Effects of obstructive sleep apnea during rapid eye movement sleep on cardiac autonomic dysfunction: Results from the Shanghai sleep health study cohort

In our large‐scale study, the correlation between obstructive sleep apnea (OSA) related to rapid eye movement (REM) sleep and cardiac autonomic dysfunction was assessed by standard polysomnography (PSG). Cardiac autonomic dysfunction was evaluated by the measurement of heart rate variability (HRV). The cardiovascular disease (CVD) risk was determined using the cross‐sectional prevalence of CVD and its overall 10 year risk according to the Framingham risk score (FRS). 4152 individuals were included in the study. A higher apnea–hypopnea index during REM sleep (AHIREM) was correlated with increased CVD risk. The adjusted odds ratios (95% CIs) for CVD prevalence and its high 10 year risk in participants having severe OSA during REM sleep (AHIREM ≥30 events/h) were 1.452 (1.012–2.084) and 1.904 (1.470–2.466) in the demographic adjusted model and 1.175 (0.810–1.704) and 1.716 (1.213–2.427) in the multivariate adjusted model, respectively, compared with the group with a AHIREM of <5 events/h. Fully adjusted multivariate linear regression models showed the independent association between AHIREM and a more elevated ratio of low‐frequency and high‐frequency (LF/HF) and LF in normalised units [LF (n.u.)] (P = 0.042, P = 0.027 in all participants and P = 0.033, P = 0.029 in participants with AHI during non‐REM sleep <5 events/h, respectively). Mediation analysis demonstrated that OSA during REM sleep and CVD risk was significantly mediated by LF/HF and LF (n.u.). OSA during REM sleep may be a marker behind CVD risk because it promotes cardiac autonomic dysfunction.

The levels of sympathetic activity and cardiovascular instability (such as heart rate, blood pressure, and myocardial metabolic demand) are greater throughout rapid eye movement (REM) sleep compared with throughout non-rapid eye movement (NREM) sleep in patients with OSA and healthy individuals (Abboud & Kumar, 2014). Consequently, REM sleep-related OSA may cause greater serious consequences than NREM-related OSA.
Ischaemic events in CVD may be triggered by acute sympathetic and haemodynamic changes during REM sleep (Gami et al., 2005). Although the correlation between REM sleep-related OSA and hypertension and non-dipping nocturnal blood pressure (i.e., markers of CVD risk) has been documented (Mokhlesi et al., , 2015, OSA during REM sleep does not affect clinical features in other studies (Conwell et al., 2012;Haba-Rubio et al., 2005). The influence of respiratory events related to REM sleep on cardiac homeostasis remains largely unexplored.
Noninvasive assessment of cardiac autonomic function is conducted by measuring heart rate variability (HRV), constituting the variations in beat-to-beat intervals affected by activities of both the sympathetic and parasympathetic nervous systems. The prognostic significance of HRV reduction which generally indicates the dysfunctionality of cardiac autonomic function, as stated by large epidemiological studies (Dekker et al., 2000;Goldberger et al., 2019). Analyses of HRV have shown that OSA patients, including patients without CVD, exhibit dysregulation of the autonomic nervous system throughout sleep and awake periods; such dysregulation includes the elevation of sympathetic activation and reduction of parasympathetic activation (Stein & Pu, 2012). Specifically, the rates of cardiovascular events and mortality increase as HRV deteriorates (Gami et al., 2005). Therefore, HRV indicating cardiac autonomic function may be regarded as a marker of potential CVD risk (Hillebrand et al., 2013).
Despite assessing the degree of OSA through the apnea-hypopnea index (AHI) (American Academy of Sleep Medicine Task Force, 1999), this parameter may not fully describe the disease severity or treatment response (Kirkham et al., 2015;Lim et al., 2020). In addition to AHI, other indices (e.g., HRV) can be used to characterise disease severity and to assess the risks of cardiovascular complications. However, we do not know much about how HRV is correlated with REM sleep-related OSA. There have been conflicting reports regarding the associations between cardiac autonomic dysfunction and OSA in REM sleep (Oh et al., 2019;Trimer et al., 2014). There was more focus on specific clinical results than on investigating the physiological processes behind the high CVD risk correlated with OSA during REM sleep. Our aim is to identify the impacts that REM sleep-related OSA exerts on CVD risk and cardiac autonomic dysfunction. Our hypothesis was that REM sleep-related OSA had an association with CVD risk and altered HRV, which indicates the progression of cardiac autonomic dysfunction.

