REM-Predominant Obstructive Sleep Apnea in Patients with Coronary Artery Disease

Obstructive sleep apnea (OSA) is common in adults with coronary artery disease (CAD). OSA that occurs predominantly during rapid-eye movement (REM) sleep has been identified as a specific phenotype (REM-predominant OSA) in sleep clinic cohorts. We aimed to examine the association of REM-predominant OSA with excessive sleepiness, functional outcomes, mood, and quality of life in a CAD cohort, of whom 286 OSA patients with total sleep time ≥ 240 min, and REM sleep ≥ 30 min, were included. REM-predominant OSA was defined as a REM-apnea-hypopnea-index (AHI) /non-REM (NREM) AHI ≥ 2. In all, 73 (25.5%) had REM-predominant OSA. They were more likely to be female (26.0% vs. 9.9%; p = 0.001), and more obese (42.5% vs. 24.4%; p = 0.003) but had less severe OSA in terms of AHI (median 22.6/h vs. 36.6/h; p < 0.001) compared to the patients with non-stage specific OSA. In adjusted logistic regression models, female sex (odds ratio [OR] 4.64, 95% confidence interval [CI] 1.85–11.64), body-mass-index (BMI; OR 1.17; 95% CI 1.07–1.28) and AHI (OR 0.93, 95% CI 0.91–0.95) were associated with REM-predominant OSA. In univariate linear regression models, there was a dose-response relationship between REM-AHI and Zung Self-rated Depression Scale but not excessive sleepiness, functional outcomes, and anxiety scores. Among the Short Form-36 subdomains, Vitality, Mental Health, and Mental Component Summary (MCS) scores were inversely correlated with REM-AHI. In multivariate linear models, only MCS remained significantly associated with REM-AHI after adjustment for age, BMI, and sex (β-coefficient −2.20, %95 CI [−0.56, −0.03]; p = 0.028). To conclude, female sex and BMI were related to REM-predominant OSA in this revascularized cohort. MCS was inversely associated with REM-AHI in the multivariate model.


Introduction
Obstructive sleep apnea (OSA) is a chronic condition characterized by repetitive episodes of cessation of airflow and arousals from sleep and intermittent hypoxemia [1]. Exhibition of OSA differs significantly between individuals in clinical cohorts; i.e., not all OSA patients demonstrate excessive daytime sleepiness (EDS) and increased risk of cardiovascular morbidity [2]. Further exploration of these differences is therefore crucial for a better understanding of the features, prognosis, and cardiovascular outcomes in patients with OSA.
Rapid eye movement (REM) sleep, which typically accounts for 20-25% of total sleep time, is associated with distinct physiological variations that influence upper airway function [3]. During REM sleep, there is a tendency for upper airway collapse due to cholinergicmediated suppression of genioglossus activity [4,5]. REM sleep is also associated with increased hemodynamic variability, sympathetic activity, and myocardial demand [6]. OSA that occurs predominantly during REM sleep has been identified as a specific phenotype (REM-predominant OSA) and has been reported to occur among 14% to 36% of all OSA cases [7]. OSA events during REM sleep are typically longer, more frequent, and are related to greater oxyhemoglobin desaturation than events during non-REM sleep [6]. Patients with REM-predominant OSA tend to be younger women and have a less severe OSA in terms of the apnea-hypopnea index (AHI) [3]. Changes in sleep architecture regarding REM sleep have been associated with depressive mood [8,9] and anxiety [10], but not with impairment in quality of life [11]. REM-predominant OSA has been independently related to prevalent and incident hypertension as well as impaired glucose metabolism [12,13]. A dose-response relationship between REM-AHI and non-dipping blood pressure has also been reported [14].
Coronary artery disease (CAD) is one of the most common conditions associated with increased morbidity and mortality [10]. The occurrence of OSA among patients with CAD is very high (50% as compared to 10-20% in the general adult population) [2], and OSA patients have an increased risk for incident CAD compared to individuals without OSA [2,15]. Moreover, major adverse cardiovascular and cerebrovascular events have been reported to be more common in patients with CAD with concomitant OSA [16]. Less is known regarding the determinants of REM predominant OSA and its associations with EDS, functional outcomes, anxiety, depression, and quality of life in adults with CAD.
The "Randomized Intervention with CPAP in Coronary Artery Disease and Obstructive Sleep Apnea" (RICCADSA) trial primarily addressed the impact of CPAP on cardiovascular outcomes in revascularized patients with CAD and concomitant OSA. In the current secondary analysis, we aimed to determine the occurrence of REM-predominant OSA and its association with daytime symptoms, depressive mood, anxiety, and quality of life at baseline in the RICCADSA cohort.

