Cardiac Sympathetic Modulation in Response to Apneas/Hypopneas through Heart Rate Variability Analysis

Autonomic dysfunction is recognized to contribute to cardiovascular consequences in obstructive sleep apnea/hypopnea syndrome (OSAHS) patients who present predominant cardiovascular sympathetic activity that persists during wakefulness. Here, we examined 1) the factors that influence sympathetic cardiac modulation in response to apneas/hypopneas; and 2) the influence of autonomic activity during apneas/hypopneas on CA. Sixteen OSAHS patients underwent in-hospital polysomnography. RR interval (RR) and RR spectral analysis using wavelet transform were used to study parasympathetic (high frequency power: HFWV) and sympathetic (low frequency power: LFWV and LFWV/HFWV ratio) activity before and after apnea/hypopnea termination. Autonomic cardiac modulations were compared according to sleep stage, apnea/hypopnea type and duration, arterial oxygen saturation, and presence of CA. At apnea/hypopnea termination, RR decreased (p<0.001) while LFWV (p = 0.001) and LFWV/HFWV ratio (p = 0.001) increased. Only RR and LFWV/HFWV ratio changes were higher when apneas/hypopneas produced CA (p = 0.030 and p = 0.035, respectively) or deep hypoxia (p = 0.023 and p = 0.046, respectively). Multivariate statistical analysis showed that elevated LFWV (p = 0.006) and LFWV/HFWV ratio (p = 0.029) during apneas/hypopneas were independently related to higher CA occurrence. Both the arousal and hypoxia processes may contribute to sympathetic cardiovascular overactivity by recurrent cardiac sympathetic modulation in response to apneas/hypopneas. Sympathetic overactivity also may play an important role in the acute central response to apneas/hypopneas, and in the sleep fragmentation.


Introduction
Epidemiological studies suggest that obstructive sleep apnea/ hypopnea syndrome (OSAHS) is common in the general population [1], and provide strong evidence that OSAHS is associated with significantly high cardiovascular morbidity and mortality [2]. Autonomic dysfunction is now recognized to contribute to these cardiovascular consequences [3] in OSAHS patients who present decrease in heart rate variability (HRV) [4] and predominant cardiovascular sympathetic activity that persists during wakefulness [5]. Moreover, sympathetic sleep fragmentation was associated with elevated nocturnal and diurnal systolic blood pressure and higher risk of systolic hypertension [6]. Studying the mechanisms that control sympathetic cardiac modulation in response to apneas/hypopneas by HRV should improve our understanding of the cardiovascular risk factor in OSAHS populations.
OSAHS is characterized by repeated episodes of total (apneas) or partial (hypopneas) upper airway occlusion during sleep, resulting in exaggerated negative intrathoracic pressure and often oxygen desaturation and carbon dioxide retention [3]. At termination, apneas/hypopneas frequently trigger cortical arousals (CA) [7], a process thought to restore pharyngeal dilator muscle tone and airflow. Apneas/hypopneas also elicit oscillations in both parasympathetic and sympathetic cardiac activities that affect RR intervals (RR), characterized by increased parasympathetic activity during apneas/hypopneas and increased sympathetic activity at apnea/hypopnea termination [8][9][10].
The factors that modulate sympathetic cardiac modulation in response to apneas/hypopneas remain unclear, and the research findings are contradictory [8][9][10]. Sympathetic cardiac change in response to apneas/hypopneas during paradoxical sleep has been reported as higher [10] or lower [8,9] than in other sleep stages, whereas in response to external [11] or internal [12] stimulations, this sympathetic cardiac modulation did not appear to differ according to sleep stage. Although this sympathetic cardiac change was elevated when CA was generated by external [11] or internal [12] stimulations, a relationship between CA and sympathetic cardiac modulation in response to apneas/hypopneas was reported [8,9] or not [10]. Only one study specifically examined the influence of the type of respiratory events and found no significant effect [8], and another that of hypoxia, and showed a correlation between concomitant minimal oxygen saturation (min SaO 2 ) and sympathetic cardiac modulation [10]. To our knowledge, the effect of the duration of respiratory events on sympathetic modulation has never been investigated. Sympathetic cardiac activity during sleep was also proposed as a potential contributor to CA occurrence following the observation in animals [13,14] and humans [15,16] that baroreflex loop stimulation during sleep, an autonomic sensitivity component, could induce CA. During apneas/hypopneas, although RR increase and parasympathetic activity dominates, Bonsignore [17] reported two typical patterns: RR continued to increase during apneas/hypopneas, and RR decreased before apnea/ hypopnea termination. However, to our knowledge, no study has examined the relationship between autonomic cardiac activity during apneas/hypopneas and subsequent CA while elevated cardiac sympathetic precede spontaneous CA occurrence [18].
Based on these findings, we first assessed autonomic cardiac modulation at obstructive apnea/hypopnea termination during all-night sleep and its relationship to electroencephalographic (EEG) cortical reactivity, sleep stage, min SaO 2 , and apnea/ hypopnea duration and type (hypopneas vs. apneas). We hypothesized that 1) sympathetic cardiac activation should be more intense when apneas/hypopneas give rise to CA or 2) when it is accompanied by a deeper hypoxia, but 3) with no difference between sleep stages. We also hypothesized 4) a rise in cardiac sympathetic modulation with longer apnea/hypopnea duration, and 5) with apnea versus hypopnea type. We also assessed the relationship between autonomic activity level during apneas/ hypopneas and subsequent CA occurrence. We hypothesized that 6) high sympathetic cardiac activity during apneas/hypopneas should be associated with facilitated CA occurrence.

