Excessive daytime sleepiness (EDS) is a debilitating complaint reported by 9–20% of the general population (Empana et al., 2009; Boden-Albala et al., 2012; Ohayon, 2012). Hypersomnolence has significant consequences for daily functioning, compromising work productivity and quality of life and increasing risk of motor vehicle and occupational accidents (Dean et al., 2010; Ohayon, 2012; Bioulac et al., 2017). Moreover, accumulating evidence suggests that EDS may adversely impact health, with studies linking hypersomnolence to diabetes (Bixler et al., 2005; Empana et al., 2009; Blachier et al., 2012; Vashum et al., 2015), coronary heart disease (Newman et al., 2000; Boden-Albala et al., 2012), stroke (Blachier et al., 2012; Boden-Albala et al., 2012; Vashum et al., 2015), depression (Bixler et al., 2005; Empana et al., 2009; Ohayon, 2012), and cancer (Ohayon, 2012; Jaumally et al., 2021). EDS has also been found to prognosticate greater risk of cardiovascular and total mortality in population-based studies (Newman et al., 2000; Empana et al., 2009; Boden-Albala et al., 2012; Li et al., 2021).

A pathological tendency to fall asleep is traditionally regarded as the cardinal symptom of obstructive sleep apnea (OSA), a common disorder affecting 34% of men and 17% of women (Peppard et al., 2013). OSA manifests with recurrent episodes of partial or complete collapse of the upper airway during sleep, leading to increased respiratory effort, hypoxemia, sympathoexcitation, and sleep fragmentation. In patients with OSA, EDS is thought to ensue primarily from sleep disruption consequent to abnormal respiratory events, although inflammation, comorbidities or even genetics may contribute (Garbarino et al., 2018). Akin to evidence from the general population, associations between hypersomnolence and excess disease risk have been reported in OSA, including hypertension, diabetes, cardiovascular disease (CVD), and depression (Koutsourelakis et al., 2008; Ronksley et al., 2009). However, other studies did not corroborate such findings (Kapur et al., 2005; Tam et al., 2019). Similarly, data on the relation between EDS and survival in OSA are limited and discordant (Young et al., 2008; Xie et al., 2018; Mazzotti et al., 2019; Trzepizur et al., 2022). Although such discrepancy may be due to several aspects, including the various definitions of EDS and assessment modalities, sex differences may play a role.

OSA exhibits a well-known sex patterning, with women making up a smaller proportion of the OSA population especially at younger ages (Peppard et al., 2013). Pathophysiology and clinical presentation of OSA vary by sex, with women being more likely to report symptoms such as fatigue, depression, and insomnia, and to exhibit lower frequency and duration of apneic events and less intense snoring (Bonsignore et al., 2019). Evidence on whether EDS associated with OSA differs between the sexes is conflicting (Kapur et al., 2005; Koutsourelakis et al., 2008; Huang et al., 2018; Nigro et al., 2018) and, importantly, the interplay between sleepiness and sex in relation to health outcomes has not been systematically studied in OSA. To this end, few investigations have found EDS to be more closely associated with poor health in men with OSA (Basta et al., 2008; Aurora and Punjabi, 2019), while others have reported stronger effects in women with OSA (Huang et al., 2018; Kendzerska et al., 2020). The implications for long-term outcomes are also largely unknown.

We therefore examined sex-specific associations between EDS and prevalence of chronic diseases in a large sample of OSA patients. In a longitudinal analysis, we investigated whether EDS predicts mortality in this group of men and women with OSA (primary endpoint).

In this large OSA sample, EDS measured by ESS showed sex-specific associations with multiple chronic diseases. Curvilinear relations with ESS scores suggest non-linear dynamics between degrees of perceived sleepiness and disease profiles. Furthermore, EDS was an independent predictor of all-cause mortality only in OSA women.

