BASIC SCIENCES

Original Investigations

Spectral Analysis of Heart Rate Variability

during Exercise in Trained Subjects
AURÉLIEN P. PICHON1, CLAIRE DE BISSCHOP1, MANUEL ROULAUD1, ANDRÉ DENJEAN1,2, and
YVES PAPELIER3
1
Laboratory of Exercise Induced Physiological Adaptations, School of Sports Science, Poitiers, FRANCE; 2Respiratory
and Exercise Physiology Department, Poitiers University Hospital, Poitiers, FRANCE; and 3Physiology and Sleep
Department, Kremlin-Bicêtre Medical School, Antoine Béclère Hospital, Clamart, FRANCE

ABSTRACT
PICHON, A. P., C. DE BISSCHOP, M. ROULAUD, A. DENJEAN, and Y. PAPELIER. Spectral Analysis of Heart Rate Variability during
Exercise in Trained Subjects. Med. Sci. Sports Exerc., Vol. 36, No. 10, pp. 1702–1708, 2004. Purpose: To investigate the effects of strenuous
exercise on heart rate variability (HRV). Methods: We evaluated the effects of exercise intensity and duration on HRV indices in 14 healthy
trained subjects. Each subject exercised for 3, 6, and 9 min at 60 and 70% of the power achieved at maximal oxygen consumption (PV̇O2max)
and for 3 and 6 min (or 3 min twice) at 80% of PV̇O2max. The electrocardiogram RR intervals were recorded then processed by fast (FFT)
and short-time (STFT) Fourier transform for determination of low-frequency (LF, 0.045– 0.15 Hz) and high-frequency (HF, 0.15–1.0 Hz)
components. Results: The LF and HF components expressed as absolute power (ms2) decreased significantly at the onset of exercise
(P ⬍ 0.05). However, with increasing exercise intensity, the HF component expressed as normalized units (n.u.) (reflecting parasympathetic
modulation) increased significantly, whereas the LF component (n.u.) and LF/HF ratio (both reflecting sympathetic modulation) decreased
significantly (P ⬍ 0.05). STFT showed that increasing exercise intensity was associated with a shift in HF peak frequency related to an
increase in respiratory rate and a marked decrease in LF power (ms2). Moreover, HFn.u. rose (r ⫽ 0.918, P ⬍ 0.01) and LFms2 fell as minute
ventilation increased (r ⫽ 0.906, P ⬍ 0.01). Conclusions: Parasympathetic respiratory control and nonautonomic mechanisms may influence
the HF-peak shift during strenuous exercise. HRV and the usual indexes of sympathetic activity do not accurately reflect changes in autonomic
modulation during exhaustive exercise. Key Words: AUTONOMIC NERVOUS SYSTEM, DYNAMIC EXERCISE, FOURIER TRANS-
FORM, BREATHING FREQUENCY

H
eart rate variability (HRV) spectral analysis (1,19)
allows noninvasive assessment of autonomic mod-
ulation of the sinoatrial node (SA) through quanti-
fication of the low- (LF) and high-frequency (HF) oscilla-
tions of RR intervals. Amara and Wolf (3) showed that HRV
spectral analysis was a reliable and reproducible technique
for evaluating cardiac autonomic balance during exercise.
However, HRV spectral analysis has yielded conflicting
results (5,7,8,19). Factors that may explain these discrepan-
cies include differences in the methods used for spectral
analysis (fast Fourier transform, coarse-grained analysis, or
autoregressive methods) and in the units used to express the
Address for correspondence: Aurélien Pichon, Laboratory of exercise in-
duced physiological adaptations, School of Sports Sciences, 4 allée Jean
Monnet, 86000 Poitiers, France; E-mail: aurelien.pichon@orange.fr.
Submitted for publication March 2004.
Accepted for publication June 2004.
0195-9131/04/3610-1702
MEDICINE & SCIENCE IN SPORTS & EXERCISE®
Copyright © 2004 by the American College of Sports Medicine
DOI: 10.1249/01.MSS.0000142403.93205.35
1702
results (4,21,29). Differences in subjects, relative exercise
intensity, and exercise duration may be involved also.
Recently, studies have shown that time-frequency distri-
butions are useful for assessing HRV at rest and under
various physiological conditions (13,17,26). With this ap-
proach, the heart signal during transition phases can be
investigated via calculation of the spectral components as a
function of time. Short-time Fourier transform (STFT) anal-
ysis also provides an evaluation of dynamic and temporal
modifications in the LF and HF components during non-
strictly stationary signals (17).
The aim of our study was to test the effects of exercise intensity
and duration on HRV using the conventional and time-frequency
approaches in a homogeneous group of trained athletes. We also
assessed the effects of ventilation on HRV indices.
METHODS
Subjects
Fourteen healthy male athletes participated in the study
(age, 19 –24 yr). All were nonsmokers and were free of
known cardiac, respiratory, or other disease. None was on
medication. The study subjects refrained from drinking cof-
fee, tea, or alcoholic beverages for at least 6 h before the
experiments. Our institutional review board approved the
study, and all subjects gave their written informed consent.

