From the Departments of Internal Medicine and Physiology, Hunter Holmes
McGuire Department of Veterans Affairs Medical Center, and Medical College of
Virginia, Virginia Commonwealth University, Richmond.
Correspondence to J. Andrew Taylor, PhD, Cardiovascular Research Laboratory HRCA Research and Training Institute, 1200 Centre St, Boston, MA 02131. E-mail ataylor{at}mail.hrca.harvard.edu
Methods and Results—We recorded RR intervals and
arterial pressures during three separate sessions, with the
patient in supine and 40 degree upright tilt positions, during
20-minute frequency (0.25 Hz) and tidal volume–controlled breathing
after intravenous injections: saline (control), atenolol
(0.2 mg/kg, ß-adrenergic blockade), atropine sulfate (0.04 mg/kg,
parasympathetic blockade), atenolol and atropine (complete autonomic
blockade), and enalaprilat (0.02 mg/kg, ACE blockade). We integrated
fast Fourier transform RR-interval spectral power at very low (0.003 to
0.03 Hz), low (0.05 to 0.15 Hz), and respiratory (0.2 to 0.3 Hz)
frequencies. ß-Adrenergic blockade had no significant effect on
very-low- or low-frequency RR-interval power but increased respiratory
frequency power 2-fold. ACE blockade had no significant effect on low
or respiratory frequency RR-interval power but modestly (
Conclusions—Although very-low-frequency heart period rhythms are
influenced by the renin-angiotensin-aldosterone
system, as low and respiratory frequency RR-interval rhythms, they
depend primarily on the presence of parasympathetic outflow. Therefore
the prognostic value of very-low-frequency heart period
oscillations may derive from the fundamental importance of
parasympathetic mechanisms in cardiovascular health.
A more recent suggestion is that very-low-frequency heart period
oscillations reflect the influence of fluctuations of renin
activity on arterial pressure. This hypothesis is supported
by short-term studies conducted in resting, unanesthetized
dogs8 and long-term studies conducted in
post–myocardial infarction patients,9 which show
that ACE blockade increases very-low-frequency heart period
variability. These studies have at least two shortcomings. First,
autonomic responses of healthy dogs may differ qualitatively from
autonomic responses of patients.10 Second,
although Holter recordings permit analysis of very slow
RR-interval fluctuations, they do not allow for control of common
factors known to affect heart period variability such as posture,
physical activity, breathing frequency, and tidal
volume.11 12 Moreover, the actual cascade of
physiological events that generates
very-low-frequency heart period variability has not been defined. For
example, it is not known whether human arterial pressure
oscillates at very low frequencies or if arterial pressure
and heart period oscillate together, independent of respiration.
We assessed autonomic and
renin-angiotensin-aldosterone system
contributions to heart period and arterial pressure
oscillations in healthy human subjects. Our paradigm
enabled us to evaluate very-low-frequency oscillations with
power spectrum analysis, without limitations inherent in Holter
recordings. We recorded 20 minutes of beat-by-beat RR
intervals and arterial pressures in healthy resting humans
during controlled breathing in the supine and 40 degree upright tilt
positions. We evaluated contributions of the autonomic nervous system
with ß-adrenergic blockade, cholinergic blockade, and complete
autonomic blockade and contributions of the
renin-angiotensin-aldosterone system with ACE
blockade. Our results support a role for the
renin-angiotensin-aldosterone system in
very-low-frequency RR intervals but not in arterial
pressure oscillations. More importantly, our data
underscore the primacy of cardiac parasympathetic activity in
generating short-term heart period oscillations and suggest
that the prognostic value of heart period variability derives from the
association between cardiac parasympathetic mechanisms and
cardiovascular health.
Measurements and Protocol
During each trial, we recorded ECG lead II, beat-by-beat
photoplethysmographic arterial pressure (Finapres, Ohmeda)
in the finger, brachial arterial pressure (Dynamap,
Critikon) once every 3 minutes, respiratory excursions (pneumobelt),
breath-by-breath tidal volume (Fleisch pneumotachograph), and
breath-by-breath end-tidal carbon dioxide concentration (infrared
analyzer connected to a face mask with a 2-way respiratory
valve). We recorded all signals continuously on FM tape for
subsequent analog-to-digital conversion.
After catheter insertion, instrumentation, and instruction, subjects
rested quietly for at least 10 minutes. Before each trial, subjects
controlled their breathing frequency in response to an auditory signal
at 0.25 Hz (15 breaths/min) to determine the most comfortable tidal
volume. Subsequently, this inspired tidal volume was displayed on an
oscilloscope to provide visual feedback so the subject could maintain a
constant tidal volume. Controlled breathing was maintained for at least
20 minutes for each trial.
