INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

DETERMINATION OF THE NATURE of the influence of sleep on cardiac activity is of considerable interest because of its potential clinical relevance. For example, there are cardiovascular consequences of the nighttime arousals that occur in disorders such as sleep apnea (29). It is generally accepted that compared with wakefulness (usually assessed just before sleep), non-rapid eye movement (NREM) sleep in humans is associated with an increase in cardiac parasympathetic nervous system (PNS) activity [as measured by noninvasive methods such as respiratory sinus arrhythmia (RSA)] (2, 23, 32-35, 37). More specifically, most agree that PNS activity progressively increases across the four stages of NREM sleep (21, 32, 34, 36), although not all have found this (35, 37). In rapid eye movement (REM) sleep, PNS activity appears to decrease compared with NREM sleep (2, 21, 23, 32-34, 36, 37).

Noninvasive cardiac measures (such as the 0.1-Hz peak derived from the spectral analysis of heart beat-to-beat intervals) indicate that sympathetic nervous system (SNS) activity decreases slightly or remains unchanged in the transition from wakefulness to NREM sleep (2, 23, 32-37). However, in humans, recordings of the sympathetic innervation of skeletal muscle blood vessels have shown a marked progressive fall in sympathetic activity with deepening NREM sleep (16, 22, 30, 31). It remains unclear whether this difference is a function of the invasiveness of the methodology or central versus peripheral autonomic activity. In REM sleep, SNS activity has been found to decrease (2, 25), increase (16, 21, 22, 30, 32, 34, 35), or show no change (23, 33, 36) relative to NREM sleep regardless of whether the measurement technique was invasive. These contradictory findings are probably due to studies failing to distinguish between tonic and phasic SNS activities. Thus SNS activity in tonic REM sleep may decrease slightly compared with NREM sleep and increase with occasional bursts during phasic REMs and myoclonic twitches (2, 16, 25).

In addition to the influence of sleep on autonomic nervous system (ANS) activity, there has been interest recently in a possible circadian influence on ANS activity. This interest stems from reports of an increased incidence of cardiovascular events in the morning (e.g., Ref. 20). However, the existence of an endogenous circadian influence on ANS activity has not yet been established because the majority of past studies have either failed to control for the confounding effects of sleep (13) or because of changes in posture and physical activity (e.g., Ref. 12). With these factors controlled for, PNS activity has shown no change (9, 32) or a decrease during the daytime (5, 14), with SNS activity remaining relatively stable (5, 14, 18, 32). In this study, ANS activity was assessed over two 26-h periods while subjects were supine and their activity was kept to a minimum. During one period, the subjects were allowed to sleep during their normal sleep period (circadian and sleep influence present), whereas in the other, they were not (circadian influence present). It was hypothesized that both the PNS and SNS would be affected by sleep but that only the PNS would be influenced by the circadian system.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Subjects

The laboratory procedures were approved by the Human Ethics Committee of the University of Melbourne (Australia), and all subjects gave written informed consent before their participation. The subjects received financial reimbursement for their time.

Design

Procedures

In each session, each subject was confined to a supine position for 26 h. In both routines, the subjects were permitted to assume an upright position for 5 min, 2 h after their normal waking time. During the sleep routine, this allowed each subject to move out of his separate bedroom into the general laboratory space and during both routines to empty his bowels if he wished. Room temperature and ambient light were kept constant, and the laboratory was well ventilated. The effects of food intake on ANS activity were minimized because the subjects received 250-calorie snacks (noncaffeinated) every 2 h and free access to water but only during those time periods when the subjects were awake. On one occasion, a subject expressed hunger after a meal and was supplied with an additional 100-calorie snack. Bottles for urination were available.

The subjects were asked to arrive at the laboratory 4 h before the start of data collection for each session. The rectal probe was inserted, the appropriate electrodes were attached, and the subjects assumed a supine position at least 2 h before the onset of data collection. Fifteen-minute data-collection periods occurred hourly (30 min during the sleep phase of the sleep routine). In the sleep routine condition, electrodes to monitor sleep-wake state were attached 2-3 h before and removed 2-3 h after the subjects' normal sleep phase.

The subjects were permitted to listen to music and watch TV and nonarousing videos throughout the times in the experiment in which they were awake. During data-collection periods, the subjects were instructed to lie quietly on their backs and not talk. In between the data-collection periods, the subjects were permitted to roll slowly along their longitudinal axis, read, and talk to their fellow subjects and laboratory staff. The subjects were scrutinized continuously throughout the experiment to prevent them from sleeping except during the allowed sleep phase in the sleep routine. During the sleep phase of the sleep routine, the lights were turned off.

