F.

Cottin1
P.-M. Leprtre1
P. Lopes1

Assessment of Ventilatory Thresholds

from Heart Rate Variability in
Well-Trained Subjects during Cycling

Y. Papelier3
C. Mdigue2
V. Billat1

The purpose of this study was to implement a new method for

assessing the ventilatory thresholds from heart rate variability
(HRV) analysis. ECG, VO2, VCO2, and VE were collected from eleven well-trained subjects during an incremental exhaustive test
performed on a cycle ergometer. The Short-Term Fourier Transform analysis was applied to RR time series to compute the high
frequency HRV energy (HF, frequency range: 0.15 2 Hz) and HF
frequency peak (fHF) vs. power stages. For all subjects, visual examination of ventilatory equivalents, fHF, and instantaneous HF
energy multiplied by fHF (HF fHF) showed two nonlinear increases. The first nonlinear increase corresponded to the first
ventilatory threshold (VT1) and was associated with the first HF
threshold (TRSA1 from fHF and HFT1 from HF fHF detection). The
second nonlinear increase represented the second ventilatory
threshold (VT2) and was associated with the second HF threshold

(TRSA2 from fHF and HFT2 from HF fHF detection). HFT1 , TRSA1, HFT2,
and TRSA2 were, respectively, not significantly different from VT1
(VT1 = 219 45 vs. HFT1 = 220 48 W, p = 0.975; VT1 vs. TRSA1 =
213 56 W, p = 0.662) and VT2 (VT2 = 293 45 vs. HFT2 = 294
48 W, p = 0.956; vs. TRSA2 = 300 58 W, p = 0.445). In addition,
when expressed as a function of power, HFT1, TRSA1, HFT2, and
TRSA2 were respectively correlated with VT1 (with HFT1 r2 = 0.94,
p < 0.001; with TRSA1 r2 = 0.48, p < 0.05) and VT2 (with HFT2 r2 =
0.97, p < 0.001; with TRSA2 r2 = 0.79, p < 0.001). This study confirms that ventilatory thresholds can be determined from RR
time series using HRV time-frequency analysis in healthy welltrained subjects. In addition it shows that HF fHF provides a more
reliable and accurate index than fHF alone for this assessment.

Physiology & Biochemistry

Abstract

Key words
Exercise respiratory components time frequency analysis
short-term fourier transform
1

Abbreviations
VE
VO2
VCO2
Vt
VE/VO2,
VE/VCO2
BF
HRV

ventilatory flow
oxygen uptake
carbon dioxide output
tidal volume
ventilatory equivalents
breathing frequency
heart rate variability

HF
fHF
VT1
VT2
HFT1
HFT2
TRSA1
TRSA2

high frequency spectral energy

frequency peak of HF-HRV
first ventilatory threshold detected from
ventilatory equivalents
second ventilatory threshold detected from
ventilatory equivalents
first ventilatory threshold detected from HF fHF
second ventilatory threshold detected from HF fHF
first ventilatory threshold detected from fHF
second ventilatory threshold detected from fHF

Affiliation
Laboratory of Exercise Physiology (LEPH), University of Evry, E. A. 3872 Genopole, Evry Cedex, France
2
French National Institute for Research in Computer Science and Control (INRIA), Le Chesnay, France
3
Laboratory of Physiology, Medicine Faculty, University of Paris XI, E. F. R., Hpital Antoine Bclre,
Clamart Cedex, France

Correspondence
Francois Cottin, PhD Department of Sport and Exercise Science University of Evry
Boulevard F. Mitterrand 91025 Evry Cedex France Phone: + 33 0169 64 48 81 Fax: + 33 0169 64 48 95
E-mail: fcottin@univ-evry.fr
Accepted after revision: December 5, 2005
Bibliography
Int J Sports Med Georg Thieme Verlag KG Stuttgart New York
DOI 10.1055/s-2006-923849 Published online 2006
ISSN 0172-4622

Introduction
Table 1 Characteristics of the subjects (n = 11)

