| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
NEUROHUMORAL CONTROL OF CIRCULATION AND HYPERTENSION
1Educational Physiology Laboratory, Graduate School of Education, University of Tokyo, Bunkyo-Ku, Tokyo 113-0033; and 2PRESTO, Japan Science and Technology Corporation, Kawaguchi, Saitama 332-0012, Japan
Submitted 28 October 2002 ; accepted in final form 5 March 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
10-3.5 Hz, while 2) below that crossover frequency,
HRV was smaller in the constant routine than in the daily routine, with the
difference becoming significant (P < 0.05) at <10-4
Hz, 3) coarse-graining spectral analysis eliminated diet-induced
peaks in generic spectral analysis-based HRV spectra during the constant
routine and emphasized the crossover at 10-3.5 Hz, and
4) CBT correction did not alter the results. Below a frequency of
10-3.5 Hz (a period >1 h), HRV is strongly influenced by
behavioral factors; above that crossover frequency, HRV is behavior
independent, possibly reflecting an intrinsic regulatory system. physical activity; food intake; sleep-awake cycle; circadian rhythm; constant routine; humans
The division between the VLF and ULF components, i.e., 0.0033 Hz (the period corresponding to 5 min), seems to follow a report by Bigger et al. (4) showing decreases in the ULF and VLF components of HRV to be more predictive of all-cause mortality in
postinfarction patients. In these ULF and VLF bands, the power spectrum of
long-term HRV exhibits 1/f
-type, power-law scaling
(12,
22), and the slope () of
the scaling was also reported to be a good predictor of patient survival after
myocardial infarction (5). Thus
the investigation into the origin(s) of the slower fluctuations in HRV is
considered important (24).
We believe, however, that such categorization of long-term HRV does not have a firm physiological basis and is, at best, arbitrary. Indeed, long-term HRV is affected by many behavioral factors, including physical activity, food intake, sleep-awake cycles, and circadian and ultradian rhythms. Consequently, it is not known whether the decreases in the ULF and VLF components observed in the previously evaluated patients (4, 5, 24) were due to impaired cardiovascular regulatory mechanisms, reduced physical activity (3, 19) caused by the disease per se, and/or attenuated circadian cardiovascular variability, perhaps resulting from, e.g., co-morbid hypertensive episodes (8).
To facilitate interpretation of past (4, 5, 9, 24) and future results using the spectral characteristics of long-term HRV in humans and to help investigate the physiological origin(s) of slower fluctuations in HRV, we studied the behavioral influences affecting lowfrequency HRV. Our specific question was as follows: What would happen to long-term HRV in healthy humans if we eliminated, or at least minimized, known behavioral modifiers of HRV? To address this question, we examined the effect of selected behavioral factors on HRV using a constant-routine protocol (6, 7, 16) that included simultaneous core body temperature (CBT) correction for HRV.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Experimental protocols. Originally used in chronobiological research, the constant-routine protocol is a technique designed to investigate the human circadian pacemaker without environmental and behavioral influences (6, 7, 16). Subjects were instructed to keep their regular sleep schedules (0000–0200 and 0700–0900 for sleep onset and awakening, respectively) and to refrain from vigorous exercise or alcohol consumption during the week before the experiment. They then reported to the laboratory at about 0800 after an overnight fast.
After placement of a rectal temperature probe and electrodes for electrocardiography (ECG), data collection was commenced at 0930–1130. During the constant-routine protocol, subjects were kept awake for 27 h in a constant semirecumbent posture with minimal physical activity. They were allowed to work on a laptop computer, read, listen to the radio, and talk to the staff (provided they were not overstimulated) in the controlled laboratory environment (24–25°C, <250 lx light intensity). Isocaloric meals, the caloric contents of which were calculated by dividing the age-estimated daily energy requirement for Japanese (2,300–2,650 kcal) by 12, were provided every 2 h to minimize diet-induced changes in HRV and CBT.
For comparison, HRV and CBT data potentially influenced by moderate and transient physical activity, larger meals, and sleep were collected by having the subjects participate in a daily-routine protocol, which was carried out in the same experimental setting, but in this case the subjects were allowed to continue their normal daily activities, although without vigorous exercise or alcohol consumption. The order of the constant-routine and daily-routine sessions was randomized.