Subjects
Sixteen untreated OSAHS patients (3 women, Table 1) were studied (Clinical Trial registration: NCT01135303). Exclusion criteria were any known cardiac abnormalities or neurological disease and being under adrenergic receptor blockers as previously published [8,10]. Seven hypertensive patients were included, since they were not under adrenergic receptor blockers but diuretic and angiotensin-converting enzyme inhibitors. No other cardiac or respiratory disorder was present ( Table 1). The study was approved by the Local Ethics Committee (CCP Saint-Etienne Sud) and performed under informed consent according to the Declaration of Helsinki. All participants provided their written informed consent to participate in this study.

Sleep analysis
To determine in which sleep stage every apneas/hypopneas occurred during the whole night, sleep stages were visually scored according to AASM 2007 criteria [19]. Hypnograms were scored based on 30-s epochs and differentiated into stage 2 (S2), slow-wave sleep (SWS), and paradoxical (REM) sleep (PS) ( Table 1). CA were defined as bursts of wakefulness-related cortical activity lasting more than 3 s. Obstructive apneas were defined as complete cessation of airflow (on airflow cannula) in the presence of respiratory effort lasting $10 s. When respiratory effort partially or totally ceased, apneas were scored as mixed apnea or central apnea, respectively, and eliminated. Hypopneas were defined as a $50% airflow reduction compared with the 3 preceding breaths. Oxygen desaturations were defined as a $3% decrease in arterial oxygen saturation.

RR interval analysis
Peak-to-peak EKG signals were analyzed to detect R waves in the QRS complex using Matlab (MathWorks, Naticks, MA, USA). EKG data were first visually selected. All undetected QRS, ectopic beats, and artifacts were corrected on EKG data. If correction was not possible for at least one heartbeat on selected periods, data were eliminated (236 excluded for artifacts and ectopic beats, 7.4% of a total of 3191 apneas/hypopneas). RR was then analyzed for each apnea/hypopnea on 32 heartbeats before and after apnea/hypopnea termination. Statistical comparisons were performed on the mean of 3 minimal RR between 2 and 7 heartbeats post-apnea/hypopnea termination (mean RR) compared with the mean of the 10 RR pre-apnea/hypopnea termination.