Immediately after the supine and tilt control trials, subjects were
given either saline for autonomic studies or enalaprilat for
renin-angiotensin-aldosterone system studies
and then allowed to rest for an additional 25 minutes. This delay was
included in the protocol to ensure full expression of the
cardiovascular effects of enalaprilat. Although our
dose reduces plasma angiotensin to less than one-tenth
basal levels for up to 4 hours after intravenous infusion
in young adults,14 the peak
cardiovascular effects of enalaprilat occur
Data Analysis and Statistics
Frequency domain analyses were performed on beat-by-beat RR
intervals and systolic and diastolic pressures. We
used a power spectrum analysis based on the Welch algorithm of
averaging periodograms.17 The 1200-second time
series of beat-by-beat RR intervals and arterial pressures
were interpolated by a cubic spline function at 4 Hz to obtain
equidistant time intervals and then were divided into 5 equal
overlapping segments. Each segment was detrended, Hanning filtered, and
fast-Fourier transformed to its frequency representation
squared. The periodograms were averaged to produce the spectrum
estimate. Our estimation with the Welch method used 400-second segments
to obtain estimates spaced at 0.0025 Hz, giving 12 estimates in the
range of 0.0025 to 0.03 Hz and allowing detection of
oscillations as slow as 0.0025 Hz. Areas under the power
spectra in very low, low, and respiratory frequencies (defined as 0.003
to 0.03, 0.05 to 0.15, and 0.20 to 0.30 Hz) were integrated and used
for statistical comparisons. Relative power (normalized units) was not
calculated because we are skeptical of its validity as an accurate
measure of cardiovascular
variability.18 (For example, in an earlier study,
fixed-rate atrial pacing eliminated all RR-interval respiratory
frequency spectral power in absolute values but did not alter spectral
power in normalized units.19 ) To examine the
strength of the relation between very-low-frequency RR-interval and
systolic pressure variabilities, we derived the coherence
estimate by cross-spectral analysis based on models described
previously.20 21 Although an estimate >0.5 has
been used to signify that two cardiovascular signals
covary significantly at a given frequency,20 our
spectral technique provided 9 degrees of freedom so that a minimum
value of 0.58 was necessary to reject the null hypothesis that the
coherence function was not different from 0 at a 0.05 significance
level (see Appendix
Effects of drugs and position on average RR intervals and
arterial pressures for each session were evaluated with
repeated-measures ANOVA and t tests with a Bonferroni post
hoc correction to identify significant differences.
Nonparametric statistics were used to examine effects on
RR-interval and arterial pressure variabilities because
spectral powers were not distributed normally, even after log
transformation.22 Spearman rank order
correlations were calculated for control supine and tilt RR intervals
and arterial pressures at very-low-frequency powers to
assess consistency across sessions. A series of
univariate sign-rank tests was used to assess drugs
effects. Differences were considered significant at P<0.05.
Measurements are reported as mean±SE.
Mean RR Interval and Arterial Pressures
Figure 1
Very-Low-Frequency Spectral Power
Figure 3
Figure 4
Figure 5
Low and Respiratory Frequency Spectral Power
Atropine with or without atenolol reduced low and respiratory frequency
systolic pressure power in the supine position. Atropine
reduced low-frequency diastolic pressure spectral power in
the supine position. Combined atropine and atenolol reduced respiratory
frequency diastolic pressure power in the supine position,
and atropine alone increased respiratory frequency
diastolic pressure spectral power in the tilted position.
ACE blockade did not affect low or respiratory frequency
arterial pressure spectral power significantly in either
position.
Thermoregulatory Mechanisms
Renin-Angiotensin-Aldosterone Mechanisms
Before our study, most published data on very-low-frequency RR rhythms
in humans came from 24-hour Holter monitor
recordings.1 2 9 26 27 Although Holter
monitor recordings provide a sufficiently long data collection
period to document fluctuations occurring as slowly as only once every
5.5 minutes (0.003 Hz.), Holter recordings are not controlled
for common factors known to affect RR-interval variability, including
posture, activity, breathing frequency, and tidal
volume.11 12 Holter recordings obtained
in post–myocardial infarction patients9 28 and
in congestive heart failure patients29 have shown
that ACE blockade increases both frequency and time domain measures of
very-low-frequency RR-interval variability. However, these findings
have not been replicated with Holter recordings in healthy
subjects.27 This discrepancy may reflect a
greater physiological role of the
angiotensin system in patients with
cardiovascular disease. Our protocol differs from the
earlier studies in patients9 28 29 in that our
subjects were young and healthy and from the Holter studies in healthy
subjects27 in that our subjects controlled their
respiratory frequency and tidal volume and we controlled body position.
Nonetheless, our findings support a role for the
renin-angiotensin-aldosterone system in
very-low-frequency RR-interval fluctuations and provide new information
regarding autonomic mechanisms.