Data collection. Beat-to-beat RRI (from the standard V5 lead of the ECG) and continuous body movements (BMs) were measured using a portable, waist-worn, long-term ambulatory monitor (2). This light-weight (200 g) and small (120 x 65 x 22 mm) device consisted of an amplifier for ECG, two shock sensors with amplifiers measuring trunk acceleration (in the vertical and horizontal axes with resolution of 0.08 g), an eight-bit CPU with 4-MHz frequency, an 8-Mbyte EEPROM, an analog-to-digital converter sampling at 250 Hz, an eight-bit parallel interface for data transfer, and a direct current power supply from two commercial dry cells. In the CPU, the analog output of the ECG amplifier was band-pass filtered to yield trigger sources corresponding to QRS spikes. The resultant RRI was stored sequentially in the memory. Signals from the shock sensors were recorded as BMs after full-wave rectification and integration over 8 s. In the present study, only BM in the vertical axis was used.
Rectal temperature, serving as an estimate for CBT (one of the standard measures used to assess the temporal characteristics of the human circadian pacemaker), was recorded with resolution of 0.01°C at 1-min intervals using a portable temperature logger (model KMC-604, Gram) (23, 25) or a telemetry device (model WEB-5000, Nihon-Koden).
Any abnormal RRI, caused by BMs or occasional extrasystoles, was corrected by omitting beats (for those <300 ms) or inserting beats (when the RRI was double or triple the length of the preceding intervals). The percentage of the corrected beats was <0.08% of the total of >90,000 beats. Abnormal rectal temperatures caused by probe slips or defecation were corrected by linear interpolation.
Data analyses. To evaluate the frequency characteristics of long-term HRV in the VLF and ULF bands, a new time series with 10-s averaged RRIs was constructed every second (i.e., at 1 Hz) and split into 20 time-shifted ensembles of 216 data points (216 s
18.2 h). The lag for the shifted ensembles was determined by dividing the
difference between the total length of data for each subject (>97,200 s =
27 h) and the length of the subset by 20. The linear trend of each subset was
eliminated by linear regression, and Bingham's data window was applied before
calculation of the power spectral density (PSD) of the subset using a fast
Fourier transform. The PSD of the entire HRV record was obtained as an
ensemble average of the spectral data for the 20 subsets. This forms the
generic spectral analysis (GSA) used in previous studies
(4,
5,
9).
Inasmuch as the effects of meals were intentionally periodic and may not have represented the intrinsic frequency character of long-term HRV, we also used coarse-graining spectral analysis (CGSA) (26,
27) to eliminate that periodic
component from the total spectral power of HRV. CGSA enables us to
discriminate fractal random walks
(15) with the
1/f-type, power-law-scaled spectra from simple
harmonic motions based on the fact that the original and the rescaled
(coarse-grained) time series had random phase relations only with fractal
signals (27). The algorithm
has been described in detail, and its efficiency in extracting periodic
components from mixed harmonic and fractal signals has been demonstrated
(27). The same 216
points of 10-s averaged, detrended, and windowed data were used for this
analysis, but because of the algorithmic nature of CGSA
(27), the lowest frequency of
the resultant PSD was twice as high as that obtained with GSA.
Further analysis was conducted to eliminate, or minimize, the effect of circadian and/or low-frequency ultradian rhythms on long-term HRV by using CBT signals as "templates." For this purpose, cross-correlation coefficients between HRV and CBT were calculated separately for each subject after the data were smoothed using 2-h moving averages (13). Once the lag time with the highest absolute correlation coefficient was obtained, least-squares linear regression was performed for the lagged HRV, with CBT as an independent variable. The residual RRI time series was used as a new HRV series, with less influence of circadian and/or low-frequency ultradian oscillatory components, and analyzed with CGSA as described above.
Statistical analysis. Values are means ± SD. The effects of the constant-routine or daily-routine protocol and frequency on the average log-PSD values for HRV and BM within a bin of log 0.1 Hz were tested by two-way analysis of variance (ANOVA) for the main effects and the interaction. Inasmuch as the BM was recorded at 0.125 Hz and we were interested in the spectral power in the ULF and VLF bands, the highest frequency of these analyses was set to 0.1 Hz for HRV and BM. When the interaction was significant, this was followed by paired t-tests for the effects of the protocol at the five lowest frequency bins, while Holm's correction was applied to keep the total error of the tests below 5 or 1%. The selection of these five frequency bins was based on the inspections of the log-PSD vs. log-frequency plots (see Figs. 3 and 4).
|
|
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Results for the representative subject shown in Fig. 1 obtained during the daily-routine and constant-routine protocols are shown in Fig. 2, A and
B, respectively. During the former, there were a few
episodes of maked tachycardia (decreased RRIs) and hyperthermia (increases in
CBT) associated with bursts of BM at 700 and 1,500 min. As expected, BM
was substantially reduced during the constant routine, and, together with the
absence of sleep effects, CBT exhibited a smooth transition manifesting the
existence of circadian rhythm. The transient tachycardia seen in the daily
routine was absent in the constant routine, but a substantial level of HRV
remained. Also, the time series of HRV during the constant routine showed less
nonstationarity over longer time scales, consistent with flatness (whiteness)
of the spectrum at the lower frequencies, as described below. CBT correction
seemed to reduce a circadian HRV component during the constant routine.
