Am J Physiol Heart Circ Physiol 296: H171–H179, 2009.
First published November 7, 20008; doi:10.1152/ajpheart.00963.2008.
Multiresolution wavelet analysis of time-dependent physiological responses
in syncopal youths
Jennifer A. Nowak, Anthony Ocon, Indu Taneja, Marvin S. Medow, and Julian M. Stewart
Pediatrics, Physiology, and Medicine, The Center for Hypotension, New York Medical College, Hawthorne, New York
Submitted 3 September 2008; accepted in final form 31 October 2008
baroreflex; syncope; variance
SYNCOPE (FAINTING) IS DEFINED by a sudden transient loss of
consciousness and postural tone due to cerebral hypoperfusion
(8). Simple postural faint is identified with upright vasovagal
syncope in which vasodilation-induced hypotension and bradycardia are relieved by recumbency (18). Upright orthostatic
stress testing and specifically upright tilt table testing have
been used to provoke postural fainting episodes. We recently
reported on the time course of physiological changes during
upright tilt using a system of fiducial points. Fiducial points
allowed for the comparison of relevant physiological events
across patients with simple postural fainting. The fiducial
points were baseline, 1-min after head-up tilt (HUT), the onset
of gradual blood pressure (BP) decline, mid decline, late
decline just prior to the actual faint, and the faint itself, which
was the final fiducial point comprising the rapid fall in BP and
heart rate (HR) (Fig. 1). These events were consistently present
in all young syncopal patients.
Address for reprint requests and other correspondence: J. M. Stewart,
Pediatrics, Physiology, and Medicine, The Center for Hypotension, Suite 1600,
19 Bradhurst Ave., New York Medical College, Hawthorne, NY 10532
(e-mail: [email protected]).
http://www.ajpheart.org
During early tilt we demonstrated a time-dependent increase
in HR coinciding with an initially stable arterial pressure (AP)
and followed by a slow decrease in AP. The slow decline
corresponded to a decrease in cardiac output (CO) and in
central hypovolemia caused by failure of splanchnic vasoconstriction and enhanced splanchnic blood pooling (34, 36).
However, total peripheral resistance (TPR) was initially increased compared with control. Since AP ⫽ CO●TPR, the
increase in TPR incompletely compensated for the fall in CO,
resulting in a slow fall in AP during the mid stages of upright
tilt. We infer that increased TPR and HR primarily result from
baroreceptor unloading with increased baroreflex-mediated
sympathetic activity and decreased reflex baroreflex-mediated
cardiovagal activity (vagal withdrawal). During the actual faint
these reflex responses are reversed: bradycardia occurs and is
attributed to increased vagotonia (37) originating in the preganglionic parasympathetic nerve centers, whereas peripheral vasodilation occurs and is attributed to withdrawal of vascular
sympathetic nerve activity from preganglionic baroreflex neurons in the rostral ventral lateral medulla (24).
Our aim in the present study was to investigate the alterations of baroreflex activity that occur in synchrony with the
beat-to-beat time-dependent changes in HR, BP, and TPR. For
this purpose we wished to employ techniques of HR and BP
variability. Often such investigations use Fourier analyses,
which are suitable for examining stationary, time-invariant
signals. Rhythms in BP and in the RR interval that pertain to
sympathetic and parasympathetic control are evident in such
analyses. Specifically, low-frequency (LF) oscillations (⬍.15
Hz) in BP, first described by Mayer (22), are highly correlated
with sympathetic nerve activity in humans (2, 11, 13) and
likely reflect sympathetic baroreflex closed-loop negative feedback (11). High-frequency (HF) oscillations (⬎0.20 Hz), most
evident in HR or RR interval time series, are synchronized with
the respiratory sinus arrhythmia and represent an index of
parasympathetic (vagal) activity (6, 7, 31). Although Fourier
techniques are completely adequate for stationary time-invariant signals, they are inherently unsuited for the large timedependent changes observed during fainting. Time-dependent
representations of RR interval and AP variation are therefore
required, and discrete wavelet transform (DWT) techniques
satisfy this requirement.
Discrete wavelets incorporate the idea of the frequency
content of a signal by partitioning it into a finite number of
different scales that can be related to a particular frequency
(28). Thus, instead of the infinite time scale of a sinusoid
signal, there are confined time and frequency responses of the
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.
0363-6135/09 $8.00 Copyright © 2009 the American Physiological Society
H171
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
Nowak JA, Ocon A, Taneja I, Medow MS, Stewart JM.
Multiresolution wavelet analysis of time-dependent physiological
responses in syncopal youths. Am J Physiol Heart Circ Physiol
296: H171–H179, 2009. First published November 7, 20008;
doi:10.1152/ajpheart.00963.2008.—Our prior studies indicated that
postural fainting relates to thoracic hypovolemia. A supranormal
increase in initial vascular resistance was sustained by increased
peripheral resistance until late during head-up tilt (HUT), whereas
splanchnic resistance, cardiac output, and blood pressure (BP) decreased throughout HUT. Our aim in the present study was to
investigate the alterations of baroreflex activity that occur in synchrony with the beat-to-beat time-dependent changes in heart rate
(HR), BP, and total peripheral resistance (TPR). We proposed that
changes of low-frequency Mayer waves reflect sympathetic baroreflex. We used DWT multiresolution analyses to measure their time
dependence. We studied 22 patients, 13 to 21 yr old, 14 who fainted
within 10 min of upright tilt (fainters) and 8 healthy control subjects.
