Obstructive sleep apnea and polycystic ovary syndrome: Clinical interactions and underlying pathophysiology

Introduction

In 1935, Dr. Stein and Dr. Leventhal first described the polycystic ovary syndrome (PCOS; Stein-Leventhal syndrome) as the etiology for irregular ovulation in young female patients presenting with hirsutism, amenorrhea, and enlarged ovaries with multiple cysts . PCOS is now a well-established syndrome in women of reproductive age, which currently represents the most frequent chronic endocrine disorder in this population, with reported prevalence rates that range from 4% to 15% or even up to 20%, depending on the applied PCOS diagnostic criteria and the demographic characteristics (e.g., age and ethnicity) of the studied cohort/population . In most women with PCOS, the initial diagnosis is made within their early reproductive years; although this may vary, with a later diagnosis not being uncommon, depending on the presenting symptoms and the predominant manifestations of the syndrome over time . Notably, the precise natural history of PCOS is still being studied, since there is a paucity of data regarding the hormonal, cardiometabolic, and anthropometric profile of women with PCOS following the menopause . Overall, the phenotypic expression and clinical manifestations of PCOS are characterized by substantial variability (hyperandrogenemia is present in 60%–80% of women with PCOS, with hirsutism typically constituting the most frequent clinical manifestation), which can vary over time and may also be linked to long-term health risks .

Moreover, PCOS is closely associated with multiple comorbidities that are part of the metabolic syndrome spectrum, including obesity (particularly central/visceral/abdominal/android obesity), insulin resistance, and type 2 diabetes mellitus (T2DM) . In clinical practice and based on consistent epidemiologic data, a substantial proportion (if not the majority) of women with PCOS are also noted to have a body mass index (BMI) in the overweight or obesity range (i.e., BMI > 25 kg/m ² or BMI > 30 kg/m ² for women of Caucasian origin, respectively), with reported rates of 30% to even more than 75%, depending on the applied diagnostic criteria of PCOS, the demographic characteristics of the studied cohort/population and the cause of clinical referral . Of note, PCOS appears to exhibit a positive association to central obesity with android body fat distribution even within the normal range of BMI (BMI of 18.5–24.9 kg/m ² for women of Caucasian origin), with increased central/visceral adiposity being present in approximately 50%–70% of women with PCOS, independently of their BMI . Furthermore, this phenotype exhibits a closer positive association with insulin resistance and hyperandrogenism compared to both women without PCOS and women with PCOS that have a predominantly peripheral body fat distribution, regardless of BMI . However, it is also noteworthy that a study that analyzed data from axial magnetic resonance imaging (MRI) scans regarding the adipose tissue distribution in women with PCOS and BMI/fat mass-matched controls (all of UK British/Irish origin) showed no significant differences in fat-depot measurements (visceral, abdominal, and gluteofemoral subcutaneous depots), suggesting that global adiposity rather than abnormal regional adipose tissue distribution characterizes these women . Moreover, although it is typically present in women with PCOS and obesity, insulin resistance can also be detected in lean (BMI within the normal range) women with PCOS . Overall, among the different phenotypes of PCOS, the one that meets the NIH PCOS diagnostic criteria appears to be linked to a more adverse cardiometabolic profile regarding central obesity, insulin resistance, and other risk factors for cardiovascular disease (CVD) . In this context, it is evident that PCOS and metabolic syndrome have a strong overlap in their manifestations and underlying pathophysiology, which appear to be driven by both central obesity and insulin resistance. Thus, additional cardiometabolic disorders that are closely related to the metabolic syndrome (despite not being included in its diagnostic criteria) , such as nonalcoholic-associated fatty liver disease (NAFLD or metabolic-associated fatty liver disease, MAFLD) and obstructive sleep apnea (OSA), may frequently coexist in women with PCOS, particularly in those with central obesity .

OSA is also a common chronic respiratory disorder in the general adult population, particularly in men, but also in women with obesity, with estimated prevalence rates of over 50% in certain countries . Indeed, approximately one billion people are considered to be affected by OSA worldwide, with estimated prevalence rates in the general population of 17%–26% and 9%–28% in men and women, respectively, depending on the demographic characteristics of the studied cohort/population, as well as the criteria and methods applied to diagnose OSA . Recurrent episodes of temporary airflow reduction (hypopnea) or cessation (apnea) during sleep are the typical characteristics of OSA, due to, respectively, partial or total upper airway collapse, which results in decreased oxygen saturation with repeated episodes of hypercapnia and hypoxemia . The obstructed airflow is restored upon arousal from sleep; hence, the normal sleep architecture/pattern is typically disrupted [e.g., loss of rapid eye movement (REM) sleep and/or deep sleep stages (stages 3 and 4)] with a negative impact on the overall sleep quality and a spectrum of additional clinical manifestations . The latter range from choking episodes during sleep, snoring, and nocturia to daytime hyper-somnolence, irritability, personality changes, restlessness, depression, and impaired concentration, as well as hypertension, decreased libido, impaired quality of life, and even increased risk for motor vehicle accidents .