Consistent with prior estimates (Kapur et al., 2005; Koutsourelakis et al., 2008; Nigro et al., 2018), approximately 40% of our patients reported ESS > 10, with similar prevalence between sexes despite women having milder OSA than men. Notwithstanding comparable EDS, sex differences emerged in the relationship between EDS and morbidity. Although crude prevalences of hypertension and CVD were lower in both OSA men and women with ESS > 10 vs. ESS ≤ 10, after adjusting for demographic and sleep covariates EDS remained associated only with decreased risk of hypertension and selectively in men. Some studies have linked EDS with higher blood pressure and greater risk of hypertension especially among severe OSA patients (Kapur et al., 2008; Koutsourelakis et al., 2008; Feng et al., 2012; Ren et al., 2016), but others did not confirm this relationship (de la Peña Bravo et al., 2007; Roure et al., 2008; Prasad et al., 2018). Recent investigations show that OSA patients with normal blood pressure have higher ESS scores than OSA patients with hypertension, with the highest score reported by normotensives with moderate to severe OSA (Martynowicz et al., 2017; Tam et al., 2019). Notably, these studies included predominantly males. Although mechanisms are unclear, the lower degree of sleepiness in those with comorbid OSA and hypertension may be a manifestation of hyperarousal (Tam et al., 2019). Co-occurrence of diseases characterized by sympathetic hyperactivation could offset OSA-induced sleepiness by stimulating alertness via central adrenergic over-excitation. Such effect could be more evident in men due to their relatively higher basal adrenergic tone (Hart et al., 2012).

Conversely, multivariable-adjusted odds for diabetes mellitus were greater in both sleepy men and women relative to their non-sleepy counterparts. Sleepiness is associated with poor health and worse glycemic control among diabetic patients with (Barceló et al., 2008; Aurora and Punjabi, 2019) and without OSA (Chasens et al., 2009; Inkster et al., 2013). EDS predicts diabetes independently of OSA symptoms (Lindberg et al., 2007), and modifies the relationship between OSA and diabetes, with only sleepy OSA patients exhibiting greater risk of diabetes (Ronksley et al., 2009). With respect to biological pathways, hypersomnolence has been associated with increased inflammation, a known precursor of insulin resistance and beta-cell dysfunction, in some (Vgontzas et al., 1997; Li et al., 2017) but not all studies (de la Peña Bravo et al., 2007; Dixon et al., 2007). Decreased cerebral glucose uptake, as observed after sleep deprivation (Wu et al., 2006), may also contribute to both sleepiness and dysregulated glucose metabolism (Tasali et al., 2009).

Modeling the ESS score as a continuous variable unmasked further independent effects, such as on the risk of depression. While female sex and EDS were predictors of depression in OSA participants of the Penn State Adult cohort (LaGrotte et al., 2016), Basta et al. (2008) found that depression was more closely associated with severe sleepiness in men than in women. Our results may aid in reconciling these discrepancies, showing that higher ratings of sleepiness were linearly associated with depression in women, while a curvilinear relation was noted among men. On the other hand, an inverse U-shaped association between ESS and cancer was apparent in women, with reduced risk of cancer at both lower and higher ESS scores. Interestingly, Ohayon et al. (Ohayon, 2012) found a significant association between cancer and moderate sleepiness but not severe sleepiness. In aggregate, these findings illustrate not only distinct patterns of disease profile in sleepy OSA men and women, but also indicate that these risk configurations are partly determined by the definition of pathological sleepiness, thus shedding light into prior conflicting evidence.

A robust sex-dependent pattern emerged also for mortality risk, with ESS independently predicting greater risk of death only in OSA women. Notably, this association emerged only after taking into account demographic characteristics, suggesting that these are important sex-specific confounders in the relation between sleepiness and mortality. Differences in the relationship between EDS and mortality as a function of sex have been noted in general population studies, albeit with discordant findings (Newman et al., 2000; Empana et al., 2009; Boden-Albala et al., 2012; Li et al., 2021). With regard to OSA, sleepiness did not modify the increased mortality risk exhibited by severe OSA participants of the Wisconsin sleep cohort (Young et al., 2008), while the excessively sleepy OSA phenotype manifested the worst survival among OSA symptom subtypes from the Sleep Heart Health Study (Mazzotti et al., 2019) – although no sex-stratified data were reported in either study. A recent study on patients with suspected OSA found that EDS predicted a composite CV endpoint (including all-cause deaths) only in women (Kendzerska et al., 2020). These data are consistent with our results on patients with confirmed OSA, and corroborate the concept that hypersomnolence may be an independent predictor of mortality in women.