Multiresolution analysis was obtained of continuous BP, HR, and
respirations as a function of time during 70° upright tilt at different
scales corresponding to frequency bands. Wavelet power was concentrated in scales corresponding to 0.125 and 0.25 Hz. A major difference from control subjects was observed in fainters at the 0.125 Hz
AP scale, which progressively decreased from early HUT. The alpha
index at 0.125 Hz was increased in fainters. RR interval 0.25 Hz
power decreased in fainters and controls but was markedly increased
in fainters with syncope and thereafter corresponding to increased
vagal tone compared with control subjects at those times only. The
data imply a rapid reduction in time-dependent sympathetic baroreflex
activity in fainters but not control subjects during HUT.
H172
WAVELETS FOR FAINTERS
METHODS
To test these hypotheses, data were extracted from a well-characterized group of fainters. The study was approved by the Institutional
Review Board of the New York Medical College. Informed consent
was obtained from all subjects.
Subjects
Fainting subjects comprised 14 patients ages 13–21 (median age 16
yr, 6 male, 8 female) who were referred for three or more episodes of
postural fainting and had a vasovagal response within 10 min of HUT
at 70°. Control subjects comprised 8 healthy volunteers (median age 17
yr, 4 male, 4 female) who had no history of orthostatic intolerance and did
not have a vasovagal response during 10 min of HUT. HUT time was
limited to 10 min because prolonged upright tilt can produce fainting in
healthy teenagers. All subjects were free of systemic illness, were nonsmokers, and were not taking any medication. There were no trained
athletes or bedridden subjects. All subjects had a normal electrocardiogram (ECG), echocardiogram, and physical examination.
Protocol
Fig. 1. Arterial pressure (AP; top), RR interval (RR; middle), and total
peripheral resistance (TPR; bottom) of a representative fainting patient during
head-up tilt (HUT) to 70°. The fiducial points “baseline,” “1 min” after HUT,
“early” onset of gradual BP decline, “mid” decline, “late” decline just prior to
the actual faint, and “faint” are indicated. A time-dependent increase in heart
rate (decrease in RR) is coupled with an initial increase in TPR, allowing for
an initially stable AP. These events are followed by a decline in TPR and AP
until faint.
analyzed signal, constrained only by the uncertainty principal.
By use of DWT, a discrete sampled biological signal is optimally and completely decomposed into a small group of
time-dependent signals. Wavelet coefficients uniquely and
completely represent that digital signal yet contain specific
scales with particular physiological significance that can be
easily examined. Particular contributions of LF and HF power
to the energy contained within any signal can therefore be
conveniently measured as a function of time.
We measured time-dependent sympathetic activity by studying the AP wavelet coefficients at a scale corresponding to the
LF Mayer wave. We examined the time-dependent RR interval
wavelet coefficients at a scale corresponding to the respiratory
arrhythmia. We used the ratio of LF RR interval power to LF
AJP-Heart Circ Physiol • VOL
Testing began at 10 AM. Subjects fasted for 4 h prior to testing and
refrained from xanthine and caffeine for at least 72 h before testing.
No subject smoked. Subjects were familiarized with procedures and
then instrumented for ECG, respiratory plethysmography, impedance
plethysmography, continuous BP recording, and capnography. For
purposes of the present study only data concerning HR (or equivalently RR interval), AP, peripheral resistance, and respirations are
reported. After a 30-min supine acclimatization period we assessed
HR, BP, and respirations for a baseline period of 5 min.
Subjects were then tilted upright to 70° for a maximum of 10 min.
Continuous HR, BP, and respirations were recorded. Subjects were
tilted back to supine if fainting ensued. Postural vasovagal fainting
was defined as a tilt-induced decrease in mean arterial pressure (MAP)
to ⬍60 mmHg or a decrease in systolic BP (SBP) ⬍80 mmHg
associated with symptoms of impending loss of consciousness, severe
lightheadedness, nausea, or diaphoresis. All of the fainters developed
a classic vasovagal faint with bradycardia as well as hypotension
while upright. A representative faint is shown in Fig. 1.
Fiducial event markers. In accordance with previous work (36), we
chose to examine and display results at fiducial time markers rather
than at specific time points. Fiducial markers related to times at which
physiological events occurred. Although the actual time between
physiological events may vary, fiducial time markers allow for the
study of events with corresponding physiology specific for each
fainter. For the present study, eight defining events were determined
for fainters from the AP trace of each subject. The first fiducial point
296 • JANUARY 2009 •
www.ajpheart.org
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
AP power [the alpha index (27, 30)] to describe changes in the
cardiovagal baroreflex sensitivity.
A priori, we hypothesized that fainters would have a more
marked increase in wavelet LF power compared with control
subjects with the onset of upright tilt corresponding to increased peripheral resistance. We proposed that there would be
a steady decrease in LF activity during late stages of tilt,
culminating in complete absence of LF activity preceding the
faint itself. These findings, if present, would reflect an initial
increase in sympathetic activity followed by subsequent sympathetic withdrawal (23). We further hypothesized that fainters
would have excessive vagal withdrawal initially, signaled by
marked reduction in HF RR interval wavelet power after tilt.
Thereafter, we expected the HF wavelet power to be reduced
compared with controls, since fainters are likely to have
increased vagal withdrawal due to increased baroreflex unloading, that is, until fainting supervenes and vagotonia ensues.