Study Design
The study required four visits to the laboratory on differ-
ent days. At the first visit, each subject completed a medical
history questionnaire and underwent a physical examination
and a 12-lead electrocardiogram (EKG). At the second visit,
we determined the maximal oxygen uptake (V̇O2max) and FIGURE 1—Constant-load exercise procedure. Five-min warm-up
(WU) and constant-load exercise in % of the power at maximal oxygen
the power reached at V̇O2max (PV̇O2max) in each subject. uptake (PV̇O2max). The mark (. . ..) shows RR interval measurement
During the two remaining visits, the subjects performed and (____) shows ventilatory parameter measurement. CD, 45-min
constant-load exercise tests (60, 70, and 80% of PV̇O2max); cool-down.
the 60% and 70% tests were each done for 3 min, 6 min, and
9 min; the 80% test was done either for 3 min and 6 min or sample rate of 1000 Hz from the thoracic EKG signal. Heart
twice for 3 min, according to the subject’s exercise capacity. rate, V̇O2, FR, and V̇E were measured and averaged over 1
RR intervals, oxygen consumption (V̇O2), and minute ven- min, starting 1.5 min before the end of exercise.
tilation (V̇E) were measured during exercise and constituted
the dependent variables. Power Spectral Analysis

Maximal Exercise Test Fast Fourier transform. Recorded RR periods were

first transferred to ASCII files. The RR interval sequences
Each subject performed a standardized maximal exercise obtained at rest or during exercise were examined visually,
test on an electrically braked stationary cycle ergometer and artifactual data were manually replaced by interpolated
(ERGOLINE 800S, Germany). After a 5-min warm-up at 90 or extrapolated data. With the RR recorder device used in
W, the test began at 100 W, and the workload was increased
at a rate of 25 W·min⫺1. The test ended when the subject
was no longer able to cycle, despite maximal voluntary
effort and encouragement. The subject breathed through a
face mask (Hans Rudolph, Kansas City, MO). Expired gas
flows were measured using a pneumotachograph (Type 3,
Hans Rudolph) and analyzed breath-by-breath using an au-
tomated system (CPX, Medical Graphics, St Paul, MN).
V̇O2, V̇E, breathing frequency (FR), and other standard re-
spiratory parameters were monitored continuously and av-
eraged every 15 s. A four-lead EKG recorded heart rate
continuously.

Constant Submaximal Exercises

Three submaximal exercise tests were carried out on two
different days (session A and B) at the same time of the day
(see Fig. 1). During session A, the subjects performed two
sequences of constant-load exercise: 3 min, 6 min, and 9
min at 60% of PV̇O2max followed by 3 min and 6 min (or 3
min twice) at 80% of PV̇O2max. The interval between the
sequences was 45 min. During session B, the subjects per-
formed a constant-load exercise at 70% PV̇O2max for 3 min,
6 min, and 9 min. Each sequence was preceded by a 5-min
warm-up. During each sequence, the exercise periods were
separated by 2-min rest intervals. The subjects breathed
freely. Sessions A and B were done in random order. FIGURE 2—Example of spectral analysis of RR interval variability
during a 6-min exercise at 70% PV̇O2max in one of the study subjects.
Recorded Variables Panel A shows the tachogram of raw RR intervals recorded during
exercise. Panel B shows the RR interval variation after linear com-
During constant-load exercise, continuous beat-to-beat pensation and Blackman windowing. Panel C shows RR-interval spec-
tral power with FFT analysis and the low-frequency (LF) and high-
recordings of the RR periods were obtained using a POLAR frequency (HF) components expressed in absolute units (ms2) and
Heart Rate Monitor (POLAR Vantage NV, Finland), with a normalized units (n.u.). The LF/HF ratio is also given.