OSA screening is relatively simple and can be performed by using well-established, validated questionnaires, such as the Berlin Questionnaire and the Epworth Sleepiness Scale, which are helpful for the initial OSA risk assessment in routine clinical practice, particularly in patients with obesity and other diseases frequently associated with obesity (e.g., in women with PCOS and obesity) . The gold standard diagnostic method for OSA is polysomnography, while home-based sleep tests can also be used for its diagnosis and severity assessment . An episode of apnea is defined as airflow cessation for a minimum of 10 s (classified as central or obstructive based on the absence or presence of respiratory effort, respectively), while hypopnea is defined based on either reduced airflow of ≥ 30% from baseline for at least 10 s with ≥ 4% desaturation from baseline or reduced airflow by ≥ 50% for at least 10 s with ≥ 3% desaturation or an arousal from sleep . Accordingly, the severity of OSA is usually defined based on the number of apnea and hypopnea episodes per hour of documented sleep, which constitute the apnea-hypopnea index (AHI—mild OSA: AHI of 5 to 15; moderate OSA: AHI of > 15 to 30; severe OSA: AHI of > 30) . Of note, the utilization of additional indices has also been suggested for assessing the severity of OSA (e.g., OSA severity based on hypoxemia), since the AHI does not always directly correspond to the OSA severity and the manifested clinical symptoms . The timely diagnosis and subsequent effective treatment of OSA is of paramount importance in the clinical practice, since undiagnosed or untreated OSA not only causes the numerous aforementioned symptoms and problems but is also further linked to significantly higher risk of CVD morbidity and mortality . Interestingly, alterations in the central control of breathing constitute an additional long-term complication of undiagnosed/untreated OSA, with central apnea episodes due to progressive desensitization of respiratory centers to hypercapnia. Although initially limited during sleep, these episodes can progressively lead to the obesity hypoventilation syndrome, which is characterized by obesity, hyper-somnolence, sleep disordered breathing, alveolar hypoventilation, chronic hypercapnia, and hypoxia, as well as right ventricular heart failure and polycythaemia . As such, prompt and sustained effective treatment is essential, not only to address the OSA-related symptoms but also to further prevent its long-term complications/risks. To date, in addition to weight loss for patients with BMI > 25 kg/m ² (e.g., lifestyle, pharmacological, or even bariatric surgery interventions for weight loss in women with PCOS and obesity), continuous positive airway pressure (CPAP) therapy constitutes the standard treatment of choice for symptomatic OSA, which is highly efficacious at preventing OSA .

Despite the aforementioned global burden and high prevalence of OSA in the general adult population, this chronic condition still remains, to a significant degree, underdiagnosed and under-reported in routine clinical practice, particularly in patient groups where there is low clinical suspicion and awareness regarding the risk of OSA, such as female patients . However, as obesity (particularly central) is increasingly recognized as a key mediating factor in the pathophysiology of both OSA and PCOS, heightened clinical suspicion is required to screen for and diagnose the coexistence of OSA in women with PCOS . Accordingly, relevant position statements and evidence-based clinical practice guidelines and recommendations for the management of women with PCOS acknowledge this link between PCOS and OSA, suggesting prompt screening for OSA in women with PCOS under certain conditions (e.g., in the presence of obesity or symptoms related to OSA, such as snoring, daytime sleepiness, and waking unrefreshed from sleep) ( Table 1 ) . Notably, what is also highlighted in these position statements and clinical guidelines/recommendations is the relatively limited evidence to inform clinical practice regarding the effective screening approaches, diagnosis tools, and management interventions for OSA in the context of PCOS. As such, in this chapter, we present the findings from pertinent studies that have explored the links and interactions between PCOS and OSA, as well as a concise overview of the potential mechanisms that appear to be implicated in the underlying pathophysiology linking these two common conditions.