Spectral analysis of heart rate
Wavelet analysis was performed on the RR signal to analyze both the time and frequency domain, using the mother function Daubechies 4. The wavelet analysis signal allows quantifying temporal changes in the signal frequency contents. Starting from the initial Daubechies 4 function, a family of functions was built by dilatation and translocation to constitute a wavelet frame. A serial list of wavelet coefficients was used to represent the correlation between the signal and the selected wavelet at different analysis levels (or different frequency ranges) along the entire signal. The first levels (2,4,8) correspond to a wavelet analysis conducted with a small dilatation factor applied to 2, 4 and 8 RR, respectively. Thus, these levels represent high-frequency variations in the signal. The last levels (16, 32, 64, etc.) correspond to a wavelet analysis conducted with a large dilatation factor applied to 16, 32, and 64 RR, respectively. Consequently, these levels represent lowfrequency variations in the signal. According to the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology [21], squared wavelet coefficients at levels 2, 4, and 8 (high frequency: HF WV ) were used to assess parasympathetic modulation. Wavelet power coefficients at levels 16 and 32 (low frequency: LF WV ) were used to assess sympathetic modulation. An LF WV /HF WV ratio was calculated as the ratio between the wavelet power coefficient at levels 16 and 32 on the one hand and the sum of the wavelet power coefficients at levels 2, 4, and 8 on the other to obtain a marker of autonomic nervous system balance (see Figure 1). This analysis was performed on 32 RR before and after each apnea/hypopnea termination [11].
To assess whether CA occurrence is independently associated with autonomic activity during apneas/hypopneas, logistic multivariate regressions were performed between each autonomic cardiac index during apneas/hypopneas (expressed in continuous values and tertiles) and the CA presence, adjusted to minimum and mean SaO 2 , type, sleep stage, interval inter-events, and subject. Risks were expressed as odds ratios (OR) with 95% confidence interval (95% CI), differences were considered significant when p,0.05, and values were expressed as mean 6 SEM.

Number of respiratory events studied
Of a total of 3191 apneas/hypopneas, 2955 were included in the analysis, with 236 excluded for artifacts and ectopic beats. For the whole group, 186664 events per subject were selected, of which 128649 were in S2, 19617 in SWS, and 40620 in PS. Due to lack of apneas/hypopneas during SWS in some OSAHS patients, statistical analysis was performed on 7 subjects for comparison between sleep stages. Due to lack of apneas/ hypopneas with deep hypoxia (SaO 2 #85%) and long duration (30 s) in some subjects, statistical analysis was performed on 10 subjects to compare min SaO 2 levels, 11 subjects to compare apnea durations, and 13 subjects to compare hypopnea durations. In the 16
Autonomic activity during apnea/hypopnea and its relationship with cortical arousal occurrence To assess whether autonomic activity during apneas/hypopneas was associated with CA occurrence, multivariate logistic regression analysis was applied to investigate the relationship between the autonomic index of HRV and subsequent CA. Results showed a significant relationship between CA occurrence and LF WV (x 2 = 7.42, p = 0.006) and LF WV /HF WV ratio (x 2 = 4.80, p = 0.029), independently of mini and mean SaO 2 , type and duration, interval inter-events, sleep stage, and subject. No significant results were noted for RR mean (x 2 = 1.35, p = 0.244) or HF WV (x 2 = 0.01, p = 0.906).

Discussion
First, the results of this study showed that sympathetic modulation at apnea/hypopnea termination is modulated by the arousal process and hypoxia rather than by sleep stage or apnea/ hypopnea type and duration in obstructive sleep apnea/hypopnea syndrome (OSAHS) patients. Greater sympathetic activation in response to apneas/hypopneas is concomitant with the presence of cortical arousal (CA) or deep hypoxia. Second, when cardiac sympathetic indexes of heart rate variability (HRV) were high during apneas/hypopneas, the risk of CA occurrence was magnified (or twice). This indicates that cardiac sympathetic dominance during apneas/hypopneas is related to facilitated CA occurrence. However, the exact relationship between sympathetic activity during apneas/hypopneas and subsequent CA remains unknown.
Sleep autonomic modulation at apnea/hypopnea termination Spicuzza and collaborators [9] used time frequency analysis to detect oscillations in parasympathetic and sympathetic cardiac activity during apneas, corroborating other studies [7,8]. These oscillations are characterized mainly by parasympathetic dominance during apneas and sympathetic modulation at apneas termination. Using a similar time frequency method, the present study confirmed that sympathetic modulation dominates at apnea/hypopnea termination (increase in both LF WV and LF WV /HF WV ratio), with a small decrease in parasympathetic activity (HF WV ). This parasympathetic withdrawal appeared significant only in comparison with the higher number of subjects, e.g. comparison with and without CA (n = 16), and according to apnea/hypopnea type (n = 16). In addition, using Fourier transform, a less sensitive method to study transit changes in RR, previous studies [8,9] showed no change in spectral HF power. This parasympathetic withdrawal at apnea/hypopnea termination appeared to be small and difficult to detect. Moreover, OSAHS patients presented a predominant cardiovascular sympathetic activity with decreased parasympathetic control [4,5]. This could explain the difficulty in detecting transit changes in the parasympathetic arm in this OSAHS population, even though we used  ratio before and after the apnea/hypopnea terminations according to hypopneas and apneas (mean ± SEM). An overall effect on RR intervals, LF WV , and LF WV /HF WV ratio appeared between before and after apnea/hypopnea terminations. Apneas/hypopneas induced a sympathetic-dependent RR decrease, which was not modulated by respiratory event type. doi:10.1371/journal.pone.0086434.g003 time frequency analysis. Thus, apnea/hypopnea termination induces a cardiac reflex activation, implying essentially sympathetic drive whereas parasympathetic withdrawal is less marked, parasympathetic activity been more related to physiological changes during apneas/hypopneas [10,22].

