INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Epidemiologic studies of obstructive sleep apnea (OSA) have consistently found a very strong male predominance of this disorder. Early clinic-based reports found male-to-female ratios of 8:1 or greater for OSA (1). Later, community-based studies found smaller differences in the prevalence of OSA, with male-to-female ratios of 2:1 or 3:1 (2). The reason for this sex difference between community- and clinic-based populations is unknown, since women and men with OSA are reported to have similar symptoms (3).

The reason for the lower prevalence of OSA in women than in men is not fully understood. Although the influence of female sex hormones has been considered, their exact role in OSA is not clear. Treatment of OSA with progesterone did not significantly affect its severity (4, 5), and OSA is known to occur in both pre- and postmenopausal women (6). The influence of obesity has also been investigated. Although initial reports emphasized the high prevalance of morbid obesity in women with OSA (7), more recent studies have described OSA in women without significant obesity (8). The greater central body fat distribution and larger neck dimensions of men account for some but not all of the sex differences found in OSA prevalence (9).

There has been no previous report of gender differences in the polysomnographic features of OSA. By analyzing differences between men and women in the severity and type of OSA and in the distribution of respiratory events during non-rapid eye movement (NREM) and rapid eye movement (REM) sleep, we hoped to provide alternative explanations for some of the previously described gender differences in OSA.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

A computerized database of all patients who had undergone overnight polysomnography (PSG) at the Wellesley Hospital sleep laboratory between June 1989 and March 1998 was reviewed. Patients were eligible for the study if a diagnosis of OSA had been made on their initial polysomnogram, and if they had spent at least 30 min in both NREM and REM sleep on the night of that polysomnogram. Patients were excluded if they had undergone a previous sleep study at another laboratory or were taking benzodiazepines, narcotic medications, or alcohol at the time of the study. OSA was defined as an apnea-hypopnea index (AHI) of 5 or more events per hour. Only the patient's initial polysomnogram was included in the study.

Protocol

Overnight PSG was performed on all patients. This included a two-channel electroencephalogram (EEG) (C3-A2, C4-A1), electrooculogram (EOG), and submental and leg electromyograms (EMGs). The ECG and heart rate were monitored through standard limb leads. Airflow was detected by monitoring expired CO₂ at the nose and mouth through cannulae adapted for the purpose and attached to a CO₂ analyzer (CD 102 Normocap; Datex Corp., Helsinki, Finland). Respiratory efforts were monitored by inductance plethysmography with transducers placed around the chest and abdomen (Respitrace; Ambulatory Monitoring, Ardsley, NY). Arterial oxygen saturation was measured continuously by pulse oximetry, using a finger probe (Biox 3740; Ohmeda, Boulder, CO) set at the fastest response. Body position was assessed continuously both through closed-circuit television and with a body position sensor (Vitalog; Respironics Inc., Redwood City, CA). All variables were recorded either on a polygraph (Model 78E; Grass Instruments, Quincy, MA) at a paper speed of 10 mm/s, which was used for sleep studies done before September 1993, or on a computerized system (Sandman; Mallinckrodt/Nellcor Puritan Bennett [Melville], Ottawa, ON, Canada), used after September 1993. Height and weight were assessed on the evening of the study.

Polysomnography

Sleep stage was scored manually according to the criteria of Rechtschaffen and Kales (10). An arousal was defined as an awakening from sleep for 3 to 15 s, manifested by simultaneous alpha activity on the EEG, eye movements, and EMG activation. We classified arousals into one of three types: respiratory arousal (apnea or hypopnea within the preceding 3 s), periodic limb-movement arousal (periodic limb movement within the preceding 3 s), or spontaneous arousal (not temporally related to a respiratory event or periodic limb movement). Obstructive apnea was defined as the absence of airflow for more than 10 s in the presence of continued respiratory effort. Central apnea was defined as the absence of airflow for more than 10 s due to loss of respiratory effort. Hypopneas were defined as a reduction in the amplitude of respiratory effort to between 10% and 50% of the baseline level during sleep and lasting more than 10 s, and were classified as central or obstructive events based on the presence of synchronous or paradoxical breathing, respectively. The AHI was defined as the number of apneas and hypopneas per hour of sleep, and was calculated separately during total sleep time (AHI_TST), during REM sleep (AHI_REM), and during non-REM sleep (AHI_NREM). The AHI was also calculated with the subject in the supine and nonsupine sleeping positions. The difference between the AHI_REM and AHI_NREM was called the "REM difference." In addition, the AHI was calculated separately for both obstructive and central respiratory events during REM and NREM sleep and during total sleep time, allowing calculation of the individual REM differences for both obstructive and central events.