Prevalence of OSA in women with PCOS

In regard to the prevalence of OSA in women, it is important to highlight again that based on the existing epidemiological evidence, there is considerable variation in the reported and estimated prevalence rates of OSA in the general adult population (men and women aged over 18 years) . This is mainly attributed to the significant methodological heterogeneity that is noted in the relevant population prevalence studies [e.g., in addition to the AHI, various other indices were utilized in such studies to assess the outcome of sleep studies for OSA, including the apnea index (AI), the oxygen desaturation index (ODI), or the respiratory disturbance index (RDI)], as well as to the differences in key characteristics of the studied cohorts/populations . Based on systematic review data, the prevalence of OSA (defined as ≥ 5 AHI/RDI) in adult women of the general population ranges from 6% to 19%, exhibiting higher prevalence rates with increasing age and BMI . This is also in accord with the reported data from individual, well-conducted studies, which had similar design/methodology and documented prevalence rates of OSA between 9% and 28% in adult women (corresponding rates of 2%–7% for moderate-to-severe OSA with AHI ≥ 15) .

This substantial variability in the reported prevalence rates of OSA noted in the general population is further heightened in relationship to PCOS, particularly given the relatively limited number of studies that have explored the presence of OSA in this specific female patient population . For example, even when considering only published cross-sectional studies that used polysomnography as the method for OSA diagnosis, the noted prevalence rates of OSA in female patients with a diagnosis of PCOS based on the 2003 ESHRE Rotterdam criteria or the NIH criteria range from 0% to approximately 72% ( Table 2 ) . Indeed, a meta-analysis from our group that included these studies, as well as additional relevant studies which published their results as conference abstracts, highlighted this large variability in the identified prevalence rates of OSA in women with PCOS . This finding can be mostly attributed to a combination of factors that primarily relate to the differences in the methods and cutoff points applied to diagnose OSA, the age/age group (i.e., adults, adolescents, or mixed age group) of the studied women, as well as their ethnicity and BMI . In addition, the small sample size of these studies, and a frequently high selection bias due to recruitment of participants from specialized healthcare services/clinics, appears to also contribute to the wide range of the noted prevalence rates of OSA in women with PCOS . Overall, our meta-analysis showed that more than one in three (36.1%; 95% confidence interval: 22.4%–51.0%) of women with PCOS in relevant studies had OSA, with this prevalence not being significantly affected by variations in the applied definition of PCOS (e.g., PCOS diagnosis based on the 2003 ESHRE Rotterdam or NIH criteria) between the studies that were meta-analyzed . As expected, the findings of our meta-analysis also confirmed the higher prevalence of OSA in women with PCOS and obesity compared to lean PCOS women, while women with PCOS also exhibited higher prevalence of OSA compared to weight-matched controls. Interestingly, approximately one-fifth of the documented variation in the noted prevalence of OSA was related to the difference regarding the age groups (i.e., adults, adolescents, or mixed age groups) that were recruited in the analyzed studies . Of note, most of the relevant studies were conducted in the USA and included participants with class 2 obesity (BMI of 35 to < 40 kg/m ² for women of Caucasian origin), were at high risk of selection bias, and did not account for key confounding factors .

Particularly in relation to the prevalence rates of OSA in adolescents with PCOS, our meta-analysis also noted an expected trend for higher prevalence rates of OSA in adult women with PCOS compared to adolescents. This is also in line with the data on the global burden of OSA in the general population, where the prevalence of OSA rises with increasing age . However, the interpretation of the existing evidence on the prevalence rates of OSA in adolescents with PCOS is even more difficult, given the marked paucity and variability of the relevant data, with a study by Nandalike et al. noting an OSA prevalence of 57% (16 out of 28 participants) in adolescents with PCOS , while another study by de Sousa et al. identified no cases of OSA in 35 adolescents with PCOS ( Table 2 ) . In addition, this lack of sufficient relevant data also highlights the existing knowledge gaps regarding the temporal association of these two conditions. As such, it is plausible that the clinical manifestations of PCOS precede the development of OSA, but cases where the onset of OSA may predate that of PCOS symptoms/manifestations cannot be excluded . Similarly, only scarce data exist regarding the prevalence of OSA in lean women with PCOS. Based on these existing data, a low prevalence of OSA appears to exist among lean women with PCOS, but this is based on small sample size studies, which cannot support any definite conclusions on this research question . Therefore, it is clearly evident that there is still a need for large and well-designed epidemiological studies, which will precisely assess the incidence of OSA among different age groups and phenotypes of women with PCOS and vice versa. Indeed, regarding the latter, the incidence and prevalence of PCOS in women with a primary diagnosis of OSA remains to be explored.