Physiology & Biochemistry

2

Nowadays, the assessment of ventilatory thresholds in athletes is

used by some coaches in order to build their specific training
programs [1,19, 25]. The measurement of the breathing components during exhaustive incremental tests allows the assessment
of two ventilatory thresholds [4, 35]. According to Wasserman et
al. [35, 36], the ventilatory thresholds are indicated by the observation of the ventilatory equivalents (VE/VO2 and VE/VCO2)
curves vs. power during an incremental exercise test on a cycle
ergometer. The first ventilatory threshold (VT1) is called adaptation ventilatory threshold resulting from the hyperpnea elicited
by the increase in the CO2 metabolic production linked to the exercise intensity above anaerobic threshold. As a result, VE/VO2
nonlinearly increases while VE/VCO2 remains constant. The second ventilatory threshold (VT2) is called: maladjustment ventilatory threshold or respiratory compensation point. Since the
hyperpnea is not sufficient to eliminate the CO2 metabolic production, VE increases whereas VCO2 remains constant leading to
a drastic increase in VE/VCO2 until exhaustion.
Furthermore, heart rate variability (HRV) has been broadly investigated during exercise [3, 6, 8, 9,12,13, 33, 38]. During short-term
recordings of exercise (5 10 minutes duration), spectral energy
is divided into two frequency bands [3, 31]:
1. A low frequency band ranges from 0.04 to 0.15 Hz (LF-HRV). It
is now admitted that this LF variability is induced by both
sympathetic and vagal cardiovascular control [31].
2. A high frequency band (HF-HRV) reflecting the amplitude of
the respiratory sinus arrhythmia (RSA), linked to the breathing rate and vagal cardiovascular control. It ranges from 0.15
to fmax Hz during exercise. fmax is the maximal frequency induced by the sampling scale of the RR signal.
It is well known that total spectral energy decreases when exercise intensity increases [9,15, 26, 33]. Recently, in healthy humans [6,12] and trotting horses [11], it has been shown that,
when the exercise intensity is lower than the intensity at ventilatory threshold (moderate exercise), the energy of LF-HRV (LF) is
prevalent compared to the energy of HF-HRV (HF). In contrast,
when the exercise intensity exceeds the intensity at ventilatory
threshold (heavy exercise) HF is prevalent compared to LF. Furthermore, it seems that the hyperpnea observed during heavy
exercise induces a mechanical electric feedback on the sinus
node [6,12, 22, 23] with the stretch increasing the spontaneous
depolarization of the cardiac myocytes [24, 28]. Both phenomena
result in an increase in HF above VT1 [6].
In addition, it has been reported that the spectral frequency peak
in the HF band (fHF) corresponds closely to the breathing frequency [6,12]. Since breathing frequency (BF) increases when
VT1 is exceeded and abruptly increases when VT2 is exceeded,
fHF could follow a behavior similar to that of the breathing frequency. Recent studies have shown that VT1 could be assessed
from BF [21] and then from fHF [2]. Furthermore, Blain et al. [5]
have demonstrated that it was possible to detect both VTs from
fHF (TRSA) when TRSA1 was corresponding to VT1 and TRSA2 was corresponding to VT2. However, in 20 % of their data it was not possible to detect VTs [5]. In addition, among the numerous papers
and methods about the ventilatory threshold assessment, only a
Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

Mean

SD

Age (years)

20.0

6.3

Height (cm)

77.5

5.0

Weight (kg)

65.8

8.7

Body mass index (%)

12.3

6.7

PVO2max (W)

353.0

53.0

MAP (W)

371.0

60.0

4.5

0.4

VO2max (ml min1 kg1)

68.2

6.3

HRVO2max (beats min1)

194.0

8.0

VO2max (l min1)

few studies have shown the reliability of the VTs assessment

from BF. The present study aims to find an alternative method
of VTs assessment from HRV analysis and to compare it with the
detection method developed by Blain et al. [5].
Therefore, it is hypothesized that during an incremental protocol,
the product of the HF energy by the HF frequency peak (HF fHF)
vs. work rate could show three identifiable phases allowing the
ventilatory thresholds assessment:
1. Since the autonomic control of heart rate decreases [6,15] and
breathing frequency remains nearly constant when work rate
increases [10,18], HF fHF could attain a minimum just before
VT1 was reached [6].
2. When work rate exceeds VT1, the resulting hyperpnea could
entail an increase in fHF and in HF energy by mechanical effect
[6,11,12]. The resulting steeper increase in HF fHF would give
the first HF threshold (HFT1).
3. When work rate exceeds VT2, the subsequent increase in
breathing frequency could result in a further increase in fHF
and in HF energy by mechanical effect [6,11, 29]. This second
abrupt increase in HF fHF would give the second HF threshold
(HFT2).

Methods
Subjects
Eleven competitive male cyclists and triathletes (20 6.3 years)
participated in this study. All subjects were free of cardiac and
pulmonary disease. The anthropometric and physiological characteristics of the subjects are summarized in Table 1. Prior to participating, each subject was familiarized with the experimental
procedure and informed of the risks associated with the protocol.
All subjects gave their written voluntary informed consent in
accordance with the guidelines of the University of Evry.
Experimental design
Two to three hours after a light breakfast, all subjects performed
an incremental exercise test in the upright position on an electronically braked cycle ergometer (Ergoline 900, Marquette-Hellige, Fribourg, Germany) in an air-conditioned room. Since the
cycle ergometer performance level was different between cy-

clists and triathletes, the chosen incremental protocol was different. After a two-minute warm-up at 60 watts for triathletes or 75
watts for cyclists, each subject performed a 15 watts min1 incremental test for triathletes or 25 watts min1 for cyclists. Seat
and handlebar heights were set for each subject and kept constant for all the tests. The pedalling frequency selected by each
subject was between 70 and 100 revolutions min1.
Data collection procedures
Time series
ECG recordings were performed with a Power lab device (ADInstruments Ltd, Chalgrove Oxfordshire, UK) with a sampling frequency of 1000 Hz. The R wave peak occurrence was assessed using a threshold technique provided with the Chart5 program
(Chart5, v5.0.2 for Power Lab, ADInstruments Ltd, Chalgrove Oxfordshire, UK). Beat-to-beat RR intervals were then extracted
from the ECG signal. The ECG sampling frequency provided an
accuracy of 1 ms for each RR period. Artifacts, cumulative RR
periods and extrasystoles were manually processed by computation of interpolated or extrapolated values.
Gas measurements
VO2, VCO2, and VE were measured throughout the test using a
Quark device (Quark Pft, Cosmed, Rome, Italy), [27]. Prior to each
test, the O2 analysis system was calibrated using ambient air
(20.9 % O2 and 0.04 % CO2) and calibration gas (12.01 % O2 and 5 %
CO2). The calibration of the turbine flow-meter of the analyzer
was performed with a 3-L syringe (Quinton Instruments, Seattle).
Anthropometric measurements
Height and weight were measured before each test. Four skinfold measurements were taken (triceps, biceps, suprailiac, subscapular) with % body fat computed using the Durnin and Womersleys formula [16].

