Flow Limitation Is Associated with Excessive Daytime Sleepiness in Individuals without Moderate or Severe Obstructive Sleep Apnea

Abstract

Rationale

Moderate-severe obstructive sleep apnea (OSA) (apnea–hypopnea index [AHI], >15 events/h) disturbs sleep through frequent bouts of apnea and is associated with daytime sleepiness. However, many individuals without moderate-severe OSA (i.e., AHI <15 events/h) also report sleepiness.

Objectives

To test the hypothesis that sleepiness in the AHI <15 events/h group is a consequence of substantial flow limitation in the absence of overt reductions in airflow (i.e., apnea/hypopnea).

Methods

A total of 1,886 participants from the MESA sleep cohort were analyzed for frequency of flow limitation from polysomnogram-recorded nasal airflow signal. Excessive daytime sleepiness (EDS) was defined by an Epworth Sleepiness Scale score ⩾11. Covariate-adjusted logistic regression assessed the association between EDS (binary dependent variable) and frequency of flow limitation (continuous) in individuals with an AHI <15 events/h.

Results

A total of 772 individuals with an AHI <15 events/h were included in the primary analysis. Flow limitation was associated with EDS (odds ratio, 2.04; 95% confidence interval, 1.17–3.54; per 2–standard deviation increase in flow limitation frequency) after adjusting for age, sex, body mass index, race/ethnicity, and sleep duration. This effect size did not appreciably change after also adjusting for AHI.

Conclusions

In individuals with an AHI <15 events/h, increasing flow limitation frequency by 2 standard deviations is associated with a twofold increase in the risk of EDS. Future studies should investigate addressing flow limitation in low-AHI individuals as a potential mechanism for ameliorating sleepiness.

Excessive daytime sleepiness (EDS)—the inability to remain fully awake or alert during the normal wake cycle—is a common symptom of obstructive sleep apnea (OSA). However, the severity of OSA reported by the apnea–hypopnea index (AHI) is poorly associated with sleepiness (1). For example, up to one-third of individuals without moderate-severe OSA may report sleepiness (2). This discrepancy highlights the requirement to better understand and quantify the factors contributing to EDS in the low-AHI group with sleep-disordered breathing and thereby facilitate the identification of individuals with EDS for therapeutic intervention. There are likely many reasons why the AHI fails to distinguish between sleepy and not sleepy individuals. First, the night-to-night variability in AHI can be substantial and appears particularly variable in low-AHI individuals (3). Also, the AHI only reports the frequency of events and does not quantify the severity of obstruction during events or flow limitation outside of scored events (4). Finally, the current assessment of sleepiness is subjective, and arbitrary cutpoints are used to define excessive sleepiness. In this paper, we hypothesize that flow limitation (i.e., objectively determined mismatch between ventilation and ventilatory demand [5–7]) is an important contributor to sleepiness in the low-AHI group.

Several mechanisms link sleep-disordered breathing to sleepiness, including nocturnal hypoxemia (8), sleep fragmentation (9), and sleep duration (10). Flow limitation has also been proposed as a “nonscored” contributor to sleepiness, where the antecedent flow limitation is expressed as upper airway resistance syndrome (UARS) (4, 11, 12). These studies demonstrate that patients with UARS frequently present with clinical symptoms of OSA such as EDS. However, although flow limitation has been implicated as a precursor to sleepiness, the degree of flow limitation has not been investigated in this context because of a lack of automated objective assessments.

We have recently developed and validated an automated tool to estimate the degree of flow limitation using nasal flow from standard diagnostic polysomnography (PSG) (13, 14). The aim of this study was to investigate the association between self-reported sleepiness and flow limitation in individuals with an AHI <15 events/h. We addressed this aim by testing the following hypothesis: The prevalence of sleepiness in people with low AHI increases with flow limitation independent of relevant covariates (i.e., age, sex, body mass index [BMI], race/ethnicity, and average sleep duration) and the AHI. We conducted this analysis in 834 participants with AHI <15 events/h who underwent overnight PSG as part of the MESA (Multi-Ethnic Study of Atherosclerosis) study.