Our results do not indicate explicitly how ACE blockade increases
very-low-frequency RR-interval spectral power; we do not know if this
observation reflects episodic increases of plasma renin activity,
potentiation of bradykinin (which accounts for a portion of the
hypotensive effects of ACE
inhibitors30 ), or modulation of some
other neurohumoral influence that fluctuates at very low frequencies.
We are intrigued by the observation that ACE blockade increased
RR-interval spectral power in the supine but not the upright tilted
position. Although we cannot exclude a ß-statistical error (that we
erred by studying too few subjects), our observation may have a
physiological basis: Increases of
angiotensin II levels occur episodically, as
arterial pressure fluctuates above and below a
threshold.31 During upright tilt,
arterial baroreceptor input may remain consistently
below the threshold for increases in plasma renin activity. Therefore
although absolute levels of plasma renin activity are increased in the
upright position, their fluctuations might be reduced.
Serial measurements of plasma renin activity have defined the capacity
of the kidney to modulate renin activity in response to sinusoidal
variations in renal arterial pressure at 0.002
Hz.32 However, even if fluctuations of plasma
renin activity occur within the very-low-frequency range, they may not
provoke systemic hemodynamic changes at these
frequencies. A study published by Cowley et al,33
conducted in dogs, suggests that responses to changes of plasma renin
activity levels occur too slowly to influence very-low-frequency
events. Arterial pressure increases provoked by step
reductions of renal artery pressure occur slowly, over
A second explanation for our observations is that steady plasma renin
activity levels modulate other neurohumoral mechanisms that fluctuate
at very low frequencies. Akselrod and coworkers8
speculated that chronic levels of renin activity and
angiotensin dampen fluctuations of peripheral
vasomotor tone and that ACE blockade increases vasomotor tone
fluctuations and (presumably by an arterial baroreflex
mechanism) corresponding RR-interval fluctuations. Our study does not
support a baroreflex mechanism; we found no increase of
arterial pressure spectral power after enalaprilat (Figure 5
Vagal Mechanisms
Limitations
The validity of our conclusion that very-low-frequency RR-interval
fluctuations in humans are not mediated by baroreflex mechanisms hinges
on the fact that our estimates of coherence between
arterial pressure and RR intervals <0.58 are not
statistically significant (see Appendix
Finally, the effects of our pharmacological interventions may have been
circumscribed by the dosages used. For example, we gave the hydrophilic
ß-adrenergic–blocking drug atenolol in a dose that was worked out
for the lipophilic ß-adrenergic–blocking drug
propranolol.38 However, we propose
that our atenolol dose is justified on three counts: Atenolol enters
the central nervous system5 ; atenolol and metoprolol,
another lipophilic ß-adrenergic–blocking drug, enhance 24-hour
RR-interval spectral power equally39 ; and none of
our conclusions are contingent on ß-adrenergic blockade being
"complete." It may also be argued that our (slightly more than)
clinical dose of enalaprilat produced no profound
physiological effects suggestive of ACE inhibition.
Yet it may not be surprising that our young (23 to 28 years old),
normotensive (Finapres-derived systolic pressure of 126
mm Hg) subjects demonstrated no profound changes in average RR
interval or arterial pressures with enalaprilat. Our dose
of enalaprilat does not alter RR interval and reduces
arterial pressure greatest in individuals with highest
basal pressure,14 describing a relation between
basal systolic pressure and pressure reduction with an
intercept
In summary, our study provides support for previous observations that
the renin-angiotensin-aldosterone system
influences very-low-frequency RR-interval variability. In addition, our
study addresses the role of the autonomic nervous system in the
generation of these RR-interval oscillations: We found that
as with respiratory and low-frequency RR-interval variabilities,
very-low-frequency RR-interval oscillations are very much
dependent on parasympathetic tone. Therefore the prognostic value of
very-low-frequency heart period oscillations may derive
from the fundamental importance of parasympathetic mechanisms in
cardiovascular health.
De Boer et al20 proposed that coherence level
indicates the strength of the linear association and provides a gauge
of variability in phase estimates between two
cardiovascular variables. After the work of de Boer
et al,20 it has become common to claim a
significant linear relation between two cardiovascular
time series when coherence values exceed 0.5. Although 0.5 was the
correct threshold for significance based on the parameters
of de Boer and colleagues' estimation procedure, other
parameters or estimation procedures will in general
determine a different threshold necessary to accept the hypothesis that
coherence significantly exceeds 0. These parameters effect
the degrees of freedom (d) of the estimate which, with a
significance level
As an example, we used the periodogram method of
Welch17 to compute our estimates of spectra and
calculated coherence to investigate whether the variance of
very-low-frequency oscillations in RR interval and
systolic pressure had a significant linear relation. Our
analysis divided N=4800 samples into segments of
length L=1600, which overlapped by half, were detrended, and
were multiplied by a Hanning window (constant=2.67). Overlapping
segments decreases the variance of the estimated spectrum and increases
the degrees of freedom17 such that in this case
the window constant effectively becomes 2.83. Thus our estimates have
d=2.83N/L=8.49, or 9 degrees of freedom. For
d=9 and
Although minimal coherence is specific to the spectral technique,
coherence estimated by other techniques (eg, autoregressive models) can
be similarly analyzed for significance.