HEART RATE VARIABILITY DURING EXERCISE Medicine & Science in Sports & Exercise姞 1703

FIGURE 3—Spectral analysis of RR-interval variability, at rest and
during the 60%, 70%, and 80% PV̇O2max exercise periods, in 14
subjects. Panel A shows the high-frequency (HF) and low-frequency
(LF) spectral powers in absolute units (ms2), panel B shows the HF
component in normalized units (n.u.), and panel C shows the LF/HF
ratio; **P < 0.01.
our study, the proportion of abnormal beats obtained is in
principle less than 2%. The recordings from one subject
were excluded because of a large number of premature beats
displayed with the RR signal during exercise. Then, suitable
RR periods at rest and at steady-state exercise were selected
for analysis. Series of 256, 512, or 1024 consecutive RR
intervals (for the 3-, 6-, and 9-min exercise periods, respec-
tively) were extracted just before the end of each exercise
period, resampled for FFT analysis, and processed with
1704 Official Journal of the American College of Sports Medicine
linear compensation and a Blackman window (17). The RR
intervals were then fast Fourier transformed with the Carl de
Boor algorithm to produce a power spectrum (Program on
MATLAB 5.3, The MathWorks, Inc., 1999). The LF
(0.045– 0.15 Hz) and HF (0.151–1.00 Hz) oscillatory com-
ponents (Fig. 2) are reported both as absolute values (ms2)
and as normalized units (n.u.) calculated as follows: HFn.u.
⫽ (HFms2)/((LFms2 ⫹ HFms2) ⫻ 100) (10,20).
Short-time Fourier transform. STFT was performed
for rest and exercise data from all the subjects. STFT of RR
intervals corresponds to a sliding fast Fourier transform
analysis. STFT processing yields a time-frequency domain
analysis that can be illustrated by a three-dimensional figure
showing HRV changes throughout the exercise period (see
Fig. 3). The signal is convolved with a constant-duration
time window, and the spectral components are calculated for
each windowed segment (17). In our study, STFT analyses
were performed using the MATLAB program (MATLAB
5.3, The MathWorks, Inc., 1999) on data processed in the
same way as for FFT analysis. After loading the ASCII file,
a RR periodogram was constructed and displayed, and the
stretch running from the time when the signal became stable
to the end of exercise was selected for STFT. This stretch
had to contain more than 320 values so that STFT could be
done on a block of 256 values. The size of the block (p) and
the optimal number of spectra (Ns, i.e., the number of
repeated FFT analyses) were determined as follows: when
the sample contained fewer than 1012 values, p was 256 and
Ns as 64, whereas when the sample contained fewer than
1022 values, p was 512 and Ns was either 128 or 256.
Before analysis, the samples were filtered using a Hamming
window (17).
Autoregressive analysis. We performed an additional
autoregressive (AR) analysis in the six subjects who exer-
cised for 6 min at 60%, 70%, and 80% of PV̇O2max (512
intervals). Data were processed as for FFT analysis, and AR
was then done using HRV Analysis Software 1.1 for Win-
dows (The Biomedical Signal Analysis Group, Department
of Applied Physics, University of Kuopio, Finland). Autore-
gressive coefficients were estimated using the forward-
backward-linear-least-squares-algorithm with a 16th order
AR model. The RR-interval time-series were interpolated at
a rate of 2 Hz and detrended before analysis.
TABLE 1. Anthropometric and physiological characteristics of subjects (N ⫽ 14).
Parameter Value
Age (yr) 22 ⫾ 2
Height (cm) 178 ⫾ 7
Weight (kg) 71 ⫾ 8
V̇O2max (mL䡠min⫺1䡠kg⫺1) 53 ⫾ 5
PV̇O2max (W) 325 ⫾ 39
HRmax (bpm) 190 ⫾ 7
RQmax 1.3 ⫾ 0.1
V̇Emax (L䡠min⫺1) 145 ⫾ 30
Training status (h䡠wk⫺1) 7⫾2
Values are mean ⫾ SD. V̇O2max, maximal oxygen uptake; PV̇O2max, power at maximal
oxygen uptake; HRmax, maximal heart rate; RQmax, maximal respiratory quotient; V̇Emax,
maximal minute ventilation.
http://www.acsm-msse.org
TABLE 2. Heart rate, oxygen uptake, ventilation, and respiratory rate corresponding to the different periods of exercise.
V̇O2 V̇E FR
Intensity Duration HR (bpm) (mL䡠minⴚ1䡠kgⴚ1) (L䡠minⴚ1) (breaths䡠minⴚ1)
60% 3 min 143 ⫾ 7 35 ⫾ 3 56 ⫾ 8 22 ⫾ 5
PV̇O2max 6 min 161 ⫾ 6 39 ⫾ 4 64 ⫾ 10 25 ⫾ 5
9 min 167 ⫾ 6 40 ⫾ 4 66 ⫾ 9 26 ⫾ 6
70% 3 min 155 ⫾ 6 39 ⫾ 4 67 ⫾ 10 25 ⫾ 4
PV̇O2max 6 min 174 ⫾ 7 44 ⫾ 3 81 ⫾ 15 29 ⫾ 6
9 min 182 ⫾ 6 45 ⫾ 3 95 ⫾ 21 35 ⫾ 7
80% 3 min (1) 166 ⫾ 8 44 ⫾ 4 84 ⫾ 14 31 ⫾ 5
PV̇O2max 3 min (2) 174 ⫾ 5 49 ⫾ 3 89 ⫾ 15 33 ⫾ 6
6 min (2) 184 ⫾ 5 49 ⫾ 2 119 ⫾ 12 42 ⫾ 9
Values are mean ⫾ SD. HR, heart rate; V̇O2, oxygen uptake; V̇E, ventilation, FR, breathing frequency.