Control of sympathetic modulation at apnea/hypopnea termination
Autonomic dysfunction is recognized to contribute to cardiovascular consequences in these OSAHS patients [3]. To better understand the relationship between apneas/hypopneas and cardiac autonomic dysfunction in OSAHS population, we specifically examined apnea/hypopnea termination to explore the mechanisms that control sympathetic cardiac modulation by HRV analysis. We did not exclude the effect of parasympathetic dysfunction in OSAHS patients on cardiovascular morbidity and mortality [3]. However, modulation of the parasympathetic arm during apneas/hypopneas has been well studied [10,22], whereas the factors that modulate sympathetic cardiac modulation at apnea/hypopnea termination have been less studied, with inconsistent results [8][9][10]. Moreover, we have previously shown that sympathetic sleep fragmentation was associated with elevated nocturnal and diurnal systolic blood pressure and higher risk of systolic hypertension [6].
Of the several factors tested in this study, sleep stage, apnea/ hypopnea duration, and respiratory event type showed no effect on sympathetic cardiac modulation: irrespective of sleep stage, type, and duration, apneas/hypopneas produced the same level of cardiac sympathetic modulation (Figure 2-4).
The absence of sleep stage difference concurs with previous studies, which showed that exteroceptive stimulations [11] or periodic legs movements [12] produce the same level of sympathetic cardiac activation regardless of sleep stage. Regarding apneas/hypopneas, sympathetic cardiac change was described as higher [10] or lower [8,9] in paradoxical sleep than in other sleep An overall effect on RR intervals, LF WV , and LF WV /HF WV ratio appeared between before and after hypopnea terminations. Hypopneas induced a sympathetic-dependent RR decrease, which was not modulated by hypopnea duration. doi:10.1371/journal.pone.0086434.g004 Figure 5. Comparison of autonomic modulation at terminations according to apnea duration: (A) RR intervals (RR), (B) low (LF WV ) and (C) high frequency (HF WV ) wavelet power, and (D) LF WV /HF WV ratio before and after apnea terminations according to apnea duration: short (10 s#apneas,20 s), medium (20 s#apneas,30 s), and long (apneas$30 s) (mean ± SEM). An overall effect on RR intervals, LF WV , and LF WV /HF WV ratio appeared between before and after apneas terminations. Apneas induced a sympathetic-dependent RR decrease, which was not modulated by apnea duration. doi:10.1371/journal.pone.0086434.g005 stages. These discrepancies could be explained by different methodological approaches (Fourier [9] vs. wavelet transform in our study), or by the previous small number of apnea/hypopnea observations (61 events [10] vs. 1524 in 7 subjects in our study). Thus, apneas/hypopneas may produce the same level of sympathetic cardiac modulation regardless of sleep stage (Figure 2).
Hierarchy in the arousal process and its close relationship with sympathetic cardiac modulation has been previously reported in several conditions of sleep fragmentation, including spontaneous arousal [12], periodic leg movements [23], and auditive [24] or nociceptive stimulation [10]. However, it remains unclear whether this relationship holds true for OSAHS patients presenting recurrent apneas/hypopneas during sleep, basal autonomic dysfunction [4], or abnormal autonomic responses to stress [25]. Moreover, previous studies showed inconsistent results: CA had an effect on sympathetic cardiac modulation [9] or not [8,10]. However, studies with negative results used a low number of apneas/hypopneas (only 61 events [10] vs. 2972 in 7 patients in the present study) or Fourier transform, known to be less effective than the time frequency method to study transient variation in RR [8]. Our results clearly showed that sympathetic cardiac modulation varied contingently with cortical reactivity. Cardiac reactivity was higher when apneas/hypopneas produced CA, although preserved in the absence of subsequent CA. Thus, the arousal process appeared to be a major modulator of sympathetic modulation to apneas/hypopneas.