Study Design and Participants
The RICCADSA cohort has been described in detail previously [15,17]. In brief, patients with CAD who underwent percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) in the Skaraborg County of West Götaland, Sweden, were recruited for the main trial between 2005 and 2010, and the follow-up for the primary outcomes was completed in May 2013 [15]. As illustrated in Figure 1, 399 of 511 patients had OSA based on an Apnea-Hypopnea-Index (AHI) ≥15/h on a home sleep apnea testing (HSAT) in the entire RICCADSA cohort. A polysomnography (PSG) was conducted in a hospital in the OSA group before the study started [17]. For the current protocol, only the patients with a REM sleep duration of at least 30 min were included. REM-predominant OSA was defined as a REM-AHI/non-REM-AHI ≥ 2 [3]. None of the patients were on treatment for OSA.

Questionnaires
Epworth sleepiness scale (ESS): The ESS questionnaire [18] was used to determine subjective daytime sleepiness. As previously described, the ESS questionnaire contains eight questions to address the possibility of dozing off under eight different conditions in the past month. A cut-off value of 10 out of 24 was used for categorizing the patients with excessive daytime sleepiness in the RICCADSA trial [17].
Zung Self-rated Depression Scale (SDS): The Zung SDS is a commonly accepted survey that provides both a total score and a categorical rating of depression [19]. In brief, 20 questions are included with a score ranging from 1 to 4 points, which gives a total raw score from 20 to 80. As previously described in detail, the raw score is multiplied by 1.25, and the total result ranges from 25 to 100. The subjects with a score below 50 were categorized as normal, and the patients with a score of 50 or more were classified as having depression [19].
Zung Self-rated Anxiety Scale (SAS): The Zung SAS is a Likert scale that compromises 20 items to measure physiological and psychological symptoms [20]. As in the Zung SDS, each item is rated on a 4-point scale from 1 (none, or a little of the time) to 4 (most, or all of the time), and the raw scores (range 20-80) are converted into index scores (range 25-100) by multiplying 1.25 [20]. Individuals with a SAS score of at least 45 points were defined as having clinically significant anxiety [21].
Functional Outcomes of the Sleep Questionnaire (FOSQ): FOSQ is a survey regarding sleep-related symptoms, divided into five dimensions, including activity level, vigilance, intimacy, and sexual relationships, general productivity, and social outcome [22]. Item responses range from no difficulty (4 points) to extreme difficulty (1 point). The total score is calculated as the sum of the subscale scores, with a total score ranging from 5 to 20. Lower scores suggest function impairment, and the clinically significant impairment in the FOSQ is defined as a total score of fewer than 17.9 points [23].
Short Form-36 Health Survey (SF-36): Data regarding the Health-Related Quality of Life (HRQoL) were collected with the SF-36, which measures eight different domains

Questionnaires
Epworth sleepiness scale (ESS): The ESS questionnaire [18] was used to determine subjective daytime sleepiness. As previously described, the ESS questionnaire contains eight questions to address the possibility of dozing off under eight different conditions in the past month. A cut-off value of 10 out of 24 was used for categorizing the patients with excessive daytime sleepiness in the RICCADSA trial [17].
Zung Self-rated Depression Scale (SDS): The Zung SDS is a commonly accepted survey that provides both a total score and a categorical rating of depression [19]. In brief, 20 questions are included with a score ranging from 1 to 4 points, which gives a total raw score from 20 to 80. As previously described in detail, the raw score is multiplied by 1.25, and the total result ranges from 25 to 100. The subjects with a score below 50 were categorized as normal, and the patients with a score of 50 or more were classified as having depression [19].
Zung Self-rated Anxiety Scale (SAS): The Zung SAS is a Likert scale that compromises 20 items to measure physiological and psychological symptoms [20]. As in the Zung SDS, each item is rated on a 4-point scale from 1 (none, or a little of the time) to 4 (most, or all of the time), and the raw scores (range 20-80) are converted into index scores (range 25-100) by multiplying 1.25 [20]. Individuals with a SAS score of at least 45 points were defined as having clinically significant anxiety [21].
Functional Outcomes of the Sleep Questionnaire (FOSQ): FOSQ is a survey regarding sleep-related symptoms, divided into five dimensions, including activity level, vigilance, intimacy, and sexual relationships, general productivity, and social outcome [22]. Item responses range from no difficulty (4 points) to extreme difficulty (1 point). The total score is calculated as the sum of the subscale scores, with a total score ranging from 5 to 20. Lower scores suggest function impairment, and the clinically significant impairment in the FOSQ is defined as a total score of fewer than 17 [24]. An SF-36 questionnaire is a subjective tool that reflects the individual's well-being preceding four weeks prior to the time of the survey. Domain and summary component scores range from 0 to 100; higher scores correspond to better health status or well-being.

