INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Excessive sleepiness has been increasingly recognized as an important public health problem, estimated to affect at least 12 to 20% of the general adult population (1) and contributing to both motor-vehicle and work-related accidents, impaired social functioning, and reduced quality of life (4). Among the many causes of excessive sleepiness is the obstructive sleep apnea syndrome (OSAS), characterized by repetitive episodes of apnea or hypopnea during sleep (8). Patients with OSAS were sleepier than control subjects when assessed by measurements of sleep latency or by sleep symptom questionnaires, and sleepiness in these patients appears to be correlated with OSAS severity, as determined by the respiratory disturbance index (RDI) (9). In community-based samples, it has been demonstrated that even mild elevations of the RDI are associated with increased sleepiness (2, 12). It has also been shown that snoring is associated with excess sleepiness (2, 12); however, as snoring is a strong predictor of an increased RDI, it is not known whether snoring is independently associated with increased sleepiness or is simply a marker for an elevated RDI. In the present study, we take advantage of polysomnography (PSG) and questionnaire data collected by the Sleep Heart Health Study (SHHS) to investigate the relation of snoring and RDI to sleepiness in a community-dwelling cohort of middle-aged and older adults selected independent of the diagnosis of sleep apnea.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Sample

All subjects are participants in the Sleep Heart Health Study (SHHS), a multicenter study of the cardiovascular consequences of sleep-disordered breathing involving subjects recruited from ongoing epidemiologic studies. The design of the SHHS has been described previously (13). Subjects were recruited independent of reported sleepiness or obstructive sleep apnea; however, subjects who reported habitual snoring were oversampled at some participating centers, and subjects were excluded if receiving supplemental oxygen therapy or current treatment for OSAS with continuous positive airway pressure or oral devices. The study protocol was approved by the institutional review board of each participating institution. Recruitment of subjects into the SHHS began in November 1995, and ended in January 1998. Subjects chosen for this analysis include all SHHS participants who had PSG of acceptable quality and who completed a Sleep Habits Questionnaire that included snoring history, the Epworth Sleepiness Scale, questions regarding sleep time and leg movements during sleep, and demographic information such as sex, race, weight, and height.

Polysomnography

A single night of unattended PSG was performed in the subject's home, using a Compumedics PS-2 system (Abbotsford, Australia) with the following montage: C₃/A₁ and C₄/A₂ electroencephalograms; right and left electrooculograms; a bipolar submental electromyogram; thoracic and abdominal excursions as detected by inductance plethysmography bands; airflow as detected by a nasal-oral thermocouple (Protec, Woodinville, WA); finger pulse oximetry (Nonin, Minneapolis, MN); electrocardiogram; and body position as detected by a mercury gauge sensor. Recordings were transferred to the SHHS Reading Center for scoring of sleep stages, arousals, and respiratory events. Details of the data acquisition and scoring protocols, and scoring reliability, have been previously published (14, 15). Acceptable studies were those with at least 4 h of recorded data of sufficient quality to allow respiratory event detection and distinction of wakefulness from sleep. For this analysis, RDI was defined as the number of apneas plus hypopneas per hour of sleep time, where apnea is defined as a reduction in the thermocouple signal to < 25% of baseline for  10 s, and hypopnea is defined as a decrease in the thermocouple signal or thoracic or abdominal excursion to < 70% of baseline for  10 s, each accompanied by a 4% decrease in oxygen saturation. In order to exclude postarousal respiratory events, apneas or hypopneas were not scored if they immediately followed an isolated large breath or movement. Arousals were scored from the EEG and chin EMG channels using American Sleep Disorders Association (ASDA) Atlas Task Force recommendations (16), modified to accommodate situations in which EMG artifact obscures the EEG channels, as previously described (14).

Snoring

Snoring was categorized as ever, never, or unknown based on the response to the question "Have you ever snored?" For those who reported ever snoring, snoring frequency was quantified as not current, less than one night per week, one to two nights per week, three to five nights per week, six to seven nights per week, or unknown based on the response to the question "How often do you snore now?"

