Our findings in this study indicate that women with PCOS are at high risk of OSA based on the Berlin questionnaire but the increased risk is only present amongst the obese women. Non-obese women with PCOS do not seem to be at increased risk of OSA. A negative Berlin questionnaire is highly predictive in ruling out OSA so the fact that the risk for sleep apnea in non-obese women in both PCOS and control was 0% strongly suggests that the prevalence of OSA in non-obese women is very low even if they have PCOS. Furthermore, the risk amongst obese women is similar for PCOS and control women though this finding is limited due to the small number of obese control women included in the study. As expected, women with PCOS had higher bioavailable testosterone levels independent of obesity. However, bioavailable testosterone was not an independent predictor of OSA risk. BMI was the only independent predictor of OSA risk for the entire group of women consisting of PCOS and control as well as amongst women with PCOS alone.
Very high rates of OSA have been reported in women with PCOS in a number of studies (2
). These studies have included mostly obese middle-aged women (2
), and in some studies women with severe obesity (4
). In a study from Taiwan that included only non-obese women (29
), women with PCOS had a higher apnea-hypopnea index (AHI) compared to control women (0.79±0.21 vs. 0.29 ± .09, P=0.041). However, the diagnosis of OSA requires an AHI above 5 and the mean AHI for women with PCOS in this study was much lower than 5 and therefore none of the women met the clinical criteria for OSA. In a recent study from Germany, De Sousa and colleagues studied 31 mildly obese adolescents with PCOS. None of the mildly obese adolescents had OSA and the mean AHI was 0.95 (30
). In another study, among 12 women with PCOS who had a BMI <32.3 kg/m2
only 1 (8.3%) had OSA which was much lower than among the obese PCOS women with BMI >32.3 kg/m2
). Taken together, these findings suggest that as women with PCOS age and gain weight their risk of developing OSA increases.
Previous studies have also indicated that obese women with PCOS who have OSA are more insulin resistant compared to obese women with PCOS who do not have OSA (4
). Some have argued that the presence of OSA identifies those women at risk for metabolic abnormalities (6
). In other studies, fasting plasma insulin was the strongest risk factor for OSA in women with PCOS (3
). We did not find a difference in fasting insulin levels or HOMAIR between obese women with PCOS who screened positive for OSA and those who screened negative for OSA. Furthermore, presence of OSA risk in women with PCOS was not a predictor of HOMAIR. The only predictor of OSA among the entire cohort consisting of women with PCOS and control as well PCOS women alone was BMI. Our findings differ from a previous study that included only obese women with PCOS in which PCOS was an independent predictor of OSA even after adjusting for BMI (5
). It may be that compared to the previous study (5
), our sample included a wider range of BMI with both non-obese and obese women.
A main limitation of our study is the lack of polysomnographic data to confirm the actual presence of OSA. Polysomnography is the gold standard for diagnosis of OSA. However, Berlin questionnaire has a good positive and negative predictive value for detection of OSA in both men and women compared to the diagnosis based on AHI obtained from polysomnography (27
). Clinic-based studies have suggested that women with OSA may have a different clinical presentation compared to men, particularly a lower proportion of women complaining of hypersomnolence and loud snoring (31
). It remains unclear whether these differences represent reporting bias or a difference in disease expression. It is possible that mild cases of OSA were missed by the Berlin questionnaire leading to underestimation of the actual prevalence of OSA in non-obese women with PCOS. If this was true, then indeed the Berlin questionnaire would be a less sensitive tool for assessing the risk of OSA in women compared to men. However, this sex difference in disease presentation did not seem to affect the accuracy of risk groupings when the Berlin questionnaire was originally validated because questions about fatigue were incorporated in the Berlin questionnaire and fatigue is more commonly reported in women with OSA (27
). It is important to note that clinic-based studies that have reported gender differences in OSA symptoms have inherent limitations due to selection bias. Several large community-based epidemiologic studies have shown that there is no gender difference in symptoms and presentation of OSA (35
). Furthermore, a recent systematic review of various screening tools for OSA reported that the Berlin questionnaire had the highest sensitivity and specificity for predicting the presence of OSA (AHI > 5) (37
Another limitation of our study is the small sample size of obese control women though even in the small sample the risk for OSA was not significantly different compared to the obese women with PCOS. It is plausible that with a larger sample size the differences in the risk of OSA between obese PCOS and obese control women may have become significant. The risk of OSA among the obese control women in our study was higher than the actual rate of OSA reported in severely obese women. In a community-based study, Vgontzas and colleagues reported that out of 194 women with BMI’s similar to our obese controls (> 40 kg/m2
) only 11 had an elevated AHI consistent with OSA (38
). Therefore, the Berlin questionnaire may have been oversensitive in our cohort in categorizing OSA risk and if polysomnograms were available the actual prevalence of OSA could have been lower in our severely obese control women. The prevalence of OSA amongst obese or severely obese women with PCOS has been reported to be significantly higher than in severely obese non-PCOS. It may be that PCOS and obesity synergistically increase the risk of OSA. For example, it is known that only obese women with PCOS have hepatic insulin resistance and hypertriglyceridemia and that obesity and PCOS act synergistically to lead to these abnormalities (39
). Furthermore, women with PCOS are centrally obese (39
) and central obesity confers a higher risk for OSA than generalized obesity (11
). Notwithstanding our limitations, we believe that a strength of our study is having a relatively large cohort of non-obese PCOS women compared to prior studies.
In summary, significant number of reproductive age women with PCOS screen positive for OSA based on the Berlin questionnaire, but the increase in risk appears to be related to the high prevalence of obesity is this population. Whether the risk also applies to non-obese women with this condition is questioned and requires further studies using polysomnography.