Time frequency analysis

Since, RR series were not stationary during the incremental protocol, classical spectral analysis was not suitable to analyze HRV.
Therefore, the Short Term Fourier Transform (STFT) was used for
computing HRV (Matlab software, 6.5.1, Mathworks Inc., Natick,
MA, USA). This method was already well described in previous
studies [13,14]. The main principle of STFT consists in choosing
a small enough window for analysis in which the signal can be
considered stationary [17]. Classical FFT can therefore be performed on this windowed signal. The same analysis window is
applied to the next signal block, and so on until the end of the
processed series. STFT is constituted by all the FFT performed on
the successive signal blocks that are determined by regular
translation of the chosen window. Therefore STFT yields a 3-D
figure made of all the spectra vs. time called spectrogram
(Fig. 1), which is a time-frequency representation of HRV
[13,17]. The three axes of the computed spectrogram are the following:
x-axis: time (s)
y-axis: frequency (Hz)
z-axis: power spectral density (ms2 Hz1).
Before the STFT processing, all RR series were resampled at 4 Hz
using a cubic spline function (Matlab software, 6.5.1, Mathworks
Inc., Natick, MA, USA). Each spectrogram window was made of
256 successive RR periods of 64 seconds since the time between
each RR period after resampling was 0.25 seconds. The successive spectrogram windows were spaced by 12 successive RR intervals corresponding to 3 seconds duration.
In order to remove the low frequency trend of the time series, a
time-varying finite impulse response (FIR) high-pass was applied on each STFT window [11]. Then a Hamming window was
applied on each STFT segment before computing FFT [20].

Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

Physiology & Biochemistry

Fig. 1 Typical example of a periodogram

from a subject (top) and the associated
computed spectrogram (bottom) using
Short Term Fourier Transform. The cycle
ergometer power increased from 110 at the
beginning of the periodogram (t = 360 s) to
450 watts at the end of both graphs. Spectrogram axis: X-axis: time (seconds). Y-axis:
frequency (Hz). Z-axis, Power Spectral
Density (P. S. D., ms2 Hz1). The spectrogram
shows both frequency bands changes versus
time. The LF frequency remains constant
whereas its power decreases and almost
disappears as soon as 800 s were reached.
The HF frequency (fHF) remains constant
from 420 to 600 s, and then it begins to
increase. Immediately after 800 s, fHF shows
a second steeper increase while the HF
energy increases.

Physiology & Biochemistry

4

Fig. 2 Typical example of both ventilatory thresholds assessment from ventilatory components
O2, V
E/V
CO2), and from HRV components (HF fHF). Each gas-exchange and HRV data point
E/V
(V
E/V
O2
corresponds to a 20-s interval. X-axis: time (seconds). Left Y-axis, ventilatory equivalents (V
E/V
CO2, solid line). Right Y-axis, HF fHF (dotted line, ms2 Hz) and HF BF (dashed line,
and V
ms2 Hz). Since BF and fHF are similar, HF fHF and HF BF are obviously similar. The first ventilatory
E/V
O2 while V
E/V
CO2 remains conthreshold (VT1) corresponds to the first substantial increase in V
E/V
O2
stant. The second ventilatory threshold (VT2) corresponds to the steeper increase in both V
E/V
CO2. The first HF-HRV threshold (HFT1) corresponds to the first increase in HF fHF, the secand V
ond HF-HRV threshold (HFT2) corresponds to the second abrupt increase in HF fHF. Since each
threshold is given as a power stage and there is one power stage by minute, each threshold was
given at the nearest power stage of the corresponding abrupt increase. However, a short double
E/V
O2 curve at
arrow was added on the graph, corresponding to each accurate threshold: VT1 on V
E/
20 s and HFT1 on the HF fHF curve at + 20 s of the power stage threshold. Also for VT2 on the V
CO2 curve at + 20 s and on the HF fHF curve also at + 20 s of the power stage threshold.
V

The spectral energy was computed in HF ranges by integrating

the power spectral density (PSD) for each spectrum of the spectrogram as following:

HF

fX
max

PSD 
f ms2

f 0:15

fmax is given by Shannon sampling theorem. The rule that All the
information in a signal, band-limited to a frequency of fmax, can
be captured in its samples taken a rate of greater than 2 fmax is
known as Shannons sampling theorem. The critical frequency
fmax is known as Nyquist rate. fmax = 1/(2 t), t being the sampling scale of the RR signal. In the present study the signal sampling was t = 0.25 s, thus fmax = 2 Hz.
In addition, the assessment of the instantaneous HF peak (fHF)
was computed from the spectrograms (Matlab software, 6.5.1,
Mathworks Inc., Natick, MA, USA).
Since the STFT provided one spectrum every 3 seconds, it was
then possible to get the HF energy and instantaneous fHF values
every 3 seconds. However, for an optimal assessment of the VTs
[19] and to synchronize HRV and ventilatory data, the HRV components were averaged every 20 seconds.
Ventilatory thresholds assessment
Breath by breath data were averaged to provide a data point for
each 20-s period. It was therefore possible to synchronize HRV
Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

and ventilatory data on the same graph ([19], Fig. 2). Since the
Wasserman method appears to be a consistent predictor for cycling performance in well-trained subjects with regard to reliability and validity [1,19], it was used to determine VT1 and VT2
[25, 36]. Therefore, VE/VO2 and VE/VCO2 were plotted vs. work
rate during the incremental exercise test (Fig. 2). VT1 corresponds to a first nonlinear increase in the VE/VO2 curve while
the VE/VCO2slope remains constant [35, 36]. In addition, VT2 is indicated by the nonlinear increase in the VE/VCO2 curve concomitant to a second strong increase in VE/VO2 with further increase
in exercise intensity [35, 36]. Based on the above criteria, two experienced researchers have independently assessed the ventilatory thresholds. When there was a disagreement, a third experienced investigator was involved in the process. When he agreed
with one investigator, the corresponding threshold was kept.
When all the investigators found different thresholds the subject
threshold could not be determined.
HF thresholds assessment
As it was indicated in the introduction, the present study compared two VTs assessment methods:
1. From fHF: fHF successive values were averaged to provide a data
point for each 20-s period synchronous with the ventilatory
data. HF thresholds were detected from the curve of fHF plotted vs. the work rate by an independent investigator. The first
HF threshold (TRSA1) corresponded to the last point before a
first increase in fHF. The second HF threshold (TRSA2) corresponded to a second nonlinear increase in fHF (Fig. 3).

2. From HF fHF. The hyperpnea elicited by the overstepping of

VTs could induce a double effect on HF-HRV: an increase in
HF energy (HF) combined with an increase in fHF. Thus, the
VTs could be assessed from the product of HF by fHF (HF fHF).
As it was mentioned above, HF fHF successive values were
then averaged to provide a data point for each 20-s period
synchronous with the ventilatory data. Therefore, HF thresholds were detected from the curve of HF fHF plotted vs. the
work rate by an independent investigator. The first HF threshold (HFT1) corresponded to the first nonlinear increase in
HF fHF after it has reached a minimum. The second HF threshold (HFT2) corresponded to a second nonlinear increase in
HF fHF (Fig. 2).
Statistical analysis
The Students t-test (Sigmastat 2.03, Jandel Scientifics, San Rafael, CA, USA, 1997) was used to compare the respective ventilatory and HF-HRV thresholds (VT1 vs. HFT1, VT2 vs. HFT2, VT1 vs.
TRSA1, and VT2 vs. TRSA2). Linear regression and the Pearson Product Moment Correlation (Sigmastat 2.03, Jandel Scientifics, San
Rafael, CA, USA, 1997) were used to test the correlation between
VT1 vs. HFT1, VT2 vs. HFT2, VT1 vs. TRSA1, and VT2 vs. TRSA2. BlandAltman [7] plots were conducted to illustrate the relationship
between ventilatory thresholds (VT1 and VT2) and their respective HRV thresholds (HFT1, HFT2, TRSA1, and TRSA2).

Results
Visual assessment of ventilatory and HF-HRV thresholds
Fig. 2 gives a typical example of ventilatory thresholds assessment in one subject from ventilatory components (VE/VO2, VE/
VCO2) and from HF fHF.
VTs assessment: For VT2 assessment, both investigators agreed
for all subjects, whereas VT1 assessment yielded conflicting re-

sults on three occasions. The third investigator, who was then involved in the assessment, always agreed with at least one of the
initial investigators.
fHF assessment: One other independent investigator assessed
TRSAs. For 18 % of the subjects (2/11 subjects), TRSA1 matched VT1
and for 36 % of the subjects (4/11 subjects) TRSA2 matched VT2 (Table 2). In the other cases TRSA1 and TRSA2, respectively, had one,
two, or more stage lags (Table 2). For one subject TRSA1 could not
be assessed (Fig. 3).
HF fHF assessment: One other independent investigator assessed
HFTs. For 81.8 % of the subjects (9/11 subjects), HFT1 matched VT1
and HFT2 matched VT2 (Table 2). HFT1 did not match VT1 for two
subjects, the difference corresponded to one stage lag for one
subject and two stage lags for the other subject (Table 2). HFT2
did not match VT2 for two subjects, the difference always corresponded to one stage lag (Table 2).
Comparison and relationships between ventilatory and
HF fHF thresholds
There were no significant differences between the absolute
power at VT1, nor at HFT1 (219 45 vs. 220 48 W, p = 0.975, Table 3) nor between the absolute power at VT2 and at HFT2 (293
45 vs. 294 48 W, p = 0.956, Table 3). When the different thresholds were expressed as a percentage of PVO2max, there were no
significant differences between the relative power at VT1 neither
at HFT1 (63 7 vs. 64 8 % PVO2max, p = 0.789, Table 3) nor between the relative power at VT2 and at HFT2 (80 6 vs. 81 7 %
PVO2max, Table 3). Linear regression analysis showed a strong correlation in absolute and relative (% PVO2max) terms between VT1
vs. HFT1 (absolute: r = 0.97, r2 = 0.94; relative r = 0.92, r2 = 0.85,
p < 0.001, Table 3) and VT2 vs. HFT2 (absolute; r = 0.98, r2 = 0.97;
relative r = 0.93, r2 = 0.87, p < 0.001, Table 3). The results of the
Bland-Altman plots are illustrated in Fig. 4. The standard deviation for the difference between VT1 vs. HFT1 and VT2 vs. HFT2
Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