| Study population
By enrolling consecutive subjects in the Shanghai Sleep Health Study Cohort, which was developed as a dynamic cohort, the associations between OSA and metabolic abnormalities were investigated (Huang et al., 2022). To ensure meaningful assessment, the exclusion criteria were: age <18 years; ongoing OSA therapy, including continuous positive airway pressure (CPAP), mandibular advancement device, or upper airway surgery; consumption of coffee or using medications (e.g., βblockers) that may influence heart rhythm; use of anxiolytics, antidepressants, hypnotics, or antipsychotic drugs; having other sleep disorders, including restless leg syndrome, narcolepsy, or upper airway resistance syndrome; <4 h recorded sleep or <30 min of recorded REM sleep; or missing HRV data (see Supplementary Material: Data S1 for details). The Ethics Committee of Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine approved the study protocols that follow the tenets of the Declaration of Helsinki. Moreover, the participants signed written informed approval to enroll in this research and to using their personal information.

| Anthropometric and biochemical measurements
The clinical assessment process in the Shanghai Sleep Health Study involved fasting blood sample collection and anthropometry. Before polysomnography (PSG), the mean values were recorded for two con-  (Catapano et al., 2016;Peppard et al., 2000). The prevalence of CVD relied on cross-sectional analysis in 4152 subjects which were determined by the presence of cardiac events, such as myocardial infarction, coronary atherosclerotic heart disease with or without coronary artery revascularisation, congestive heart failure, and arrhythmia (Aurora et al., 2018). The estimated CVD risk depended on the Framingham risk score (FRS) in 3796 subjects without prevalent CVD. The FRS, a widely used risk assessment tool developed in the Framingham Heart Study, was used in assessing CVD risk (D'Agostino Sr. et al., 2008). The FRS was computed for each individual by ageing, HDL, TC, using antihypertensive medication, systolic blood pressure, diabetes mellitus, and smoking habit as sexspecific adverse outcomes. Values varied from 1 to 30; the higher the scores, the more elevated the risk. The FRS was divided into tertiles; subjects in the 2nd and 3rd tertiles were considered to be at a higher CVD risk within 10 years (Goff Jr. et al., 2014). The Epworth sleepiness scale (ESS) was required to be completed by all participants, and a score >10 indicated excessive daytime sleepiness.

| Polysomnography
During the 48 h before the sleep study and the data collection, we asked the participants to refrain from caffeine, daytime naps, and alcohol. Full-night, hospital-based PSG (Alice 4,5,6; Philips Respironics, Murrysville, PA, USA) was performed for nocturnal monitoring; this included standard electroencephalography, electrooculography, mandibular electromyography, nasal airflow pressure, thoracic and abdominal respiratory effort, finger pulse oximetry, and body position measurements. The standard two-lead configuration conducted three-electrode electrocardiography with a 512 Hz sampling rate. Registered sleep technologists manually staged and scored sleep using standard evaluation metrics recommended by the American Academy of Sleep Medicine (2007) (Berry et al., 2012). Participants were encouraged to turn the lights off at their discretion; the lights were turned on again at 06:00.  The above data were utilised to determine HRV variables. ARTiiFACT 3.0 (Psychonomic Society, Inc.) determined R-peaks and derived interbeat intervals. Two experienced technologists, who were blinded to the PSG results, manually reviewed each electrocardiography recording to identify the following characteristics: unusual QRS wave morphology, ectopic heartbeats, and motion artifacts. The review was also conducted to ensure that the HRV analysis program properly labelled R-waves, thus permitting accurate R-R interval measurement.

| HRV analysis and cardiac autonomic dysfunction
Fast Fourier transform analysis was then applied to the processed information to determine frequency-domain HRV variables.    (Table S1). Figure S1 shows the flow chart.    showed the independent association between AHI REM and increased LF/HF (β = 0.018; P = 0.042) and LF (n.u.) (β = 0.011; P = 0.027) after full adjustment for confounders, including age, sex, BMI, alcohol consumption, smoking, hypertension, DM, ESS, and AHI NREM (Table 4). We also performed sensitivity analyses by exploring only the associations between AHI REM and HRV in subjects with AHI NREM <5 events/h (Table 4 and Table S4). AHI REM remained significantly associated with increases in LF/HF (β = 0.032; P = 0.033) and LF (n.u.) (β = 0.017; P = 0.029) after full adjustment for potential confounders (Table 4).