Sleepiness

Sleepiness was defined as the score on the Epworth Sleepiness Scale (ESS), a popular subjective sleepiness scale that consists of an 8-item self-completion questionnaire that asks the subject to rate his/her likelihood of falling asleep in a variety of commonly encountered situations (17, 18). The ESS has been demonstrated to have good test- retest reliability and internal consistency (18). Possible scores range from zero (the least sleepy) to 24 (the most sleepy), with a score  11 commonly considered indicative of pathologically excessive sleepiness, based on Johns' finding (17) that in a group of 30 subjects with a normal sleep history without snoring, the mean ESS score was 5.9 (SD 2.2), with no subjects having a score greater than 10. A slight modification of the original questionnaire was utilized in the SHHS, differing primarily in that the introduction to the questions was abbreviated. A pilot study demonstrated comparable ESS scores between the original questionnaire and the modification used in this study (13). The ESS score has been shown to correlate positively with RDI (11) and inversely with objective sleep latency (19) in patients with OSAS, and it was previously shown to be correlated with RDI in the SHHS (12).

Statistical Analysis

All analyses were performed using SPSS data analysis software. Analysis of variance and analysis of covariance were used to assess differences in mean ESS score between categories of snoring and RDI severity, to test for interaction between snoring and RDI as predictors of ESS score, and to assess potential effect modification by age, sex, race, BMI, duration of usual sleep period, difference between usual sleep time on weekdays (or usual work days) versus weekends (or usual nonwork days), frequency of awakening with nocturnal leg jerks or cramps, and SHHS center from which the subject was recruited. Contingency table analysis was used to compare the prevalence of excessive sleepiness, defined as a score  11 on the ESS, among categories of snoring and RDI.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subject Characteristics, Snoring History, and RDI

Between November 1995 and January 1998, 6,440 subjects underwent in-home PSG that was of acceptable quality. Of these, 663 were excluded from analysis for failure to complete all required questionnaire items or for missing demographic data or body mass index (BMI). The remaining 5,777 subjects (2,737 men and 3,040 women) form the basis of this report. Subject characteristics and snoring history are shown in Table 1. For most analyses, subjects were divided into four groups based on RDI < 1.5, 1.5 to 5, 5 to 15, and  15; these groups comprised 27, 26, 29, and 18%, respectively, of the study sample. These cutpoints reflect RDI thresholds of 5 and 15 events/h commonly used in clinical practice, with the lowest RDI group further divided into two groups. The mean RDI was greater in men (11.5 [SD 13.9] events/h) than in women (6.4 [10.1] events/h), and RDI was positively correlated with BMI (r = 0.30, p < 0.001) and age (r = 0.10, p < 0.001).

Sleepiness

The mean score on the ESS was 7.7 (SD 4.4). Men had higher mean scores on the ESS than did women (8.2 [4.5] versus 7.3 [4.2], p < 0.001), and subjects  65 yr of age had higher mean scores than did those > 65 yr of age (7.9 [4.5] versus 7.5 [4.3], p = 0.002). The ESS score was inversely correlated with questionnaire-reported usual total sleep time on weekdays (r = 0.09, p < 0.001), and positively with a marker of the insufficient sleep syndrome, the difference of usual total sleep time on weekends minus usual total sleep time on weekdays (r = 0.06, p < 0.001) (20). The ESS score was lower in subjects reporting that they were awakened by leg cramps or leg jerks "never," "rarely," or "sometimes" than in those reporting these symptoms "often" or "almost always" (7.6 versus 9.1, p < 0.001).

Relation of Snoring and RDI to Sleepiness

Subjects who reported never snoring were less sleepy than those who reported ever snoring (mean ESS score, 6.3 [4.1] versus 8.2 [4.4], p < 0.001) (Table 2). Subjects who did not know if they ever snored were similar to never-snorers (mean ESS score, 6.4 [4.1]). When snoring was classified by frequency, sleepiness increased progressively with snoring frequency, from 6.4 (4.2) in those who report no current snoring to 9.3 (4.8) in those who snored six to seven nights per week (p < 0.001) (Table 3). Using a threshold of ESS score  11 to define excessive daytime sleepiness (EDS), 28% of ever-snorers had EDS, as compared with only 15% of never-snorers (p < 0.001). Among current snorers, the prevalence of EDS increased progressively with increasing frequency of snoring, from 18% in those who reported snoring less than once per week to 39% in those who reported snoring six to seven nights per week (p < 0.001) (Figure 1). Although the percentage of subjects identified as having EDS varied as the ESS threshold used to define EDS was varied from 9 to 13, the strength of the relation of snoring to EDS was little affected by varying this threshold.

View larger version (17K): [in this window] [in a new window]   Figure 1.   The relation of current snoring frequency to prevalence of excessive daytime sleepiness (EDS), defined as a score  11 on the Epworth Sleepiness Scale. The prevalence of EDS increases with increasing snoring frequency (p < 0.001).