Statistical Analysis Effects of intensity and duration of exercise. The

Data are expressed as means ⫾ SD. Statistical analysis
was performed using the STATISTICA 5.1 package (Stat-
soft, Inc., Tulsa, OK). The effects of exercise intensity and
duration on frequency components were evaluated using
analysis of variance for repeated measures (ANOVA-
MANOVA) with, if necessary, Greenhouse and Geisser’s
adjustment. Two ANOVA were performed: two-way
ANOVA with two intensities (60% and 70% PV̇O2max) and
three durations (3, 6, and 9 min) in 14 subjects and two-way
ANOVA with three intensities (60%, 70%, and 80%
PV̇O2max) and two durations (3 and 6 min) in six subjects.
The Newman-Keuls test was used for post hoc analysis.
Student’s t-tests for paired observations were used when
appropriate. P values lower than 0.05 were considered sta-
tistically significant.
RESULTS
Physiological characteristics of subjects. Table 1
reports the characteristics of the subjects. Of the 14 subjects,
eight exercised twice for 3 min at 80% PV̇O2max, and six
exercised once for 3 min and once for 6 min. Heart rate,
V̇O2max, and V̇E during the exercise periods are shown in
Table 2.
Effects of exercise. The HF and LF components (ms2)
decreased significantly with exercise [F(6;66) ⫽ 2.75, P ⬍
0.05; and F(6;66) ⫽ 14.6, P ⬍ 0.001; respectively] (Fig. 3).
The LF/HF ratio was also significantly lower during exer-
cise than at rest [F(6;66) ⫽ 25.5, P ⬍ 0.001]. However,
when expressed in normalized units, the HF component was
significantly higher during exercise than at rest [F(6;66) ⫽
27.1, P ⬍ 0.001].
effects of exercise intensity and duration on HRV are re-
ported in Table 3. ANOVA on the data (ms2) from the 14
subjects (60% and 70% PV̇O2max) showed a significant
interaction for the HF component between intensity and
duration of exercise [F(2;24) ⫽ 3.9, P ⬍ 0.05]. At 60%
PV̇O2max, HFms2 was significantly higher during the 3-min
exercise period than during the other exercise periods.
LFms2 decreased significantly with exercise intensity [F(1;
12) ⫽ 17.5, P ⬍ 0.005] and with exercise duration [F(2;24)
⫽ 4.3, P ⬍ 0.05]. The same statistical analysis on data in
normalized units showed a significant increase in HF [F(1;
12) ⫽ 31.7, P ⬍ 0.001] and a significant decrease in the
LF/HF ratio as exercise intensity increased [F(1;12) ⫽ 9.2,
P ⬍ 0.05] (see Fig. 3). Moreover, significant effects of
intensity on LFms2, LFn.u., HFn.u., and the LF/HF ratio
were found also in the six subjects who exercised at 60%,
70%, and 80% PV̇O2max during 3 min and 6 min.
Figure 4 depicts the results of STFT on a stretch of RR
intervals from one of the six subjects who performed 6 min
of constant-load exercise at 60%, 70%, and 80% PV̇O2max.
In this example, LFms2 decreased sharply with increasing
exercise intensity, whereas HFms2 remained unchanged and
peak HFms2 frequency increased. This increase in peak
HFms2 frequency was associated with an increase in the
respiratory rate recorded during the last minute of exercise
(0.35 Hz and 21.2 breaths·min⫺1 at 60% PV̇O2max; 0.45 Hz
and 27 breaths·min⫺1 at 70% PV̇O2max; and 0.64 Hz and
38.2 breaths·min⫺1 at 80% PV̇O2max).
As with FFT and STFT, AR showed that increasing the
intensity of exercise was associated with i) a sharp decline
in the LF component, ii) stability of the HF component (see