However, the cutoff for minimal coherence should not be applied
indiscriminately. The level of 0.5 can be defended because it suggests
a relation between two signals based on 50% shared variance, whereas
an F test with sufficient degrees of freedom can indicate
that a low shared variance (ie, low coherence value) significantly
differs from 0. Nonetheless, it is important to note that spectral
techniques with few degrees of freedom can produce high coherence
values that do not significantly differ from 0. Thus meaningful
interpretation of coherence should consider the level of confidence in
the estimate.
Received November 10, 1997;
revision received March 20, 1998;
accepted March 26, 1998.
© 1998 American Heart Association, Inc.
Clinical Investigation and Reports
Mechanisms Underlying Very-Low-Frequency RR-Interval Oscillations in Humans
Abstract
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
Background—Survival of post–myocardial infarction patients
is related inversely to their levels of very-low-frequency (0.003 to
0.03 Hz) RR-interval variability. The physiological
basis for such oscillations is unclear. In our study, we
used blocking drugs to evaluate potential contributions of sympathetic
and vagal mechanisms and the
renin-angiotensin-aldosterone system to
very-low-frequency RR-interval variability in 10 young healthy
subjects. 
21%)
increased very-low-frequency power in the supine (but not upright tilt)
position (P<0.05). The most profound effects were
exerted by parasympathetic blockade: Atropine, given alone or with
atenolol, abolished nearly all RR-interval variability and decreased
very-low-frequency variability by 92%.
Key Words: receptors, adrenergic, beta • renin • heart rate • vagus nerve
Introduction
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
Reductions of
very-low-frequency RR-interval oscillations (with periods
between 30 and 330 seconds or 0.03 to 0.003 Hz) are associated with
increased risk for cardiac and dysrhythmic
death1 2 and possibly
syncope.3 Two mechanisms for these very slow
heart period oscillations have been proposed:
thermoregulation and the
renin-angiotensin-aldosterone system. Because
slow oscillations of peripheral vascular tone
can be entrained by thermal stimuli at low
frequencies,4 5 Hyndman6
and Kitney7 suggested that they (and
corresponding RR-interval rhythms) are in fact caused by
thermoregulation.
Methods
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
Subjects
Ten healthy subjects (7 men and 3 women), 23 to 28 years of age,
participated in this study. Volunteers were nonsmokers without
histories of cardiovascular or other major diseases who
were taking no cardioactive medications. Subjects refrained from
alcohol or caffeine ingestion and strenuous physical activity for 24
hours preceding the study sessions. This research was approved by the
human research committees of the Hunter Holmes McGuire Department of
Veterans Affairs Medical Center and the Medical College of Virginia.
All volunteers gave their written informed consent to
participate.
Three sessions were conducted on separate days. For each
session, drugs were administered through an antecubital vein catheter
in a fixed order: (1) saline (control), atenolol (0.2 mg/kg,
ß-adrenergic blockade), and atropine sulfate (0.04 mg/kg, combined
ß-adrenergic and muscarinic cholinergic blockade); (2) saline,
atropine, and atenolol; and (3) saline, enalaprilat (Merck, Sharp &
Dohme, 0.02 mg/kg [slightly higher than the recommended clinical dose
of 1.25 mg for a 70 kg patient], ACE blockade), and saline (placebo).
Responses to each drug administration were assessed with the patient in
the supine and 40 degree upright tilt position. (We used 40 degree tilt
to increase sympathetic outflow because healthy young humans have low
levels of sympathetic outflow in the supine
position.13 ) Thus each of the 3 sessions
comprised 6 trials (3 drug administrations in 2 positions). Session
order and position order after drug administration were randomized, and
subjects were not told which drugs they would be given. 1.5 to 2
hours after injection.15 16 Therefore in addition
to the 25 minutes of rest, only data from the third set of supine and
tilt trials (which were at least 90 minutes after enalaprilat
administration) were used to assess the effects of ACE blockade on
cardiovascular variability. After administration of the
autonomic blocking drugs, a 7-minute drug effect period was allowed
before measurements during the supine and tilt trials.
All data were digitized at a rate of 500 Hz with commercial
hardware and software (CODAS, Dataq Instruments). ECG R waves were
identified to derive beat-by-beat RR intervals. Arterial
pressure peaks and valleys were identified to derive beat-by-beat
systolic and diastolic pressures. Mean and standard
deviations for RR interval and systolic and
diastolic pressures were calculated from the beat-by-beat
values for each 20-minute trial. 