OSA Type

On the basis of the PSG findings, each patient with OSA was classified as having one of three mutually exclusive types of OSA, as follows: (1) REM OSA, was designated as mild OSA that occurred almost exclusively during REM sleep (Figure 1). This was defined as an AHI_TST of between 5 and 25 events/h, AHI_REM/AHI_NREM > 2, and AHI_NREM < 15 events/h. (2) S OSA, designated as OSA of any severity that occurred almost exclusively in the supine position (Figure 2). (3) A OSA, designated as OSA of any severity that was not dependent on sleep stage or sleep position (Figure 3). A OSA and S OSA were determined prospectively by a single investigator (P.H.) after semiquantitative assessment of each patient's polysomnogram.

View larger version (34K): [in this window] [in a new window]   Figure 1.   Example of REM OSA. Graphic summary of overnight polysomnography in a single patient. (Top panel  ) Hypnogram with stages of wakefulness, REM sleep (dark lines) and non-REM sleep (stages 1, 2, 3, and 4). (Second panel  ) Respiratory events profile with classification of apneas (APN) and hypopneas (HYP) as central (CN), obstructive (OB), and mixed (MX). (Third panel ) Oximetry, showing changes in oxygen saturation (SaO2). (Fourth panel ) Body position with time spent on right side, prone, on left side, and supine. Note that apneas and hypopneas occur almost exclusively during REM sleep. AHITST = 18 events/h, AHINREM = 6 events/h, AHIREM = 54 events/h.

View larger version (32K): [in this window] [in a new window]   Figure 2.   Example of S OSA. Format as described in Figure 1. Note that apneas and hypopneas occur almost exclusively in the supine position. AHITST = 14 events/h, AHIsupine = 32 events/h, AHInon-supine = 2 events/h.

View larger version (34K): [in this window] [in a new window]   View larger version (43K): [in this window] [in a new window]   Figure 3.   (A) Example of mild A OSA. Format as described in Figure 1. Note that apneas and hypopneas are not sleep stage- or posture-dependent. AHITST = 24 events/h, AHINREM = 23 events/h, AHIREM = 28 events/h. (B) Example of severe A OSA. Format as described in Figure 1. Note that apneas and hypopneas are not sleep stage- or posture-dependent. AHITST = 68 events/h, AHINREM = 71 events/h, AHIREM = 55 events/h.

OSA Severity

The severity of OSA was determined by the AHI_TST, and was classified as mild (5 to 25 events/h), moderate (26 to 50 events/h), or severe (> 50 events/h). This arbitrary classification, although consistent with one previously published recommendation (11), was primarily used to aid the presentation of our data over a wide range of AHI values, and was not intended to be a commentary on the specific polysomnographic features of OSA that represent mild, moderate, and severe disease. The severity of OSA during REM or NREM sleep was determined on the basis of AHI_REM and AHI_NREM, respectively, and was classified as very mild (0 to 5 events/h), mild (6 to 25 events/h), moderate (26 to 50 events/h), or severe (> 50 events/h).

Statistical Analysis

Differences in proportions of categorical variables were assessed through the chi-square test. Differences in means of continuous variables were assessed with Student's t test. Univariate logistic regression analysis was used to assess the effect of body mass index (BMI) and age on the REM difference, and to assess the effect of age on the severity of OSA. All data were analyzed with SPSS version 7.5 software for Windows (12) (SPSS, Inc., Chicago, IL). All data are reported as mean ± SEM. A two-tailed value of p < 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Database Characteristics

A total of 2,702 patients underwent overnight PSG at the Wellesley Hospital sleep laboratory during the study period. A diagnosis of OSA was made in 1,272 patients, of whom 838 were eligible for the present study. Of the 434 patients excluded from the study, 314 did not spend at least 30 min in both REM and NREM sleep, 89 had previous PSG at another sleep laboratory, and 31 were taking an anxiolytic or narcotic medication. The male-to-female (M:F) ratio for all patients in the database was 2.1:1, and that for patients with OSA who were eligible for the study was 3.1:1. Women in the study were heavier than men (BMI = 35.1 ± 0.6 kg/m² versus 32.1 ± 0.3 kg/m², p < 0.01) and were slightly older (average age = 50.8 ± 0.9 yr versus 48.6 ± 0.5 yr), but this difference was not statistically significant.