| Mediation analysis
Based on multivariate linear regression analysis of OSA during REM sleep and HRV, we examined if OSA during REM sleep and CVD risk were mediated by the significant HRV indices, LF/HF, and LF (n.u.). Table 5 shows each step of the mediation analysis. All paths showed statistical significance (P < 0.05), demonstrating that LF/HF and LF (n.u.) significantly mediated the correlation of OSA during REM sleep to CVD risk. LF/HF and LF (n.u.) significantly mediated the association between AHI REM and prevalence of CVD, respectively (path β1 = 0.044, P = 0.012; β2 = 0.044, P = 0.008) (Figure 3a). Meanwhile, LF/HF and LF (n.u.) were also separate significant mediators for the correlation between AHI REM and high 10 year CVD risk (β3 = 0.068, P < 0.001; β4 = 0.0.067, P < 0.001) (Figure 3b). These mediation analyses estimated that LF/HF and LF (n.u.) explained 4.3% T A B L E 4 Association of AHI REM with differential HRV indices Note: Demographic adjusted model was adjusted for age, sex, and BMI. Multivariate adjusted model was adjusted for variables included in demographic adjusted model and alcohol consumption, smoking, hypertension, diabetes, ESS, and AHI NREM . All variables were log transferred. Abbreviations: AHI REM , apnea-hypopnea index during rapid eye movement sleep; AHI NREM , apnea-hypopnea index during non-rapid eye movement sleep; BMI, body mass index; CI, confidence interval; ESS, Epworth sleepiness scale; HF (abs), absolute values of high frequency; HF (n.u.), normalised high frequency power; HRV, heart rate variability; LF/HF, low frequency to high frequency power ratio; LF (n.u.), normalised low frequency power. and 4.9% of the association between AHI REM and the prevalence of CVD, respectively (Figure 3a). LF/HF and LF (n.u.) explained 3.1% and 3.6% of the relationship between AHI REM and a high 10 year CVD risk, respectively (Figure 3b). Adjustment of all analyses was performed for age, sex, BMI, alcohol consumption, smoking, hypertension, DM, ESS, and AHI NREM .