Methods

Study Participants

We analyzed PSG data from the MESA study (15, 16). Briefly, MESA is a large multicenter longitudinal study with a focus on investigating factors associated with incident subclinical and clinical cardiovascular disease (17). The study includes Black, White, Hispanic, and Chinese American men and women. Participants were enrolled between 2000 and 2002, when they were aged 45–84 years, and were free of known clinical cardiovascular disease. Participants were studied at examinations conducted every 2 to 4 years thereafter. In conjunction with exam 5 (2010–2012), participants were enrolled in a sleep examination (2010–2013), when they underwent unattended PSG. The details of the sleep study and scoring of respiratory and arousal events have been described previously (15); however, in the present study, we have adopted the American Academy of Sleep Medicine 2012 recommended guidelines (18) (i.e., hypopneas required an associated ⩾3% desaturation and/or arousal) for scoring of respiratory events.

Flow Limitation

We previously reported an automated method to estimate the proportion of breaths considered flow limited (14). Briefly, this method uses an ordinal regression model to categorize flow limitation (certain, possible, or not flow limited) on a breath-by-breath basis using flow-shape features (e.g., flattening, scooping) extracted from nasal pressure signal, providing a flow limitation category for every breath within a sleep study and not restricted to event-based scoring. In the present study, we retrained the previously validated model specifically for MESA data that were sampled at 10 Hz rather than the previously validated 25 Hz (see the Methods section in the data supplement), and after determining non-inferiority in prediction based on 10 Hz flow-shape feature calculation (see Figures E1–E4), applied the existing categorical flow limitation model to the available polysomnography dataset. Here, we investigate associations between the proportion of sleep breaths (excluding arousal breaths) assessed as certain flow limitation and self-reported sleepiness.

EDS

The Epworth Sleepiness Scale (ESS) subjectively captures general sleepiness through eight questions, each addressed via a 4-point Likert-type scale (19). The total sum is used to indicate sleepiness, with values 11–12, 13–15, and 16–24 indicting mild, moderate, and severe categorizations of EDS, respectively. In this study, we adopted the originally described cutpoint of ESS score ⩾11 to indicate EDS (19).

Prevalence of EDS and Associations with Flow Limitation

To investigate the prevalence of EDS across the individuals with AHI <15 events/h, we initially divided the sample into tertiles on the basis of flow limitation. Logistic regression modeling was used to investigate the association between EDS (binary dependent variable) and flow limitation. The main model included standard covariates (age, sex, BMI, race/ethnicity, and average sleep duration), and secondary models testing sensitivity were based on the main model with additional covariates and precision variables.

Model 1, the main model, was as follows: flow limitation + age + sex + BMI + race/ethnicity + average sleep duration. The sensitivity models were as follows:

• Sensitivity model 1: M1 + AHI

• Sensitivity model 2: M1 + hypoxic burden and arousal index

• Sensitivity model 3: M1 + insomnia, depression, and diabetes

Continuous parameters were standardized by subtracting the mean and dividing by the standard deviation (SD). Sex, race, insomnia, depression, and diabetes were treated as nominal dichotomous categorical variables (0, 1). Sex (male, female) and race/ethnicity (Black, White, Hispanic, and Chinese American) were self-reported. Habitual sleep duration was determined by weighted weekly self-reported values [(2/7) × weekend + (5/7) × weekday]. Insomnia was considered present if either the Women’s Health Initiative Insomnia Rating Scale was ⩾9 (20) or a self-reported physician diagnosis of insomnia. Depression was defined by the Center for Epidemiological Studies–Depression score ⩾16 (21). Hypoxic burden was estimated by established methods (22).

Primary analyses included the main model and sensitivity models in the subset of individuals with AHI <15 events/h. In secondary analyses, we applied these models within data available from individuals with AHI >15 events/h for comparison purposes.

In supplementary analyses, we also investigated the effect of substituting alternative covariate measures. For example, in lieu of habitual self-reported sleep duration, we considered sleep duration measured by PSG. Similarly, we tested alternative measures of exposure to hypoxemia (oxygen desaturation index, mean overnight oxygen saturation as measured by pulse oximetry, and desaturation severity [23] in place of hypoxic burden). We also tested alternative metrics capturing the impact of arousals (arousal intensity [24] and arousal severity, a novel arousal metric that incorporates both intensity [24] and duration [25]). Finally, we tested alternative definitions for insomnia classification in addition to testing other comorbidities (e.g., hypertension) in place of diabetes.