Distribution of Respiratory Events during Sleep

In order to assess differences between men and women in the distribution of respiratory events during sleep, we calculated the AHI_TST, AHI_NREM, and AHI_REM for all eligible OSA patients in the study database (Table 1). Although values of AHI_TST and AHI_NREM were significantly greater in men than in women, AHI_REM was virtually identical in the two sexes. This was reflected by an almost threefold greater REM difference in women than in men. The effect of OSA severity on the distribution of respiratory events during sleep was determined by evaluating the REM difference in mild, moderate, and severe OSA, as previously defined. The REM difference was greater in women than in men with OSA of all degrees of severity (Table 2).

Other Potential Influences on the REM Difference

In addition to sex, other potential influences on REM difference included weight, age, and duration of apnea. The adjusted r² of a univariate regression analysis of BMI versus REM difference was less than 0.002 for both sexes, and did not attain statistical significance. Furthermore, we compared the REM difference for men and women within different ranges of BMI (< 30, 30 to 40, and > 40 kg/m²) and found no significant effect of BMI. A univariate analysis of age and REM difference in both men and women yielded an r² of less than 0.01, which did not attain statistical significance in women (p = 0.2) but did attain significance in men (p = 0.02). The potential for the relatively higher AHI during REM sleep in women to be caused by shorter episodes of apnea was assessed by calculating the mean apnea duration in men and women during REM sleep. Women had slightly shorter episodes of apnea than did men during REM sleep (17.3 ± 0.5 s versus 21.2 ± 0.3 s, p < 0.01).

A number of other factors could have contributed to the REM difference in men and women. Although the lower AHI during NREM sleep in women could have been due to a postural effect rather than to a sleep-stage effect, this is highly unlikely for two reasons. First, sleep position had only a small effect on OSA in women, as reflected by the low prevalence of posture-dependent OSA (S OSA) in women (Table 4). Second, the mean time spent by women and men in the supine position was almost identical both during NREM (95.3 min in women versus 96.5 min in men) and REM sleep (20 min in women versus 20.8 min in men).

The potential impact of alcohol and antidepressant medications (e.g., serotonin-specific reuptake inhibitors [SSRIs]) on upper airway function was also considered. Our patients were routinely asked to abstain from alcohol for 48 h before their sleep study, but were allowed to take antidepressant medications, including SSRIs, on the night of the study. The prevalence of antidepressant medication use was higher among women than among men (13.6% versus 7.2%, p < 0.05). However, the REM difference was similar in women taking and those not taking antidepressants (27.8 ± 2.6 versus 28.2 ± 1.9 events/h, p = NS) and in men taking and those not taking antidepressants (9.8 ± 1.6 versus 10.5 ± 1.3 events/h, p = NS). Consequently, we do not feel that the use of these medications was responsible for our findings.

Respiratory events were categorized as central or obstructive, as previously described. This enabled us to determine whether the REM difference was due to central or obstructive events. Because patients with central sleep apnea were excluded from the study, the frequency of central events was very low (3.0 ± 0.5 events/h for women versus 6.9 ± 0.6 events/h for men, p < 0.05). The "central REM difference" was virtually the same in women and men (0.1 ± 0.5 versus 1.7 ± 0.6 events/h, p = NS), reflecting a minimal effect of sleep stage on the distribution of central events. The REM difference was entirely the result of the "obstructive REM difference", which was significantly higher in women than in men (28.0 ± 1.5 versus 12.0 ± 1.2 events/h, respectively, p < 0.001). These data indicate that the overwhelming majority of respiratory events in the study were obstructive, that the frequency of central apneas and hypopneas was similar in women and men, and that the REM difference in AHI was due to obstructive rather than to central events.