H173
WAVELETS FOR FAINTERS
AJP-Heart Circ Physiol • VOL
transformation (MODWT).” The MODWT fills all time points at each
scale, allows precise alignment of the signal and its wavelets, allows
for any sample size (not just a power of 2), has zero phase-shifted
details (representing each scale’s contributions to the original signal
as a function of time; a detail is the inverse DWT of the Wavelet
coefficients at given scale), and zero phase-shifted smooths (representing averages over all scales larger than the scales of interest as a
function of time). Wavelet coefficients can be used to perform an
analysis of variance. Averages over wavelet coefficients at any scale
are zero. The square of the wavelet coefficient at each scale in the time
sequence corresponds to the power or variance contained at that scale
as a function of time. By Parseval’s theorem, the sum of all power
over all scales equals the total variance as a function of time. By this
approach, the first (finest) wavelet scale corresponds to 1 Hz, the
second wavelet scale corresponds to 0.5 Hz, the third scale corresponds to 0.25 Hz, the fourth to 0.125 Hz, the fifth to 0.0625 Hz, the
sixth to 0.03125 Hz, and larger scales (lower frequencies) are combined
into a smooth. The first and second scales correspond to the HR and are
simply removed from the analysis. The fifth and sixth scales corresponding to 0.0625 and 0.03125 Hz contain considerable power only at end
points and little power during the tilt. Multiresolution analysis for the
systolic AP from a representative healthy subject is shown in Fig. 2. The
original signal, the pertinent details, corresponding power, and smooth
are shown. We focused attention on the third and fourth scales corresponding to 0.25 and 0.125, which contain the wavelet power classically
related to the respiratory rhythm (0.25 Hz) and to sympathetic baroreflex
activity (⬃0.1 Hz). Multiresolution analyses were performed for SBP,
diastolic BP, MAP, RR interval, and respiration.
Data Analysis and Statistics
Data were digitized and stored in a personal computer and were
analyzed offline with custom software. Wavelet power at the LF band
(4th dyadic scale ⬃0.125 Hz) and HF band (3rd dyadic scale ⬃0.25
Hz) for RR interval, BP, diastolic BP, MAP, and respiration were
determined at each fiducial point and averaged over subjects. Data are
expressed as the logarithm base 10 of averages over 15-s intervals
centered at these time markers to accommodate the wide dispersion of
values. The alpha index, a measure of cardiovagal baroreflex sensitivity, was computed as the square root of the ratio between LF RR
interval power and LF systolic AP power.
There were two groups of subjects, fainters and control subjects,
whose data are represented over time. We used repeated-measures
ANOVA to compare group and time-dependent differences at fiducial
points. This defined “time effects” or significant differences within a
group and “group effects” or significant differences between the
groups. If significant, a Bonferroni test was used as a post hoc test for
multiple comparisons. To compare data between fainters and control
subjects, results were calculated by using SPSS (Statistical Package
for the Social Sciences) software version 14.0 and graphed by using
GraphPad Prism (San Diego, CA) software version 4. Graphic results
are reported as means ⫾ SE.
RESULTS
Results are depicted in Table 1 and Figs. 3–7. Power
(variance) is expressed in arbitrary units (au).
Table 1 shows age and resting vital signs in fainters and control
subjects. There are no significant differences in age; gender
distribution; systolic, diastolic, and mean BPs; or respiratory rate.
Arterial Pressure
Figure 3 shows a multiresolution analysis of systolic AP
from a typical fainter in contrast to a similar analysis shown for
a control subject in Fig. 2. The LF (0.125 Hz) and HF (0.25 Hz)
power is contained within a box. There is a decrease in LF and
296 • JANUARY 2009 •
www.ajpheart.org
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
was baseline pretilt. The second fiducial point was chosen to be 1 min
after HUT. Immediately after a tilt-up younger subjects are particularly likely to develop a transient and sometimes large decrease in BP
and a reciprocal increase in HR (33). BP and HR recover quickly and
within 1 min in healthy volunteers, thus defining the second fiducial
point. The third point was defined in fainters by a gradual decline in
BP relative to the second point. Subsequently, BP falls off abruptly
which defined the fifth point, just prior to faint. Objective criteria for
fiducial points 3 at the early BP fall and 5 at the late BP fall were
determined by using scale 6 smooth of the multiresolution analysis.
Lower frequency oscillations are thereby removed from the computation.
The extrema of the second derivative define “corner” or “edge” points
that correspond visually to fiducial points 3 and 5. The fourth point was
defined as a point midway between points 3 and 5. The sixth point
occurred at the end of tilt (at faint for fainters, at 10 min for control
subjects). We included seventh and eighth fiducial points representing
immediate and later recovery of HR and BP from upright tilt. Early
recovery started 5 s after tilt-down. Late recovery occurred when BP
had returned to baseline. Thus the eight fiducial event markers are 1)
“baseline” (before tilt), 2) “1 min” (1 min after tilt), 3) “early decline”
(early fall in BP), 4) “mid decline”, 5) “late decline,” 6) “faint,” 7)
“early recovery,” and 8) “late recovery” and are shown for a representative subject in Fig. 1. Healthy control subjects do not have a fall
in BP and do not faint. Therefore, we defined equivalent fiducial time
points for healthy control subjects by averaging the actual time of
fiducial markers 3, 4, and 5 in the fainters. Point 6 was defined by end
of HUT. By this means, control subjects and fainting subjects were
compared at equivalent average times to allow for a uniform evaluation of cardiorespiratory parameters across groups.
HR and BP monitoring. HR was monitored by a single-lead
electrocardiogram. Upper extremity arterial BP was continuously
monitored with a Finometer placed on the right index or middle finger
which measured finger AP using height correction (Finometer, FMS,
Amsterdam). Finometer data were calibrated to a brachial artery
oscillographic pressure.