Statistical Analysis

All modeling and statistical analyses were performed using MATLAB (R2022b, MathWorks). P < 0.05 was considered statistically significant. Average values are presented as mean and SD. Statistical assessment for the prevalence of EDS was approached using the chi-square test of homogeneity, with the null hypothesis being that EDS versus non-EDS would present with the same expected prevalence. In logistic regression modeling, we report the adjusted odds ratio (OR) and 95% confidence interval (CI_95%), describing the odds of EDS per 2-SD change in predictor variable, allowing a clear distinction between high (+1 SD) and low (−1 SD) expression of the relevant predictor variable.

Results

Study Participants

PSG data from 2,060 individuals were analyzed; however, clinical scoring (per American Academy of Sleep Medicine recommended guidelines) was absent for 25, and airflow signal quality was inadequate for determination of flow limitation for 149. Thus, a total of 1,886 individuals were included in subsequent analysis (Figure 1). A total of 772 individuals with AHI <15 events/h were included in primary analyses. Characteristics of the individuals studied are detailed in Table 1, with additional details provided in Table E1 in the data supplement. Overall, the prevalence of EDS in this community-based cohort was low compared with similarly aged clinic-based cohorts that were recruited on the basis of suspicion of sleep-disordered breathing (26). Although there was nearly even representation of males and females in the full dataset (53.0% females), there was a higher proportion of females in the AHI <15 events/h sample (66.2%), and a female predominance of the lower AHI sample was observed in both the EDS and non-EDS subgroups.

Figure 1.

Flow diagram of patient exclusion reasons and derivation of the study sample. AHI = apnea–hypopnea index; EDS = excessive daytime sleepiness; MESA = multi-ethnic study of atherosclerosis; PSG = polysomnography.

Table 1.

Participant characteristics

Characteristics	All (N = 1,886)	AHI <15 (n = 772)	AHI <15 and Non-EDS (n = 678)	AHI <15 and EDS (n = 94)
Demographics
Age, yr	67 (14)	65 (14)	66 (14)	63 (15)
Sex, % female	53.0%	66.2%	65.9%	68.1%
Race/ethnicity
Black*	28.3%	31.9%	30.5%	41.5%
White	36.1%	37.0%	37.3%	35.1%
Chinese American	11.9%	11.0%	11.7%	6.4%
Hispanic	23.7%	20.1%	20.5%	17.0%
Body mass index, kg/m2	28.6 (5.5)	27.0 (5.1)	26.9 (5.1)	27.4 (4.9)
ESS score†	6.0 (4.1)	5.7 (3.9)	4.6 (2.8)	13.5 (2.3)
Proportion with ESS score ⩾11	13.7%	12.2%	0%	100%
Comorbidities
Insomnia*	37.1%	37.4%	36.1%	46.8%
Depression*	13.5%	13.7%	12.7%	21.3%
Diabetes	19.5%	15.2%	15.7%	11.7%
Polysomnography
AHI, events/h	18.2 [23.8]	8.0 [6.8]	8.0 [6.8]	8.3 [7.1]
Arousal index, events/h	22.3 (12.1)	15.7 (7.0)	15.8 (6.9)	14.8 (7.7)
Total sleep time, min	361 (81)	371 (81)	371 (78)	370 (98)
Non-REM 1, % total sleep time†	14.4 (9.2)	10.5 (5.4)	10.6 (5.4)	9.3 (4.6)
Non-REM 2, % total sleep time	57.6 (10.2)	58.8 (9.8)	58.8 (9.8)	58.6 (10.1)
Non-REM 3, % total sleep time	9.9 (8.9)	11.2 (9.3)	11.2 (9.3)	11.3 (9.5)
REM, % total sleep time†	18.1 (6.7)	19.5 (6.3)	19.4 (6.3)	20.8 (6.9)
Flow limitation, % breaths during sleep‡	8.5 [16.5]	3.8 [6.9]	3.5 [6.7]	5.1 [8.0]

Definition of abbreviations: AHI = apnea–hypopnea index; EDS = excessive daytime sleepiness; ESS = Epworth Sleepiness Scale; REM = rapid eye movement.

Values are median [interquartile range], mean (standard deviation), or percent, as appropriate. EDS is defined as ESS score ⩾11. Symbols indicate statistically significant (P < 0.05) difference between non-EDS and EDS grouping.

^*Chi-square test.

^† t test.

^‡ Wilcoxon rank-sum test.

Prevalence of EDS per Flow Limitation Tertile

Figure 2 shows the proportion of individuals with EDS per tertile of flow limitation (individuals with AHI <15 events/h shown; χ² = 6.53; P = 0.038).

Figure 2.