Assessment of sleep-wake state. Assessment of the sleep-wake state was conducted during the sleep phase of the sleep routine beginning at the time of each subject's usual sleep onset. Each subject's sleep-wake state was assessed by a left central (C₃) electroencephalogram referenced to the right earlobe (A₂), an electrooculogram (left and right outer canthi displaced vertically), and an electromyogram (submental) and visually scored according to standardized procedures (27). These measures were used to confirm that the subjects were asleep.

Assessment of respiratory activity. Respiration was measured to ensure that off-line data analysis was conducted on periods of stable ventilation, which was necessary because the RSA measure is sensitive to large variations in respiratory rate (3). Respiration was assessed by measuring temperature changes in airflow via a Mon-a-therm tympanic temperature sensor inserted into a general purpose gas mask.

Assessment of rectal temperature. Rectal thermistors (Mallinckrodt Hi-Lo temp general purpose disposable probes) attached to Mini-Logger series 2000 ambulatory temperature recorders were used to measure the rectal temperature of each subject every 2 min. The thermistors were inserted by each subject 10 cm into his anus, and the lead from the thermistor was taped to the skin until it reached the waist where it was attached to the recorder.

Assessment of heart rate and T wave amplitude. Heart rate (HR) and T wave amplitude (TWA) were assessed with an electrocardiogram (ECG) measured through Meditrace Ag-AgCl spot electrodes. One electrode was placed on each subject's lower left and right rib cage and right clavicle.

Assessment of preejection period. Preejection period (PEP) was measured by impedance cardiography (Minnesota impedance cardiograph model 304B). The impedance signal was measured through four Meditrace Ag-AgCl spot electrodes. The two inner (recording) electrodes were placed 4 cm above the clavicle on the front of the neck and over the sternum at the fourth rib. The two outer (current) electrodes were placed over the fourth cervical vertebra, on the back of the neck, and on the back over the ninth thoracic vertebra (6). The electrodes were placed in the same position for each experimental session. A 4-mA AC current at 100 kHz was passed through the two outer electrodes and basal impedance and rate of change in the impedance waveform on a given beat (dZ/dt; /s) were estimated from the two inner electrodes.

The respiratory, ECG, and dZ/dt signals were analyzed off-line by computer software developed within the laboratory to calculate HR, RSA, 0.1-Hz peak, TWA, and PEP for every cardiac cycle (see Data Analysis).

Data Analysis

RSA and 0.1-Hz peak. Spectral analysis of the beat-to-beat intervals was carried out in each of the 5-min epochs. The differences between the time points of each R wave were calculated (R_n  R_n1) and analyzed with a power spectrum estimation (periodogram method). The frequency of the spectrum was expressed in units equivalent to cycles per second ("EqHz"), calculated as the inverse of the average R-R interval length (24). The program calculated the power spectral density estimate for each frequency bin, and these were combined to form frequency bands 0.02 EqHz wide. The power spectrum ranged from 0 to 0.50 EqHz. A program searched the frequency spectrum for two major components: a high-frequency component in the region of the respiratory rate (RSA), defined as the largest peak > 0.16 EqHz, and a low-frequency component at ~0.10 EqHz, defined as the largest peak < 0.16 EqHz but > 0.03 EqHz (<0.03 EqHz was designated as the DC component). The subject's respiration rate was calculated for each epoch to confirm that it corresponded to the peak that the program identified as the RSA component. Each frequency bin of each power spectrum was recalculated as a proportion of the adjusted total power of the spectrum (total power minus DC noise; Refs. 12, 24). Thus, from each power spectrum, the normalized areas of the RSA and 0.1-Hz peaks were determined. HR was also determined from the R-R intervals and averaged across each 5-min epoch.

TWA and PEP. The positions in time of the B point in the dZ/dt signal and the Q and T waves in the ECG were detected. The epochs were visually scanned and edited when the algorithm failed to correctly detect these points. PEP was then calculated for each cardiac cycle as the time interval between the Q wave on the ECG signal and the B point on the dZ/dt signal. The amplitude of the T wave for each cardiac cycle was calculated as the difference between the baseline (0 V) and the peak of the T wave. The PEP and TWA for each cardiac cycle were then averaged across each 5-min epoch.

Statistical analysis. Three forms of analyses were conducted. First, analysis of variance (ANOVA) methods assessed the effect of sleep on each dependent variable. The subjects' data were separated into "wake time" and "sleep time" in both routines according to their regular sleep-wake schedule. Thus, in the nonsleep routine, sleep time was the period in which the subjects were awake but would normally be asleep. A 2 × 2 ANOVA with repeated measures on each variable was used to compare state (wake time vs. sleep time) with type of routine (nonsleep vs. sleep). Second, the presence of a circadian component was tested by fitting functions to the time series ordered according to time from normal sleep onset. The data in each routine were fitted with 24-h (simple) and 24-h fundamental with 12-h harmonic (complex) curves, with the significance of an F-test indicating goodness of fit. The complex curves were viewed to be a better fit than the simple curves if they significantly increased the variance accounted for as determined with a stepwise regression. The third analysis compared the time series derived from the cardiac variables with the body temperature time series with cross-correlational methods.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Preliminary Analyses

