Fatigue and sleepiness determine respiratory quality of life among veterans evaluated for sleep apnea

This cross-sectional retrospective study of prospectively collected clinical data was undertaken in a cohort of military veterans referred due to OSA symptoms for a standardized assessment protocol that included both a multi-battery questionnaire and confirmatory polysomnography. We intentionally included a respiratory-specific HRQL measure in the questionnaire so that we could clinically assess breathing problems as well as carry out an analysis with this as our central patient-centered outcome.

Of the 1578 persons analyzed, 823 (52.1%) met criteria for moderate to severe OSA (AHI ≥15/h), whereas 755 (47.9%) manifested mild (n = 451, AHI < 15/h and ≥5/h) or no OSA (n = 304, AHI < 5/h) (Table 1). Just under one-third of the study participants met criteria for severe OSA (30.5%). Demographic data by OSA status are presented in Table 1. Overall, the group was predominantly male (93.9%) and obese (median BMI 32.2 [IQR 8.1]), with a low prevalence of current tobacco smoking (12.6%). There were 262 (16.6%) who carried a diagnosis of concomitant lung disease at the time of the referral. By diagnosis, these were: COPD (N = 94; 6.0%); asthma (N = 181; 11.5%); and interstitial lung disease (N = 8; 0.5%). Those with moderate to severe OSA compared to those with mild or no disease on average were 6-years older, had a higher BMI, were more likely to be male, more likely to be married or partnered, and were less likely currently employed (all p < 0.05, Table 1). There were no statistically significant differences by OSA status for race-ethnicity or educational attainment. In terms of concomitant lung disease, those with moderate to severe OSA were statistically more likely to have concomitant ILD but not COPD or asthma.

The median overall ESS score was 11 [IQR 9], with 838 (53.1%) falling above the 10-point score cut off for increased daytime sleepiness. The overall FSS score was 42 [IQR 23], with 968 (61.3%) falling above the 36-point score cut off for increased fatigue. The ESS and FSS scores were moderately intercorrelated (r = 0.40; p < 0.01). The correlations between ESS and FSS when stratified by OSA status were: r = 0.37 among those with mild to no OSA and r = 0.46 among those with moderate or severe OSA.

We tested sleepiness (ESS), fatigue (FSS), and the degree of OSA (AHI) as predictors of physical and respiratory specific HRQL, taking into account other covariates (Table 2). Both ESS and FSS were significantly associated with poorer SF-12v2 PCS (p < 0.01), although the point estimate of effect was greater for the latter (−3.3 and −7.6 point lower PCS per standardized units of ESS and FSS, respectively). A similar pattern was evident for respiratory-specific HRQL measured by the AQ-20R. In contrast, AHI was not associated with SF-12v2 PCS or AQ-20R.

In analyses stratified by the presence or absence of OSA (Fig. 1), increased fatigue was associated with poorer respiratory HRQL in both the none-to-mild and the moderate-to-severe OSA groups. In comparison, daytime sleepiness was not associated with poorer respiratory-specific HRQL in those with none-to-mild OSA and demonstrated an attenuated effect (standardized beta 0.6 for ESS compared to 2.2 for FSS) in those with moderate-to-severe OSA. Also as shown in Fig. 1, the pattern was similar when stratified by the presence or absence of obesity.

We further explored the associations of ESS and FSS with respiratory-specific HRQL, taking into account generic HRQL (SF-12v2 PCS), OSA severity, comorbid chronic lung conditions, and the covariates included in the previous multivariate analyses (Table 3). Both ESS and FSS remained significantly associated with AQ-20R, the FSS manifesting a somewhat greater effect. SF-12v2PCS and comorbid lung conditions were associated with poorer respiratory specific quality of life measured by the AQ-20R, whereas presence or severity of OSA was not. Excluding 1275 subjects with portable sleep study testing (remaining n = 303) had no substantial effect on these multivariate analyses except that the associations with COPD (p = 0.06) and ILD (p = 0.94) were no longer statistically significant.