However, in the daily routine, the residual RRI time series still resembled
HRV before the correction.
|
When GSA was used (Fig. 3), the group mean PSD of HRV during the daily routine exhibited an almost linear decay on log-log axes, suggesting the existence of power-law scaling (5,
9,
12,
22) for the entire range of
frequencies. The ANOVA showed the significant (P < 0.01) main
effects of frequency and protocol, as well as the "protocol x
frequency" interaction. The mean PSD of HRV during the constant routine
showed power-law scaling "roughly" similar to that during the
daily routine at frequencies above 10-4 Hz. Below that
frequency, however, the spectral power of HRV during the constant routine was
significantly (P < 0.05) lower than that in the daily routine.
The mean PSD of BM in the daily routine also showed the power-law scaling at frequencies above 10-3.5 Hz, below which the spectral power was less sloped. The mean PSD of BM during the constant routine was consistently lower than that during the daily routine. There were significant
(P < 0.01) main effects of protocol and frequency and the protocol
x frequency interaction.
As shown in Fig. 3, there was a small "peak" at 10-3.8 Hz in the GSA-based HRV spectra for the constant routine. This frequency corresponds to 2 h, i.e., the regular meal interval. When CGSA was used to eliminate the periodic
influence of regular meals during the constant routine
(Fig. 4, top traces),
the peak disappeared, and the similarity in the power-law scalings for the
constant routine and daily routine above a crossover frequency of
10-3.5 Hz was emphasized. The ANOVA showed the significant
(P < 0.01) main effects of frequency and protocol and the protocol
x frequency interaction. Because the systematic changes seemed to occur
below that crossover frequency, post hoc comparisons for two protocols were
made at these frequencies. Consequently, the spectral power of HRV during the
constant routine was significantly (P < 0.05) lower than that
during the daily routine at this range of frequencies
(Fig. 4, top traces).
Further analysis with the CBT correction did not alter these results
(Fig. 4, bottom
traces). During the daily routine, however, the mean PSD of HRV remained
scaled over the entire range of frequencies, even after minimization of the
effects of BM and circadian rhythm.
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
In the present study, we analyzed the effects of such behavioral factors on long-term HRV in humans using a constant-routine protocol, which is a chronobiological research technique used to investigate human biological rhythms while minimizing environmental and behavioral influences (6, 7, 16). With the use of this technique, the effects of physical activity and sleepawake cycles were substantially reduced, and the effects of "large" meals were minimized by providing subjects with small isocaloric meals every 2 h. Furthermore, residual effects of the regular meals were eliminated by CGSA, and those of circadian rhythm were minimized by subtracting components of HRV covaried with ongoing CBT fluctuations. Thus, after the application of CGSA and the CBT correction (Fig. 4, bottom traces), the PSD of HRV during the constant routine was considered to be virtually free from known behavioral modifiers. We then compared the HRV spectrum obtained in the constant routine with that obtained in the daily routine, which contained all the aforementioned behavioral influences.
We found that, for healthy young individuals at frequencies below 10-3.5 Hz (a period longer than 1 h; dashed-dotted line in Fig. 4), the PSD of HRV in the constant routine was lower than that in the daily routine, indicating
that, in this frequency range, HRV is at least in part dependent on behavior.
At higher frequencies, by contrast, the power spectra of HRV in the constant
routine and daily routine were similar, suggesting that HRV with a periodicity
less than 1 h is relatively independent of the behavioral effects,
including those of usual daily activities, possibly reflecting an intrinsic
regulatory system.
Comparison of results obtained with different experimental protocols and analytic techniques enabled us to gain insight into the factor(s) affecting HRV at frequencies below 10-3.5 Hz. First, circadian rhythm seemed to play only a minor role in HRV at these frequencies, inasmuch as the
CBT correction did not substantially affect the results obtained during the
constant routine (Fig. 4). Second, the effect of physical activity appears not to be as strong as one
might expect, because in the daily routine the mean PSD of BM was less sloped
than that of HRV in this frequency range
(Fig. 3). If that is the case,
the CBT correction would largely eliminate components of HRV covaried with
movement-induced, low-frequency hyperthermic episodes
(Fig. 2A).