View larger version (21K): [in this window] [in a new window]   Fig. 1.   Change in () respiratory sinus arrhythmia (RSA; in proportion of power spectrum), preejection period (PEP; in ms), heart rate (HR; in beats/min), 0.1-EqHz (frequency spectrum expressed in units equivalent to cycles/s) peak (in proportion of power spectrum), and T wave amplitude (TWA; in µV) ordered according to 24-h-clock time and averaged across 12 subjects. Solid lines, nonsleep routine; dashed lines, sleep routine. Values are means ± SE (see RESULTS for explanation).

View larger version (19K): [in this window] [in a new window]   Fig. 2.   RSA (in proportion of power spectrum), PEP (in ms), HR (in beats/min), 0.1-EqHz peak (in proportion of power spectrum), and TWA (in µV) ordered according to time from normal sleep onset and averaged across 12 subjects. Solid lines, nonsleep routine; dashed lines, sleep routine. Values are means ± SE (see RESULTS for explanation).

Ideally, for each subject in each routine, there were 25 hourly measurements of each dependent variable (RSA, PEP, HR, 0.1-EqHz peak, and TWA). Because of a range of technical difficulties, 5% of the hourly measurements was lost. If the missing data consisted of one sample between two available samples, then it was replaced with the mean of the before and after samples. If there were more than one missing consecutive sample (this occurred on only three occasions), then the values of the missing samples were estimated from the group trend adjusted by the particular subject's average level for the variable.

Of the 12 subjects, 9 in the nonsleep and 8 in the sleep routine yielded reliable temperature data with no probe slippage. The recordings taken every 2 min were averaged to yield hourly estimates of rectal temperature, recalculated to reflect the average change from the mean, and plotted according to clock time and sleep onset time (Fig. 3).

View larger version (26K): [in this window] [in a new window]   Fig. 3.   Core body temperature (in °C) ordered according to time from normal sleep onset (A) and 24-h-clock time (B) averaged across subjects. Solid lines, nonsleep routine (n = 9 subjects); dashed lines, sleep routine (n = 8 subjects). Values are means ± SE (see RESULTS for explanation).

RSA

PEP

HR

0.1-EqHz Peak

TWA

Group Differences

In terms of group differences in the phase and amplitude of the autonomic variables across routines, there were two differences. Group 21 showed a higher amplitude in their RSA measurement during the nonsleep routine (P < 0.05). Because the two groups showed similar trends across time, this difference appeared to be due to group 21 being further in time from their last sleep period than group 12. Thus group 21 had a larger presleep increase and postsleep decrease in RSA. A more critical group difference occurred with TWA in the nonsleep routine where the position of the phase of group 21 was more advanced in time compared with that of group 12 (P < 0.05). This finding is reflective of a general difference in the TWA pattern between the two groups (Fig. 4). Considering that PEP and 0.1 EqHz did not show such dramatic group differences, this suggests that TWA was more sensitive to the experimental situation than the 0.1-EqHz peak and PEP.

View larger version (30K): [in this window] [in a new window]   Fig. 4.   TWA (in µV) ordered according to time from normal sleep onset in group 12 (A) and group 21 (B). Solid lines, nonsleep routine; dashed lines, sleep routine. Values are means ± SE (see RESULTS for explanation).

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The results from this study indicate that the PNS is primarily influenced by the circadian system and not by sleep, whereas the SNS is primarily influenced by sleep and not by the circadian system. RSA and PEP are believed to be relatively pure measures of PNS and SNS activity, respectively. Physiologically, RSA is the shortening and lengthening of the beat-to-beat intervals during inspiration and expiration, respectively. RSA is viewed by most as a "pure" measure of PNS activity because SNS efferents cannot operate at the respiratory frequency (1, 3, 6). This interpretation has been supported by many experimental manipulations (e.g., Refs. 6, 26). PEP is the time interval from the onset of ventricular depolarization to the opening of the semilunar valves. Because of the predominance of SNS innervation of the ventricular myocardium, PEP is regarded as being inversely proportional to cardiac SNS activity (6).

HR, the 0.1-EqHz peak, and TWA were influenced by both the circadian and sleep systems. This was anticipated because all three measures are influenced by the PNS and SNS. Although some regard the 0.1-EqHz peak as a primary measure of SNS activity (e.g., Refs. 12, 32, 34), it most likely represents both PNS and SNS afferents and efferents in the baroreflex control of blood pressure (e.g., Refs. 1, 17, 26). TWA is also a disputed measure of SNS activity. It represents myocardial ventricular repolarization, and because the PNS influence on the ventricular myocardium is believed to be slight, TWA has been regarded as an index inversely proportional to cardiac SNS activity (e.g., Refs. 10, 11). Others (7, 28), however, have criticized its use. Indeed, in this study, it appeared to be more reflective of mental stress rather than of tonic SNS activity and so highlighted the group differences in time from the last sleep period (Fig. 4).