Physiology & Biochemistry

Fig. 3 Typical example of a non-detection of TRSA1 from the same subject used in Fig. 2. Each HRV data point
corresponds to a 20-s interval. X-axis: time (seconds). Left Y-axis, fHF (solid line, Hz). Right Y-axis: HF fHF (dashed
line, ms2 Hz). Since fHF vs. time is quasi linear (r2 = 0.931) TRSA1 could not be assessed whereas TRSA2, HFT1, and
HFT2 could be assessed. Since each threshold is given as a power stage and there is one power stage by minute,
each threshold was given at the nearest power stage of the corresponding abrupt increase. However, a short
double arrow was added on the graph, corresponding to each accurate threshold: HFT1 and HFT2 on the HF fHF
curve at + 20 s and for TRSA2 on the fHF curve at 20 s of the power stage threshold.

Table 2 VT1 vs. HFT1, VT2 vs. HFT2, VT1 vs. TRSA1, and VT2 vs. TRSA2 matching
N = 11

Matching

1 stage difference

2 stages difference

3 or more

No detection

VT1 vs. HFT1

81.82%
9 subjects

9.09%
1 subject

9.09%
1 subject

0.00%
0 subject

0.00%
0 subject

VT2 vs. HFT2

81.82%
9 subjects

18.18%
2 subjects

0.00%
0 subject

0.00%
0 subject

0.00%
0 subject

VT1 vs. TRSA1

18.18%
2 subjects

18.18%
2 subjects

27.27%
3 subjects

18.18%
3 subjects

9.09%
1 subject

VT2 vs. TRSA2

36.36%
4 subjects

36.36%
4 subjects

18.18%
2 subjects

9.09%
1 subject

0.00%
0 subject

Physiology & Biochemistry

Table 3 Comparison between ventilatory vs. HF fHF thresholds and ventilatory vs. fHF thresholds by Students t-test and linear regression analysis. Significance (p) for t-test, linear regression significance (p), and correlation coefficients (r) between: VT1 vs. HFT1, VT2 vs. HFT2,
VT1 vs. TRSA1, and VT2 vs. TRSA2
Ventilatory thresholds

HRV thresholds
from HF fHF

Students
t-test

Linear regression
analysis

VT1 (W)

219 45

HFT1 (W)

220 48

0.975

0.97

VT2 (W)

293 45

HFT2 (W)

HRV thresholds
from fHF

p
< 0.001

TRSA1 (W)

213 56

Students
t-test

Linear regression
analysis

0.662

0.69

< 0.05

294 48

0.956

0.98

< 0.001

TRSA2 (W)

300 58

0.445

0.89

< 0.001

VT1 (% PVO2max)

64 7

HFT1 %

64 8

0.789

0.92

< 0.001

TRSA1 %

60 11

0.321

0.26

0.47

VT2 (% PVO2max)

80 6

HFT2 %

81 7

0.703

0.93

< 0.001

TRSA2 %

85 7

0.008

0.40

0.22

Fig. 4 Bland-Altman plots: difference against average of threshold power output. Top: first and bottom: second ventilatory thresholds. Left:
HFTs detection from HF fHF index and right: TRSAs detection from fHF alone. Center dashed line equals mean difference between ventilatory and
HRV threshold power output. The upper and lower dashed lines represent mean difference 1.96 times the standard deviation of the difference.

Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

Fig. 5 Typical example of

the non-detection of the
first ventilatory threshold
from HF index alone. Each
HRV data point corresponds
to a 20-s interval; X-axis:
time (seconds); left Y-axis:
HF fHF (dotted line, ms2Hz);
right Y-axis: (HF, solid line,
ms2). The determination of
VTs from the HF curve could
not be assessed whereas
HFT1 and HFT2 could be detected from the HF fHF
curve.

Comparison and relationships between ventilatory and fHF

thresholds
There were no significant differences between the absolute
power at VT1 and at TRSA1 (236 46 vs. 226 56 W, p = 0.662, Table 3) nor between the absolute power at VT2 and at TRSA2
(217 51 vs. 315 58 W, p = 0.445, Table 3). When the different
thresholds are expressed as a percentage of PVO2max, there were
no significant differences between the relative power at VT1 and
at HFT1 (63 7 vs. 60 11 % PVO2max,, Table 3) nor between the
relative power at VT2 and at HFT2 (80 6 vs. 85 7 % PVO2max,
p = 0.100, Table 3). Linear regression analysis showed a correlation between VT1 vs. HFT1 (r = 0.69, r2 = 0.48, p < 0.05, Table 3)
and VT2 vs. HFT2 in absolute terms (r = 0.89, r2 = 0.79, p < 0.001,
Table 3), but when expressed as a percentage of PVO2max, neither
VT1 nor VT2 could be predicted by respectively TRSA1 and TRSA2
(VT1% vs. TRSA1%: r = 0.26, r2 = 0.07 and VT2% vs. TRSA2%: r = 0.40,
r2 = 0.16, n. s., Table 3). The results of the Bland-Altman plots are
illustrated in Fig. 4. It reveals that TRSA1 underestimates VT1
whereas TRSA2 overestimates VT2. The standard deviations for
the difference between VT1 vs. TRSA1 and VT2 vs. TRSA2 were respectively 41.4 and 26.6 watts.

