Published ahead of print on December 27, 2002, doi:10.1164/rccm.200201-006OC
© 2003 American Thoracic Society
Autonomic Modulation of Heart Rate during Obstructive versus Central Apneas in Patients with Sleep-disordered BreathingDepartment of Internal Medicine, University of Catania, Catania; and Department of Internal Medicine, IRCCS Policlinico S. Matteo and University of Pavia, Italy Correspondence and requests for reprints should be addressed to Lucia Spicuzza, M.D., Istituto di Malattie Respiratorie, Via Passo Gravina 187, 95125 Catania, Italy. E-mail: luciaspicuzza{at}tiscalinet.it
Sleep-disordered breathing is associated with an altered sympathovagal balance determined by the nocturnal cyclic alternating of apneas and hyperventilation. The aim of this study was to determine whether the autonomic modulation of heart rate during obstructive apneas (OA) and central apneas (CA) in patients with sleep-disordered breathing is different. Therefore, by using the time-varying Wigner-Ville transform spectral analysis we described, in 17 patients, the time course of the low-frequency (LF) and the high-frequency (HF) components of the interbeat interval (R-R interval) reflecting, at large, respectively, the sympathetic and the parasympathetic modulation, during OA (n = 185) and CA (n = 51) and during the postapneic hyperventilation. In both types of apneas we found cyclic lengthening/shortening in R-R interval, during apneas/postapneic hyperventilation, respectively, with more marked bradycardia during OA (R-R: 1,011 ± 23 versus 893 ± 30 ms2, p < 0.01). In OA the HF oscillations decreased from the apnea to the postapneic hyperventilation (from 1,964 ± 244 to 387 ± 98 ms2, p < 0.0001), whereas the LF oscillations increased (from 2,649 ± 230 to 9,820 ± 716 ms2, p < 0.0001). Conversely, in CA the HF oscillations increased from the apnea to the postapneic hyperventilation (from 452 ± 177 to 1,485 ± 406 ms2, p < 0.0001), whereas the LF component remained unchanged. These results show markedly different autonomic alterations during and after OA versus CA, suggesting a surge in sympathetic modulation after the obstructive episodes.
Key Words: obstructive apneas • central apneas • autonomic nervous system Sleep-disordered breathing is associated with a significantly high cardiovascular morbidity and mortality (1). It is now recognized that the causal link between nocturnal respiratory disorders and cardiovascular risk consists in the altered sympathovagal balance reported in these patients during sleep, which may persist also during the daytime (2). In fact, an increased sympathetic drive in patients with sleep-disordered breathing, while awake, has been shown in terms of decreased heart rate variability, increased muscle sympathetic nerve activity, and high levels of circulating norepinephrine (2–4). Repetitive episodes of apnea, postapneic hyperventilation, and simultaneous hypoxia and hypercapnia seem to be the key factors determining the enhanced sympathetic traffic through a tonic activation of the chemoreflex activity (3). During obstructive apneas (OA) the increased respiratory effort, which follows the upper airways occlusion, determines an increase in the negative intrathoracic pressure and in the afferents inputs from lung and chest wall, thus affecting the autonomic function (5). Instead, central apneas (CA), which often occur in concomitance with OA in patients with sleep-disordered breathing, are characterized by the lack of both central and peripheral respiratory activity. These different mechanisms may therefore produce different patterns of autonomic activity, but so far a comparison of the autonomic changes between CA and OA has never been performed in humans. Indeed, there are difficulties in describing the instant changes in nocturnal autonomic function in patients with sleep-disordered breathing due to the continuous changes in ventilation during the apnea–hyperventilation cycles. Spectral analysis technique is an extremely useful tool to quantitatively describe the components of heart rate and cardiovascular variability, which are under autonomic control (6). To apply this method correctly, it should be recognized that (1) this is an indirect method for the study of autonomic modulation, (2) spurious effects of respiration and other possible artifacts on the spectral components must be excluded (7), (3) the low- frequency (LF) and high-frequency (HF) components are relative indices of sympathetic and parasympathetic modulation (hence these should be analyzed in relative and not absolute terms) (8). Within these limitations, the LF component can be considered as a relative marker of sympathetic activity, and the HF component, which is an expression of respiratory sinus arrhythmia, is a marker of parasympathetic activity (6). However, classic methods of spectral analysis, like the Fast Fourier transform or parametric modeling can only be applied to stable conditions and cannot be used when fast cardiovascular and respiratory changes occur, as during episodes of apnea/hyperventilation. To cope with this problem, in the present study we applied the Wigner-Ville transform, joined time-frequency analysis, which is able to describe even beat-to-beat changes of the input time series (9–11). The aim of this study was to investigate the instantaneous changes in the autonomic modulation of heart rate during OA and CA in patients with sleep-disordered breathing and to determine whether these changes may be influenced by factors such as the degree of hypoxemia or the sleep stage. We therefore used the Wigner-Ville transform to continuously follow the time course of the LF and HF components of the interbeat interval (R-R interval) during cycles of OA/CA and postapneic hyperventilation.