Respiratory volumes. Relative respiratory volumes were obtained
by using a respiratory inductance plethysmograph (Respitrace, NIMS
Scientific) normalizing the respiratory volume on startup to an internal
scale. After normalization we calibrated normalized supine Respitrace
volumes against a pneumotachograph (model RSS100-HR, Hans
Rudolph, Kansas City, MO), which also yielded tidal volumes. This
enabled the use of the Respitrace as a comparative measure of relative
changes in tidal volume throughout testing.
Upright tilt-table testing. An electrically driven tilt table (Collin
Medical, San Antonio, TX) with footboard was used. After supine
measurements the table was tilted up in 6 s to 70° and remained at that
angle of tilt for 10 min or until syncope ensued. Syncope was operationally defined as a decrease in SBP to ⬍60 mmHg or a decrease in SBP to
⬍80 mmHg associated with symptoms of impending loss of consciousness, severe light-headedness, nausea, or diaphoresis.
Signal processing and wavelet analysis. HR and RR interval were
derived from ECG. ECG, Respitrace, and Finometer pressure data
were interfaced to a personal computer through an analog-to-digital
converter (DI-720 DataQ Ind, Milwaukee, WI).
BP, RR interval, and respiratory signals were low-pass filtered
below 1.0 Hz. Filtered data were resampled at 2 Hz. For the present
analysis we used Mallat’s dyadic pyramid algorithm for computing
the fast DWT (20, 21) to obtain a multiresolution analysis (also known
as the multiscale approximation).
We used the Dabauchies 4 mother (generating) wavelet function
(4). The mother wavelet is used to generate an orthonormal basis,
where each term represents a doubling of scale (dyadic dilation). This
provides a unique decomposition of each biological signal into subbands of wavelet coefficients, each representing the contribution of
the scaled and translated wavelet function at the given scale. We
employed an extended version of the DWT as described by Percival
and Walden (28) to produce a “maximal overlap discrete wavelet
H174
WAVELETS FOR FAINTERS
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
Fig. 2. Representative multiresolution analysis of systolic blood pressure (BP; SBP) of a healthy control subject during HUT is shown. Left panels from top
down: original mean arterial pressure (MAP) signal, a “smooth” version of this signal representing larger scales corresponding to frequencies below 0.03125 Hz,
and signal details derived from wavelet scale coefficients corresponding to 0.03125, 0.0625, 0.125, and 0.25 Hz as a function of time. Right panels from top down:
original signal, total power summed over all scales corresponding to frequencies of 0.03125 Hz or higher, and individual power (wavelet variance) as a function
of time at scales corresponding to 0.03125, 0.0625, 0.125, and 0.25 Hz. Low frequency (LF; 0.125 Hz) and high frequency (HF; 0.0625 Hz) contribute the large
majority of total wavelet power. Tilt-up and Tilt-down are demarcated by solid lines. LF power (0.125 Hz) and HF power (0.25 Hz) are shown within a box.
HF power from the middle of upright tilt until fainting occurs.
Such a decrease is not observed in control subjects.
Figure 4, top compares the systolic and diastolic pressures
for fainters with control subjects. The figure shows a small
(P ⬍ 0.05) increase in systolic pressure during upright tilt.
There was an abrupt decrease in systolic and diastolic pressure
at the time of the faint.
Averaged LF (0.125 Hz) variance (power). Figure 4, bottom
shows systolic, mean, and diastolic wavelet power for fainters
AJP-Heart Circ Physiol • VOL
and control subjects. Systolic, diastolic, and MAP LF variances
(Mayer waves) were similarly increased during HUT compared
with baseline (P ⬍ 0.01) for both fainters and control subjects.
Baseline variance of MAP and diastolic pressure were significantly decreased for fainters compared with control subjects.
Although systolic, diastolic, and MAP variances were not
different between fainters and control subjects at 1 min postHUT, a group effect reduction in systolic, diastolic, and MAP
variances was present in fainters compared with control sub-
296 • JANUARY 2009 •
www.ajpheart.org
H175
WAVELETS FOR FAINTERS
Table 1. Demographic and resting hemodynamics in fainters
and healthy control subjects
Parameters
Fainters
Controls
N
Age, yr
Sex, %female
Resting SBP, mmHg
Resting DBP, mmHg
Resting MAP, mmHg
Resting HR, beats/min
Resting RR, min⫺1
14
16.2⫾1.0
61%
123⫾3
65⫾2
85⫾2
66⫾2
18⫾1
8
17.9⫾1.1
58%
122⫾6
67⫾3
85⫾3
65⫾4
19⫾1
Alpha Coefficient
Data are depicted in Fig. 7.
The alpha coefficient decreased significantly in both fainters
and control subjects as a function of time compared with
baseline (P ⬍ 0.05). During HUT the alpha coefficient was
significantly higher in fainters than controls (P ⬍ 0.001).
Although this difference was greatest during recovery, it was
present throughout HUT.
DISCUSSION
jects and significant differences were present at the early BP
fall. This was most evident in diastolic pressure and MAP but
was also present in systolic pressure and corresponds closely to
the time course of decrease in TPR, which we have previously
reported (36). Although somewhat obscured by large interpatient variability, an intrapatient decrease in AP variance was
always present in fainters and was easily seen. This contrasts
with the rather indolent early (prefaint) decreases in BP.
Averaged HF (0.25 Hz) variance (power). Systolic, diastolic, and MAP HF variances were increased during HUT
compared with baseline (P ⬍ 0.01) for both fainters and
control subjects. Significant differences between subject
groups could not be measured.