Impact of OSA in women with PCOS

Our current knowledge on the potential impact of the presence of OSA in women with PCOS is also based on a small number of relevant studies. As such, preliminary insight on the potential impact of OSA on the hormonal and cardiometabolic profile of women with PCOS can be gained by the differences in such characteristics between PCOS women with and without OSA and sleep disordered breathing that were noted in pertinent cross-sectional studies ( Table 3 ). Notably, in the study by Chatterjee et al. in South Asian women, the participants with both PCOS and sleep disordered breathing ( N = 33) exhibited significantly higher BMI (29.8 ± 3.4 versus 24.4 ± 2.29 kg/m ² , p -value < 0.001), waist circumference (WC: 95.6 ± 6.47 versus 85.1 ± 4.34, p -value < 0.001), systolic blood pressure (SBP: 129.3 ± 10.93 versus 119.2 ± 8.03 mmHg, p -value: 0.002), and diastolic BP (DBP: 78.6 ± 9.07 versus 73.5 ± 6.22 mmHg, p -value: 0.044), as well as hirsutism (Ferriman-Gallwey score: 9.8 ± 2.78 versus 8.0 ± 2.50, p -value: 0.028) compared to those with PCOS but without sleep disordered breathing ( N = 17) . Furthermore, following adjustment for BMI in a logistic regression analysis, only the associations of fasting plasma glucose and DBP with sleep disordered breathing remained significant in this study.

Moreover, in a study by Tasali et al. in the USA (62% of the participants were of African American or Hispanic origin), women with both PCOS and OSA ( N = 29) had a higher BMI (42.2 ± 1.1 versus 35.3 ± 1.4 kg/m ² , p -value < 0.001) and were both older (age: 31.6 ± 1.0 versus 27.3 ± 0.7 years, p -value: 0.002) and more insulin resistant based on the homeostatic model assessment of insulin resistance (HOMA-IR: 5.7 ± 0.4 versus 3.5 ± 0.4, p -value: 0.006) compared to women with PCOS without OSA ( N = 23) . In this study, following adjustment for age, ethnicity, and BMI, the AHI was a significant predictor for HOMA-IR and the blood glucose levels 2 h after an oral glucose tolerance test, as well as for the fasting plasma levels of both glucose and insulin. Based on the findings of this study, it also appears that the severity of glucose intolerance and insulin resistance in women with PCOS may be related mostly to the degree of sleep fragmentation, rather than the severity of the underlying hypoxemia. However, the small sample size of this study, as well as the relatively large BMI difference (7.1 kg/m ² ) between study participants with and without OSA, cannot support firm conclusions about whether these women with PCOS and OSA represent a truly different subgroup of patients at higher metabolic risk compared to those without OSA. Interestingly, in a short-term (8 weeks) interventional study by Tasali et al. that explored the effects of home CPAP treatment with daily usage monitoring in 19 young women with PCOS, OSA, and obesity (age ± standard error of the mean: 31.2 ± 1.2 years; BMI: 46.4 ± 2.4 kg/m ² ), insulin sensitivity improved (approximately 7% relative increase), while DBP decreased by approximately 2.3 mmHg . Furthermore, following this 8-week CPAP treatment, a reduction in both nighttime and daytime norepinephrine plasma levels was also noted. Whether these improvements following this short-term CPAP treatment may subsequently confer clinically meaningful benefits requires additional long-term studies with large sample sizes, since in this pilot study there was no control group, while only a per protocol analysis was reported that included only nine of the study participants (10 of the 19 study patients were excluded from the performed analysis due to inadequate compliance to the CPAP treatment based on average use of CPAP for < 4 h per night) .

In another study in the USA by Vgontzas et al., women with PCOS and sleep disordered breathing (defined as OSA or upper airway resistance syndrome; N = 9) had higher BMI (45.7 ± 2.6 versus 37.2 ± 1.1 kg/m ² , p -value < 0.003) and higher fasting insulin plasma levels (306.5 ± 52.4 versus 176.1 ± 18.5 pmol/L, p -value < 0.01), as well as lower ratio of glucose-to-insulin (0.02 ± 0.006 versus 0.04 ± 0.003, p -value < 0.05) compared to women with PCOS without sleep disordered breathing ( N = 44) . In this study, logistic regression analysis revealed that the ratio of glucose-to-insulin and the fasting insulin plasma levels exhibited a stronger association with sleep disordered breathing than age, BMI, or free and total testosterone plasma levels. However, an effect of obesity on the noted metabolic differences between the two groups of this study cannot be conclusively ruled out, due to the relatively large BMI difference (8.5 kg/m ² ) between these two groups, despite the performed relevant statistical adjustment.