Subjects We studied a total of 17 patients (all males, 49 ± 17 years, body mass index 31 ± 6, all normoxic and eucapnic during the day) experiencing both OA and CA in whom sleep-disordered breathing was newly diagnosed. The inclusion criterion was the presence of at least five CA and five OA per hour of sleep. The severity of the disorder, from moderate to severe, was based on the apnea–hypopnea index (the number of apneas/hypopneas per hour of sleep). Chronic hypertension was present in seven subjects and four of these were treated with antihypertensive drugs (diuretics and angiotensin-converting enzyme inhibitors). No other cardiac or respiratory disorder was present and no other medication was reported. Two patients were current smokers. From the polysomnographic records of these patients a total of 185 OA and 51 CA, of similar duration and associated with a similar fall in the percentage of oxyhemoglobin saturation (SaO2), were chosen and analyzed (Table 1) .
Polysomnography Overnight standard polysomnographies were performed in our sleep laboratory using a Compumedic S-Series Sleep System (Abbotsford, Australia) according to standard criteria (12). OA and CA were defined as a complete cessation of airflow (for at least 10 seconds associated with a decrease in SaO2  4%) in the presence and in the absence, respectively, of respiratory effort (see online supplement for further details).
Spectral Analysis Figure 1 shows that changes in the power of two oscillatory components at different frequencies can be continuously and accurately tracked over time. Due to the use of a moving average of five data points (see online supplement) the onset and the termination of the apneas were assessed more precisely on the raw data. Using the Wigner-Ville spectral estimation, the absolute and the relative (normalized) power of the LF and HF components were calculated (see online supplement) from the R-R interval. In addition, the frequency of the respiratory component was obtained from the abdominal and thoracic data during OA and during the postapneic phase. An approximate estimation of the degree of hyperventilation (in arbitrary units) after OA and CA was obtained by the total power of spectral analysis of the respiratory flow.
Statistical Analysis The results are given as means ± SE. To compare each variable during and after the apneas and to compare CA and OA we used a two-factor repeated-measures analysis of variance (18). Linear regression analysis was used to test the relationship between cardiac and polysomnographic parameters. Probability values less than 0.05 were considered statistically significant.
The mean apnea–hypopnea index was 51 ± 8 and the relative frequencies (per hour of sleep) of OA, CA, mixed apneas, and hypopneas were 25 ± 5, 10 ± 2, 3 ± 1 and 13 ± 2, respectively.
Heart Rate in OA and CA during Non-REM Sleep
We found that the mean fall in SaO2 induced by OA and CA was 14% ± 0.5 and 14% ± 1.3 (p = not significant), respectively, and the mean lowest SaO2 reached was 81% ± 0.7 and 83% ± 0.9 (p = not significant), for OA and CA, respectively. For OA, the mean shortest R-R interval during the postapneic hyperventilation significantly correlated with the lowest SaO2 value that occurred after the apnea (r = 0.32). Examples of the data obtained in OA and CA are shown in Figures 2 and 3 .