Although women had a lower AHI during NREM sleep than did men, we considered the possibility that women had a similar amount of sleep fragmentation as men as the result of a subtle form of sleep-disordered breathing, such as upper airway resistance syndrome. The frequency of combined respiratory and spontaneous arousals during NREM sleep, as defined in the METHODS section, was much lower in women than in men (15 ± 1.8 versus 25 ± 1.6 events/h, p < 0.01), whereas the frequency of these arousals during REM sleep was similar in women and men (23 ± 1.5 versus 28 ± 1.3 events/h, p < 0.05). The REM difference for combined respiratory and spontaneous arousals was two- to threefold higher in women than in men (8 ± 1.7 versus 3 ± 1.4 events/h, p < 0.01), which is almost identical to the REM difference in AHI between women and men. Consequently, we conclude that the REM difference in AHI was accompanied by a similar REM difference in sleep disruption reflected by arousal frequency.

A further possibility is that the REM difference in AHI was due to a gender difference in upper airway anatomy. Although we did not systematically evaluate bony morphology of the upper airway, it is unlikely that this accounted for our findings, since it would have to entail a major sex difference in the prevalence of upper airway abnormalities, which has not been reported. Neck circumference was systematically measured in all patients and was standardized by correction for the patient's height. We found no sex difference in this measurement (0.20 ± 0.1 in men versus 0.21 ± 0.2 in women) and consequently we do not believe that the gender differences we observed in AHI were due to structural differences in the upper airway.

Gender Effect on OSA Severity

In order to assess differences between men and women in the severity of OSA, all eligible patients in the database were categorized by severity of OSA during total sleep time, during NREM, and during REM sleep, as described in METHODS. Women had a higher proportion of mild apnea during total sleep time. This was reflected by an increasing male-to-female ratio from 2:1 for mild OSA to 7:1 for severe OSA (Figure 4). This was entirely due to a higher proportion of mild OSA during NREM sleep in women. During NREM sleep, the male-to-female ratio increased from 1:1 for very mild OSA to 8:1 for severe OSA. This trend was not seen during REM sleep, for which the male-to-female ratio fell from 6.3:1 for very mild OSA to 3.3:1 for severe OSA. The effect of age on OSA severity was assessed by a univariate linear regression analysis of AHI_TST versus age in men and women. For men, r² for AHI versus age was 0.001 (p = 0.23), and for women r² for AHI versus age was 0.015 (p = 0.06), reflecting an insignificant effect of age on the severity of OSA in both sexes.

View larger version (23K): [in this window] [in a new window]   Figure 4.   Comparison of the number of male and female patients with OSA of different degrees of severity during total sleep time (top graph), NREM sleep (middle graph), and REM sleep (bottom graph). The male-to-female-ratio increases as OSA severity increases during both total sleep time and NREM sleep, whereas the opposite trend is seen during REM sleep, with the male-to-female ratio generally decreasing as OSA severity increases (p < 0.001).

OSA Types

The prevalence of different types of OSA in men and women is outlined in Table 3. This varied significantly by gender. Women with OSA had a much higher prevalence of REM OSA than did men, whereas men had a higher prevalence of A OSA than did women. S OSA occurred almost exclusively in men. In order to determine whether these differences reflected a higher prevalence of mild OSA among women, we performed this analysis for patients with mild OSA only (Table 4). The overall prevalence of REM OSA was higher when patients with moderate and severe OSA were eliminated, but the male-to-female ratios remained essentially unchanged. Women still had a much higher prevalence of REM OSA than did men, and had S OSA very infrequently as compared with men.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study constitutes the first analysis of gender differences in the polysomnographic features of OSA in a large group of patients. The major finding of the study was that during REM sleep the severity of sleep apnea was similar in women and men, but that during NREM sleep apnea was less severe in women than in men. In terms of the polysomnographic features of OSA, this is seen in three ways. First, women with OSA have milder OSA than do men. Second, women with OSA have a greater proportion of their respiratory events during REM sleep than do men, and third, women have a higher prevalence of OSA occurring almost exclusively during REM sleep (REM OSA) than do men. These differences between the two sexes do not appear to be age- or weight-dependent.