The assessment of HRQL, a patient-centered outcome encompassing health, functional, and social factors, is critical to inform individual patient management decisions and, in the aggregate, health care policy. Asthma, COPD, and interstitial lung disease all are chronic respiratory conditions associated with poorer HRQL [24]. Given the potential combined effects from comorbid OSA and chronic respiratory disorders such as these, evaluating generic and respiratory HRQL in populations with multiple morbidities is important.

Our analysis, carried out in a relatively large sample of military veterans undergoing sleep evaluations for suspected OSA, yielded two key findings. First, sleepiness and fatigue were important determinants of respiratory-specific HRQL, independent of OSA, respiratory comorbidity, or even generic physical HRQL. Second, OSA severity per se was not associated with a significant decrement in either generic physical HRQL or respiratory-specific HRQL. Although the lack of a strong association of OSA with physical HRQL has been reported [11, 30], respiratory-specific HRQL previously had not been studied in parallel. Thus taking these two findings together, among those with symptoms or clinical findings suggesting OSA, regardless of the ultimate diagnosis, fatigue and sleepiness negatively impact the key patient-centered outcome of respiratory-specific HRQL.

Although untreated OSA may be associated with poorer sleep-specific quality of life, both general population and sleep clinic-based studies have reported inconsistent associations between OSA severity and generic physical HRQL [5‐7, 9‐11, 30, 31]. Additionally, studies evaluating symptomatic patients referred for OSA assessment report similar findings to our investigation. Among 109 patients with suspected OSA referred for PSG testing (mean age 53 years, 77% male, BMI 31, AHI 33.6), ESS was negatively associated with physical HRQL measured by the SF-12 PCS (p < 0.001), whereas AHI was not associated with that outcome [6]. In another series of 135 consecutive patients with suspected OSA, ESS was associated with generic physical HRQL, but not the AHI or EEG arousal index [10]. Other referral series also fail to show a relationship between AHI and physical HRQL [5, 7, 11]. Given these findings and our own, there appears to be no clear association between OSA severity (as measured by AHI) and generic physical HRQL, even though subjective sleepiness (measured by the ESS) indeed is correlated consistently with HRQL. This interpretation is consistent with another study in which an association of general HRQL with ESS was observed, but not with OSA [10].

Our study goes further still by also taking into account other respiratory co-morbidities as well as other symptoms such as fatigue. Even when patients with active medical and psychiatric disorders are excluded, ESS but not AHI, respiratory disturbance index or oxygen desaturation index is still negatively associated with physical HRQL measured by the SF-12 PCS [30].

Similar to our results, studies in patients with chronic respiratory disorders such as COPD report no correlation between AHI and respiratory-specific HRQL measured by the St. George Respiratory Questionnaire (SGRQ) [12]. It is well established that both generic and respiratory-specific HRQL are poor among patients with asthma and COPD [15, 16, 19, 32]. However, the interrelationships among OSA, respiratory comorbidity, sleepiness, and respiratory-specific HRQL are less clear-cut. This interrelationship is likely to be complex given that asthma and COPD are associated with poor sleep quality that may be manifested as daytime sleepiness, fatigue, and poorer general HRQL, independent of OSA [2, 12, 16, 33, 34]. Our findings support those links, but adds to picture by taking into account both OSA and other chronic respiratory comorbid conditions, yet still observing a potent negative relationship between increased daytime sleepiness and fatigue and respiratory-specific HRQL. Standardizing the scoring metric, we observed a greater negative impact of fatigue relative compared to daytime sleepiness on generic and respiratory-specific HRQL. Studies suggest subjective fatigue and sleepiness are indeed distinct entities [35], albeit not wholly independent consistent with our finding that FSS and ESS scores were moderately intercorrelated (r = 0.38). Some studies have reported that AHI is poorly correlated with fatigue [36, 37], whereas other chronic respiratory disorders are linked to fatigue [38, 39]. Thus, our analysis supports a growing consensus that fatigue negatively impacts generic and respiratory-specific HRQL while specifically showing that this relationship transcends OSA and even other comorbid respiratory disease.

