Female sex hormones and symptoms of obstructive sleep apnea in European women of a population-based cohort.

Obstructive sleep apnea (OSA) is characterized by repeated cessations of breathing, caused by upper airway obstruction [ 1 , 2 ]. Apart from substantially lowering quality of life through poor sleep, OSA may increase the production of free radicals with corresponding oxidative stress and have a widespread negative impact on the body, particularly the cardiovascular system [ 3 , 4 ]. OSA is more common in men than in women and epidemiological studies suggest the ratio to be around 2:1 male to female [ 5 – 7 ]. However, the prevalence and severity of OSA increases among women after menopause [ 8 , 9 ]. The prevalence of sleep apnea in postmenopausal women not using hormone replacement therapy is significantly lower than in men after adjusting for age and BMI, namely 2.7% versus 3.9% using the Sleep Disorders Clinic criteria and 5.5% versus 7.2%, using the sleep laboratory criteria [ 10 ]. The decreasing sex difference in OSA prevalence with increasing age [ 11 ] suggests that the hormonal shift attributed to menopause might play a role in the pathology of OSA. This hypothesis is supported by the finding that women with polycystic ovary syndrome, a disease characterized by comparatively low levels of female sex hormones, are at greater risk for developing sleep apnea [ 12 ]. Therefore the main female sex hormone 17β-estradiol, likely plays a role in developing OSA, possibly through its antioxidant properties [ 13 , 14 ]. Additionally, 17β-estradiol receptors are associated with the development of the upper respiratory tract musculature [ 15 , 16 ] and progesterone acts as respiratory stimulant. Experiments in rodents indicate progesterone to exert a protective function against OSA through its role in the hypoxia reflex [ 14 ]. Although all these observations are suggestive for a potential impact of female sex hormones on the risk of OSA, to date there is no population-based or large clinical study available; only a handful of smaller studies have reported lower levels of endogenous female sex hormones [ 17 – 19 ] in women with OSA, and there is limited evidence regarding a possible beneficial effect of exogenous sex hormones (e.g. hormone replacement therapy) [ 9 ]. With the aging of the general population the time women expect to live after menopause is becoming an ever-increasing portion of their total lifespan. Considering that OSA may lead to cardiovascular conditions (ischemic heart disease and stroke) [ 20 ], which are among the greatest contributors to mortality worldwide [ 21 ], gaining a broader understanding of the reasons for the increased propensity of postmenopausal women to develop OSA is important. Therefore, we aimed to investigate the association between serum concentration of female sex hormones and sleep apnea symptoms in a large population-based cohort including well-characterized subjects and high-sensitivity hormone measurements.

Anthropometric characteristics of the study population are presented in Table 1 . Out of the 774 women, 551 women reported they had been told they snored, of whom 411 responded positively to at least one of the three OSA sub questions ( Fig 2 ). Characteristics of the study population and by reported snoring are presented in Table 2 . 1 Lower limit of quantification (LLQ)–Upper limit of quantification (ULQ). n/a: Not applicable. N (SD) for continuous variables, N (%) for categorical variables and average [interquartile range] for measured serum concentrations. Among the included study population, a doubling of the concentrations of estrone and progesterone in the bloodstream was associated with 19% respectively 9% decreased odds of snoring. Among women who were told to be snorers, a doubling of the concentrations of 17β-estradiol, estrone and estrone 3-sulfate was associated with 18%, 23% and 17% decreased odds of being told that they stop breathing or breathed irregularly while being asleep (irregular breathing). A doubling of progesterone concentrations was further associated with a 12% decreased chance of waking up suddenly with a choking sensation or not being able to breathe (gasping). Other evaluated associations were not statistically significant ( Fig 3 ). The quartile analyses confirmed the linearity of the associations and while the included sex hormones highly correlate with each other (pairwise correlations: p<0.0001) mutual adjustment resulted in very similar findings (see Table 3 ) as seen for the main analyses. This was also the case when we applied multilevel model with subjects nested into centers. The sensitivity analyses on a) the subset of women who reported to be married and/or cohabiting, and b) additional adjustment for the frequency of consumption of beer, wine and spirits yielded very similar results as the main model ( S1 and S2 Figs). The results of the evaluation of the dynamics using sex hormone ratios showed significant associations between higher ratios of 17β-estradiol to estrone and 17β-estradiol to progesterone with lower odds of irregular breathing. ( S3 Fig ). A higher ratio of Estrone 3-sulfate to Estrone was associated with lower odds of gasping and a higher ratio of Estrone 3-sulfate to progesterone was associated with higher odds of gasping and disturbing snore.

We observed that in middle-aged women lower levels of female sex hormone were associated with higher risk of suffering from OSA symptoms. It is crucial to develop strategies to decrease the high prevalence and associated morbidity of OSA and adjusting female sex hormones levels might be the key to accomplish this, but further longitudinal studies with repeated measures of sex hormones and objective characterization of OSA are required to confirm our findings in other settings.