Current understanding of the pathogenesis of OSA is incomplete (13), but the condition is thought to be due to a combination of structural abnormalities of the upper airway, either congenital or acquired, with abnormalities of upper airway function during sleep. The reason for the higher prevalence of OSA in men is only partly understood (14), but most likely reflects gender differences in both the structure and function of the upper airway. Although upper airway morphology is a strong predictor for the development of OSA (15), no studies have examined gender differences in airway size either in obese men and women or in men and women with OSA. It is known that men have a greater tendency than women for android fat distribution, resulting in a larger neck size, and that this accounts for some of the increased prevalence of OSA in men (9). The remaining gender difference in OSA prevalence most likely results from differences in upper airway muscle function during sleep. During wakefulness, women have greater genioglossus activity than do men (16), and persistence of this difference into NREM sleep could prevent upper airway collapse. This hypothesis is supported by the finding in a recent study that men had a greater increase in airway resistance and a greater susceptibility to flow limitation during NREM sleep than did women (17). These findings are consistent with our results, which imply that during NREM sleep, women have a functional advantage over men that is protective against airway collapse, and that this protective mechanism is lost with the transition to REM sleep.

Age did not have a significant effect on any of our findings, as reflected by the very small r² values for the correlations between age and severity of OSA and between age and REM difference in both sexes. The median age of our female subjects was 51 yr, indicating that approximately 50% were postmenopausal. This suggests that menopause has little effect on sleep-disordered breathing in women. This is consistent with previous reports of OSA in both pre- and postmenopausal women (6), and with a previous study showing no difference in response to nasal occlusion in pre- and postmenopausal women (18). Consequently, we do not feel that levels of sex hormones at the time of PSG were responsible for the differences that we observed in the polysomnographic features of sleep apnea in men and women.

A potential weakness of our study is referral bias resulting from our clinic population not being representative of the community. However, it is unlikely for many reasons that referral bias affected our results. First, our population sample size was very large, and the population was accumulated over a long period, with more patients with moderate and severe OSA than in the largest community-based studies (2). Second, patients were referred to our sleep laboratory predominantly by their primary care physicians, and were excluded if a previous PSG had been done elsewhere. Third, it is known that men and women with OSA have similar symptoms, making unlikely the preferential referral of a subgroup of women with OSA (3). Fourth, several studies have reported similar age, BMI, and symptoms of OSA in both community- and sleep clinic-based populations (19). Moreover, the male-to-female ratio in our study was similar to that in community-based studies (22).

Our most robust finding was the clustering of respiratory events during REM sleep in women, which was seen consistently with OSA of all degrees of severity and in patients of different age and weight. It is difficult to determine what, if any, sort of referral bias could have led to this pervasive finding. In addition, our findings of milder OSA in women than in men are consistent with what has previously been reported in the literature (23). Studies of mild OSA and the upper airway resistance syndrome, have reported similar male-to-female ratios to that found in our study for subjects with mild OSA (24, 25). The perception that OSA is predominantly a male disease (even in community-based studies, men outnumber women by two to one) would make overreferral of women with mild OSA unlikely. The converse is more likely: that only women with the most severe symptoms and disease would be referred. The possibility that our study reflects scoring bias is also unlikely, since there was no awareness of possible differences between men and women when the polysomnographic data were scored and reported.

The identification of REM OSA in a patient with sleep apnea may carry implications for the patient's immediate management. It is known that the correlation between AHI_TST and daytime symptoms, such as excessive sleepiness, is not strong, and that a more specific polysomnographic predictor of daytime symptoms is needed. It has been reported that in patients with mild OSA (AHI_TST < 10 events/h), the severity of daytime sleepiness correlated best with the AHI during REM sleep, and that an AHI_REM > 15 events/h was predictive of reduced sleep latency on the multiple sleep latency test (24). Furthermore, in the Wisconsin Sleep Cohort Study (3), women became symptomatic at a lower AHI_TST than did men, which may have been the result of greater disruption of REM sleep through clustering of apneas and hypopneas within that sleep stage. Although the long-term effects of isolated disruption of REM sleep are unknown, short-term studies have shown that isolated REM awakenings can cause reduced sleep latency, as measured by the multiple sleep latency test (26). The possibility that REM OSA is unique in its clinical presentation and response to treatment is intriguing and needs further study.

In summary, this article is the first report of differences in the polysomnographic features of OSA in men and women. The greater clustering of respiratory events during REM sleep in women in our study probably reflects gender differences in upper airway control during sleep. The potential influence of this on the clinical features, natural history, and treatment of sleep apnea needs further study.