Our study supports the utility of the AQ20-R in measuring respiratory-specific HRQL in OSA. The AQ20 was developed as a simple respiratory HRQL instrument that correlates well with other respiratory (SGRQ) and general (SF-12 and SF-36) HRQL measures in populations with various types of respiratory disorders [24, 25]. Additionally, the AQ20-R is an independent predictor of healthcare utilization outcomes in those with chronic respiratory disease [24]. Our study is novel in assessing the impact of subjective fatigue and sleepiness on respiratory HRQL measured by the AQ20-R. Consistent with our findings, fatigue (measured by the Multidimensional Fatigue Inventory) strongly correlated with respiratory-specific HRQL measured by the SGRQ (r = 0.75, p < 0.01) [20]. In patients with chronic respiratory disorders, there appears to be a complex inter-relationship among fatigue, sleep quality, anxiety, depressive symptoms, and awake-dyspnea [40]. Interestingly, the AQ20 appears moderately correlated with subjective sleep disturbance in asthmatics (r = 0.51, p ≤ 0.001) [34], while both non-respiratory medical and psychiatric comorbidities are independent modulators of HRQL measured by the AQ20-R [41].

Taken together, the clinical implications of our observations are that the evaluation and treatment of respiratory complaints among those with suspected OSA is important, particularly in those reporting fatigue. The AQ20-R assesses respiratory-specific HRQL. Taking this or comparable disease-specific patient-centered measures of health into account in models of health care delivery is likely to be important for clinical management and improved patient satisfaction among those with suspected OSA, whether or not they ultimately prove to have that condition.

Our study has potentially important limitations. These limitations could argue against the generalizability of our findings, the consistency of our observations with those of other relevant studies, our emphasis on respiratory–specific or even general physical status as opposed to mental HRQL, or the validity of our definition of disease. We evaluated generally older obese male military veterans referred for suspected OSA, which should temper any general population inferences. However, studies in populations consisting of older, primarily male non-veterans with suspected OSA report similar results to ours [6, 10, 30]. Second, the applicability of general physical HRQL measures such as the SF-12 in OSA populations has been questioned, particularly when comorbid medical and psychiatric conditions are not controlled for [42], and because sleep and OSA-specific HRQL batteries correlate only modestly with generic HRQL [43‐45]. Additionally, generic mental health rather than physical HRQL may correlate with AHI [46, 47]. We limited our analysis of the SF-12 to the PCS and did not investigate associations with the SF-12 Mental Component Scale (that measures psychological HRQL), nor did we study psychological comorbidity. Third, in our bivariate analysis, pooling mild OSA together with no OSA to compare this categorization against moderate-severe OSA could lead to disease misclassification. The Sleep Heart Health Study, however, reported no clinically relevant differences in SF-36 and ESS scores between those with mild vs. no OSA [9, 48] and the recent Apnea Positive Pressure Long-term Efficacy Study similarly reported no significant differences in ESS and Sleep Apnea Quality of Life Index between those with mild vs. no OSA [49]. Finally, given limited spirometric data, we could not analyze the association of respiratory impairment (as opposed to the presence or absence of a diagnosed respiratory condition) with HRQL. Considering these potential limitations as whole, they are modest at most and are unlikely to explain the associations that we observed, argue strongly for an alternative explanation of the findings of this analysis, or seriously undermine the credibility of our findings.

In summary, daytime sleepiness and fatigue contribute to poorer respiratory-specific HRQL, taking into account the effects of OSA, respiratory comorbidity, and generic HRQL. The clinical implications of this analysis are that the evaluation and management of non-OSA causes of fatigue, sleepiness, and daytime breathing problems are relevant to patients with complaints suggesting a diagnosis of OSA. This is the case whether or not such patients ultimately prove to have OSA or, if they do, the degree of its severity. Furthermore, we believe the AQ20-R provides additional utility in HRQL assessments as well as being a simple patient-centered measure of breathing problems in those suspected of having OSA. Future longitudinal assessments should evaluate whether respiratory-specific HRQL, with or without concomitant lung disease, predicts subsequent health outcomes in OSA populations and the extent to which fatigue or sleepiness account for such relationships.