Nevertheless, after the CBT correction, the PSD of HRV during the daily
routine was still much higher than that during the constant routine
(Fig. 4). Indeed, we recently
reported that the contribution of BM to very-long-term HRV in humans was
observed only within a narrow range of frequencies with periods approximating
90 min (2). Third, the effects
of food intake were considered to be small, because during the constant
routine the size of the diet-induced peak in the GSA-based HRV spectra was
also small (Fig. 3). It is
therefore considered likely that the effects of sleep-awake cycles are a
potential cause of the greater HRV in this frequency range. Sleepawake
differences in HRV, for instance, were reported to be smaller in hypertensive
patients (8), and the effects
of co-morbid hypertension, or other co-factors affecting the sleep-awake
differences, on the PSD of long-term HRV in cardiac patients at the lowermost
frequencies should be studied further.
For HRV at frequencies higher than 10-3.5 Hz, we need to consider behavior-independent and possibly intrinsic mechanism(s) of the regulatory system, because the power-law spectra for the constant routine and the daily routine were similar in this frequency range
(Fig. 4). This is a frequency
region where researchers have recently observed "complex" and/or
"multifractal" HRV dynamics
(1,
10,
11). Using the same dataset
used in the present study, Amaral et al.
(1) recently observed that the
multifractality in this frequency range was preserved even during the constant
routine; moreover, pharmacological blockade of vagal and sympathetic
influences greatly reduced the multifractal complexity of HRV. Thus challenges
to the intrinsic autonomic mechanisms might shed light on HRV dynamics in this
frequency range.
The power spectrum of long-term HRV in free-running humans exhibits 1/f-type, power-law scaling, as shown by the daily-routine data in the present study. This phenomenon was first described more than a decade ago (12,
22), and subsequent clinical
studies have revealed that the power in the VLF and ULF bands
(4,
5) and the slope () of
the scaling (5) are good
predictors of patient survival after myocardial infarction.
More recent studies have focused on a "crossover" phenomenon in the power-law scaling of long-term human HRV, seen in log-log plots with multiple linear relations. For example, by using a detrended fluctuation analysis (18), Peng et al.
(17) found such a crossover at
10 heartbeats. Inasmuch as we used 10-s averaged RRI data in the present
study, we were unable to look at the effects of the constant routine on this
high-frequency crossover; moreover, interpretation of this phenomenon is
difficult because of the existence of highly periodic fluctuations at around
this frequency (14,
21).
On the other hand, not much attention had been paid to the possibility of a low-frequency crossover phenomenon. We found that the power-law scaling in longterm human HRV could be categorized into two frequency components, which were divided at 10-3.5 Hz and had different physiological
meanings. A similar result was recently reported by Sakata et al.
(20). These investigators
found that, when calculated by CGSA, the power-law behavior in the PSD of 24-h
HRV during the daily routine had a crossover at 10-3 Hz; they
also showed that the slope below that frequency correlated negatively with
age. We therefore believe that one should be cautious about using a uniform
scaling exponent (5,
9) for analyzing long-term
HRV.
Finally, our findings might suggest a need for reevaluation of the frequency divisions described in the Task Force report (24). The Task Force division between the VLF and ULF bands is shown in
Fig. 4 to emphasize the
apparent lack of positive reasons for such categorization; the spectral
characteristics around that boundary frequency are quite uniform. Instead, the
results of the present study suggested that the division of ULF and VLF
components at 10-3.5 Hz (a period corresponding to 1 h),
which potentially probe different physiological mechanisms, would be more
physiological and thus suitable for research targeting the origin(s) of
low-frequency fluctuations in human HRV.
|
ACKNOWLEDGMENTS |
|---|
This work was supported in part by Ministry of Education, Culture, Sports, Science, and Technology Grants-in-Aid for Scientific Research 10480005 and 11694135.
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. Grossman, F. H. Wilhelm, and M. Spoerle Respiratory sinus arrhythmia, cardiac vagal control, and daily activity Am J Physiol Heart Circ Physiol, August 1, 2004; 287(2): H728 - H734. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. M. Stauss Heart rate variability Am J Physiol Regulatory Integrative Comp Physiol, November 1, 2003; 285(5): R927 - R931. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
P < 0.05; 
P < 0.01, constant routine
vs. daily routine.