).
Results
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
Respiration
Subjects controlled their respiration very well (average
respiratory frequency for all trials: 0.24±0.002 Hz, 4.11 seconds per
breath); they maintained average tidal volumes within 15% of target
volumes in 176 of the 180 trials and within 20% of target volumes in
the 4 remaining trials. Although end-tidal CO2
concentrations tended to decrease from the beginning to the end of the
20-minute paced breathing trials, the average decline was <5%.
Table 1 lists average RR intervals
and arterial pressures for all trials. Upright tilt
decreased RR intervals consistently (P<0.05) except
after complete autonomic blockade. Upright tilt decreased
systolic and mean arterial pressures significantly
only after complete autonomic blockade (P<0.05). As
expected, atenolol increased and atropine, with or without atenolol,
decreased mean RR intervals (all P<0.05). Enalaprilat did
not alter mean RR intervals and arterial pressures
significantly. Atropine and atropine with atenolol increased average
systolic pressure in the supine position and average
diastolic and mean pressures in both the supine and tilted
positions (P<0.05).
View this table:
[in a new window]
Table 1. Average RR Intervals and Arterial
Pressures During Each Trial of the 3 Experimental Sessions shows average RR-interval
spectral power for three conditions for all subjects. The very sharp
peaks at the imposed respiratory frequency (0.25 Hz) after saline and
enalaprilat support the contention that our subjects controlled their
breathing very well. The middle panel suggests that enalaprilat
increased very-low-frequency RR-interval spectral power modestly, and
the right panel indicates that atropine nearly abolished RR-interval
spectral power at all frequencies. We discuss these changes below.
View larger version (17K):
[in a new window]
Figure 1. Average RR-interval spectral power for all
subjects. The small increase of very-low-frequency spectral power after
enalaprilat (middle panel, extreme left) was statistically
significant.
Figure 2, left, shows raw and
filtered (low and high frequency cutoffs: 0.003 and 0.03 Hz) RR
interval and systolic pressure recordings from one
supine subject. The right panels of this figure show very-low-frequency
spectral power calculated from the time series shown on the left. This
and most other subjects had substantial RR-interval spectral power in
the very-low-frequency range (averages for all subjects were supine:
23±3% and 40 degree tilt: 37±2% of total power). These measures
tended to be consistent both within and across sessions
(r2=0.67, P<0.08;
r2=0.72, P<0.05). In
contrast, although very-low-frequency spectral power accounted for a
large portion of total systolic and diastolic
pressure power (averages for all subjects were supine: 51±2% and
47±2%, tilt: 47±2% and 39±2% of total power), very-low-frequency
arterial pressure powers were consistent only
within sessions (systolic
r2=0.82, diastolic
r2=0.61; P<0.05) and not
across sessions (r2=0.44 and 0.23,
P>0.20). Table 2 lists
average integrated spectral power at very low frequencies before and
after administration of blocking drugs for all subjects.
View larger version (29K):
[in a new window]
Figure 2. Twenty minutes of data from a
representative supine subject. Beat-by-beat and 0.003
to 0.03 Hz filtered RR intervals and arterial pressure are
shown on the left. Spectral power of the beat-by-beat data in the
very-low-frequency range are shown on the right.
View this table:
[in a new window]
Table 2. Very-Low-Frequency (0.003 to 0.03 Hz) Variability in
RR Interval and Arterial Pressures During Each Trial of the
3 Experimental Sessions shows average (±SE) coherence
between systolic pressure and RR intervals over the
very-low-frequency range for all subjects for one trial. Average
coherence was <0.58 over almost all of the very-low-frequency range.
The low reproducibility of arterial pressure spectral
power, discussed above, may reflect the lack of coherence between RR
intervals and systolic pressure at very low frequencies.
Coherence in the very-low-frequency range was quite variable both
within and among subjects. Six subjects had significant coherence
between RR intervals and systolic pressure in at least 1 of the
6 control trials (supine and tilt); no control trial demonstrated
coherence >0.58 in more than 2 subjects. Thus we found no
consistent relation between RR intervals and systolic
arterial pressure in the very-low-frequency range.