than TRSA1 (r = 0.97 vs. r = 0.69, Table 3). When expressed as a percentage of PVO2max, if VT1 can be predicted by HFT1 (r = 0.92, Table 3), VT1 cannot be predicted by TRSA1 (r = 0.26, Table 3). Secondly, since the difference with VT1 is lower for HFT1 than TRSA1
(HFT1VT1: 0.45 12.3 W vs. TRSA1VT1: 13.5 41.4 W, Fig. 4),
the Bland-Altman analysis reveals a better accuracy in the HF fHF
detection than fHF alone. Thirdly, VT1 matches HFT1 in 81.8 % of
the subjects (9/11 subjects) whereas VT1 match TRSA1 in 18.0 % of
the subjects (2/11 subjects). In addition, TRSA1 could not be detected for one subject (Fig. 3). Therefore, from these results, the
HRV assessment of VT1 is more accurate when using the HF fHF
index than the fHF alone.
In relation to VT2 detection and in absolute terms, VT2 can be detected by HFT2 more accurately than TRSA2 (r = 0.98 vs. r = 0.45, Table 3). Firstly, in terms of absolute value VT2 can be predicted by
HFT2 more accurately than TRSA2 (r = 0.98 vs. r = 0.45, Table 3).
When expressed as a percentage of PVO2max, if VT2 can be predicted by HFT2 (r = 0.93, Table 3), VT2 cannot be predicted TRSA2
(r = 0.4, Table 3). Secondly, the Bland-Altman analysis reveals a
better accuracy in the HF fHF detection than the fHF alone because
the difference with VT2 is lower for HFT2 than TRSA2 (HFT2 VT2:
0.91 9.2 W vs. TRSA2 VT2 = 18.18 26.6 W, Fig. 4). Thirdly, VT2
matches HFT2 in 81.8 % of the subjects (9/11 subjects) whereas
VT2 matches TRSA2 in 36.4 % of the subjects (4/11 subjects). Therefore, regarding these results, as for VT1, the HRV assessment of
VT2 is more accurate when using HF fHF index than fHF alone.

Discussion
The present study shows no significant difference between the
ventilatory thresholds assessed from ventilatory signals and
from the ECG signal (VT1 vs. HFT1 and TRSA1; VT2 vs. HFT2 and
TRSA2, n. s.). Thus, the HRV thresholds can be detected from fHF
[2, 5] but also from HF fHF. The discussion will now focus on two
main parts. The first part will compare the two HRV assessment
methods developed in this paper (fHF vs. HF fHF assessment). The
second part will discuss the advantage of using HF fHF for detecting HFT1 and HFT2 (HF fHF) and the physiological mechanisms
linked to its behavior during the incremental exhaustive test.
HRV assessments: fHF vs. HF fHF in relation to VT1 detection and
in absolute terms, VT1 can be predicted by HFT1 more accurately

The choice of HF fHF index for the ventilatory thresholds detection: The second point of the discussion considers the choice of
HF fHF index which enabled the assessment of the ventilatory
thresholds. For a better understanding, the discussion about the
assessment of VT1 will be dissociated from the VT2 assessment.
What could be the physiological mechanisms involved in the first
increase in HF fHF allowing HFT1 detection? And why this index
is more adequate than fHF alone to detect VT1? The detection of
HFT1 is linked to the two concomitant increases in HF and fHF.
The former considers the HF energy changes. It has been shown
that during an incremental exercise, when RR decreases the
overall HRV decreases [15, 26, 32, 34, 37]. As a result, just before
VT1 was reached, HF energy is minimal [6]. In contrast, when
VT1 is exceeded, HF increases progressively (Fig. 1). During heavy
Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

Physiology & Biochemistry

were respectively 12.3 and 9.2 watts. No hysteresis effect of the

average output on the differences between VTs and HFTs thresholds was observed.

Physiology & Biochemistry

8

exercise cardiac vagal control is no longer effective [30]. While

vagal withdrawal is conflicting with an increase in HF energy,
the hyperpnea (concomitant increase in Vt and BF) [10,18] induced by the overstepping of VT1 could be the result of a mechanical effect on the sinus node, inducing an increase in HF synchronous with VT1 [6, 9,11,12]. Blain et al. [6] found that, when
expressed as a percentage of VO2peak, HF energy is minimal
around 61 62 %. This result is consistent with the present study
results: HF is minimal around 60 % PVO2max. However, beyond
64 % PVO2max (beyond the detected HFT1 and VT1 of this study)
the hyperpnea induces a double effect on HF fHF: on the one
hand, an increase in HF by mechanical effect and on the other
hand an increase in fHF linked to BF increase. There is no physiological argument that the minimal point of HF would be concomitant with VT1. However, the first increase after HF had reached a
minimum is probably linked to the overstepping of VT1. Unfortunately, this increase in HF is sometimes not very marked and the
VT1 detection is then difficult (Fig. 5). However, when VT1 is exceeded, fHF increases, but as for the HF concomitant increase, the
expected increase in fHF is sometimes linear (Fig. 3) and the VT1
detection is then difficult or even impossible. In contrast, when
multiplying HF by fHF the concomitant increase in both HF and
fHF is amplified just after the overstepping of VT1. Consequently,
the first ventilatory threshold assessment is easier and more accurate from HF fHF than from fHF or HF alone.