The proportion of individuals without moderate-severe obstructive sleep apnea who report excessive daytime sleepiness (EDS) (Epworth Sleepiness Scale score ⩾11) per tertile of flow limitation. The prevalence of EDS within individuals with the higher degrees of flow limitation (upper tertile, 6.5–85.5%) is approaching double that with low degrees of flow limitation (lower tertile, 0–2.2%). χ² = 6.53; P = 0.038. Note that the distribution of flow limitation in these individuals was positively skewed and very similar to the “normal” group presented by Palombini and colleagues (54) (see Figure E5).

Associations between EDS and Flow Limitation

In primary analyses (i.e., individuals with AHI <15 events/h), our main model revealed a 2.04 increase in odds for EDS (CI_95%, 1.17–3.54; P = 0.011) for each 2-SD increase in flow limitation (+19.2% of sleep in this group). In all three sensitivity models, flow limitation remained significantly associated with EDS. In sensitivity model 1, which included both flow limitation and AHI, only flow limitation was associated with EDS (OR, 2.08; CI_95%, 1.19–3.64; P = 0.010). In sensitivity model 2, which included additional covariates (hypoxic burden and arousal index), flow limitation remained associated with EDS (OR, 1.95; CI_95%, 1.10–3.46; P = 0.022). Finally, in sensitivity model 3, which included additional precision variables (insomnia, depression, and diabetes), flow limitation remained associated with EDS (OR, 2.08; CI_95%, 1.19–3.64; P = 0.010). Table 2 and Figure 3 show the key results from primary analyses, and Table E2 provides detailed model results.

Table 2.

Main model results

Parameter	Main Model (M1)
	Estimate	P Value	Odds Ratio	Lower, Upper
Flow limitation	0.713	0.011	2.040	1.174, 3.545
Standard covariates
Age	−0.334	0.213	0.716	0.423, 1.212
Sex (male)	−0.314	0.203	0.730	0.449, 1.186
BMI	0.050	0.850	1.051	0.626, 1.764
Black	0.273	0.313	1.314	0.773, 2.234
Hispanic	−0.179	0.589	0.836	0.436, 1.604
Chinese	−0.728	0.129	0.483	0.188, 1.238
Average sleep duration	−0.562	0.022	0.570	0.353, 0.922

Definition of abbreviation: BMI = body mass index.

Estimate is the model-estimated coefficient for each variable. P value indicates significance level for each estimated coefficient, with bold typeface indicating those satisfying the significance threshold. The odds ratio is presented for each variable, together with the corresponding confidence interval as lower and upper bounds. Average sleep duration is the average of weekly self-reported sleep duration.

Figure 3.

Forest plot representation of sleepiness models in primary analyses, showing only the results for flow limitation. Individual lines demonstrate model outcome odds ratio (marker) and 95% confidence interval (lines) of flow limitation as a risk for sleepiness (Epworth Sleepiness Scale score ⩾11). Top to bottom: main model, followed by sensitivity models, as described in the figure and in text. Full model results are shown in Table E2. AHI = apnea–hypopnea index; BMI = body mass index.

In secondary analyses (i.e., individuals with AHI ⩾15 events/h), flow limitation was not significantly associated with EDS (main model result, 0.95; CI_95%, 0.65–1.38; P = 0.783; per 2-SD increase in flow limitation frequency; i.e., +34.6.2% of sleep within individuals with AHI ⩾15 events/h). Similarly, flow limitation was not significantly associated with EDS in any of the sensitivity models in the secondary analyses. See Table E3 for comprehensive details of each secondary model. In all supplementary analyses, where we tested alternate covariate measures, there was no change to the interpretation of the main results.

Discussion

This study investigated the association between EDS and flow limitation in a large community sample. In individuals with AHI <15 events/h, we observed an increasing prevalence of sleepiness with increasing proportions of flow limitation. In adjusted analysis, we observed that flow limitation was associated with EDS (OR, 2.04) when controlling for standard covariates (age, sex, BMI, race/ethnicity, and average sleep duration). Flow limitation remained significantly associated with EDS in all three sensitivity models (which included additional sleep-related precision variables (i.e., AHI, hypoxic burden, arousal index, insomnia, depression, and diabetes status). In contrast, flow limitation was not associated with EDS in the AHI ⩾15 events/h group. Overall, in individuals with few to mildly elevated scored respiratory events (i.e., non-OSA and mild OSA), a 2-SD increase in flow limitation was associated with a twofold increase in the risk of sleepiness.