Sleep Recordings
For the HSATs, the Embletta ® Portable Digital System device (Embla, Broomfield, CO, USA) was used. As explained previously [15], the HSAT system included a nasal pressure detector, two respiratory inductance plethysmography belts (RIP) for thoracoabdominal movements and body position, and a finger pulse-oximeter for heart rate and oxyhemoglobin saturation (SpO 2 ). Apnea was defined as at least ≥90% cessation of airflow and hypopnea was defined as a 50% reduction, at least, in thoracoabdominal movement and nasal pressure amplitude for ≥10 s [25]. Additionally, the total number of significant falls in SpO 2 (≥4% from the immediately preceding baseline) was scored and the oxygen desaturation index (ODI) was determined as the number of significant desaturations per hour of estimated sleep. Obstructive events with a clear reduction in RIP belts and in the nasal pressure amplitude for at least 10 s were also recorded as hypopneas if there was a significant desaturation [25]. For the overnight PSG in-hospital for the OSA group, before the study started, a computerized recording system (Embla A10 ® , Embla, Broomfield, CO, USA) was used. As previously described [17], the PSG system included sleep monitoring through three-channel electroencephalography (EEG [C4/A1, C3/A2, CZ/A1]), two-channel electrooculography (EOG), one-channel submental electromyography (EMG), bilateral tibial EMG and two-lead electrocardiogram (ECG) in addition to the cardiorespiratory channels as described for the Embletta system above [17]. PSG recordings were scored by an observer blinded to clinical data and baseline screening results from the previous HSAT recordings. Obstructive events on the PSGs were scored according to the same criteria applied to the HSATs.

Comorbidities
As described previously, baseline anthropometrics, smoking habits, and medical history of the entire study population were obtained from the medical records [15]. Obesity was defined as a body mass index (BMI) ≥ 30 kg/m 2 , and abdominal obesity was defined as the waist-to-hip ratio (WHR) ≥ 0.9 for men and WHR ≥ 0.8 for women, respectively [26].

Statistical Analysis
The Shapiro-Wilk test was used to test the normality assumption of the current data for all variables. Descriptive data are shown as means and standard deviation (SD), or median with interquartile ranges (IQR) for continuous variables, and as a percentage for categorical variables. For comparison between groups, an independent sampled student t-test or, when appropriate, the Mann-Whitney U test was used. The Chi-square test was used for the comparison of categorical variables. Pearson's correlation analysis was used to determine the relationship between REM-AHI levels and the other continuous variables. A logistic regression model was applied to determine variables associated with REM predominant OSA. Age, BMI, sex, percentage of stage 3 of non-REM-sleep (slow-wave sleep [SWS]) and AHI (Model 1) or ODI (Model 2), as well as significant associates (if any) in the univariate analysis, were entered into the multivariate models. Linear univariate regression analysis was used to evaluate the association between REM-AHI and the continuous variables of the ESS, FOSQ, SF-36, Zung SDS, and SAS scores, respectively, and the significant results were entered into the multivariate models with age, BMI, and female sex covariates. The significant variables associated with REM-AHI were tested separately against non-REM-AHI to address if the REM-AHI-related associations were exclusive for the REM sleep stage. All statistical tests were two-sided, odds ratios (ORs) with 95% confidence interval (CI) were reported, and a p-value < 0.05 was considered significant. Statistical analysis was performed using the Statistical Package for Social Sciences, version 22.0 for the Windows ® system (SPSS ® Inc., Chicago, IL, USA).