RR Interval
Figure 5 shows that the RR interval decreased significantly
compared with baseline with HUT (top left) for both fainters
and control subjects. It was further decreased in fainters compared with control subjects before fainting occurred. With
fainting, RR interval was increased in fainters compared with
control subjects, which persisted through during recovery.
The LF RR interval power (top right) in fainters was not
different from control subjects throughout the tilt. The wide
variation in power reflects interpersonal differences in size and
shape of the decrease in slope, the nadir, and the subsequent
increase in slope and recovery. Intrapersonal differences were
less variable and with far less intraindividual error.
The HF RR interval (bottom right) was not decreased in
fainters compared with control subjects before fainting occurred. It was substantially increased compared with control
subjects at the time of faint and during recovery.
The ratio of LF to HF power (bottom left) tended to be
increased in fainters during early tilt and was significantly
decreased at faint and during early recovery in these patients in
association with the large increase in HF power.
Our data show that RR interval, arterial BP, and respirations
can be readily analyzed at dyadic scales by using the DWT to
produce time-dependent multiresolution analyses of these biological signals (Fig. 2 and 3). These produce analytic products
capable of distinguishing physiological changes in young patients with simple vasovagal faint. The typical and often
reported findings in fainters of increased HR (decreased RR
interval) and initially increased SBP compared with control
subjects are present, as is the gradual decrease in SBP and TPR
(Fig. 1 and Ref. 36) until fainting ensues and HR and BP fall
precipitously.
RR Interval Power
The decrease in LF RR power shown in Fig. 5 was commensurate in fainters and control subjects. LF RR power is
often attributed to sympathetic activity. Sympathetic activity is
known to increase during early tilt (3, 15). However, investigators have demonstrated that the transduction of BP variation
(Mayer waves) into HR variation is primarily vagal and, thus,
parasympathetic (7, 10). Figure 4 shows that LF BP power is
reduced in fainters compared with control subjects. Figure 7
demonstrates that cardiovagal gain, represented by the alpha
index, shows a similar decrease from baseline and thereafter
for fainters and control subjects during HUT. However, the
alpha index is larger in fainters at all times. Increased gain/
sensitivity implies increased transfer gain. In fainters gain is
increase such that significantly reduced AP power is transduced to an unreduced RR interval power.
HF RR interval power increases to a significantly greater
degree in fainters compared with control subjects only
around the time of faint, implying an increase in vagal tone
that has its onset at, but not prior to, the time of faint. Furlan
and colleagues (9) noted a similar increase in HF RR
interval in a study of young healthy volunteers who experienced fainting for the first time during a 90° head-up tilt.
Differences in overall HR variability were confined to faint
and the postfaint period.
Respiration
Data are depicted in Fig. 6.
Respiratory variance (power) was assessed at the scale
corresponding to 0.25 Hz, which approximates the respiratory
frequency. Respiratory frequency is not disturbed by HUT
(36). There was no difference in baseline respiratory variance
between fainters and control subjects.
AJP-Heart Circ Physiol • VOL
Arterial Pressure Power
Although controversial, we propose that LF AP variance,
representing the well-known Mayer wave phenomena (13, 22),
is an adequate and sufficient time-dependent index of sympathetic baroreflex sensitivity/gain. It increases in both control
subjects and fainters with upright tilt (5) corresponding to the
296 • JANUARY 2009 •
www.ajpheart.org
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
Age and resting vital signs in fainters and control subjects. There are no
significant differences in age; gender distribution; systolic (SBP), diastolic
(DBP), and mean (MAP) blood pressures; or respiratory rate (RR). HR, heart
rate.
Overall respiratory variance was significantly increased
from baseline by HUT in both fainters and control subjects.
Respiratory variance was significantly greater than controls
(P ⬍ 0.01) leading up to the time of fainting.
H176
WAVELETS FOR FAINTERS
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
Fig. 3. Representative multiresolution analysis of systolic AP of a fainter during HUT is shown. Left panels from top down: the original MAP signal, a smooth
version of this signal representing larger scales corresponding to frequencies below 0.03125 Hz, and signal details derived from wavelet scale coefficients
corresponding to 0.03125, 0.0625, 0.125, and 0.25 Hz as a function of time. Right panels from top down: the original signal, total power summed over all scales
corresponding to frequencies of 0.03125 Hz or higher, and individual power (wavelet variance) as a function of time at scales corresponding to 0.03125, 0.0625,
0.125, and 0.25 Hz. LF (0.125 Hz) and HF (0.25 Hz) contribute the large majority of total wavelet power. Tilt-up and Tilt-down are demarcated by solid lines.
LF power (0.125 Hz) and HF power (0.25 Hz) are shown within a box.
increase in sympathetic nerve activity (9) and is (appropriately)
directionally opposite to changes in cardiovagal gain as observed in muscle sympathetic nerve activity (MSNA)-tilt experiments (26). The progressive decrement observed in vasovagal fainters from the first minute of tilt resembles the falloff
in LF oscillations of sympathetic nerve activity (14). LF
AJP-Heart Circ Physiol • VOL
diastolic pressure is the most markedly affected, which is
comparable to the relationship between diastolic BP and
MSNA during peroneal microneurography (35). LF AP power
in fainters is always less than or equal to control, consistent
with observations of sympathetic baroreflex dysfunction in
patients with vasovagal syncope (1). The time course of fall in
296 • JANUARY 2009 •
www.ajpheart.org
H177
WAVELETS FOR FAINTERS
LF wavelet AP power precedes and predicts a fall in TPR at a
later fiducial point as is shown in prior work (36). We interpret
these findings to signify that a consistent and progressive
central process produced by a progressively reduced sympathetic baroreflex is detected by changes in Mayer wave variance and accounts for the time course of faint.