TABLE 3. Variance, high-frequency (HF), and low-frequency (LF) components in absolute units (ms2) and normalized units (n.u), and LF/HF ratio calculated from FFT analysis at
rest and during three constant-load exercises (60%, 70% and 80% of power at maximal oxygen uptake) and three durations (3, 6, and 9 min); results from AR analysis are
given in brackets for the 6-min exercise period.
60% PV̇O2max 70% PV̇O2max 80% PV̇O2max
Rest 3 min 6 min 9 min 3 min 6 min 9 min 3 min(1) 3 min(2) 6 min(2)
Variance ms2 1726 ⫾ 1870 41 ⫾ 46 21 ⫾ 29 15 ⫾ 27 22 ⫾ 27 8⫾7 5⫾2 13 ⫾ 12 10 ⫾ 8 9⫾6
HFms2 393 ⫾ 838 1.3 ⫾ 0.9 0.8 (0.6) ⫾ 0.5 0.7 ⫾ 0.4 1.2 ⫾ 0.9 0.8 (0.8) ⫾ 0.6 0.9 ⫾ 0.7 1.0 ⫾ 0.7 1.6 (1.1) ⫾ 2.6 1.3 ⫾ 0.9
HFn.u. 15 ⫾ 11 56 ⫾ 25 59 (29.7) ⫾ 21 65 ⫾ 21 76 ⫾ 21 78 (31) ⫾ 24 85 ⫾ 12 84 ⫾ 13 82 (34) ⫾ 15 95 ⫾ 2
LFms2 1333 ⫾ 1136 1.2 ⫾ 1.2 0.7 (0) ⫾ 0.9 0.4 ⫾ 0.3 0.3 ⫾ 0.2 0.2 (0) ⫾ 0.3 0.1 ⫾ 0.1 0.2 ⫾ 0.2 0.1 (0) ⫾ 0.04 0.1 ⫾ 0.04
LFn.u. 84 ⫾ 11 44 ⫾ 25 41 (0) ⫾ 21 35 ⫾ 21 24 ⫾ 21 22 (0.0) ⫾ 24 15 ⫾ 12 16 ⫾ 13 18 (0.0) ⫾ 15 5⫾2
LF/HF ratio 8.4 ⫾ 5.3 1.4 ⫾ 1.6 1.2 ⫾ 1.7 0.7 ⫾ 0.8 0.6 ⫾ 1.0 0.6 ⫾ 1.2 0.2 ⫾ 0.2 0.2 ⫾ 0.2 0.3 ⫾ 0.3 0.1 ⫾ 0.03
Values are mean ⫾ SD; 3 min(1) and 3 min(2), first and second 3-min exercise periods at 80% PV̇O2max; 6 min(2): 6-min exercise period at 80% PV̇O2max performed after 3 min(1)
[6 subjects].

HEART RATE VARIABILITY DURING EXERCISE Medicine & Science in Sports & Exercise姞 1705