The relation of RDI to sleepiness was assessed by dividing the subjects into four categories of RDI severity as described above. Sleepiness increased progressively with increasing RDI, from a mean of 7.1 (4.2) in subjects with RDI < 1.5 to 8.8 (4.8) in subjects with RDI  15 (p < 0.001) (Table 2). When the highest RDI category was further subdivided, subjects with RDI  30 (n = 354) were found to have a mean ESS score of 9.7 (4.9).

There was a strong association between snoring and RDI, with the percentage of ever-snorers increasing progressively from 68% of those with RDI < 1.5 to 86% in those with RDI  15 (p < 0.001) (Table 2). Two-factor analysis of variance was used to determine whether the relation of snoring to sleepiness was due to confounding by RDI. This analysis revealed that the relation of ever-snoring to sleepiness was seen across all categories of RDI, and that there was no significant interaction between snoring and RDI as predictors of ESS score (p < 0.001 for each main effect; p = 0.92 for interaction). Similarly, the relation of snoring frequency to ESS score was not significantly modified by RDI category (p < 0.001 for each main effect; p = 0.88 for interaction). In order to exclude the possibility of residual confounding by RDI within the four RDI categories, analyses of the relation of ever-snoring and snoring frequency to ESS score were repeated, adjusting for RDI or its natural logarithm as continuous variables. These analyses again showed no significant modification by RDI of the relation between snoring and sleepiness.

Multivariate analyses were performed to assess whether the relation of snoring to sleepiness was modified by age, sex, race, BMI, usual total sleep time on weekdays, or self-reported nocturnal awakenings caused by leg jerks or leg cramps. Although each of these variables was significantly associated with ESS score, they had little impact on the relation between ESS score and ever-snoring, snoring frequency, or RDI. For example, the mean ESS scores for subjects who snored zero, less than one, one to two, three to five, and six to seven nights per week, adjusted for RDI alone, were 6.5, 7.1, 7.6, 8.5, and 9.2, respectively; when the above covariates were included in the model, the corresponding values were 6.7, 7.2, 7.6, 8.4, and 9.1.

Among current snorers, the relations of additional questionnaire-derived snoring characteristics to sleepiness were also evaluated, including loudness of snoring and change in snoring over time. The mean ESS score increased from 7.1 in 850 subjects whose snoring was "slightly louder than heavy breathing," to 8.2 in 1,429 whose snoring was "loud as mumbling or talking," 9.1 in 715 whose snoring was "louder than talking," and 9.6 in 446 whose snoring was "extremely loud" (p < 0.001). Subjects who reported that their snoring was increasing over time were sleepier (mean ESS score, 9.1; n = 507) than those who reported no change (8.3, n = 1,982) or a decrease (7.9, n = 265) in their snoring (p < 0.001). These relations were little affected by adjustment for RDI, age, sex, BMI, hours of usual sleep time on weekdays, and self-reported nocturnal awakenings caused by leg jerks or leg cramps; however, inclusion of snoring frequency in the models substantially weakened the association of these additional snoring characteristics with sleepiness, although the association remained statistically significant.

Relation of Arousal Index and Sleep Stage Distribution to RDI and Snoring

The relation of arousal frequency to RDI and snoring was investigated in order to assess whether snoring-related arousal might explain the association between snoring and sleepiness. For these analyses, we included only those subjects with PSG studies reported by the scorers to be entirely free of problems in determining sleep stage and free of problems identifying arousals in both REM and NREM sleep (n = 2,856), using previously described criteria (14). Both RDI and snoring were significantly associated with the arousal index (ArI, defined as the number of arousals per hour of sleep). The product-moment correlation between RDI and ArI was 0.51 (p < 0.001). Spearman's rank correlation between snoring frequency and ArI was 0.13 (p < 0.001). As shown in Table 4, ArI increased across all four categories of RDI, was higher in ever-snorers than in never-snorers, and increased progressively with frequency of snoring; however, after adjusting for RDI, there was no significant difference in ArI between never- and ever-snorers, and no significant difference in ArI among snoring frequency groups.