View larger version (47K):
[in a new window]
Figure 3. Mean and standard error of the coherence between
systolic pressure and RR interval across the very-low-frequency
range. When the coherence exceeded 0.5 within a frequency range, the
two signals were considered to covary significantly at that
frequency. shows changes of
very-low-frequency RR-interval spectral power before and after blocking
drugs for all subjects. Although atenolol (top panel) appeared to
increase very-low-frequency RR-interval spectral power, the range of
changes was large, and the spectral powers in the supine and tilted
positions were not statistically different from those measured after
saline administration (P=0.16). [The apparent increase of
RR-interval spectral power after atenolol (top panel, left) reflected
an inordinately large increase in 1 subject. Without this subject (whom
we had no other reason to exclude), the median increase of
very-low-frequency RR-interval spectral power after enalaprilat was
66% (P=0.30).] Enalaprilat (bottom panel) exerted only a
modest effect on very-low-frequency RR-interval spectral power [the
increase averaged 21% in the supine position], which was,
nonetheless, statistically significant (P<0.05). Increases
of very-low-frequency RR-interval spectral power after enalaprilat in
the tilted position were not statistically significant. The most
striking changes were exerted by atropine (top two panels, right).
Atropine with or without atenolol nearly eliminated very-low-frequency
RR-interval spectral power in both supine and tilted positions
(decreases averaged 92% to 99% of control).
View larger version (23K):
[in a new window]
Figure 4. Change from control for very-low-frequency
RR-interval power (median, 25th and 75th percentiles) for all blockade
conditions. Double blockade and atropine decreased power; enalaprilat
increased power in the supine position only
(P<0.05). shows and Table 2
lists average
changes of very-low-frequency systolic pressure spectral power
before and after blocking drugs for all subjects. Neither atenolol nor
enalaprilat significantly altered very-low-frequency systolic
pressure spectral power. In contrast, atropine decreased
very-low-frequency systolic pressure power (all
P<0.05 except atropine in the tilt position,
P=0.11 saline versus atropine). Very-low-frequency
diastolic pressure power (not shown) was not affected
significantly by any blockade.
View larger version (26K):
[in a new window]
Figure 5. Change from control for very-low-frequency
systolic pressure power (median, 25th and 75th percentiles) for
each blockade condition in supine and tilted positions. Atropine and
double blockade decreased power (P<0.05, except tilt
after atropine P=0.11).
Tables 3 and 4 list mean low and respiratory frequency
RR-interval and arterial pressure spectral power for all
trials. Atenolol increased RR-interval spectral power only at the
respiratory frequency (P<0.05). Enalaprilat had no effect
on either low or respiratory frequency RR-interval spectral power. As
Figure 1 indicates, atropine with or without atenolol nearly abolished
RR-interval spectral power in all frequency bands. Upright tilt
significantly reduced low-frequency RR-interval spectral power in only
1 of the 3 saline trials (P<0.05). Upright tilt after
atropine with or without atenolol led to a small but significant
(P<0.05) increase of low-frequency RR-interval spectral
power. Upright tilt decreased respiratory frequency RR-interval
spectral power except after parasympathetic or complete autonomic
blockade (P<0.05).
View this table:
[in a new window]
Table 3. Low-Frequency (0.05 to 0.15 Hz) Variability in RR
Interval and Arterial Pressures During Each Trial of the 3
Experimental Sessions
View this table:
[in a new window]
Table 4. Respiratory Frequency (0.2 to 0.3 Hz) Variability in
RR Interval and Arterial Pressures During Each Trial of the
3 Experimental Sessions
Discussion
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
We studied 10 healthy young adult volunteers in the supine and 40
degree upright tilted positions, before and after autonomic or ACE
blockade, to explore mechanisms responsible for very-low-frequency
cardiovascular rhythms. Our measurements were not
confounded by changes of respiration, posture, or physical activity,
which markedly alter RR-interval and arterial pressure
variability, and our recording periods were long enough to
permit us to draw meaningful inferences regarding very-low-frequency
rhythms. Our study supports 3 primary conclusions. First, although
healthy humans have substantial RR-interval and arterial
pressure spectral power in the 0.003 to 0.03 Hz range, there appears to
be no consistent linkage between the two rhythms. Thus
very-low-frequency RR-interval rhythms cannot be explained simply in
terms of baroreflex mechanisms. Second, as others before
us,9 we document a contribution from the
renin-angiotensin-aldosterone system to
very-low-frequency cardiovascular rhythms; however, our
results indicate that this contribution is small and that it involves
RR intervals but not arterial pressures. Third and most
important, our study shows that contributions from parasympathetic
activity dominate very low (as well as higher) frequency RR-interval
rhythms.
Two primary (not mutually exclusive) mechanisms have been proposed
to explain very-low-frequency RR-interval variability. The mechanism
proposed first was that very-low-frequency
cardiovascular rhythms reflect thermoregulation. In our
view, published evidence supports this possibility, but only
indirectly. It is clear that cutaneous blood flow oscillates
slowly23 and that cutaneous blood
flow4 and RR-interval5
oscillations can be entrained by externally applied
oscillatory temperature changes. However, a direct link between
cutaneous thermoregulatory rhythms and cardiovascular
rhythms has not been established; to our knowledge, no one has actually
documented changes of body core temperature at very low frequencies and
shown with coherence analysis that cutaneous blood flow,
arterial pressures, and RR intervals fluctuate together.