FIGURE 4 —Three-dimensional spectral analysis of RR interval vari-
ability from short-term Fourier transform (STFT) during 6 min of
constant-load exercise at 60% (A), 70% (B), and 80% (C) of the power
at maximal oxygen uptake (PV̇O2max). Example from one subject.
Table 3), and iii) gradual resetting of peak HFms2 to a
higher frequency.
In addition, we found that HFn.u. was positively corre-
lated with V̇E and with the percentage of V̇O2max [r ⫽ 0.92
1706 Official Journal of the American College of Sports Medicine
and r ⫽ 0.97, P ⬍ 0.01] (see Fig. 5A and 5B). In contrast,
LFms2 was negatively correlated with V̇E and with the
percentage of V̇O2max [r ⫽ 0.82 and r ⫽ 0.88, P ⬍ 0.01] (see
Fig. 5C and 5D).
DISCUSSION
In this study, we showed that the HFn.u. component of
HRV, believed to reflect parasympathetic influences mod-
ulating the sinoatrial node, increased significantly with the
exercise load. In addition, the LF/HF ratio, which is the
classic index of sympathetic modulation, and the LF com-
ponent (ms2 and n.u.) decreased significantly with the ex-
ercise load. Also, ventilation had a major influence on HRV
indices measured during exercise. These results were con-
firmed by STFT analysis, which allows for both a time and
a frequency analysis of HRV changes during exercise.
Methodological aspects. The RR interval recordings
were obtained using the POLAR Vantage NV device, which
has been validated previously (18,24). For measuring the
instantaneous heart rate during exercise, the maximum error
with our sampling frequency of 1000 Hz was about 1
beats·min⫺1 [Error ⫽ (2 HR2)/(60SF ⫹ HR)]. However,
despite the high sampling frequency, the signal-to-noise
ratio was reduced by 1000 during exercise because of the
reduction in total variance (Var) of the signal from about
5000 ms2 at rest to less than 5 ms2 during exercise (signal-
to-noise ratio ⫽ 6 Var·SF2 ⫺ 1) (2).
To improve the validity of LF and HF component esti-
mates, we analyzed the longest possible signal stretch dur-
ing each exercise period. Thus, we selected the last 256,
512, and 1024 RR intervals during the 3-, 6-, and 9-min
exercise periods, respectively. The data were obtained using
both FFT and STFT analyses. STFT analysis is an original
technique (17,26) that has rarely been used for exercise
studies (13). It allows assessment of HRV in both the time
and the frequency domain for nonstrictly stationary RR
signals. Using the AR method to eliminate a potential effect
of wide-band noise (white noise) on FFT analysis (15), we
found similar results to those provided by FFT for all
exercise intensities during 6 min. Nevertheless, a potential
limitation of this study may be the repetition of the exercise
periods. Indeed, each exercise period might affect the HRV
values obtained during the subsequent period(s).
Effects of exercise and ventilation. As expected
(4,10,16,21), the variance of the RR signal was significantly
reduced during dynamic exercise as compared with rest.
Studies have shown that this effect is chiefly ascribable to
parasympathetic withdrawal occurring in the first part of
exercise and affecting the entire range of heart rate changes
(1,22,27).
In all our study subjects, we found that exercise was
associated with apparent inversion of the parasympathetic/
sympathetic balance as assessed by the LF/HF ratio. At the
highest exercise intensity (80% PV̇O2max), the HFn.u. com-
ponent contributed more than 80% of the overall variance,
http://www.acsm-msse.org