The analyses were based on 774 complete case observations from the second follow-up (2010–2012) of the European Community Respiratory Health Survey (ECRHS, response rate 49%). Examinations consisted of a structured interviewer-led questionnaire on respiratory health and lifestyle factors; a validated sleep questionnaire [ 22 ] and an in-clinic examination where blood samples for hormone analyses were drawn and anthropometric measurements taken. In addition, the female study participants were asked to complete a questionnaire on women’s health. In order to form hormonally well-defined groups we excluded women answering affirmatively to any of the following questions, indicating gynecological conditions: 1)“Have you ever had excessive growth of body hair?”; 2)“Has a doctor or health professional ever told you have polycystic ovaries or polycystic ovary syndrome?”; and 3)“Has a doctor or health professional ever told you have endometriosis”. We further excluded all women taking hormone replacement therapy, hormonal contraception or other exogenous hormone treatment and/or modulators of the genital system for menopausal symptoms, fertility treatment, gynecological disorders or undisclosed reasons. Additional exclusion criteria were current pregnancy, lactation and persistent irregular menstruation, independent of the menopausal transition. We carried out complete case analyses, therefore women with missing data on hormone levels or covariates were also excluded ( Fig 1 ). Data were missing at random. A comparison between the excluded and included study population is presented in S1 Table . The distribution of snorers over traditionally used menopausal categories can be found in S2 Table . Ethical approval was obtained from the appropriate ethical committee of each study center and all participants provided informed written consent before being enrolled in the study. The specific committees were for Aarhus: Scientific ethical committee for Region Midtlylland; Albacete: Comité Ético de Investigación Clínica del Hospital Universitario de Albacete; Galdakao: Comité Ético de Investigación Clínica del Hospital de Galdakao-Usansolo; Huelva: Comisión de Ética de Investigación Sanitarias del Hospital Juan Ramón Jiménez de Huelva; Bergen: Regional Committee for Medical and Health Research Ethics in Western Norway (2010/759); Bordeaux, Grenoble, Montpellier, Paris: Comite De Protection Des Personnes (2011-A00013-38); Erfurt, Hamburg: Ethikkommission der Bayerischen Landesärztekammer (Reg Nr. 10015); Uppsala, Umeå, Göteborg: Ethics Committee of Uppsala University (2010/068); Reykjavik: National Bioethics Committee of Iceland (VSN 11–121); Tartu: Research Ethics Committee of the University of Tartu (209/T-17); The serum samples of the participants were analyzed at the Core Facility for Metabolomics at the University of Bergen (Norway) for concentrations of the steroid hormones 17β-estradiol, estrone, estrone 3-sulfate and progesterone, using a high-sensitivity liquid chromatography-tandem mass spectrometry method [ 23 ]. The measurement range for estrogens is well in agreement with the reference range in females, only slightly constrained by the analytical sensitivity in the very low range. If a measurement exceeded the upper limit of quantification (ULQ) it was included as 1.5*ULQ. If a measurement was lower than the lower limit of quantification (LLQ) it was included as LLQ/2 [ 24 ]. Progesterone could not be evaluated well in the lower range, which was considered in the subsequent statistical analyses. In the sleep questionnaire women were asked “Have you been told that you snore”, which we used as main outcome (1. Snoring). An affirmative answer lead to three subsequent questions: “In the last 12 months have you been told that you stop breathing or have irregular breathing while you are sleeping?” (2. Irregular breathing); “Have you woken up all of a sudden with a choking sensation or not being able to breathe in the last 12 months?” (3. Gasping); “Have you been told that you snore loudly or that your snoring disturbs other people in the last 12 months?” (4. Disturbing snore). The response options to these questions were: “never”, “seldom”, “sometimes”, “frequently” or “every time”. If women answered “no” to the question “Have you been told that you snore?” they were allotted the answer “never” to all three subsequent questions for the analysis. In order to transform the outcome into a dichotomous variable we merged all affirmative answers (“seldom”, “sometimes”, “frequently” or “every time”). We fitted logistic regression models with female sex hormones (log 2 -transformed) as predictors (one at a time) and snoring, irregular breathing, gasping and sleep disturbance as outcomes (one at a time). We adjusted the analysis for a predetermined set of covariates: age (continuous), BMI (continuous), smoking habits (lifelong non-smoker, ex-smoker and smoker), age at completed full time education (20 years) as a socioeconomic proxy, study center and a novel reproductive score (RAS) based on fuzzy logic, which describes the progress of reproductive aging in a continuous manner, taking into account menstrual frequency and age [ 25 ]. The distribution of the RAS over the included study population (density plot) and its univariate association with age can be found in S4 Fig . To evaluate linearity, we carried out sensitivity analyses for quartiles of 17β-estradiol, estrone and estrone 3-sulfate. As the study population contained many postmenopausal women, various progesterone measurements were below LLQ, therefore analyses concerning this hormone were carried out using it as dichotomous variable (below vs. above LLQ). We further evaluated models (based on the main model) that additionally were mutually adjusted for all included hormones (17β-estradiol, estrone, estrone 3-sulfate, progesterone) as well as the use of multi-level models (subjects nested within study centers). We conducted sensitivity analyses on a) the subset of women who reported to be married and/or cohabiting, and b) adjusting for the frequency of consumption of beer, wine and spirits (Options: “rarely/never”, “1–3 times per month”, “once per week”, “2–4 times per week”, “5–6 times per week”, “once a day”, “2–3 times per day”, “4 or more times per day”). To evaluate the dynamics of the respective sex hormones we additionally conducted sensitivity analyses on hormone ratios. Analyses were performed using R (Version 3.1.0, The R Foundation for Statistical Computing).

Mean (SD) for continuous variables, N (%) for categorical variables and mean [interquartile range] for measured hormone concentrations. (DOCX) Click here for additional data file. (DOCX) Click here for additional data file. (DOCX) Click here for additional data file. (DOCX) Click here for additional data file. (DOCX) Click here for additional data file. (DOCX) Click here for additional data file.