Our data confirm the reports of Webb (1
) and Schmidt-Nowara and Jennum (15
) showing that sleep-disordered breathing is pervasive among middle-aged U.S. adults. In the Schmidt-Nowara and Jennum Hispanic population sample (15
) and our mixed-ethnicity sample, prevalences of ≥5 events per hour were almost identical for similar ages. Indeed, the sometimes recommended criterion for abnormality of RDI ≥5 would select more than half of all men ages 40–64 and almost half of the women 40–64 in the U.S. population.
Though the RDI prevalences reported by Young and colleagues were considered remarkably high, our estimated prevalences for the general U.S. population were almost exactly twice the comparable estimates obtained from Wisconsin state employees (6
). Whereas our colleagues estimated roughly 10.0% of men and 3.8% of women in the general U.S. population ages 40–60 years would have an apnea-hypopnea index ≥ 15 (6
), we would estimate that ODI4 ≥ 15 is found among 20.3% of men and 7.6% of women of comparable ages. Our estimates exceeded the 95% confidence intervals of Young et al. for the entire U.S. population (for males, conversely, their prevalence estimate was below our confidence interval), but our estimates for Whites were virtually identical to their overall estimates. Note that the somewhat more lenient criterion of ODI4 ≥15 yields higher estimated population prevalences than we described for ODI4 ≥20 in the results above. We doubt that disparities between the San Diego and Wisconsin results lie in recording methodologies, since our validation data demonstrated that the polysomnographic RDI and the oximetric ODI4 (with our methodology) are virtually equivalent. Others have previously reported similar validations (27
). We doubt that our colleagues’ focus on governmental employees could account for the disparity, since our data showed that employment status, income, and education were not significant correlates of ODI4, and we identified no recruitment biases in our sample that could account for such disparity. Rather, it seems likely that our sleep-disordered breathing prevalences were higher because our U.S. estimates were adjusted for ethnic status. Perhaps the Wisconsin sample included too few minority subjects for adjustment.
Our data may be contrasted with a report that only 5% of men ages 35–65 years had ODI5 ≥ 5 in an English general practice (29
). Part of this disparity results from use of different desaturation criterion (ODI5 vs. ODI4), less obesity in Britain (30
), and the some-what younger English age group, but ethnic or racial factors may also have been operative. The BMIs of our subjects were comparable to U.S. national prevalences—slightly below national norms for women and slightly above national norms for men (31
). Our prevalences for ODI4 ≥ 5 are at least twice those of Bear-park et al. for ODI3 ≥ 5 among Busselton, Australia men (here the difference in criteria should have produced higher Australian prevalences) (32
). Although our mean BMI for males, 27.0, was almost identical to that of Bearpark et al. (26.9) (32
), the Australian sample may have been almost entirely of British extraction. On the other hand, in another Australian sample ages 35–69 years, the prevalence of ODI4 ≥20 among men (16.6%) was actually higher than in our sample and among women (4.6%) about the same (33
). Subjects ages 65–69 years tended to raise the averages in this sample by their particularly high prevalence. In a northern Italian population sample of men ages 40–65, ODI5 >20 was found in 4.8% (34
). This lower prevalence, compared with San Diego, might be partly explained by the desaturation criterion of >4% (essentially, ODI5) and the slightly lower mean BMI in the Italian study; however, correlations between ODI5 and snoring or BMI were similar to those in San Diego. In summary, it appears that factors of ethnicity, age, and obesity contribute to differing prevalence estimates in diverse contemporary surveys, but a relatively high prevalence of sleep-disordered breathing is consistently found.
The very high prevalence of ODI4 ≥20 among our minority subjects is a source of concern. It has been suggested that sleep apnea might be a factor causing more frequent hypertension among African-Americans (35
). Our sample was not large enough to reliably contrast prevalence of ODI4 ≥20 between distinct minority groups, but the prevalence for Hispanics, Blacks, and our other minorities (largely Asian) seemed quite similar. Such findings might actually reflect social, ethnic, or developmental influences on sleep-disordered breathing rather than genetic racial factors, especially since about half of San Diego Hispanics are classified by the census as White. We might expect minority ethnic groups to have ODI4 ≥20 prevalence distinctly different from each other, as well as from Whites, if the differences were predominantly of genetic etiology.
Like the Wisconsin sample (6
), our results indicated considerable sleep-disordered breathing among women and demonstrated substantial ODI4 rates for women at all ages. Both surveys suggested that there may have been a bias toward men in referrals to sleep clinics, where very high male: female apnea ratios are reported. We could not confirm any significant age-BMI-independent increase in sleep-disordered breathing after menopause as reported in the Wisconsin sample and previous studies (38
); however, menopause effects were not significant in Wisconsin for the higher criterion levels.