Results
A total of 399 participants with OSA were eligible for the current protocol. After excluding patients with REM sleep duration less than 30 min (n = 107), total sleep time less than 4 h (n = 1), and AHI less than 5 events/h on the PSG (n = 5), 286 (mean age 63.7 ± 7.9 years, 86.1% men) remained as the final study population (Figure 1).
In all, 73 (25.5%) of the study cohort had REM-predominant OSA. As shown in Table 1, there were more women among the participants with REM-predominant OSA compared to those with non-stage-specific OSA (26.0% vs. 9.9%; p = 0.001). Average BMI was significantly higher, and obesity was more common (42.5% vs. 25.4; p = 0.003) in the REMpredominant group, whereas there were no significant differences between the groups regarding age and comorbidities. Definitions of abbreviations: AMI = acute myocardial infarction; BMI = body mass index; CABG = coronary artery by-pass grafting; OSA = obstructive sleep apnea; REM = rapid eye movements. Table 2 demonstrates that sleep efficiency, as well as the duration and proportion of slow-wave sleep (SWS), was significantly higher in the REM-predominant group, whereas the non-stage specific OSA group had more severe OSA in terms of total AHI and oxygenation indices. Per selection criteria requiring at least 30 min of REM sleep, the REMdependent OSA group tended to have longer REM sleep time and significantly more severe REM-AHI than the non-stage specific group. As shown in Table 3, there were no significant differences between the groups regarding EDS, functional outcomes, mood, and HRQoL measures at baseline. Table 3. Comparison of excessive sleepiness, functional outcomes, depression, anxiety, and quality of life measures of the study participants (n = 286) polysomnographic findings of the study population (n = 286). In unadjusted logistic regression analyzes, there were significant positive relationships between REM-predominant OSA and female sex, obesity, BMI, and SWS, whereas AHI and ODI were inversely correlated (Table 4). In the adjusted models, female sex, obesity, and BMI, as well as AHI and ODI, remained significant determinants of REM-predominant OSA (Table 4).  As illustrated in Figure 2A, there was significant inverse relationship between REM-AHI and age (r = −0.124, p = 0.035), whereas the correlation with BMI was positive ( Figure 2B; r = 0.351, p < 0.001). ponent summary; REM = rapid eye movements; SAS = self-rating anxiety scale; SDS = self-rating depression scale; SF = short form; SWS = slow wave sleep (stage 3 of non-REM sleep).