| DISCUSSION
Several strategies were used to improve the quality and effect of samples in the present study, including the largest known scale of sampling in an HRV study of OSA, strict data acquisition, determining the dose-effect correlation, and adequate control of potential confounding factors; the results revealed the correlation between severe REM sleep-related OSA and higher rates of CVD risk and cardiac autonomic dysfunction. HRV during wakefulness immediately before sleep allows readily accessible measurements of cardiovascular characteristics in OSA subjects and can facilitate clinical translation.
Severe REM sleep-related OSA shows an independent correlation with hypertension prevalence and incidence , a higher incidence of a composite cardiovascular endpoint (Aurora et al., 2018), and impairment of glucose metabolism (Chami et al., 2015;Grimaldi et al., 2014). Hypertension and metabolic dysfunction are correlated with an increased CVD risk. Our study demonstrated a dose-response influence of REM sleep-related OSA on increased CVD risk. OSA leads to cardiac autonomic dysfunction, as indicated by changes in HRV components. OSA patients posed higher LF oscillations (baroreflex function) and LF/HF during REM sleep than controls; they also exhibited decreased HF oscillations (vagal input) during REM sleep (Narkiewicz et al., 1998;Somers et al., 1995). Multiple studies of patients with CVD have shown that mortality risk increases in relation to alterations in HRV parameters; such changes include increases in sympathetic tone and decreases in vagal tone (Kleiger et al., 1987). Successful treatment of OSA by using CPAP (Efazati et al., 2020), mandibular advancement devices (Coruzzi et al., 2006), as well as upper airway surgery (Choi et al., 2012) (Oh et al., 2019;Reynolds et al., 2007). Our study demonstrated that REM sleep-related OSA had a correlation with increasing levels of LF/HF and LF (n.u), which indicates sympathetic activity.
The pathophysiological mechanism that underlies the role of OSA during REM sleep to CVD remains unknown. The solitary tract nucleus and ventrolateral medulla have been shown to share many respiratory and cardiovascular control circuits (Calandra-Buonaura et al., 2016). CVD in OSA is presumed to reflect maladaptive autonomic neuroplasticity in episodic sympathetic responses to respiratory incidents (Cortelli et al., 2012). In addition, more wake-related electroencephalography and cardiac electrophysiology changes occur in REM compared with in NREM sleep (Calandra-Buonaura et al., 2016).
Although how REM-related OSA is correlated with CVD is speculative, this relationship may reflect the REM-specific vulnerability of the baroreflex control system to respiratory disorders while sleeping through enhanced wake-input involvement in cardiorespiratory con- sympathetic nerve activity and blood pressure, and also stimulate respiration; all of these processes cause sleep disturbance and awakening (Eckert et al., 2009). Because apnea occurs more frequently and lasts for a long time during REM compared with the NREM sleep, REM sleep apnea produces significantly greater oxygen deficiency (Findley et al., 1985), as well as rapidly enhanced autonomic activity, heart rate, blood pressure, and cardiac output on awakening; these changes lead to greater increases in cardiac oxygen and myocardial metabolic demand with low arterial mean oxygen saturation (Eckert et al., 2009).
Furthermore, REM sleep was found to decrease coronary blood flow in stenosed coronary artery areas in a canine model (Kirby & Verrier, 1989 (Grimaldi et al., 2014).
Therefore, using CPAP for 3-4 h (the most recommended) is unlikely to offset the influences of OSA during REM sleep (Weaver et al., 2007), which occurs in later sleep cycles when patients are unlikely to continue using CPAP (Grimaldi et al., 2014;Mokhlesi et al., 2014). This may largely show why, in randomised controlled trials, CPAP has demonstrated a modest impact on hypertension, which may contribute to CVD progression (Fava et al., 2014). In addition, the use of CPAP has not reduced cardiometabolic burden. The randomised controlled trials which examined the CPAP impact on incident hypertension or cardiovascular events determined the nonimproved outcomes due to CPAP according to the intention-to-treat approach (Barbé et al., 2012;McEvoy et al., 2016). Nonetheless, in post hoc analyses, patients with higher CPAP adherence (median of 6 h per night) exhibited significant reductions in the incidences of hypertension or cardiovascular events (Barbé et al., 2012;Martínez-García et al., 2013). Additionally, up to 20% of individuals with mildto-moderate OSA may have REM sleep-related OSA, which is more prevalent among women and young individuals, including children (Goh et al., 2000;Haba-Rubio et al., 2005;O'Connor et al., 2000).  LF/HF and LF (n.u.) were also significant mediators of the association between AHI REM and high 10 year risk of CVD (β3 = 0.068, P < 0.001; β4 = 0.067, P < 0.001). All analyses were adjusted for age, sex, BMI, alcohol consumption, smoking, hypertension, diabetes mellitus, ESS, and AHI NREM . AHI NREM , apnea-hypopnea index during non-rapid eye movement sleep; AHI REM , apnea-hypopnea index during rapid eye movement sleep; BMI, body mass index; CVD, cardiovascular disease; ESS, Epworth sleepiness scale; LF (n.u.), normalised low-frequency power; LF/HF, low frequency to high frequency power ratio the associations of AHI REM , CVD, and cardiac autonomic dysfunction are biologically plausible, definitive conclusions regarding causal relationships, or the directions of such relationships, could not be reached because this study was observational. Herein, subjects were only evaluated during their waking hours, which is why we were not able to compare HRV values during sleeping hours. Because of the low proportion of women included in the study population, we cannot generalise our findings to women. REM sleep-related OSA occurs more frequently in women than in men; therefore, those women may have a disproportionate influence on the CVD risk (O'Connor et al., 2000).

AUTHOR CONTRIBUTIONS
The corresponding authors are responsible for the authenticity of the data. All authors made a significant contribution to the work reported (i.e., in the conception design or execution of the study, acquisition, analysis, or interpretation of the data, or in all of these areas