Relationship between Flow Limitation, AHI, and Sleepiness

The notion that flow limitation is not adequately quantified within the AHI has previously drawn considerable research interest (27). Historically, this has been described as UARS. First reported in children (28) and later in adults (29), UARS describes the occurrence of EDS that is not explained by other causes (i.e., diagnosed chronic illness) and is associated with nonapneic and/or nonhypopneic respiratory events (e.g., respiratory effort–related arousals). Such respiratory events typically demonstrate flow limitation (30, 31), whereupon positive airway pressure treatment with the intention of eliminating flow limitation (32) has provided improvements in clinical outcomes (33–35). However, flow limitation may also persist in sustained periods of so-called stable flow limitation that occurs during event-free sleep, clearly distinct from respiratory effort–related arousals (36). The present study confirms and extends the previous work demonstrating a significant positive relationship between flow limitation and sleepiness in low-AHI individuals, but not necessarily in individuals with moderate or more severe OSA (see Figure E6 for sensitivity analysis adjusting AHI cutpoint). When considering individuals with AHI ⩾15 events/h, age and BMI were consistently significant risk factors for EDS, and insomnia and diabetes were also significant when included, providing some rationale for why flow limitation does not appear significantly associated with EDS in this group. Indeed, across the full MESA cohort, there was considerable variation in the association between AHI and EDS, consistent with previous work (37). Although there was a significant association between flow limitation and ESS across the full MESA cohort (tested by applying model 1 but substituting a dependent variable with continuous ESS; flow limitation P = 0.018), this model only explained ∼5% of the variability in the ESS. Importantly, flow limitation in the present study is not only event based but also derived from all flow-limited breaths within the sleep study and is independently associated with EDS in individuals with AHI <15 event/h when controlling for covariates.

Physiological Insights

There are a several physiological pathways that may explain the relationship between flow limitation and sleepiness. In this study, we reveal that flow limitation (currently unscored in clinical practice) is associated with EDS in individuals with low AHI independent of sleep duration and other covariates (age, sex, BMI, race/ethnicity). We also show that this relationship remains significant when including additional variables within sensitivity models. These results demonstrate that flow limitation is a robust measure of sleep-disordered breathing in the AHI <15 events/h group and also indicate that the flow limitation component of sleep-disordered breathing may contribute to sleepiness in a way that is not captured by the AHI or other conventional metrics (e.g., arousal index). As such, within individuals with relatively few scored events (i.e., AHI <15 events/h), we hypothesize that flow limitation likely contributes to sleepiness through novel pathways, such as 1) compromised sleep continuity and quality and/or 2) response to hypercapnia.

First, flow limitation is associated with altered K-complex morphology (increased α, akin to arousal) (38), which diminishes capacity for sleep continuity, in turn adversely impacting both sleep duration (39) and sleep architecture (40) necessary for restorative and restful sleep, ultimately contributing to daytime sleepiness. Although such changes in sleep continuity may be present in traditional sleep staging, persistent flow-limited breathing likely also contributes to electroencephalogram “microarchitecture” (i.e., less than traditional 30-s epoch staging) fragmentation. Indeed, several reviewed studies (41) have described significant increases in daytime sleepiness as a result of increased sleep fragmentation, even when controlling for sleep deprivation and sleep stage parameters, demonstrating that disrupting sleep continuity adversely impacts daytime sleepiness. Interestingly, we did not see a significant relationship with the arousal index; however, it is possible that some individuals (particularly those with EDS) reduce their likelihood of terminating events with arousals through raising their arousal threshold (42, 43), thus making arousal-related associations difficult to quantify. In addition, disruption to sleep continuity may also occur in response to flow limitation through increased ventilatory effort (44).

Second, hypercapnia during sleep has been associated with daytime sleepiness (45) and has even been proposed as a more significant factor than hypoxia with regard to neurocognitive and neurobehavioral outcomes, including sleepiness (46). Flow limitation—the mismatch between ventilatory drive and achieved airflow—typically presents as either 1) maintained ventilatory drive with resultant hypoventilation or 2) increased ventilatory drive and apparent eupneic ventilation. Both presentations increase the propensity toward hypercapnia through either reduced ventilation (47) or increased work of breathing. Furthermore, flow limitation can occur during cyclical respiratory events, leading to acute episodic hypercapnia, but also throughout periods of sustained flow limitation (currently unscored) leading to chronic hypercapnia (48). In some individuals, the hypercapnia-induced increase in ventilatory drive may resolve apneic events (49); however, when not resolved, this may also contribute to ongoing flow limitation and a potentially perpetuating a hypercapnic state that further contributes to increased work of breathing (50) and increased arousal propensity (51), ultimately leading to EDS.

