criteria (26), while another reported no such variations across PCOS definitions (23). Consistent with He et al. (26), our subgroup analysis also demonstrated that the overall prevalence of OSAS was higher in those of white ethnicity with PCOS, but not in those of Asian ethnicity with PCOS. Finally, BMI variations were also evident in our pooled analyses, whereby overall risk of having OSAS in PCOS compared with non-PCOS was greater in those populations who were overweight/obese (OR: 7.62) and mixed weight (i.e., studies with sample populations that range across normal, overweight and obese) (OR: 39.09) compared with the normal weight (OR: 0.57) population. This finding is consistent with prior meta-analyses which have reported a higher risk of OSAS in individuals with both PCOS and obesity compared with those without obesity (23, 26). It should be noted that the very wide 95% confidence interval in the mixed weight group (95% CI: 13.31 to 114.77) imply imprecision in the overall certainty of the effect estimate. The normal weight group in our analysis also consisted of only Yang et al. (40), with a very small sample size (n = 28) and absence of AHI ≥5 in all participants. Of note, the BMI classification as per the World Health Organization (WHO) guideline (35) may not apply to Asian populations given that these were generated mostly from the White populations (46). Further, only two studies in our meta-analyses were from Asian populations (18, 40), thus this result should be interpreted accordingly.

PCOS may also be an independent contributor to the risk of OSAS. Our sensitivity analysis that was limited to studies that matched or controlled for confounders including BMI demonstrated a higher risk of OSAS in those with PCOS compared to non-PCOS (OR: 5.67). This finding suggests that other mechanistic factors besides weight such as hyperinsulinemia, hyperandrogenism and inflammation may also influence the OSAS-PCOS relationship (26). Insulin resistance and hyperinsulinemia, either directly or indirectly, through factors such as adiposity can lead to increase circulating androgen levels (6). The inflammatory and oxidative stress responses from recurrent episodes of apnea, coupled with cycles of hypoxia and reoxygenation in those with OSAS can also exacerbate the above pathophysiological features underpinning PCOS (41). It should be noted that despite prior meta-analysis reporting worse metabolic parameters (e.g., insulin resistance) in those with PCOS and OSAS compared to without OSAS (26), other common mechanistic factors like hyperandrogenism and insulin resistance linking PCOS and OSAS were not invariably measured independently of confounders like BMI. Many of their included studies were not included in our meta-analysis due to varying exclusion criteria (e.g., lack of PSG data) (16, 21). Further research is needed to understand the associations between pathophysiological risk factors of PCOS and the onset and progression of OSAS, including potential interactions with obesity.

To our knowledge, this meta-analysis is the first to comprehensively examine the prevalence and severity of OSAS and its related symptoms in PCOS. We included peer-reviewed published studies that utilized the PSG/polygraph tool along with related symptom assessments to ensure OSAS severity was well-defined and to minimize outcome detection bias. We followed rigorous, internationally endorsed methodology, guided by experts in the field as part of the 2023 PCOS Guideline update. We also assessed the certainty of the evidence at the outcome-level using the validated GRADE framework (33). Except for the composite OSA outcome where certainty of evidence was low, other symptom categories of OSAS outcomes (i.e., AHI≥5 only, AHI≥5 with symptoms, and AHI≥10 with symptoms) were deemed to have moderate certainty evidence. Similarly, publication bias assessments and rigorous quality and sensitivity analyses were conducted to enhance accuracy and minimize potential over- or under-estimation of results.

The main limitation of this review is that studies had a moderate to high risk of bias due to detection and selection biases. For example, some participants with PCOS were recruited from specialized clinics while healthy controls were drawn separately from the general population. The relatively small pooled sample sizes in this review (n = 280 PCOS and n = 662 non-PCOS) could have reduced the statistical power and generalizability of the results. Other limitations include, variable adjustment for important confounders such as age, BMI and ethnicity (16, 19, 20), the heterogenous BMI distribution of the study population, the varied definitions and assessment methods for OSA across studies that can lead to inconsistencies in reported outcomes. In addition, under-representation of ethnic and geographic groups (e.g., Asian or mixed-ethnicity) can restrict the generalizability of the findings. Finally, the heterogeneity in OSAS prevalence estimates (e.g., I² = 96.8% for OSAS in PCOS) signify variability population characteristics. Hence, the findings should be interpreted in view of these points. Nevertheless, the impact of the included studies is likely to be minimal, since the exclusion of these studies in sensitivity analysis did not influence the overall meta-analysis results. Due to the limited studies, single-paper analyses as well as subgroup analyses where only one study belonged to the comparison group should be interpreted with caution as this may lead to chance findings. Studies exploring PCOS population subgroups with varying symptomatic OSAS profiles are warranted, as are studies exploring ethnically diverse and adolescent populations, to enhance generalization to the global PCOS population. These are crucial evidence gaps as morbidity of OSAS is related to age and racial differences in both clinical and metabolic profiles of PCOS.

The present systematic review and meta-analysis found that, independent of PCOS diagnostic criteria, the overall risk of OSAS is markedly higher in PCOS compared with non-PCOS populations, especially in adult women, those with overweight/obesity and of white ethnicity, and was more pronounced with increasing symptomatic OSAS severity. Using the methodology from the 2023 International Evidence-based PCOS Guideline, our findings recommend that women with PCOS be screened for symptoms of OSAS using validated questionnaires combined with a thorough sleep apnea history to target OSA treatment to those more likely to benefit (1). Recognizing the heightened risk of OSAS in PCOS will aid in identifying those most susceptible, to enable the development of more effective and targeted strategies for OSAS treatment.