The percentage of sleep time in each of Stages 1, 2, slow-wave (combined Stages 3 and 4), and REM were compared across RDI and snoring frequency groups. Higher RDI was significantly associated with increased percentages of Stages 1 and 2 sleep, and decreased slow-wave and REM sleep (all p < 0.001). Increased snoring frequency was associated with trends in the same direction; however, after adjustment for RDI, there was no significant association between snoring frequency and the percentage of sleep time spent in any sleep stage.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Excessive daytime sleepiness is a common condition that is associated with substantial morbidity, including adverse effects on job performance (4, 21), family relationships (5), and quality of life (5, 7, 22), and that contributes to motor vehicle and job-related accidents (6, 23, 24). In the present study, we observe a 25% prevalence of excessive sleepiness, defined as a score  11 on the ESS, which is somewhat higher than the prevalence of 12 to 20% noted in earlier population surveys that utilized different questionnaire instruments (1, 2). Although oversampling of snorers in the present study may contribute to this finding, the prevalence of excessive sleepiness in our study is nearly identical to the 26% prevalence observed in a recent national poll using the same instrument employed in the present study (3). Obstructive sleep apnea syndrome, an important cause of excessive sleepiness, has a prevalence estimated at 2 to 4% of middle-aged state government employees (2), with even higher prevalences estimated in other U.S. adult populations (25, 26). Moreover, community-based epidemiologic studies have demonstrated excess sleepiness in the much larger portion of the middle-aged population with mildly elevated RDI but without diagnosed OSAS (2, 12). In the present study, we extend these findings by demonstrating that in a large, ethnically diverse population of community-dwelling, middle-aged and older adults, snoring is associated with excess sleepiness independent of the effect of an elevated RDI.

This observation suggests that even the mildest form of sleep-disordered breathing, snoring without elevated RDI, may cause sleepiness. This finding is consistent with the findings of Young and colleagues (2) that snorers with RDI < 5 were substantially more likely to report excessive daytime sleepiness or awakening unrefreshed than were nonsnorers with RDI < 5. We observed a dose response in the relation between snoring and sleepiness as the degree of sleepiness increases with the frequency and intensity of snoring. This association between snoring and sleepiness is seen in both men and women, and is similar in those  65 and those > 65 yr of age. Sleepiness was increased in subjects who reported sleeping for fewer hours per night, who had evidence of insufficient sleep syndrome as evidenced by a longer sleep period on weekends than on weekdays (20), or who had questionnaire evidence of periodic limb movement disorder or nocturnal leg cramps. The effect of snoring on sleepiness was, however, independent of these other causes of sleepiness.

The magnitude of the relation between snoring and sleepiness observed in this study, coupled with the high prevalence of snoring in the general adult population, suggests that snoring-related sleepiness may have an important public health impact. While threshold levels of sleepiness that result in adverse health effects, such as motor-vehicle accidents or impaired quality of life, have not been established, there are data to suggest that even mild increases in sleepiness due to sleep-disordered breathing may increase the risk of accidents. For example, it has recently been reported that among subjects participating in the Wisconsin Sleep Cohort Study, snorers with an RDI < 5 had a threefold increased odds of experiencing multiple motor vehicle accidents during the preceding 5 yr than were subjects with no evidence of sleep-disordered breathing (27).

Several potential limitations of this study should be considered. The study sample was recruited from among community-dwelling adult participants in ongoing longitudinal cohort studies. The sample differs from the adult U.S. population in its racial and ethnic composition, and may differ in other, unmeasured, characteristics; however, there is no reason to suspect that these differences would introduce a bias into the snoring-ESS association. Adjustment by analysis of variance for the parent cohort from which the subject had been recruited into the SHHS confirmed that center effects did not bias the relation between snoring and ESS score. Sleepiness is multifactorial, and we were unable to measure all possible causes of sleepiness in this study; however, unless these unmeasured causes of sleepiness were associated with snoring, they would not bias the observed association of snoring and sleepiness.

Sleepiness was measured using the Epworth Sleepiness Scale, a popular self-completion questionnaire that assesses subjective sleepiness. The ESS is a well-validated questionnaire, which has been shown to reliably measure persistent daytime sleepiness in adults (17, 18). Although there is only a moderate inverse correlation between ESS score and objective sleepiness as measured by the multiple sleep latency test (r = 0.42) and the maintenance of wakefulness test (r = 0.48) (19, 28), a positive correlation between RDI and ESS score has been observed by us in a general adult population and by others in OSAS patients (10, 11, 17). In two recent placebo-controlled OSAS treatment trials, the ESS score fell significantly with active treatment but not with placebo (29, 30). At present, there is little literature relating any of the sleepiness measures to functional morbidity, although in one study the ESS score was significantly related to the likelihood of falling asleep while driving (31), and in another it was correlated with the vitality subscale of the Medical Outcomes Study SF-36 health-related quality of life questionnaire (7). It thus appears that the ESS, although a subjective measure of sleepiness, is sensitive to differences in sleepiness among groups with differing severity of sleep-disordered breathing. Misclassification resulting from underestimation of sleepiness in the setting of chronically elevated sleep drive would, in any case, bias toward a null result. Snoring was also measured using a novel self-report instrument. The reliability of this instrument upon repeat administration has not been studied, nor has the instrument been validated against objectively measured snoring, although criterion validity is demonstrated by its detection of known variation in snoring frequency with age, sex, obesity, and RDI.