The reasons underlying this observation remain unclear, and unmeasured factors may be involved. Among these, growing evidence implicates insufficient sleep as a contributor to increased morbidity and mortality (Covassin and Singh, 2016; Itani et al., 2017). However, studies on differences in habitual sleep duration between sleepy and non-sleepy individuals are mixed, including among OSA (Kapur et al., 2005; Koutsourelakis et al., 2008; Prasad et al., 2018). Because excluding early deaths attenuated the association between ESS > 10 and mortality in women, sleepiness could be an indicator of subclinical diseases or overall poor health status. Nevertheless, estimates were unaltered when using continuous ESS, suggesting limited confounding influence of any fatal undiagnosed conditions.

Mechanistically, although women with OSA have higher levels of inflammatory markers (Gaines et al., 2015), the inflammatory burden is not associated with sleepiness among them (Svensson et al., 2012). Kritikou et al. (2014) noted instead a stronger association between sleepiness and inflammation in OSA men. Other candidates plausibly implicated in our results include increased oxidative stress and vascular dysfunction (Wang et al., 2015; Lira and de Sousa Rodrigues, 2016). However, data on sex differences on these potential mechanisms are scarce and inconclusive, as is their relation with sleepiness in OSA (de la Peña Bravo et al., 2007; Šiarnik et al., 2014). Further research targeting the pathophysiological pathways underlying the heightened risk of death in sleepy OSA women is warranted.

Our study has important clinical implications. Because our study shows that OSA men with hypertension were less likely to report EDS, a negative ESS should not be used to rule out OSA among hypertensive patients. On the other hand, an elevated ESS is a sentinel for diabetes in both sexes, thus underscoring the need to consider screening for diabetes in those complaining of hypersomnolence. From a therapeutic perspective, despite OSA therapy improves EDS (Patil et al., 2019), residual sleepiness persists in a substantial proportion of treated patients. In our study, while OSA treatment could have attenuated the strength of the relation between sleepiness and mortality by ameliorating EDS, further adjusting for PAP acceptance did not affect the results. Although only crude information on therapy was available, it is possible that suboptimal resolution of EDS in OSA women may be implicated in their survival disadvantage.

A strength of our study is its large sample size enabling adequate female representation. Sleepiness was determined by ESS, a validated instrument broadly used to quantify symptoms of sleepiness, thus enhancing applicability of our results. On the other hand, the ESS is inherently prone to recall bias and misperception, and additional studies including objective sleepiness measures are needed. As the relation between EDS and chronic diseases was assessed cross-sectionally, causality cannot be determined. Generalizability may be limited as our sample comprises mostly of White individuals and was drawn from the patient population evaluated at a sleep clinic, and thus subject to referral bias. Because the acceptable definition of hypopnea (i.e., 4% oxygen desaturation scoring standard) was applied when scoring respiratory events, this likely led to lower estimates of OSA severity than if the recommended standard (i.e., 3% desaturation or arousal) were used instead. Last, as the near totality of the sample underwent split-night studies, evaluation of OSA features associated with sleep architecture could not be performed.

In summary, our study shows sex-specific patterns of associations between perceived sleepiness and chronic disease and mortality in OSA patients. Recognizing that the predictive value of EDS is sex-dependent is critical to better understanding its health implications and to develop targeted therapeutic approaches.

The data analyzed in this study is subject to the following licenses/restrictions: Deidentified participant’s data will be available for scientific research upon request submitted to the corresponding author. Requests will be reviewed for suitability. Data will be made available providing IRB approval and a data sharing agreement, in accordance with data sharing policies and Mayo Clinic IRB requirements, are obtained. Requests to access these datasets should be directed to somers.virend@mayo.edu.

The studies involving human participants were reviewed and approved by Mayo Clinic Institutional Review Board. The ethics committee waived the requirement of written informed consent for participation.

NC, DL, AC, ES, and VS conceived and designed the study. NC, DL, PS, and JL analyzed the data. NC, DL, ES, and VS drafted the manuscript. NC, DL, and VS have direct access and verified the data. All authors revised the manuscript for important intellectual content and approved the final version of the manuscript.

Financial support to this study was provided by National Institutes of Health grants RO1 HL114676 and RO1 HL65176 and by a grant from Sleep Number to Mayo Clinic for studies of sleepiness. The contents of this article are solely the responsibility of the authors and do not necessarily represent the official view of the NIH. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

EO is a member of the Board of Directors of the American Academy of Sleep Medicine. VS serves as a consultant for ResMed, Jazz Pharmaceuticals, Bayer, Lilly, Sleep Number, Zoll, Respicardia, and Huxley.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.