of the product of HF fHF accentuates the expected two successive

nonlinear increases when the exercise intensity oversteps the
ventilatory thresholds. Consequently, HF fHF is considered to be
a more effective index to assess the ventilatory thresholds from
HRV than HF or fHF alone.

Conclusion
This study has confirmed that the ventilatory thresholds can be
detected from the cardiac RR series using HRV time frequency
analysis during an incremental exercise test in athletes. In addition, it has been shown that HF fHF provides a reliable index for
this assessment. Therefore, this study proposed a noninvasive
method of VTs detection from a simple ECG without any use of
expensive ventilatory device. Thus, the assessment of both HFT1
and HFT2 from HRV could provide a substitute for the respiratory
methods when the breathing analysis is not available. Although,
HRV thresholds have been detected by different independent experts, further studies could be conducted to implement algorithms for the automated detection of the HRV thresholds.

References
1

Similar questions can be addressed for the VT2 assessment. What

could be the physiological mechanisms involved in the second
nonlinear increase in HF fHF allowing HFT2 detection? And why
this index is more adequate than fHF alone to detect VT2? Regarding fHF, at this exercise intensity, the further increase in minute
ventilation (VE) is a consequence of an increasing breathing frequency while Vt remains maximal and constant [10,18]. Consequently, when VT2 is exceeded fHF increases abruptly (Fig. 3). It
is then possible to detect VT2 from fHF alone [5]. Regarding HF,
Perlini et al. [29] have shown an increase in HF energy when
breathing frequency increases and tidal volume remained constant, in anesthetized, vagotomized, and mechanically ventilated
rabbits. This effect was enhanced at higher tidal volume. Thus,
the increase in breathing frequency during heavy exercise could
induce an increase in the RSA amplitude and consequently an increase in HF energy synchronous with BF. According to the above
mentioned studies [6,11,12, 29], it seems possible that the increase in HF energy observed when VT2 is exceeded could be
the result of a mechanical effect of the increasing breathing frequency on the heart. Briefly, the overstepping of VT2 entails a
combined increase in both HF and fHF. Thus, the curve of the
product of HF by fHF plotted vs. the work rate provides a more
pronounced increase at VT2 than the curve of fHF alone. Therefore,
the detection of VT2 from HF fHF is easier and more accurate than
from fHF alone (Fig. 4, Tables 2, 3).

10

11

12

To sum up, during an incremental protocol, the increase in HF energy and fHF frequency, together with the increasing exercise intensity could induce two successive nonlinear increases corresponding to VT1 (HFT1) and VT2 (HFT2) (Figs. 1, 2). However, at
high work loads, neither the expected increase in HF, nor the
two nonlinear increases in fHF were clearly observed in a few subjects ([5], Figs. 3, 5), whereas the product of HF fHF vs. work load
showed the two expected nonlinear increases. Therefore, the use
Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med

13

14

Amann M, Subudhi AW, Walker J, Eisenman P, Shultz B, Foster C. An

evaluation of the predictive validity and reliability of ventilatory
threshold. Med Sci Sports Exerc 2004; 36: 1716 1722
Anosov O, Patzak A, Kononovich Y, Persson PB. High-frequency oscillations of the heart rate during ramp load reflect the human anaerobic threshold. Eur J Appl Physiol 2000; 83: 388 394
Aubert AE, Seps B, Beckers F. Heart rate variability in athletes. Sports
Med 2003; 33: 889 919
Beaver WL, Wasserman K, Whipp BJ. A new method for detecting
anaerobic threshold by gas exchange. J Appl Physiol 1986; 60:
2020 2027
Blain G, Meste O, Bouchard T, Bermon S. Assessment of ventilatory
thresholds during graded and maximal exercise test using time varying analysis of respiratory sinus arrhythmia. Br J Sports Med 2005;
39: 448 452
Blain G, Meste O, Bermon S. Influences of breathing patterns on respiratory sinus arrhythmia during exercise. Am J Physiol 2005; 288:
H887 H895
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307
310
Buchheit M, Richard R, Doutreleau S, Lonsdorfer-Wolf E, Brandenberger G, Simon C. Effect of acute hypoxia on heart rate variability at
rest and during exercise. Int J Sports Med 2004; 25: 264 269
Casadei B, Moon J, Johnston J, Caiazza A, Sleight P. Is respiratory sinus
arrhythmia a good index of cardiac vagal tone in exercise? J Appl
Physiol 1996; 81: 556 564
Clark JM, Hagerman FC, Gefland R. Breathing patterns during submaximal and maximal exercise in elite oarsmen. J Appl Physiol 1983;
55: 440 446
Cottin F, Mdigue C, Lopes P, Petit E, Papelier Y, Billat VL. Heart rate
variability analysis in horse trotting during exercise training. Int J
Sports Med 2005; 26: 859 867
Cottin F, Mdigue C, Leprtre PM, Papelier Y, Koralsztein JP, Billat VL.
Heart rate variability and dynamic cardio-respiratory interactions
during exercise. Med Sci Sports Exerc 2004; 36: 594 600
Cottin F, Durbin F, Papelier Y. Heart rate variability during cycloergometric exercise or judo wrestling eliciting the same heart rate level.
Eur J Appl Physiol 2004; 91: 177 184
Cottin F, Papelier Y, Durbin F, Koralsztein JP, Billat VL. Effect of fatigue
on spontaneous velocity variations in human middle-distance running: use of Short Term Fourier Transform. Eur J Appl Physiol 2002; 87:
17 27