After carefully collecting quality of well-being (QWB) scale data to assess potential influences of sleep-disordered breathing on overall health status and morbidity, we could not demonstrate any significant associations of adjusted ODI4 or ODI4 ≥20 with QWB (). Lack of association of ODI4 with a question on general health status and numerous other items from the National Health Interview, mood inventories, sleep symptoms, and medication inventories documented surprisingly little association of sleep-disordered breathing with medical co-morbidities once factors of age, gender, obesity, and ethnicity were controlled. The sample size would have been sufficient to detect small effects. There is some question whether or not sleep-disordered breathing might be a cause of obesity (39
); therefore, control for obesity might negatively bias the partial correlations of log10
[ODI4] with symptoms. Indeed, controlling only for age, gender, and ethnicity, significant partial correlations were found between log10
[ODI4] and QWB (rp
= −0.17, p = 0.001), reported hypertension (rp
= 0.15, p = 0.004), the CES-D depression scale (rp
= 0.14, p = 0.011), and reported poor health (rp
= 0.21, p < 0.001), but even without controlling for BMI, these correlations remained small.
Even specific sleep complaints such as poor sleep satisfaction and reported daytime sleepiness had only statistically equivocal (and clinically trivial) associations with ODI4, with the exception of an appreciable association with reported snoring. Only very weak associations were found between sleep-disordered breathing and questions regarding hypersomnolence. Similarly, the majority of Wisconsin subjects with complaints of hypersomnolence had apnea-hypopnea scores < 5 (6
), so hypersomnolence was not a sensitive or specific indicator of sleep-disordered breathing in Wisconsin subjects. Reported snoring was significantly associated with ODI4 in San Diego subjects, but snoring was found to be only a modest predictor of ODI4 ≥20 and was not strongly associated with ODI4 at any level. Although we hope to present our objective sound-pressure recordings separately, we note here that these recordings supported the conclusion that snoring sounds and desaturation events are only weakly associated. Prediction of ODI4 could be done mainly by obesity, age, gender, and ethnicity, irrespective of symptoms. Subjects with high ODI4, e.g. ODI4 ≥20, were not a clinically and symptomatically distinctive group. Although these three-night recordings might be expected to be more reliable, as contrasted to similar one-night recordings from elderly San Diegans (14
), the associations of questionnaire-reported symptoms with sleep-breathing disturbances were found to be minimal below age 65 years, as they were above age 65.
In these San Diego data, clustering of high ODI4 (at any level), reports of hypersomnolence, and reports of snoring were too weak to confirm that in this population there is a distinct sleep-apnea syndrome. Only modest co-variance of symptoms with sleep-disordered breathing was also found in studies in Australia (40
) and England (29
). The question arises whether sleep-disordered breathing, snoring, and hypersomnia might be relatively independent co-morbidities in the population. Because it is obese men with symptoms of snoring and daytime sleepiness whose referral to sleep clinics has become popularized, patients in whom such symptoms are associated (perhaps fortuitously) are often represented in clinic sleep-apnea case series. Patients without symptoms would rarely be referred to fee-for-service sleep clinics, since asymptomatic patients would be unmotivated to undergo a very expensive diagnostic and treatment regimen. Evidently, women have been rarely referred, especially premenopausal women, so the sleep-clinic literature has reported an extraordinary predominance of sleep-apnea syndromes among men, inconsistent with the population samples. Thus, preconceptions of sleep-disordered breathing may have pre-selected clinic patients with a clustering of symptoms that we have been unable to confirm as a clustering in the population.
The high population prevalence of sleep-disordered breathing that we have observed strengthens the argument that sleep apnea, determined by home oximetry, is a widespread population health problem but tends to minimize our impression of the associated morbidity. Whereas associations of the number of desaturations with increased time-in-bed and wake-after-sleep-onset suggest that sleep-disordered breathing interfered mildly with the efficiency of nocturnal sleep, the daytime sleepiness associated with sleep-disordered breathing often seemed trivial. Significant associations of ODI4 with modest increases in systolic and diastolic pressures [similar to those reported by Hla et al. (41
)] do show that sleep apnea, as indicated by home oximetry, may cause mild blood-pressure elevations, but snoring independent of ODI4, age, gender, ethnicity, and obesity was not associated with hypertension. In summary, these analyses suggest that widespread sleep-disordered breathing may produce definite but rather limited morbidity in the population as a whole.
Randomizing treatment trials and mortality-risk studies could provide additional assessment of the consequences of sleep-disordered breathing. No convincing long-term, randomized-treatment trials have yet been published, but randomized trials that distinguish symptom reduction from spontaneous remission and placebo effects might help estimate the type and amount of morbidity caused by sleep-disordered breathing. Examining mortality, several studies have found that sleep-disordered breathing is associated with increased mortality, but sleep-disordered breathing may not be an independent mortality predictor. In a San Diego City sample with age >65 years, AI > 20 was associated with reduced survival but was not independent of age and histories of cardiovascular and pulmonary disease (42
). Although multivariate analyses have not yet been fully summarized, results from a Veterans Affairs hospital male inpatient sample (44
) also leave uncertain the extent to which sleep-disordered breathing is an independent risk factor. Bliwise et al., Mant et al., and Lavie et al. had similar findings (45
We would caution that there are some patients with extreme sleep apnea who should indubitably be promptly treated. Our data do not argue against treatment of extreme cases but do suggest that the value of treating modest degrees of sleep apnea (e.g. with ODI4 of 5–20, as determined by home oximetry) should be examined critically when associated symptoms are not compelling.