Power Spectral Analysis of the R-R Interval during Non-REM Sleep Obstructive apneas. The HF oscillations, related to respiration and expression of respiratory sinus arrhythmia, were significantly higher during the apnea as compared with the postapneic hyperventilation (Table 2). In contrast, the LF component of the spectrum, a marker of sympathetic cardiac modulation, was significantly higher, particularly in normalized units, during the postapneic hyperventilation as compared with the apnea (Table 2, Figure 2). The decrease in HF power and the increase in LF power during the postapneic hyperventilation were not correlated to respiratory variables, such as the lowest SaO2 reached after the end of the apnea, the mean fall in SaO2, or the duration of the apnea. However, both the normalized LF power in the hyperventilatory phase and its increase with respect to the apneic phase correlated with the lowest SaO2 (r= 0.19, p < 0.05 and r = 0.20, p < 0.05, respectively). The rate of chest/abdominal respiratory movements was significantly higher during the apneic phase (Table 2) than during the postapneic phase, although the absolute difference was small. Figures 4 and 5 show the R-R interval and the respiratory spectra and cross spectra (R-R interval versus respiratory data) obtained in one case of OA.
Central apneas. Changes in the HF power during CA were the opposite of those observed during OA. In fact, only a minor HF component of the spectrum was present during the apnea, which then significantly increased during the postapneic hyperventilation, both in absolute and normalized units (p < 0.001 for both, Table 2). Due to the absolute and relative increase in respiratory sinus arrhythmia, the normalized LF component decreased from the apnea to the postapneic hyperventilation (Table 2), although in absolute terms the decrease was not significant (Table 2). During the apneic phase, the LF component was significantly higher in CA as compared with OA in both absolute (p < 0.05) and normalized units (p < 0.001); however, during the postapneic hyperventilation the LF became significantly lower in CA as compared with OA, again both in absolute (p < 0.01) and normalized units (p < 0.001, Table 2). The respiratory rate, in the postapneic period, was similar to that of OA (Table 2). The degree of hyperventilation, within the limitations of the indirect technique, was similar after CA and OA (Table 1). No correlation was found between changes in the R-R power spectrum and the lowest SaO2 value reached after the end of the apnea, the mean fall in SaO2, or the duration of the apnea.
Effect of the arousals.
Changes in the Power Spectrum of the R-R Interval during Non-REM and REM Sleep
The main finding of this study is that in patients with sleep-disordered breathing, CA and OA are associated with a different autonomic control of heart rate. We found that the respiratory component of the R-R interval variability (the respiratory sinus arrhythmia) is highest during the apneic phase of OA and is blunted during hyperventilation. Conversely, in the CA, the respiratory component of R-R interval variability (HF) is nearly absent during the apneic phase and recovers with resumption of ventilation. The LF component of the spectrum, particularly if analyzed in normalized units, shows a reciprocal behavior, as it is higher during the postapneic hyperventilation in the obstructive episodes and in the apneic phases of the central episodes. The marked reduction in respiratory sinus arrhythmia in the presence of hyperventilation after the obstructive episodes indicate a relative increase in sympathetic activity to the heart with respect to the apneic phase, which is confirmed by the increase in normalized power of the LF component seen after the end of the apnea. It is therefore clear that the simple cyclic alternating of bradycardia/tachycardia during the apnea/hyperventilation reflects a complex pattern of autonomic control. These findings may suggest that the autonomic changes initiated by the OA may prolong into the postapneic phase, whereas the changes induced by the CA passively reflect the changes in ventilation. In addition, this study shows the possibility to instantaneously track changes in heart rate variability during sleep recordings.