15

16

17

18

19

20

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

Pathak CL. Autoregulation of chronotropic response of the heart to

pacemaker stretch. Cardiology 1973; 58: 45 64
Perlini S, Solda PL, Piepoli M, Sala-Gallini G, Calciati A, Finardi G, Bernardi L. Determinants of respiratory sinus arrhythmia in the vagotomized rabbit. Am J Physiol 1995; 269: H909 H915
Rowell LB. Human Cardiovascular Control. New York, USA: Oxford
University Press, 1993; 5: 172 175
Task force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Heart rate variability.
Standards of measurement, physiological interpretation, and clinical
use. Circul 1996; 93: 1043 1065
Tulppo MP, Mkikallio TH, Seppnen T, Laukkanen RT, Huikuri HV. Vagal modulation of heart rate during exercise: effects of age and physical fitness. Am J Physiol 1998; 274: H424 H429
Tulppo MP, Mkikallio TH, Takala TES, Seppnen T, Huikuri HV. Quantitative beat-to-beat analysis of heart rate dynamics during exercise.
Am J Physiol 1996; 271: H244 H252
Warren JH, Jaffe RS, Wraa CE, Stebbins CL. Effect of autonomic blockade on power spectrum of heart rate variability during exercise. Am J
Physiol 1997; 273: R495 R502
Wasserman K, Whipp BJ, Koyal SN, Beaver WL. Anaerobic threshold
and respiratory gas exchange during exercise. J Appl Physiol 1973;
35: 236 243
Wasserman K, Mc Ilroy MB. Detecting the threshold of anaerobic metabolism in cardiac patients during exercise. Am J Cardiol 1964; 14:
844 852
Yamamoto Y, Hughson RL, Nakamura Y. Autonomic nervous system
responses to exercise in relation to ventilatory threshold. Chest
1992; 10: 206S 210S
Yamamoto Y, Hughson RL, Peterson JC. Autonomic control of heart
rate during exercise studied by heart rate variability spectral analysis.
J Appl Physiol 1991; 71: 1136 1142

Physiology & Biochemistry

21

Cottin F, Papelier Y, Escourrou P. Effects of exercise load and breathing

frequency on heart rate and blood pressure variability during dynamic exercise. Int J Sports Med 1999; 20: 232 238
Durnin JV, Womersley J. Body fat assessment from total body density
and its estimation from skinfold thickness: measurements on 481
men and women aged 16 to 76 years. Br J Nutr 1974; 32: 77 97
Gabor D. Theory of communication. J Inst Electr Engin 1946; 93: 429
457
Gallagher CG, Brown E, Younes M. Breathing pattern during maximal
exercise and during submaximal exercise with hypercapnia. J Appl
Physiol 1987; 63: 238 244
Gaskill SE, Ruby BC, Walker AJ, Sanchez OA, Serfass RC, Leon AS. Validity and reliability of combining three methods to determine ventilatory threshold. Med Sci Sports Exerc 2001; 33: 1841 1848
Harris FJ. On the use of windows for harmonic analysis with the discrete Fourier Transform. Proc Inst Electr Electron Engin 1978; 66: 51
83
James NW, Adams GM, Wilson AF. Determination of anaerobic
threshold by ventilatory frequency. Int J Sports Med 1989; 10: 192
196
Kohl P, Hunter P, Noble D. Stretch-induced changes in heart rate and
rhythm: clinical observations, experiments, and mathematical models. Prog Biophys Mol Biol 1999; 71: 91 138
Kohl P, Kamkin AG, Kiseleva S, Streubel T. Mechanosensitive cells in
the atrium of frog heart. Exp Physiol 1992; 77: 213 216
Lange G, Lu HH, Chang A, Brooks CM. Effect of stretch on the isolated
cat sinoatrial node. Am J Physiol 1966; 211: 1192 1196
Londeree BR. Effect of training on lactate/ventilatory thresholds: a
meta-analysis. Med Sci Sports Exerc 1997; 29: 837 843
Macor F, Fagard R, Amery A. Power spectral analysis of RR interval and
blood pressure short-term variability at rest and during dynamic exercise: comparison between cyclists and controls. Int J Sports Med
1996; 17: 175 181
Mc Laughlin JE, King GA, Howley ET, Basset DR, Ainsworth BF. Validation of the Cosmed K4b2 portable metabolic system. Int J Sports Med
2001; 22: 280 284

Cottin F et al. Ventilatory Thresholds Assessment from HRV Int J Sports Med