Respiratory Power
Results in the Young
Of significance is the age group of our subjects. In the
Prevention of Syncope Trial (POST), Sheldon and colleagues
(32) reported a mode age of 13 yr and a median age of 18 yr
as the most frequent ages of onset of vasovagal syncope. Since
simple faint most commonly begins during adolescence, it may
be important to study children’s autonomic responses to tilt
since they are physiologically different from adults. A variety
of time-frequency methods have been employed in adults and
have yielded different results. Thus Novak and coworkers (18,
25) used Wigner-distribution based time-frequency analysis
and found no differences between fainters and control subjects
at Mayer wave frequencies. In contrast to our data, those
investigators suggested that very-low-frequency waves (0.01–
0.05 Hz) could represent sympathetic activity and showed that
an increase in the amplitude of LF systolic and diastolic
oscillations during the first 5 min of tilt followed by a decrease
and eventual disappearance of the very-low-frequency (0.01–
0.05 Hz) oscillations in fainters. Age differences or different
analytic methods may account for these discrepancies.
Other researchers have found no decrease in BP variability
at all. For example, Lipsitz et al. (20) found that systolic and
diastolic BP variabilities were similar comparing fainter and
nonfainters. They found that although BP decreased during
HUT there was no significant change in high or LF BP during
the 3 min before syncope.
Limitations
Fig. 4. Top: systolic and diastolic blood pressure during HUT in fainters
and controls. Bottom from top down: systolic, mean, and diastolic pressure
low-frequency (low freq.) power. ■, Fainters; 䊐, healthy controls. Systolic
pressure initially increases above control and then gradually decreases until
syncope when it abruptly falls. Diastolic pressure changes are less marked.
With the exception of 1 min posttilt, LF systolic, mean, and diastolic power
are significantly lower in fainters than in controls (P ⬍ 0.05). The
difference in power widens as HUT progresses. Bsl, baseline; ERec, early
recovery; LRec, late recovery. Pgroup, probability of group differences;
Ptime, probability of time differences.
AJP-Heart Circ Physiol • VOL
A coherence-transfer function analysis is not a natural part
of the discrete wavelet approach. Cross correlation-covariance
analyses can provide somewhat similar information but have
not been well developed for discrete wavelets applied to
nonstationary signals. Continuous wavelets can be used to
produce a gain-phase approach but cannot supply a finite
orthonormal basis for signal approximation.
The use of fiducial time markers allowed us to compare
subjects at similar physiological time points but removed the
absolute time dependence of the observed phenomena.
We did not directly measure sympathetic or parasympathetic
activity nor did we seek to manipulate them. However, the
preponderance of evidence suggests that Mayer waves take
their origin in sympathetic baroreflex feedback loops whereas
HF HR variability and the alpha index can serve as adequate
surrogate indexes of cardiac parasympathetic activity.
296 • JANUARY 2009 •
www.ajpheart.org
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
It is now well known that hyperpnea and hypocapnia precede simple faint in a majority of cases (16). Our data and
those of Bernardi’s group (29) demonstrate increased respiratory variance in fainters compared with control subjects, which
may adversely affect cerebral blood flow and potentiate syncope.
H178
WAVELETS FOR FAINTERS
We studied the subjects during 10 min of HUT, which was
considered an appropriate time on the basis of prior studies
(12). Increasing the time of HUT may give different information; however, healthy control subjects may faint during prolonged standing.
Our study focused on syncope in the young. The peak
incidence of syncope occurs during adolescence. Therefore our
findings may not completely translate to adults. For example, a
gradual decrease in BP followed by a rapid increase is characteristic of young patients whereas older subjects may have
only an abrupt fall in BP, a gradual fall in BP, or both.
The study describes the pattern of changes in oscillations of
BP and RR interval that are in some ways similar to those
reported by Kamiya et al. (15), who also documented accompanying MSNA changes. Dr. Kamiya focused on MSNA but also
RR interval and AP comparing data from healthy male subjects
who fainted during a 60° upright tilt to healthy male subjects who
did not faint. Such healthy fainters would be regarded as “false
positives” and could not be assumed to be equivalent to the
episodic fainters that we studied. Currently, the American Heart
Fig. 6. Changes in respiratory power (variance) during HUT in fainters (closed
boxes) and control subjects (open boxes). Power started from a similar baseline
and increased with HUT for fainters and control subjects. Power was greater
in fainters throughout HUT (P ⬍ 0.01).
AJP-Heart Circ Physiol • VOL
Association uses clinical criteria to determine the diagnosis of
reflex neurocardiogenic syncope (simple faint). One reason for
this is the plethora of “false positive” results especially in the
young. Had Dr. Kamiya studied female false positives as well,
he might have noted a diversity of responses including vasodepressor responses and sinus tachycardia that would have
fallen outside the range of an investigation of postural syncope.
In addition he studied a somewhat different age group using the
technique of complex frequency demodulation, which differs
from multiresolution wavelet analysis used here. The demodulation technique employs classic Fourier-based filtering techniques to low-pass filter a frequency-shifted signal. Although it
is similar in some ways to wavelets, which can be interpreted
as band-pass filters with interesting scaling properties, there
remain differences. Nevertheless, the general trends in Dr.
Kamiya’s results and our own are similar. This may imply that
false positive fainters are similar to episodic fainters.