as compared with only 15% at rest. Gradually increasing
parasympathetic modulation seems an unsatisfactory expla-
nation to this predominance of HFn.u. at a time when heart
rate was increasing in response to strenuous exercise. We
will therefore discuss a number of alternative hypotheses.
First, the predominance of HFn.u. during exercise
(5,10,11,23) may be related to exacerbation of respiratory
sinus arrhythmia occurring as a response to increased oscil-
lations in venous input to the left cardiac pump due to
increasing tidal pressure and volumes (5,6,11,14,17). Ber-
nardi et al. (5,6) suggested that an intrinsic heart mechanism
(atrial stretch), in synchrony with ventilation, may increase
the HF component during strenuous exercise. In our study,
we found that minute ventilation increased with exercise
intensity, from about 63 L·min⫺1 at 60% PV̇O2max to about
88 L·min⫺1 at 80% PV̇O2max (see Table 2). This may in
itself explain the increase in HFn.u. Moreover, the STFT
analysis of the 6-min exercise periods (see Fig. 4) showed
that increasing exercise intensity resulted in an elevation in
peak HF frequency that was concomitant to the increase in
respiratory rate. Keselbrener and Akselrod (17) obtained
similar results from a STFT analysis of the effects of ven-
tilation parameters on HRV. In addition, the significant
positive correlation noted during exercise in our study be-
tween HFn.u. and V̇E supports evidence reported by Ber-
nardi et al. (5,6) and Casadei et al. (10,11) that nonneural
mechanisms influence HRV (see Fig. 5A). Casadei et al.
(11) used ganglion blockade to show that the contribution of
respiratory sinus arrhythmia to the total variance increased
HEART RATE VARIABILITY DURING EXERCISE
FIGURE 5—Linear relationship between
heart rate variability index and physiolog-
ical parameters for all exercise periods in
the 14 subjects. A. Relation between the
mean high-frequency component ex-
pressed in normalized units (HFn.u.) and
minute ventilation (V̇E). B. Relation be-
tween HFn.u. and exercise intensity in %
of the maximal oxygen uptake
(%V̇O2max). C. Relation between the mean
low-frequency component expressed in
absolute units (LFms2) and V̇E. D. Rela-
tion between LFms2 and % V̇O2max.
by 32% during supine exercise performed at 25% V̇O2max.
The sizable increase in respiratory drive may explain the
change in the sympathetic/parasympathetic balance associ-
ated with exercise (11), and nonautonomic mechanisms may
be physiologically useful by regulating the heart rate in the
absence of effective autonomic modulation during strenuous
exercise. Thus, the presence of a nonautonomic component
of respiratory sinus arrhythmia and the increase in signal-
to-noise ratio seen during strenuous exercise may distort
estimates of the index of vagal modulation (i.e., the HF
component).
The increase in HFn.u. in our study appears to contradict
earlier reports that the HF component decreases when ex-
ercise intensity increases (4,8,9,12,28). However, in contrast
to HFn.u., HFms2 decreased at the onset of exercise then
showed no change with increasing exercise intensity. This
result suggests a paradoxical stability of parasympathetic
modulation during strenuous exercise. Moreover, we found
significant negative correlations between LFms2 and both
V̇E and the percentage of V̇O2max (see Fig. 5C and 5D), as
well as a nonsignificant correlation between HFms2 and
both V̇E and the percentage of V̇O2max. Thus, the HFn.u.
increase observed during strenuous exercise may be merely
a consequence of the LFms2 decrease (10,25,29) in combi-
nation with stability of HFms2. These results invite a partial
reappraisal of the ventilatory hypothesis put forward to
explain the HFn.u. increase associated with strenuous dy-
namic exercise.
Medicine & Science in Sports & Exercise姞 1707
CONCLUSION
The results of this study using both FFT and STFT
indicate a major influence of ventilation and of nonauto-
nomic mechanisms on the HF and LF components of HRV.
These influences may challenge the validity of standard
HRV indices for assessing autonomic control of heart rate
during strenuous exercise. Our findings also establish that
REFERENCES
1. AKSELROD, S. D., F. A. GORDON, D. C. UBEL, A. C. SHANNON, A. C.
BERGER, and R. J. COHEN. Power spectrum analysis of heart rate
fluctuation: a quantitative probe of beat-to-beat cardiovascular
control. Science 213:220 –222, 1981.
2. ALTIMIRAS, J. Understanding autonomic sympathovagal balance
from short-term heart rate variations: are we analyzing noise?
Comp. Biochem. Physiol. A Mol. Integr. Physiol. 124:447– 460,
1999.
3. AMARA, C. E., and L. A. WOLFE. Reliability of noninvasive meth-
ods to measure cardiac autonomic function. Can. J. Appl. Physiol.
23:396 – 408, 1998.
4. ARAI, Y., J. P. SAUL, P. ALBRECHT, et al. Modulation of cardiac
autonomic activity during and immediately after exercise. Am. J.
Physiol. 256(1 Pt 2):H132–H141, 1989.
5. BERNARDI, L., F. SALVUCCI, R. SUARDI, et al. Evidence for an
intrinsic mechanism regulating heart rate variability in the trans-