Thus, as hypothesized, cardiac PNS, but not SNS, activity was found to be influenced by the circadian system. Previous suggestions of a circadian influence on PNS activity, but not on SNS activity, have come from previous studies that controlled for posture and changes in physical activity. They found PNS activity (RSA) decreased (5, 14), whereas SNS activity (0.1-Hz peak or microneurography) remained relatively stable during the daytime (5, 14, 18, 32). The one study showing little circadian influence on PNS activity (5) collected only three assessments of ANS activity in a 24-h period, in contrast to the hourly measurements taken in this study. Contrary to expectations, PNS activity was not influenced by the sleep system. There have been previous findings of changes in cardiac PNS activity between sleep stages (21, 32, 34, 36), but in these studies, changes in the influence of the circadian system were not controlled for. Furthermore, in dogs, increases in HR with arousal from NREM sleep have been mainly attributed to an increase in SNS activity rather than to a decrease in PNS activity (15).

Given the circadian influence on PNS activity, it was interesting to compare the variations in RSA and core body temperature across the 24-h period. Although RSA was found to correlate with body temperature and was fit significantly with complex curves in both routines, it is important to note that there were marked phase differences between the two variables (Figs. 2 and 3). Although it is possible that PNS activity is influenced by an oscillator separate from the circadian oscillator that influences body temperature, it is important to note that other "markers" of the endogenous circadian oscillator believed to control body temperature, such as cortisol, do not show a similar pattern to body temperature. Thus it is not possible to determine from these results the nature of the circadian control of PNS activity. The finding of a circadian influence on PNS activity and the fact that PNS activity began to increase in anticipation of sleep onset suggest that there is some circadian control of ANS functioning in preparation for sleep onset. Thus the circadian system appears to downregulate ANS activity as sleep onset approaches by increasing PNS activity rather than by decreasing SNS activity. Indeed, it may be that a failure of PNS activity to increase in anticipation of sleep onset might be as important as SNS arousal in causing sleep onset insomnia.

The decrease in SNS activity with sleep, as measured here for the first time with PEP, has implications for the earlier findings based on the 0.1-EqHz peak. Previously, invasive and thus necessarily peripheral measurements of SNS activity such as microneurography in humans have found marked decreases in SNS activity with sleep (16, 22, 30, 31), whereas the more central and noninvasive measure, the 0.1-EqHz peak, has not consistently shown this effect (2, 23, 32-36). With the present results, it now seems likely that this discrepancy is due more to the dubious nature of the 0.1-EqHz peak as a pure measure of SNS activity rather than to either the invasiveness or peripheral nature of the microneurographic technique. Because of the design of this experiment, the focus of the study was not on sleep stage differences in ANS activity. However, the decrease in SNS activity with sleep onset rather than in anticipation of sleep onset is consistent with previous findings (5, 9, 14, 32). Thus the sleep influence on SNS activity may have implications for the cardiovascular sequelae of sleep disorders such as obstructive sleep apnea, which is associated with repetitive arousals from sleep (29).

In terms of the methodological limitations of this study, it is important to note that although the constant-routine procedure was originally designed to control for the masking effects of food intake, sleep, and changes in posture and physical activity, the increase in fatigue and stress due to the desire to be more active than the procedure permitted and/or the effect of a maintained supine posture may have produced additional masking effects (19). However, as mentioned in Design, this was controlled for by having two groups of subjects who started the experiment at different times during the day. It is also possible that in the week between the two routines, the subjects may have phase shifted slightly from the sleep-wake schedule they had just before their first routine. Nevertheless, this would have been controlled for by the counterbalancing of routines across subjects.

The results from this experiment also indicate that evidence for a circadian influence on SNS activity most likely reflects postural, physical activity, and sleep confounds in the design of previous studies. Indeed, of the three previous studies that did control for these factors (5, 14, 32), SNS activity was not found to vary markedly during the daytime, consistent with the findings of this experiment. Therefore, the recent interest in identifying the cause of the increased incidence of cardiovascular events in the morning may be better aimed at examining the influence of increased physical activity and/or postural change during morning awakening. Another important factor may be the phasic bursts of SNS activity or general autonomic disruption in REM sleep (2, 16, 25).

In conclusion, the activity of the PNS was found to be predominantly influenced by the circadian system and SNS activity by the sleep system. Future research should focus on examining more closely these separate influences, for example, changes in PEP in different sleep stages, the importance of decreased SNS activity for the initiation of sleep (e.g., Ref. 4), and the mechanisms of the circadian influence on PNS activity.