Fig. 7. Alpha index computed as the square root of the ratio between LF RR
interval power and LF systolic AP. Alpha index was used as an index of
cardiovagal baroreflex sensitivity. ■, Fainters; 䊐, healthy controls. The index
decreased significantly with HUT for fainters and control subjects. The index
was larger in fainters compared with control subjects during HUT, at the faint
itself and during recovery.
296 • JANUARY 2009 •
www.ajpheart.org
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
Fig. 5. RR interval (RRI) as a function of time
is shown at top left, LF wavelet power as a
function of time at top right, HF wavelet power
(variance) as a function of time at bottom right,
and LF-to-HF ratio at bottom left. ■, Fainters;
䊐, healthy controls. RR interval was smaller
(heart rate was higher) during tilt in fainters
until fainting supervened, after which RR interval was larger in fainters than in control subjects. LF power changed similarly with time for
fainters and controls alike. HF power decreased
significantly and similarly with time in fainters
and controls until fainting supervened, after
which HF power was larger in fainters than in
control subjects. LF-to-HF ratio tended to be
increased in early HUT in fainters and was
significantly decreased at the time of faint and
during early recovery. Ptime*group, probability of
group and time differences.
H179
WAVELETS FOR FAINTERS
To conclude, fainting in young patients is associated with an
initial rise and then a steady falloff of lower frequency AP
wavelet power (Mayer wave power) as a function of time
during upright tilt. This occurred in every fainter and the initial
decrease in power was absent in control subjects and predictive
of fainting. There are reasons to believe that LF power reflects
sympathetic baroreflex activity. The diminished baroreflex activity is best related to diastolic BP. Differences in cardiovagal
baroreflex activity are limited to the acute faint and recovery.
Disturbed respirations occur. However, sympathetic baroreflex
activity is typically enhanced by hypocapnia, creating a question of causality.
ACKNOWLEDGMENTS
GRANTS
This research was supported by 0735603T from the American Heart
Association and RO1 HL074873, RO1HL087803, and 1R21HL091948 from
the National Heart, Lung, and Blood Institute of the National Institutes of
Health.
REFERENCES
1. Bechir M, Binggeli C, Corti R, Chenevard R, Spieker L, Ruschitzka F,
Luscher TF, Noll G. Dysfunctional baroreflex regulation of sympathetic
nerve activity in patients with vasovagal syncope. Circulation 107: 1620 –
1625, 2003.
2. Chapuis B, Vidal-Petiot E, Orea V, Barres C, Julien C. Linear modelling analysis of baroreflex control of arterial pressure variability in rats.
J Physiol 559: 639 – 649, 2004.
3. Cooke WH, Hoag JB, Crossman AA, Kuusela TA, Tahvanainen KU,
Eckberg DL. Human responses to upright tilt: a window on central
autonomic integration. J Physiol 517: 617– 628, 1999.
4. Daubechies I. Orthonormal bases of compactly supported wavelets.
Comm Pure Appl Math 41: 909 –996, 1988.
5. Ducla-Soares JL, Santos-Bento M, Laranjo S, Andrade A, DuclaSoares E, Boto JP, Silva-Carvalho L, Rocha I. Wavelet analysis of
autonomic outflow of normal subjects on head-up tilt, cold pressor test,
Valsalva manoeuvre and deep breathing. Exp Physiol 92: 677– 686, 2007.
6. Eckberg DL. Human sinus arrhythmia as an index of vagal cardiac
outflow. J Appl Physiol 54: 961–966, 1983.
7. Eckberg DL. Physiological basis for human autonomic rhythms. Ann Med
32: 341–349, 2000.
8. Folino AF, Russo G, Porta A, Buja G, Cerutti S, Iliceto S. Modulations
of autonomic activity leading to tilt-mediated syncope. Int J Cardiol 120:
102–107, 2007.
9. Furlan R, Piazza S, Dell’Orto S, Barbic F, Bianchi A, Mainardi L,
Cerutti S, Pagani M, Malliani A. Cardiac autonomic patterns preceding
occasional vasovagal reactions in healthy humans. Circulation 98: 1756 –
1761, 1998.
10. Furlan R, Porta A, Costa F, Tank J, Baker L, Schiavi R, Robertson D,
Malliani A, Mosqueda-Garcia R. Oscillatory patterns in sympathetic
neural discharge and cardiovascular variables during orthostatic stimulus.
Circulation 101: 886 – 892, 2000.
11. Grasso R, Schena F, Gulli G, Cevese A. Does low-frequency variability
of heart period reflect a specific parasympathetic mechanism? J Auton
Nerv Syst 63: 30 –38, 1997.
12. Hammer PE, Saul JP. Resonance in a mathematical model of baroreflex
control: arterial blood pressure waves accompanying postural stress. Am J
Physiol Regul Integr Comp Physiol 288: R1637–R1648, 2005.
13. Jardine DL, Ikram H, Frampton CM, Frethey R, Bennett SI, Crozier
IG. Autonomic control of vasovagal syncope. Am J Physiol Heart Circ
Physiol 274: H2110 –H2115, 1998.
AJP-Heart Circ Physiol • VOL
296 • JANUARY 2009 •
www.ajpheart.org
Downloaded from http://ajpheart.physiology.org/ by 10.220.33.1 on June 18, 2017
We thank members of the Department of Pediatrics, especially its
Chairman, Dr. Leonard Newman, and the Division of Pediatric Cardiology,
especially its Director, Michael H. Gewitz M.D., for support. We also wish to
acknowledge our intellectual debt to our mentors, Dr. Thomas H. Hintze and
Dr. Gabor Kaley, for constant inspiration and stimulation.