Similarly, a worse hormonal and cardiometabolic profile of women with both PCOS and OSA was noted in another study by Tock et al. in Brazil , where these participants ( N = 12) had higher BMI (37.8 ± 4.8 versus 30.7 ± 7.7 kg/m ² , p -value: 0.006), waist-to-hip ratio (WHR: 1.0 ± 0.1 versus 0.9 ± 0.1, p -value: 0.029), WC (114.4 ± 12.0 versus 98.1 ± 19.9 cm, p -value: 0.013), HOMA-IR (4.4 ± 3.2 versus 2.3 ± 1.4, p -value: 0.009), total cholesterol (205.0 ± 28.7 versus 172.3 ± 35.8 mg/dL, p -value: 0.009), low density lipoprotein-cholesterol (LDL-C: 128.6 ± 21.6 versus 98.9 ± 29.6, p -value: 0.004), and free testosterone plasma levels (1.9 ± 1.3 versus 1.1 ± 0.8 ng/dL, p -value: 0.014), as well as higher prevalence of NAFLD (83.3% versus 26.9%, p -value: 0.001) compared to those without OSA ( N = 26). Of note, following adjustment for obesity in the multivariate logistic regression analysis of this study, raised free testosterone serum levels ≥ 1.07 ng/dL increased the risk of OSA in women with PCOS by 8.2 times, suggesting that hyperandrogenemia may be a predisposing risk factor for OSA in women with PCOS. Interestingly, a multiple logistic regression analysis in this study, which included OSA (AHI ≥ 5), insulin resistance (HOMA-IR ≥ 2.7) and obesity (BMI ≥ 30 kg/m ² ) as the independent variables and NAFLD as the dependent variable, showed that only OSA was an independent predictor for the presence of NAFLD. Indeed, the presence of OSA in women with PCOS increased the odds of NAFLD by 7.63 times, suggesting that OSA is a predictor of NAFLD independently of obesity and insulin resistance .

Finally, the retrospective study in adolescent girls with PCOS by Nandalike et al. also showed that those with OSA ( N = 16) exhibited higher SBP (128.4 ± 12.8 versus 115.6 ± 11.4 mmHg, p -value: 0.009), higher triglyceride plasma levels (149.7 ± 87.7 versus 93.3 ± 25.8 mg/dL, p -value: 0.03), and lower high-density lipoprotein plasma levels (HDL: 38.6 ± 8.7 versus 49.0 ± 10.9 mg/dL, p -value: 0.01), as well as higher prevalence rates of both insulin resistance (HOMA-IR > 4: 81.3% versus 41.6%, p -value: 0.03) and metabolic syndrome (56.3% versus 8.3%, p -value: 0.03), compared to those without OSA ( N = 12) .

Despite the low number and the various methodological limitations of the existing relevant studies, when taken together the available data on the impact of OSA in women with PCOS suggest that OSA and the related sleep disordered breathing may exhibit a positive association with a more severe PCOS phenotype, particularly in respect to the metabolic profile and the presence of increased BMI (overweight or obesity) and insulin resistance. This is also supported by a systematic review from our group, which included a meta-analysis that showed that women with both PCOS and OSA had significantly higher BMI (mean difference: 6.01 kg/m ² , 95% confidence interval: 4.69–7.33), WC (mean difference: 10.93 cm, 95% confidence interval: 8.03–13.83), insulin resistance, SBP and DBP, as well as impaired glucose regulation and a worse lipidemic profile compared to women with PCOS without OSA . Our systematic review of the available studies also showed that free and total testosterone plasma levels were not significantly different between these two groups; however, a meta-analysis was not feasible for this specific outcome due to the different methods and units used to, respectively, measure and report testosterone plasma levels. Of note, most of the existing studies that were eligible for this systematic review were only from one country (i.e., the USA), lacked accounting for key confounders and exhibited a high risk of selection bias . Overall, OSA appears associated with obesity and a worse cardiometabolic profile in women with PCOS, but it remains unclear whether these effects are truly independent of obesity. Indeed, the lack of adjustment for significant pertinent variables (e.g., accounting for ethnicity and central/abdominal adiposity) prohibits definite conclusions based on these existing studies. This is even more apparent regarding the potential impact of OSA on the hormonal profile (e.g., the potential associations with hyperandrogenemia) in women with PCOS, where additional methodological parameters (e.g., differences in the measurements of testosterone plasma levels) may also play a confounding role. Therefore, further to the aforementioned additional research needed to clarify the prevalence of OSA in PCOS, well-designed studies are also clearly required to delineate the effects of OSA on the hormonal and cardiometabolic profile of women with PCOS. To that aim, interventional studies (e.g., studies on long-term CPAP therapy in women with PCOS and OSA) can be also designed to study these potential effects and further explore whether ameliorating (if not eliminating) these through effective OSA treatment can result in long-term clinically significant outcomes in these women.