Thus the linked hypotheses that thermoregulatory skin blood flow
rhythms translate into arterial pressure rhythms and that
arterial pressure rhythms translate into
baroreflex-mediated RR-interval fluctuations7
have not been validated. We did not measure body core temperature in
our study; however, we did measure arterial pressure and RR
intervals and failed to find the baroreflex linkage that is a critical
element of the thermoregulatory hypothesis.
The mechanism proposed second was that very-low-frequency
RR-interval rhythms reflect influences of the
renin-angiotensin-aldosterone system. As
mentioned, Akselrod and coworkers8 reported that
ACE blockade increases very-low-frequency RR-interval spectral power.
However, data from other studies, also conducted in conscious dogs,
challenge Akselrod and colleagues' conclusions (which were based on
results obtained during 5-minute recordings from only 3 dogs).
Brown et al24 and Rimoldi et
al25 reported no change or an actual reduction of
RR-interval (and systolic pressure) spectral power after ACE
blockade. Our study confirms the findings of Akselrod et al in the
sense that we found that ACE blockade increases very-low-frequency
RR-interval spectral power (but modestly, not dramatically). 15 to 30
minutes. ) and no significant coherence (>0.58; see Appendix) between RR
intervals and arterial pressure at very low frequencies
(Figure 3). Alternatively, angiotensin blockade may enhance
cardiac vagal outflow34 35 and thus increase
cardiac vagal oscillations at very low frequencies (see
discussion below).
Parasympathetic blockade exerted the most dramatic effect of all
the pharmacological interventions we used. Atropine nearly abolished
very-low-frequency RR-interval power (Figure 1 and Table 2). Atropine
also nearly abolished low and respiratory frequency power, as described
previously.33 34 35 Our study suggests that the
parasympathetic nervous system is prepotent in the generation of all
the RR-interval oscillations we studied, including those
occurring at very low, low, and respiratory frequencies. We advance a
simple, economical explanation for our findings-that efferent vagus
nerve traffic to the human heart fluctuates over very low to
respiratory frequencies and that large-dose atropine blocks sinoatrial
node responses to those fluctuations. Atropine does not alter RR
intervals when vagus nerve traffic is absent.36 A
corollary of this is that the adverse prognostic significance of low
levels of very-low-frequency RR-interval fluctuations in postinfarction
patients1 2 is tied to reductions of efferent
vagal-cardiac nerve traffic.
We believed that it was important for our subjects to control
their breathing so they would avoid the huge (10-fold) variations of
RR-interval spectral power that result from variations of breathing
frequency.12 Although our subjects maintained
constant respiratory rates and tidal volumes longer (20 minutes) than
those of any other study to date, our study would have been
strengthened if our subjects had maintained constant breathing for even
longer periods. However, our confidence in our power estimates
increases as the frequency of interest increases. For example, a
20-minute breathing period includes 3.6 cycles of 0.003 Hz but 12
cycles of 0.01 Hz. (More than half of the RR variability within the
very-low-frequency range was >0.01 Hz in 70% of the trials.) We used
20-minute periods because pilot studies showed that even dedicated
volunteers have difficulty maintaining constant breathing continuously
for >20 minutes and because our protocol required 6 trials for each
experimental session. Because our measurement periods were only 20
minutes, we can say nothing about ultralow-frequency rhythms. This may
not pose a problem, however, because a study of Bigger et
al1 showed that very-low-frequency RR-interval
spectral power has major prognostic significance in postinfarction
patients, even when it is calculated over epochs lasting <20
minutes. ). Although our subjects
controlled such input variables as respiration and physical
activity, we cannot exclude the possibility that arterial
pressure and RR intervals are related nonlinearly at very low
frequencies. Moreover, studies in cats36 and
dogs37 document baroreflex influences at very low
frequencies. 120 mm Hg. Last, it should be noted that
physiological effects of clinical doses of
enalaprilat and atenolol may not be absolutely analogous to those of
complete blocking doses of atropine.