Then, we hypothesized that severity of respiratory events could be expressed as length or type of respiratory events, but our result rather showed that both type and duration of respiratory events do no bring information on cardiac sympathetic modulation. Our results confirm previous data [8] showing no direct effect of respiratory event type, i.e. apneas vs. hypopneas, on sympathetic cardiac modulation (Figure 3-4). Concerning the duration of respiratory event, although we hypothesized that longer duration Figure 6. Comparison of autonomic modulation at apnea/hypopnea terminations according to oxygen desaturation: (A) RR intervals (RR), (B) low (LF WV ) and (C) high frequency (HF WV ) wavelet power, and (D) LF WV /HF WV ratio before and after apnea/ hypopnea terminations according to light oxygen desaturation (min SaO 2 $90%) medium (90.min SaO 2 .85%) and deep oxygen desaturation (min SaO 2 #85%) (mean ± SEM). An overall effect on RR intervals, LF WV , and LF WV /HF WV ratio appeared between before and after apnea/hypopnea terminations. Apneas/hypopneas induced a sympathetic-dependent RR decrease, which was higher when apneas/hypopneas produced deep oxygen desaturation. doi:10.1371/journal.pone.0086434.g006 Figure 7. Comparison of autonomic modulation at apnea/hypopnea terminations according to cortical arousals: (A) RR intervals (RR), (B) low (LF WV ) and (C) high frequency (HF WV ) wavelet power, and (D) LF WV /HF WV ratio before and after apnea/hypopnea terminations during no cortical and cortical arousals (mean ± SEM). An overall effect on RR intervals, LF WV , and LF WV /HF WV ratio appeared between before and after apnea/hypopnea terminations. Apneas/hypopneas induced a sympathetic-dependent RR decrease, which was higher when apneas/hypopneas produced cortical arousals. doi:10.1371/journal.pone.0086434.g007 would increase cardiac sympathetic modulation, this was not the case.
Finally, recurrent nocturnal hypoxia, which is suspected to be a major factor of cardiac disease in OSAHS patients [3,26], appears to have a strong influence on cardiac sympathetic modulation in response to apneas/hypopneas in our study. Indeed a min SaO 2 decrease lesser than 90% was able to induce a significant cardiac sympathetic modulation increase ( Figure 6). These results concur with those of a previous study showing a correlation between concomitant min SaO 2 and sympathetic cardiac modulation in response to apneas/hypopneas [10]. However, several authors [27,28] found no difference between hypoxia (mean SaO 2 : 88.860.4%) and normoxia (mean SaO 2 : 97.960.2%) in cardiac response to auditive stimulations during sleep in healthy volunteers. This discrepancy between their and our results could be explained by the elevated chemoreflex sensitivity in OSAHS patients, which causes cardiac sympathetic modulation in response to decrease in SaO 2 [29]. In turn, this exposure to recurrent hypoxia and cardiac sympathetic modulation during sleep may also contribute to chemoreflex impairment in OSAHS patients [30]. Thus, elevated cardiac sympathetic activation in response to apneas/hypopneas is concomitant with strong chemoreflex activation, potentially leading to the increase in diurnal chemoreflex sensitivity in OSAHS population [30,31].
This cardiac sympathetic modulation in response to respiratory events influenced by the arousal and hypoxia processes could represent a pathway contributing to cardiovascular disease risk in OSAHS. Recurrent hypoxia, known to be a strong factor of cardiac disease in OSAHS patients, can also act by other pathways, such as oxygen free radical production, inflammatory pathways, and impairment in the vascular endothelial function independently of activation of the sympathetic arm [3]. EEG sleep fragmentation, proposed as a potential contributor to hypertension, was related to an approximately 3 mmHg rise in diurnal systolic blood pressure in a population without sleep-disordered breathing [32] and recently, sleep fragmentation based on autonomic markers (pulse transit time) have been shown to be related to diurnal blood pressure increase and hypertension risk [33]. Our study supports the idea that the recurrent arousal and hypoxia processes, by cardiac sympathetic modulation in response to apneas/hypopneas, could contribute to a decrease in HRV [4,33] and a predominant cardiovascular sympathetic activity that persists during wakefulness [5], increasing the cardiovascular disease risk in OSAHS populations.