Changes in Respiratory Modulation of Heart Rate in OA and CA However, although it is well known that OA are associated with an increased intrathoracic pressure, one possible limitation of our study is that this has not been directly measured. Further studies with quantitative estimation of the inspiratory efforts may show a correlation with the degree of respiratory sinus arrhythmia during the obstructive phases, however, it is clear from this study that the respiratory sinus arrhythmia was present in coincidence with such efforts, as it was almost abolished during CA. In contrast to OA, we found a significant reduction in respiratory sinus arrhythmia during the apneic phase in CA. Because CA are characterized by the lack of central respiratory outflow and consequently by the lack of inspiratory effort (abolishment of intrathoracic negative pressure swings), it is conceivable that both central and peripheral mechanisms may lead to a blunted respiratory sinus arrhythmia. These findings are in accordance with data obtained during voluntary apneas in awake humans (22). Although it is well known that in awake subjects the respiratory sinus arrhythmia increases during hyperventilation, we found that in OA the HF component abruptly drops during the postapneic hyperventilation. This finding, apparently paradoxical, is nevertheless supported by previous reports in patients with obstructive sleep apnea syndrome (23, 24). It has been suggested that the disappearing of respiratory sinus arrhythmia during hyperventilation may be due to a reflex mechanism originating from lung inflation with consequent stimulation of lung stretch receptors (23). If this hypothesis is correct, then the question arises as to why during the hyperventilation that follows CA respiratory sinus arrhythmia shows an opposite pattern. We can exclude that this different pattern of respiratory sinus arrhythmia (RSA) changes in OA and CA was due the different degree of postapneic hyperventilation, as ventilation was not different in the two types of apnea (Table 1). Alternatively, it is possible that different from CA, in OA the sudden transition from negative to normal intrathoracic pressure at the end of the obstructive respiratory effort may contribute to the sudden impairment in parasympathetic modulation of heart rate after resumption of ventilation. In addition, the sympathetic activation seen during postapneic hyperventilation in OA (see below) may have overridden the changes in parasympathetic activity. Although these explanations are only speculative, the comparison with our results on CA strongly suggests that different mechanisms, other than simple lung inflation, must be involved.
Changes in Nonrespiratory Modulation of Heart Rate during OA and CA It is noteworthy that, at first sight in contrast to our finding, Somers and colleagues reported that during the OA, the muscle sympathetic nerve activity progressively increases from the beginning toward the end of the apnea and suddenly declines with the resumption of ventilation (2). This decrease was thought to be due to the stimulation of pulmonary afferents that inhibit sympathetic discharge and to the increase in blood pressure, which activate vagal baroreflex, further suppressing sympathetic activity. However, a close look at the figures of that study shows that muscle nerve sympathetic activity often peaks after termination of the apnea or remains elevated several seconds after the apnea; in addition, another study has shown that blood pressure peaks after termination of apnea (31). These findings suggest one or more of the following hypotheses: (1) a possible phase lag between sympathetic efferent activity and its effects on blood pressure/heart rate, such that the effect of sympathetic activity prolongs into the postapneic phase; (2) the reduction in baroreflex activity may contribute to the increase in blood pressure, heart rate, and sympathetic activity occurring in certain conditions such as hyperventilation during dynamic exercise or chemoreceptors activation (32); these factors may have been active also at the resumption of OA and contribute to the observed changes in R-R interval; (3) although available data on sympathetic efferent activity are obtained with measurements at the peroneal nerve, one should not exclude that the efferent activity to the heart may be more prolonged. The interpretation that sympathetic activity increases in the postapneic phases of OA is in contrast with two other findings: (1) although the mean R-R interval shortens from the apneic to the postapneic phases, its postapneic value is similar to the postapneic value of the central episodes, (2) the variance (a measure of overall heart rate variability) increases from the apneic to the postapneic phases. Because the variance is considered a marker of parasympathetic activity, this is in contrast with the previous interpretation. Particularly with the mean R-R interval, a solution would be to be able to assess a reliable baseline (preapneic data), i.e., a period of absence of any type of apnea. This can be easily obtained in the awake state, but it is much more problematic during sleep, as most of these subjects go from one episode to the next without a real stable baseline. About the variance, there is evidence that this does not always follow the time course of vagal activity. We and others have shown (33, 34) that the increase in sympathetic activity that occurs during early morning is associated with an increase in heart rate variability (HRV) and a decrease in mean R-R interval. We subsequently showed (35) that variance is also increased during "changes" in activity (again with a reduction in mean R-R interval). It is therefore possible that this paradox increase in variance (with a decrease in mean R-R interval and