Appendix 1
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
The coherence of time series x(t) and y(t)
measures the variance of y linearly predictable from
x at each frequency f.40
Coherence serves as the frequency domain analog of
r2, the coefficient of
determination. Like r2, its value
lies between 0 and 1, with values near 1 indicating a strong linear
relation between the two series. Coherence is defined as the ratio of
the squared covariance of the 2 series to the product of
their individual variances. If Pxx and
Pyy denote the autospectra of x
and y, and Pxy denotes their
cross-spectrum, then coherence
2(f) is given at each frequency
f by
(1) 
, can be used to determine the minimal coherence
(2min) to reject the
null hypothesis with an F test by
The degrees of freedom derive from the relations between
sample size, segment or window length, and window shape. Sampling a
data series of duration T seconds at an interval of
(2) 
t
seconds, or a sampling rate of fs=1/t
Hz, produces N=T/t samples. The averaged periodogram
method of power spectrum estimation (eg, Welch17 )
divides these N samples into segments of length L
(a duration of Lt), whereas the correlogram method (eg,
Blackman and Tukey41 ) applies a single window of
length L. The optimal length depends on the desired balance
between low variance (obtained by small L) and high
resolution (obtained by large L) but must contain at least
one full cycle of the lowest frequency oscillation of
interest; for example, at least 300 seconds (1/0.0033 Hz) is required
for the very-low-frequency band, 0.0033 to 0.03 Hz. The window used to
smooth the spectral estimate determines a multiplicative constant that
increases the degrees of freedom.40 Nonetheless,
the degrees of freedom are roughly proportional to N/L. For
a sufficiently large N, the estimated coherence can be
approximated by a 
2
distribution.17 40 Thus equation 2 merely applies
an F test with 2 and d-2 degrees of
freedom.40 =0.05, we have
F2,7(0.05)=4.74. Using equation 2, we can reject
the null hypothesis that coherence equals 0 when our estimate exceeds
0.58 at the =0.05 level. At the =0.10 and =0.01 levels,
minimal coherence values are 0.48 and 0.73.
Acknowledgments
This research was supported by the National Institute on Aging
(AG05636 to 01), the National Heart, Lung, and Blood Institute
(HL-22296), the Department of Veterans Affairs, and the National
Aeronautics and Space Administration (NAS9 to 16046 and NAG9 to 412).
We thank John M. Karemaker, PhD, and Tom A. Kuusela, PhD, for their
review of the manuscript.
References
Top
Abstract
Introduction
Methods
Results
Discussion
Appendix 1
References
This article has been cited by other articles:
 |
|
 |
|
  V. Vaccarino, R. Lampert, J. D. Bremner, F. Lee, S. Su, C. Maisano, N. V. Murrah, L. Jones, F. Jawed, N. Afzal, et al. Depressive Symptoms and Heart Rate Variability: Evidence for a Shared Genetic Substrate in a Study of Twins Psychosom Med, July 1, 2008; 70(6): 628 - 636. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. K. Lahiri, P. J. Kannankeril, and J. J. Goldberger Assessment of autonomic function in cardiovascular disease: physiological basis and prognostic implications. J. Am. Coll. Cardiol., May 6, 2008; 51(18): 1725 - 1733. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. E. Claydon and A. V. Krassioukov Clinical correlates of frequency analyses of cardiovascular control after spinal cord injury Am J Physiol Heart Circ Physiol, February 1, 2008; 294(2): H668 - H678. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Kanbar, V. Orea, B. Chapuis, C. Barres, and C. Julien A transfer function method for the continuous assessment of baroreflex control of renal sympathetic nerve activity in rats Am J Physiol Regulatory Integrative Comp Physiol, November 1, 2007; 293(5): R1938 - R1946. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Stewart, I. Taneja, and M. S. Medow Reduced central blood volume and cardiac output and increased vascular resistance during static handgrip exercise in postural tachycardia syndrome Am J Physiol Heart Circ Physiol, September 1, 2007; 293(3): H1908 - H1917. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. D. Pinna Assessing baroreflex sensitivity by the transfer function method: what are we really measuring? J Appl Physiol, April 1, 2007; 102(4): 1310 - 1311. [Full Text] [PDF]
|
|
|
|
 |
|
 R. Hanss, B. Bein, P. Turowski, E. Cavus, M. Bauer, M. Andretzke, M. Steinfath, J. Scholz, and P. H. Tonner The influence of xenon on regulation of the autonomic nervous system in patients at high risk of perioperative cardiac complications Br. J. Anaesth., April 1, 2006; 96(4): 427 - 436. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Julien The enigma of Mayer waves: Facts and models Cardiovasc Res, April 1, 2006; 70(1): 12 - 21. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Martinmaki, H. Rusko, L. Kooistra, J. Kettunen, and S. Saalasti Intraindividual validation of heart rate variability indexes to measure vagal effects on hearts Am J Physiol Heart Circ Physiol, February 1, 2006; 290(2): H640 - H647. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Latka, M. Turalska, M. Glaubic-Latka, W. Kolodziej, D. Latka, and B. J. West Phase dynamics in cerebral autoregulation Am J Physiol Heart Circ Physiol, November 1, 2005; 289(5): H2272 - H2279. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. L Eckberg and T. A Kuusela Human vagal baroreflex sensitivity fluctuates widely and rhythmically at very low frequencies J. Physiol., September 15, 2005; 567(3): 1011 - 1019. [Abstract] [Full Text] [PDF]
|
|