Clinical Impact and Implications

The clinical presentation and diagnosis of sleepiness and OSA are inherently entwined; however, excessive sleepiness may present in isolation. Those individuals who report substantial sleepiness with an AHI <15 events/h present as perplexing in current clinical practice, whereupon the clinician shall investigate other causes and/or potentially recommend a standard treatment for OSA (e.g., continuous positive airway pressure, mandibular advancement splint, positional modification) on the basis of available evidence. However, an optimal response to treatment within patients with EDS without clinical OSA requires an understanding of the underlying cause of sleepiness, such that an appropriate treatment modality is recommended.

It has been proposed that treating flow limitation, beyond treating the apneas and hypopneas, in mild OSA could improve individual clinical outcomes (33–35). However, this requires tools available to identify the individuals who are likely candidates to benefit from treatment paired with the appropriate treatment mechanisms to resolve sleepiness. Flow limitation typically occurs in the natural history of disease before snoring becomes evident (52), and therefore early identification via conventional methods is unlikely. Importantly, the method (14) used in this study facilitates the objective automated quantification of flow limitation from nasal pressure recorded as part of a standard diagnostic PSG. The therapeutic implication is that flow limitation appears to explain the sleep-disordered breathing–related component of an individual’s sleepiness. Subsequent investigations should assess this theory and the extent to which one’s sleepiness may be resolved by treating their sleep-disordered breathing, specifically flow limitation. Although further prospective studies are required to establish whether cross-sectional association in this study extends to predicting treatment response, our results suggest that this tool could be used in the clinic to identify individuals more judiciously in whom sleepiness is most likely to resolve with treatment. In particular, whether there are individuals with AHI <15 events/h and substantial flow limitation who warrant treatment and, conversely, whether there are individuals with similarly low AHI but negligeable flow limitation in whom sleepiness is unlikely to improve with treatment.

Limitations and Further Work

There are several methodological considerations. First, we modeled sleepiness as a binary dependent variable, using a predefined ESS score cutpoint (ESS, ⩾11) to categorize individuals as either presenting with EDS or not. Although this was necessary to handle nonnormally distributed data, it also serves to partially overcome inherent limitations in this subjective assessment. We conducted a sensitivity analysis adjusting the ESS cutpoint and found the relationship between flow limitation and sleepiness remained significant when adjusting the threshold from ESS ⩾9 through ⩾13 (see Figure E7). We suggest that future studies should consider more holistic outcome measures such as the Sleep Apnea Quality of Life Index. Second, it is possible that the self-reported sleep durations we have used in this study were not precisely recorded by participants, because individuals often overestimate sleep duration. However, we justify the use of subjective sleep duration estimates recorded over 7 days because it more likely captures habitual sleep duration than PSG-based sleep duration, which is influenced by “first-night effect.” In supplementary analysis, we tested the impact of PSG sleep duration, and we found no change in results. Third, in the present study, we did not adjust for chronic disease (e.g., epilepsy, dementia, Parkinson’s disease, multiple sclerosis, myotonic dystrophies) or medication use (antihistamines, muscle relaxants, antidepressants, β-blockers, narcotics, benzodiazepines) that may be associated with sleepiness; however, many of these exposures were rare in the cohort and, because they may influence neuromuscular function or ventilation, may have overadjusted for the association with flow limitation. Fourth, the MESA cohort contains a comparatively low prevalence of individuals with EDS relative to similarly aged clinic-based cohorts seeking evaluation for sleep disorders (Table E4). Generalizability of the results should be tested in other community-based cohorts, particularly those with lower ages, and further studies could investigate clinical populations. Previous research has shown that EDS with sleep-disordered breathing is a risk factor for mortality in older adults (53). A similar analysis could be conducted to investigate mortality risk in those individuals with EDS and increased prevalence of flow limitation.

Conclusions

In individuals without moderate or more severe OSA (i.e., AHI <15 events/h), flow limitation is a significant risk factor for sleepiness. This study highlights the importance of considering alternative measures of sleep-disordered breathing, particularly in those with lower AHI values, because factors beyond the AHI appear to contribute independently to sleepiness.