with an increase in the LF components) reflects some similar mechanism, whose nature remains to be elucidated. Alternatively, it is possible that contrasting autonomic stimulations (like the increase in parasympathetic activity elicited by the hyperventilation, and the increase in sympathetic activity induced by hypoxia) may act on different aspects of heart rate variability, thus inducing an increase in LF, a marked drop in respiratory sinus arrhythmia, a relative tachycardia, and an increase in variance. The pattern of LF modulation that we found in CA is different from OA. In fact, as discussed previously the respiratory oscillations in R-R interval modulation drop when respiration stops and restart when respiration resumes in the postapneic phase, with minor changes in the power of LF. The differences in the autonomic pattern that we described in OA and CA were not related to the degree of hypoxemia because a similar mean fall in SaO2 was shown in the two groups. The changes in spectral variables may suggest an increased parasympathetic activity during the postapneic hyperventilation after CA. However, the heart rate did not decrease, opposite to what could be predicted by increased parasympathetic tone or modulation. This finding may have different explanations: one possibility is that in CA there is also an increase in sympathetic activity that is masked and that maintains a high heart rate. It is also well possible that the increase in HF simply follows the changes in ventilation without a major increase in parasympathetic activity and with only limited autonomic changes confined in the apneic phase. Arousals from sleep, which occur more often at the end of OA compared with OA, might be a factor determining the enhanced sympathetic cardiac outflow during postapneic hyperventilation in OA (36). However, we found that the presence of the arousals did not modify the pattern of R-R interval modulation. This is supported by previous data showing that the increase in blood pressure at the end of the apnea is not affected by the occurrence of an arousal but mainly depends on oscillations in ventilation (31, 37). Finally, it seems unlikely that the difference in timing of changes in HF and LF might be due to the different duration of OA and CA as none of the spectral variables correlated with the apneas' duration.
Autonomic Pattern during REM and Non-REM Sleep in OA
Limitations of the Study It is misleading to assume that the absolute power of LF is directly related to sympathetic activity, as all spectral variables change with the variance, which is influenced by both vagal (directly) and sympathetic (inversely) activity. Therefore, it is only by analyzing the relative change in LF and HF, both in absolute and relative terms, that an indication of sympathetic activity is obtained. Furthermore, a correct evaluation of spectral components can be obtained only when the influence of respiration or other artifacts can be excluded (35). In the present study, we have simultaneously analyzed the changes in the R-R interval and respiration, and, due to the presence of the other polysomnographic sensors, other sources of error could be excluded. It is clear that spectral analysis, even performed with the best possible methods available, as we did in the present study, is an indirect method, but a continuous measurement of cardiac sympathetic and parasympathetic activity cannot be otherwise obtained in humans. Direct measurement of sympathetic nerve activity by microneurography is the gold standard for the measurement of sympathetic activity, but this method also has limitations, as it is limited to a specific region, typically the peroneal nerve, so it is not obvious that changes at this site will always reflect in amplitude and phase those occurring at the heart. Nevertheless, to the extent that data could be compared, the results obtained in the very few studies in which such data were obtained (2, 3) are not in conflict with the present findings. The particular conditions of these subjects (high number of apneic episodes of either type) prevented us from obtaining a reliable "baseline" reference during each phase of sleep to be compared against the apneic and postapneic phases of either type. Because the use of awake data was questionable due to the absence of sleep, we limited our analysis to comparisons between apneic and postapneic phases.
Conclusion Because chronic autonomic imbalance seems to be the main factor determining the increased cardiovascular risk in patients with sleep-disordered breathing, we believe that further studies are necessary to correlate the autonomic changes that we observed during sleep apneas occurring with the degree of chronic sympathetic activation during daytime. Our results and the comparison with data obtained by microneurography (2) indicate that either a different sympathetic control is acting at the heart level, prolonging the effects of sympathetic activity initiated in the obstructive apneic phase or a delay of the cardiac and vascular effector response is in fact responsible for the observed changes, compatible with sympathetic activation seen mainly during hyperventilation. Further studies are necessary to solve this problem, but our present finding may suggest some of the mechanisms leading to sustained sympathetic stimulation in patients with sleep-disordered breathing.
The authors gratefully thank Dr. Ing. Alberto Macerata, Bioengineering and Medical Computing Department, CNR Institute of Clinical Physiology, Pisa, Italy for his helpful criticisms and technical support in revising this manuscript.
This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Received in original form January 4, 2002; accepted in final form December 17, 2002
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