Sleep fragmentation caused by repeated arousals in response to apneas and hypopneas is thought to be the cause of excessive sleepiness in patients with OSAS (32). The mechanism underlying the relation of snoring to excess sleepiness is uncertain, although snoring-related arousal caused by increased upper airway resistance or acoustic stimulation is a likely mechanism. The failure to demonstrate an independent association between arousal frequency and snoring in this study therefore merits discussion. Sleep studies were performed unmonitored in the home, not in the controlled environment of the sleep laboratory. Technical factors might be expected to impair the ability to reliably detect arousals, with signal quality degradation reducing the sensitivity of arousal detection, and electrical interference reducing the specificity of arousal detection. For these reasons, we have restricted our analysis of arousal frequency to the highest quality studies, excluding any in which there was evidence of unreliability of arousal scoring, making it unlikely that such technical factors explain the lack of an observed relation between snoring and arousal frequency. An alternative explanation of our failure to observe an independent relation of snoring to arousal frequency, after adjusting for RDI, is that snoring may result in arousals that are insufficiently intense to elicit cortical EEG changes meeting the ASDA definition of arousal. This explanation is consistent with the repeated observation that airway resistance or acoustic stimuli insufficiently intense to elicit EEG evidence of arousal may nonetheless be associated with activation of the sympathetic nervous system (33). Such electroencephalographically occult arousals have been found to be associated with increased sleepiness (36). It is possible that an alternative measure of arousal such as frequency of body movements or increases in pulse rate might provide an index of sleep fragmentation that is more closely correlated with daytime sleepiness (11).

Several alternative explanations of the association between snoring and sleepiness should also be considered. It is possible that snoring is a cause of subjective sleepiness through arousal-independent mechanisms such as fatigue caused by increased work of breathing accompanying the higher airway resistance in snorers compared with nonsnorers (37). Sleep deprivation is an important cause of sleepiness, and has also been shown to worsen the severity of snoring (38). Sleep deprivation might thereby lead to a correlation between snoring and sleepiness. It is unlikely, however, that this explains the association between snoring and ESS score observed in this study, as adjustment for usual sleep time or for the difference between usual sleep time on weekends and weekdays does not diminish the strength of the association.

Finally, we have considered the possibility that methodologic rather than biologic factors might account for the observed association between snoring and sleepiness. It is possible that the observed association is due to report bias, whereby those who are likely to overestimate snoring are also more likely to overestimate sleepiness. Although snoring as measured by this questionnaire shows the expected relations to age, sex, obesity, and RDI, in the absence of an objective assessment of snoring the possibility of report bias cannot be excluded. Respiratory disturbance index is known to vary somewhat from night to night, an observation that has been confirmed in the SHHS (39). This will cause some degree of inaccuracy in the estimation of "usual" RDI from a single night PSG. Snoring frequency, in contrast, is estimated from a subject's typical behavior over a period of time. Therefore, snoring history may provide additional information regarding the usual frequency of apneas and hypopneas during sleep that is not captured by a single-night PSG. The complementary information provided by snoring history and single-night PSG might lead to an improved ability to predict daytime sleepiness resulting from a subject's "usual" RDI, compared with the use of either measure individually. We observed a fairly strong correlation between snoring history and single-night RDI, however, and would expect some overlap between these measures in their ability to predict "usual" RDI. This should cause the effects of snoring history and single-night RDI on ESS score to be less than additive. As the observed effects are additive, it is unlikely that improved prediction of "usual" RDI by snoring history explains our results. Thus, snoring-related occult arousal remains the most likely explanation of the observed association of snoring with sleepiness, despite the lack of confirmatory arousal data in this study.

In summary, we have shown that in a sample of community-dwelling adults there is a significant association between snoring and sleepiness that is independent of the association between RDI and sleepiness. Snoring may therefore be an independent cause of excess sleepiness, not merely a proxy for obstructive sleep apnea, through mechanisms that remain to be elucidated. In this community-based sample, the association between snoring and sleepiness is similar in magnitude to the association between elevated RDI and sleepiness. This suggests that snoring, because of its very high prevalence, may contribute substantially to the population burden of excessive sleepiness, an important cause of accidents, impaired social performance, and reduced quality of life.