Discussion
The current analysis demonstrated that REM-predominant OSA was significantly associated with female sex and obesity in this revascularized CAD cohort. Despite a lower degree of OSA in terms of AHI and oxygenation indices, the patients with REMpredominant OSA had similar characteristics regarding comorbidities, EDS, functional outcomes, mood, and QoL measures as did the patients with non-stage specific OSA. Further analyzes revealed significant associations of the REM-AHI with depressive mood scores as well as SF-36 subdomains Vitality and Mental Health, which all, however, disappeared in the multivariate models. Only the SF-domain MCS remained significantly associated with REM-AHI independent of age, BMI, and female sex.
To our knowledge, this is the first study to evaluate the association of REM-predominant OSA with excessive sleepiness, functional outcomes, anxiety, and depression, as well as the quality of life measures in a CAD cohort. In prior literature, Zinchuk and colleagues have reported that adults with REM-predominant OSA in sleep clinic cohorts are usually younger women with more efficient sleep and longer total sleep time compared to patients with non-stage specific OSA [3]. Our results suggesting an inverse relationship between REM-AHI and age as well as the dominance of female sex among the REM-predominant OSA supports those findings to be valid also in patients with CAD. Our results are in line with the findings from prior sleep-clinic-based studies, which showed that REM-predominant OSA was not related to excessive sleepiness as measured by multiple sleep latency tests [27,28], as well as with the findings from the large population-based Sleep Heart Health Study, which showed no relationship between REM-AHI and ESS scores after adjustment for confounding factors such as age, sex, and race [11].
Few studies addressed the functional outcomes in patients with REM-predominant OSA. Though the definition of hypopneas, as well as REM-predominant OSA, differed, the total FOSQ score tended to be lower among patients with REM-predominant OSA compared to non-stage specific OSA in 2 studies [29,30]. Our results do not support those findings, which might reflect the differences in clinical characteristics of sleep clinics vs. cardiac cohorts, as the patients from sleep clinics are usually symptomatic and have lower FOSQ scores at baseline. The average total FOSQ score was around 12 units in both studies [29,30], whereas our cohort demonstrated values of around 18 units, which suggests that many patients with CAD are with concomitant OSA are asymptomatic and do not have impairment in sleep-related functional outcomes.
Changes in sleep architecture regarding REM sleep have been associated with depressive mood [8,9] and anxiety [10]. In general, sleep pattern changes among patients with depression consist of impaired sleep onset and maintenance, early morning awakening, as well as reduced deep sleep and disinhibition of REM sleep [8]. Sleep disturbances are prevalent in some mental diseases, such as anxiety disorders, and anxiety disorders have been reported to be more common in REM-related OSA in a sleep clinic cohort [10], but we did not observe any difference in the anxiety scores in our CAD cohort. Of note, the Hospital Anxiety-Depression Scale was used in that study [10], whereas Zung SDS and Zung SAS scores were conducted in the current protocol. We found a significant association between REM-AHI and the Zung SDS scores, but this association disappeared after adjustment for the female sex.
Prior studies regarding the association of REM-predominant OSA with the QoL measures have not been consistent. The findings from the Sleep Heart Health Study [11] suggested that NREM-AHI but not REM-AHI was significantly associated with impairment in the PCS and MCS after adjustment for the confounding factors. The inclusion of individuals with a REM sleep time of less than 30 min, as well as the statistical analysis including both REM-AHI and NREM-AHI in the same model (collinearity), might explain some of the differences between our findings and the results of that population-based study [11]. We found significant linear correlations between REM-AHI and Mental Health and Vitality scores, but these associations were not significant anymore in the multivariate analysis. Notwithstanding, the association of the REM-AHI with the MCS remained significant after adjustment for age, BMI, and female sex. Given the absence of a significant association between non-REM-AHI and MCS, our findings suggest that the decline in MCS is exclusive to the REM sleep stage.
It should be noted that there have been several definitions of REM-dependent OSA in literature: (I) overall AHI ≥ 5 and REM-AHI/NREM-AHI ratio ≥ 2; (II) overall AHI ≥ 5, REM-AHI/NREM-AHI ratio ≥ 2, and NREM-AHI < 15; or (III) overall AHI ≥ 5, REM-AHI/NREM-AHI ratio ≥ 2, NREM-AHI < 15, and at least 10.5 min of REM sleep duration [31]. There have also been studies including only subjects with at least 30 min of recorded REM sleep [32] to reduce the possibility of exaggerating the effect of REMpredominant OSA in individuals for short REM duration, which is also applied in the current study. However, the strict criterion brings a potential bias, as the excluded patients are probably the ones with severe OSA (overall AHI ≥ 30), who otherwise would belong to the non-stage-specific OSA.
Notwithstanding, the significant association of REM-AHI with the MCS despite the lower degree of OSA severity highlights the need for a more complete characterization of each patient's pathophysiology and a sex-stratified approach in the management of patients with CAD with concomitant OSA. This may be particularly important for increasing adherence to CPAP treatment to cover the early morning hours before awakening, thus treating the whole REM sleep period for better MCS outcomes. Given that there is a decrease in upper respiratory muscle tone in the REM period [33], which results in lower respiratory drive, and that longer periods of respiratory events cause more severe hypoxemia [34], a dose-response relationship between REM-AHI and hypertension has also been suggested [35]. However, more research with larger studies considering individual risk factor profiling in refining treatment strategies in OSA phenotypes is still urged.
We should acknowledge certain limitations. Firstly, the power estimate for the entire RICCADSA cohort was conducted for the primary outcome and not for the secondary outcomes assessed in this subprotocol. Secondly, the EDS definition was based on the ESS threshold, which may not be precise in a CAD population. Thirdly, our results are not generalizable to adults with OSA in the general population or in sleep clinic cohorts.

Conclusions
To conclude, female sex and BMI were related to REM-predominant OSA in this revascularized cohort. MCS was inversely associated with REM-AHI in the multivariate model. Further research is needed to establish better whether these patients will obtain any benefit from long-term CPAP therapy and whether sex-stratified treatment strategies should be applied for patients with CAD and REM-predominant OSA.
Author Contributions: Y.P. designed the study and performed the patient recruitment and clinical follow-ups. B.B., Y.C. and Y.P. performed the statistical analysis. B.B., Y.C. and Y.P. prepared the manuscript and drafted the article. All authors (B.B., Y.C., J.N., P.J.S.J. and Y.P.) interpreted the data. Y.P. obtained study funding and takes full responsibility for the work, including the study design, access to data, and the decision to submit and publish the manuscript. All authors have read and agreed to the published version of the manuscript. Conflicts of Interest: Y.P. received institutional grants from ResMed for the main RICCADSA trial. P.S. reports personnel fees from Emmi Solutions, Jazz Pharmaceuticals, Itamar Medical, Separation Design Group, Belluscura, grants and personal fees from ResMed, Inspire Medical Systems, grants from Philips-Respironics, National Football League, and PinMed, other from Biologix, outside the submitted work; B.B., Y.C. and J.N. report no conflict of interest.