planted and the intact heart during submaximal dynamic exercise?
Cardiovasc. Res. 24:969 –981, 1990.
6. BERNARDI, L., C. PORTA, A. GABUTTI, et al. Modulatory effects of
respiration. Auton. Neurosci. 90:47–56, 2001.
7. BRENNER, I. K., S. THOMAS, and R. J. SHEPARD. Autonomic regu-
lation of the circulation during exercise and heat exposure: infer-
ences from heart rate variability. Sports Med. 26:85–99, 1998.
8. BRENNER, I. K., S. THOMAS, and R. J. SHEPARD. Spectral analysis of
heart rate variability during heat exposure and repeated exercise.
Eur. J. Appl. Physiol. Occup. Physiol. 76:145–156, 1997.
9. BREUER, H. W., A. SKYSCHALLY, R. SCHULZ, C. MARTIN, M. WEHR,
and G. HEUSCH. Heart rate variability and circulating catechol-
amine concentrations during steady state exercise in healthy vol-
unteers. Br. Heart. J. 70:144 –149, 1993.
10. CASADEI, B., S. COCHRANE, J. JOHNSTON, J. CONWAY, and P. SLEIGHT.
Pitfalls in the interpretation of spectral analysis of the heart rate
variability during exercise in humans. Acta Physiol. Scand. 153:
125–131, 1995.
11. CASADEI, B., J. MOON, J. JOHNSTON, A. CAIAZZA, and P. SLEIGHT. Is
respiratory sinus arrhythmia a good index of cardiac vagal tone in
exercise? J. Appl. Physiol. 81:556 –564, 1996.
12. COTTIN, F., Y. PAPELIER, F. DURBIN, and P. ESCOURROU. Effects of
exercise load and breathing frequency on heart rate and blood
pressure variability during dynamic exercise. Int. J. Sports Med.
20:232–238, 1999.
13. COTTIN, F., F. DURBIN, and Y. PAPELIER. Heart rate variability
during cycloergometric exercise or judo wrestling eliciting the
same heart rate level. Eur. J. Appl. Physiol. 91:177–184, 2003.
14. DE MEERSMAN, R. E. Respiratory sinus arrhythmia alteration fol-
lowing training in endurance athletes. Eur. J. Appl. Physiol. Oc-
cup. Physiol. 64:434 – 436, 1992.
1708 Official Journal of the American College of Sports Medicine
STFT is a useful tool for assessing HRV in time and fre-
quency domains during exercise. It would be of great inter-
est to quantify the nonautonomic component of HRV
changes during strenuous exercise.
We thank the physiology laboratory technicians for their help
during the experiments and the subjects for their participation.
15. HAYANO, J., Y. SAKAKIBARA, A. YAMADA, et al. Accuracy of as-
sessment of cardiac vagal tone by heart rate variability in normal
subjects. Am. J. Cardiol. 67:199 –204, 1991.
16. KAMATH, M. V., E. L. FALLEN, and R. MCKELVIE. Effects of steady
state exercise on the power spectrum of heart rate variability. Med.
Sci. Sports Exerc. 23:428 – 434, 1991.
17. KESELBRENER, L., and S. AKSELROD. Selective discrete Fourier
transform algorithm for time-frequency analysis: method and ap-
plication on simulated and cardiovascular signals. IEEE Trans.
Biomed. Eng. 43:789 – 802, 1996.
18. LOIMAALA, A., H. SIEVANEN, R. LAUKKANEN, et al. Accuracy of a
novel real-time microprocessor QRS detector for heart rate vari-
ability assessment. Clin. Physiol. 19:84 – 88, 1999.
19. NAKAMURA, Y., Y. YAMAMOTO, and I. MURAOKA. Autonomic con-
trol of heart rate during physical exercise and fractal dimension of
heart rate variability. J. Appl. Physiol. 74:875– 881, 1993.
20. PAGANI, M., F. LOMBARDI, S. GUZZETTI, et al. Power spectral analysis of
heart rate and arterial pressure variabilities as a marker of sympatho-vagal
interaction in man and conscious dog. Circ. Res. 59:178–193, 1986.
21. PERINI, R., C. ORIZIO, G. BASELI, S. CERUTTI, and A. VEICSTEINAS.
The influence of exercise intensity on the power spectrum of heart
rate variability. Eur. J. Appl. Physiol. 61:143–148, 1990.
22. POMERANZ, B., R. J. MACAULAY, M. A. CAUDILL, et al. Assessment
of autonomic function in humans by heart rate spectral analysis.
Am. J. Physiol. 248:H151–H153, 1985.
23. RIMOLDI, O., R. FURLAN, M. R. PAGANI, et al. Analysis of neural
mechanisms accompanying different intensities of dynamic exer-
cise. Chest 101:226S–230S, 1992.
24. SEAWARD, B. L., R. H. SLEAMAKER, T. MCAULIFFE, et al. The
precision and accuracy of a portable heart rate monitor. Biomed.
Instrum. Technol. 24:37– 41, 1990.
25. SHIN, K., H. MINAMITANI, S. ONISHI, et al. The power spectral
analysis of heart rate variability in athletes during dynamic exer-
cise: part I. Clin. Cardiol. 18:583–586, 1995.
26. TAYLOR, J. A., C. W. MYERS, J. R. HALLIWILL, et al. Sympathetic
restraint of respiratory sinus arrhythmia: implications for vagal-
cardiac tone assessment in humans. Am. J. Physiol. Heart. Circ.
Physiol. 280:H2804 –H2814, 2001.
27. WARREN, J. H., R. S. JAFFE, C. E. WRAA, and C. L. STEBBINS. Effect
of autonomic blockade on power spectrum of heart rate variability
during exercise. Am. J. Physiol. 273(2 Pt 2):R495–R502, 1997.
28. YAMAMOTO, Y., R. L. HUGHSON, and Y. NAKAMURA. Autonomic
nervous system responses to exercise in relation to ventilatory
threshold. Chest 101(Suppl. 5):206S–210S, 1992.
29. YAMAMOTO, Y., R. L. HUGHSON, and J. C. PETERSON. Autonomic
control of heart rate during exercise studied by heart rate variabil-
ity spectral analysis. J. Appl. Physiol. 71:1136 –1142, 1991.
http://www.acsm-msse.org