14. Julien C. The enigma of Mayer waves: facts and models. Cardiovasc Res
70: 12–21, 2006.
15. Kamiya A, Hayano J, Kawada T, Michikami D, Yamamoto K,
Ariumi H, Shimizu S, Uemura K, Miyamoto T, Aiba T, Sunagawa
K, Sugimachi M. Low-frequency oscillation of sympathetic nerve activity decreases during development of tilt-induced syncope preceding sympathetic withdrawal and bradycardia. Am J Physiol Heart Circ Physiol
289: H1758 –H1769, 2005.
16. Kamiya A, Kawada T, Yamamoto K, Michikami D, Ariumi H,
Uemura K, Zheng C, Shimizu S, Aiba T, Miyamoto T, Sugimachi
M, Sunagawa K. Resetting of the arterial baroreflex increases orthostatic
sympathetic activation and prevents postural hypotension in rabbits.
J Physiol 566: 237–246, 2005.
17. Lagi A, Cencetti S, Corsoni V, Georgiadis D, Bacalli S. Cerebral
vasoconstriction in vasovagal syncope: any link with symptoms? A transcranial Doppler study. Circulation 104: 2694 –2698, 2001.
18. Lepicovska V, Novak P, Nadeau R. Time-frequency dynamics in neurally mediated syncope. Clin Auton Res 2: 317–326, 1992.
19. Lewis T. A lecture on vasovagal syncope and the carotid sinus mechanism: with comments on Gower’s and Nothnagel’s syndrome. Br Med J 1:
873– 876, 1932.
20. Lipsitz LA, Hayano J, Sakata S, Okada A, Morin RJ. Complex
demodulation of cardiorespiratory dynamics preceding vasovagal syncope. Circulation 98: 977–983, 1998.
21. Mallat SG. A theory for multiresolution signal decomposition: the wavelet representation. IEEE Trans Pattern Anal Mach Intell 11: 674 – 693,
1989.
22. Mallat SG. Multiresolution approximations and wavelet orthonormal
bases of L2(R). Trans Am Math Soc 315: 69 – 87, 1989.
23. Mayer S. Studien zur physiologie des herzens und der blutgefässe. Sitz
Kaiser Akad Wiss 75: 281–307, 1876.
24. Morillo CA, Eckberg DL, Ellenbogen KA, Beightol LA, Hoag JB,
Tahvanainen KU, Kuusela TA, Diedrich AM. Vagal and sympathetic
mechanisms in patients with orthostatic vasovagal syncope. Circulation
96: 2509 –2513, 1997.
25. Novak V, Novak P, Kus T, Nadeau R. Slow cardiovascular rhythms in
tilt and syncope. J Clin Neurophysiol 12: 64 –71, 1995.
26. O’Leary DD, Kimmerly DS, Cechetto AD, Shoemaker JK. Differential
effect of head-up tilt on cardiovagal and sympathetic baroreflex sensitivity
in humans. Exp Physiol 88: 769 –774, 2003.
27. Pagani M, Somers V, Furlan R, Dell’Orto S, Conway J, Baselli G,
Cerutti S, Sleight P, Malliani A. Changes in autonomic regulation
induced by physical training in mild hypertension. Hypertension 12:
600 – 610, 1988.
28. Percival DB, Walden AT. Wavelet Methods for Time Series Analysis.
Cambridge, UK: Cambridge University Press, 2000.
29. Porta C, Casucci G, Castoldi S, Rinaldi A, Bernardi L. Influence of
respiratory instability during neurocardiogenic presyncope on cerebrovascular and cardiovascular dynamics. Heart 94: 1433–1439, 2008.
30. Robbe HW, Mulder LJ, Ruddel H, Langewitz WA, Veldman JB,
Mulder G. Assessment of baroreceptor reflex sensitivity by means of
spectral analysis. Hypertension 10: 538 –543, 1987.
31. Saul JP, Berger RD, Albrecht P, Stein SP, Chen MH, Cohen RJ.
Transfer function analysis of the circulation: unique insights into cardiovascular regulation. Am J Physiol Heart Circ Physiol 261: H1231–H1245,
1991.
32. Sheldon R, Connolly S, Rose S, Klingenheben T, Krahn A, Morillo C,
Talajic M, Ku T, Fouad-Tarazi F, Ritchie D, Koshman ML. Prevention
of Syncope Trial (POST): a randomized, placebo-controlled study of
metoprolol in the prevention of vasovagal syncope. Circulation 113:
1164 –1170, 2006.
33. Stewart JM. Transient orthostatic hypotension is common in adolescents.
J Pediatr 140: 418 – 424, 2002.
34. Stewart JM, McLeod KJ, Sanyal S, Herzberg G, Montgomery LD.
Relation of postural vasovagal syncope to splanchnic hypervolemia in
adolescents. Circulation 110: 2575–2581, 2004.
35. Sundlof G, Wallin BG. Human muscle nerve sympathetic activity at rest.
Relationship to blood pressure and age. J Physiol 274: 621– 637, 1978.
36. Taneja I, Medow MS, Glover JL, Raghunath NK, Stewart JM.
Increased vasoconstriction predisposes to hyperpnea and postural faint.
Am J Physiol Heart Circ Physiol 295: H372–H381, 2008.
37. Van Lieshout JJ, Wieling W, Karemaker JM, Eckberg DL. The
vasovagal response. Clin Sci (Lond) 81: 575–586, 1991.