Relationship between sympathetic cardiac activity and subsequent cortical arousal
In the last part of this study, we examined the relationship between autonomic indexes of HRV during apneas/hypopneas and subsequent CA. Results showed that the spectral LF WV power and LF WV /HF WV ratio during apneas/hypopneas, sympathetic criteria of HRV, are related to CA occurrence at apnea/hypopnea termination, indicating that cardiac sympathetic dominance during apneas/hypopneas is related to a facilitated CA occurrence.
These results are in accord with studies that proposed sympathetic cardiac activity during sleep as a contributing factor to the CA occurrence process, suggested by observations obtained in both animals [13,14] and humans [15,16]. Other studies showed that spontaneous CA [17] and periodic leg movements [34] or bruxism events [35] are preceded by a rise in heart rate. Moreover, during apneas/hypopneas, although RR increase and parasympathetic activity dominates, Bonsignore [3,17] reported two typical patterns: RR continued to increase during the apneas, and RR decreased before apnea termination. Although our study indicates a relationship between sympathetic dominance during apneas/hypopneas during and subsequent to CA, we did not identify the mechanism by which autonomic stimulation contributes to arousal or the stimuli at the origin of this sympathetic disequilibrium.
It therefore appears likely that variations in peripheral afferents information may play an important role in the acute autonomic response as part of a multifactorial process [36], such as hypoxemia [22] and hypercapnia [37], thoracic pressure swings [17], baroreflex control [16], or pulmonary afferent stimuli arising from the chest wall or the lung in response to intrathoracic pressure changes or from the upper airway in response to airflow obstruction. Limitations Several limitations of this study should be considered. The method is limited for study of sympathetic activity: the LF W is related to both sympathetic and parasympathetic activity. It is only by studying the relative changes in LF W and HF W , and normalized indexes as LF W /HF W ratio that an indication of sympathetic activity is obtained. Several studies have showed that LF/HF ratio and LF% allow approaching relative sympathetic activity, in response to tilt test under atropine [38] as well as to experimental pain [11], auditory stimulations [24] and bruxism episodes [35] during sleep. These studies are consistent with our interpretation of the present results. Second, we compared autonomic changes at terminations from autonomic level during apneas/hypopneas. This can be adapted to compare with stable baselines. Although stable baselines can be easily obtained in wakefulness or stable sleep periods in healthy subjects, but is more problematic in OSAHS patients, as most of these subjects go from one episode to the next one without real stable baseline. Third, we cannot exclude that CO 2 retention, a factor that was not considered here, could contribute to these autonomic changes. However, a previous study showed that CO 2 level did not affect cardiac modulation in response to spontaneous arousals in healthy volunteers [27]. At last, number of apnea/hypopnea observations during SWS, statistical analysis was performed on a small number subjects (7 subjects for comparison between sleep stages), because its remains difficult to obtain hypopnea/apneas during SWS, due to the small numbers. These issues need to be addressed in wide OSAHS populations in future studies.

Conclusions
Apneas/hypopneas induce a cardiac sympathetic modulation influenced by the arousal and hypoxia processes rather than by sleep stage, total or partial upper airway occlusion, or occlusion duration. Higher sympathetic activation in response to apneas/ hypopneas is concomitant with the presence of cortical arousal or a deep hypoxia compared to no cortical arousal or no hypoxia. Furthermore, cardiac sympathetic dominance during apneas/ hypopneas is related to facilitated CA occurrence. Both the recurrent arousal and hypoxia processes by cardiac sympathetic modulation in response to apneas/hypopneas, could contribute to predominant cardiovascular sympathetic activity that persists during wakefulness, increasing the cardiovascular disease risk in OSAHS populations. On the other hand, sympathetic overactivity also may play an important role in the acute central response to apneas/hypopneas, and so in the sleep fragmentation.