Content uploaded by Gregoire P Millet
Author content
All content in this area was uploaded by Gregoire P Millet on Feb 08, 2015
Content may be subject to copyright.
ORIGINAL ARTICLE
A pilot study on quantification of training load: The use of HRV in
training practice
DAMIEN SABOUL
1,2
, PASCAL BALDUCCI
1
, GRÉGOIRE MILLET
3
, VINCENT PIALOUX
1
,
& CHRISTOPHE HAUTIER
1
1
CRIS, Center of Research and Innovation on Sport, University Claude Bernard Lyon 1, France,
2
Almerys, 46, rue du Ressort
63967 Clermont –Ferrand, France,
3
ISSUL, Institute of Sport Sciences, University of Lausanne, Lausanne, Switzerland
Abstract
Recent laboratory studies have suggested that heart rate variability (HRV) may be an appropriate criterion for training load
(TL) quantification. The aim of this study was to validate a novel HRV index that may be used to assess TL in field
conditions. Eleven well-trained long-distance male runners performed four exercises of different duration and intensity. TL
was evaluated using Foster and Banister methods. In addition, HRV measurements were performed 5 minutes before
exercise and 5 and 30 minutes after exercise. We calculated HRV index (TL
HRV
) based on the ratio between HRV decrease
during exercise and HRV increase during recovery. HRV decrease during exercise was strongly correlated with exercise
intensity (R=−0.70; p< 0.01) but not with exercise duration or training volume. TL
HRV
index was correlated with Foster
(R= 0.61; p= 0.01) and Banister (R= 0.57; p= 0.01) methods. This study confirms that HRV changes during exercise and
recovery phase are affected by both intensity and physiological impact of the exercise. Since the TL
HRV
formula takes into
account the disturbance and the return to homeostatic balance induced by exercise, this new method provides an objective
and rational TL index. However, some simplification of the protocol measurement could be envisaged for field use.
Keywords: Heart rate variability, monitoring, fatigue, athletes, autonomic nervous system
Introduction
Quantification of training load (TL) is considered as
an important topic in sport sciences since it is a
major tool for training follow-up (Borresen & Lam-
bert, 2009). Initially, questionnaires and diaries were
used to measure TL but they were replaced by more
objective methods based on physiological measure-
ments (Borresen & Lambert, 2009). Some studies
suggested evaluating TL through biological markers
like oxygen uptake or blood lactate concentration
(Hopkins, 1991). However, these measures require
specific equipment and are mainly used in the con-
text of scientific research. Conversely, other methods
based on heart rate (HR) or rating of perceived
exertion (RPE) appear more suitable for daily use
and practical application (Karvonen & Vuorimaa,
1988). From these markers, several indices of train-
ing stress, like training impulse (TRIMP) or session
RPE, have been developed to assess TL (Banister,
Good, Holman, & Hamilton, 1986; Foster, 1998).
Especially, Banister method consists in calculating
the ratio of HR means during exercise and HR reserve
multiplied by weighting coefficient and exercise
duration (equation 1; Banister et al., 1986; Borresen
& Lambert, 2009). Conversely, Foster method is
based on the multiplication of a score (from 1 to 10)
representing exercise difficulty perceived by athlete by
exercise duration (Borresen & Lambert, 2009; Foster,
1998). However, these two methods present some
limitations. For instance, TRIMP method cannot
discriminate intermittent or continuous training ses-
sions (TSs) of same duration since mean HR is
equivalent. In addition, this method is not valid for
high-intensity interval training, especially during
shorter repetitions with strong anaerobic contribution
because of the shifted kinetics of HR in response to
Correspondence: D. Saboul, Centre de Recherche et d’Innovation sur le Sport, Université Claude Bernard Lyon 1, 69622 Villeurbanne
Cedex, Lyon, France. E-mail: dsaboul@free.fr
European Journal of Sport Science, 2015
http://dx.doi.org/10.1080/17461391.2015.1004373
© 2015 European College of Sport Science
Downloaded by [UQ Library] at 07:32 08 February 2015
exercise (Buchheit & Laursen, 2013). Similarly,
Foster method is mainly based on a self-evaluation
of the TS intensity which is likely a limitation
for subjects not familiarised with the RPE scale and
also for expert subjects which may under or over
evaluate the results (Foster, Heimann, Esten, Brice, &
Porcari, 2001). Nonetheless, for practical reasons, the
Banister and Foster methods are widely used by
coaches and athletes to assess TL and monitor
endurance training but it is important to note that
there is no real “gold standard”for TL evaluation
(Hellard et al., 2006; Kaikkonen, Hynynen, Mann,
Rusko, & Nummela, 2010).
For 20 years, heart rate variability (HRV) has been
widely used as a non-invasive method to estimate
cardiac autonomic regulation, which may reflect
the activity of the autonomic nervous system (ANS;
Task-Force, 1996). This indicator is sensitive to
homeostatic perturbations like fatigue, physiological
and psychological stress (Aubert, Seps, & Beckers,
2003; Chandola, Heraclides, & Kumari, 2010). After
stimulus, ANS regulates homeostatic function of the
body (Aubert et al., 2003; Buchheit, Laursen, &
Ahmaidi, 2007) and therefore plays an important
role in the individual exercise training responses
(Chalencon et al., 2012; Hautala, Kiviniemi, &
Tulppo, 2009). More specifically, it was shown that
exercise induced parasympathetic withdrawal and
sympathetic excitation and that these effects were
reversed during recovery phase (Buchheit et al., 2007).
Recent studies focused on the relationship between
training content (e.g. intensity, duration, etc.) and
post-exercise HRV changes (Casties, Mottet, & Le
Gallais, 2006; Kaikkonen, Hynynen, Mann, Rusko,
& Nummela, 2012; S. Seiler, Haugen, & Kuffel,
2007). It is now clearly established that intensity is
related to immediate post-exercise HRV (Buchheit
et al., 2007; Kaikkonen, Rusko, & Martinmäki,
2008; Kaikkonen et al., 2010). On the other hand,
it has been shown that immediate post-exercise
HRV was not affected by the increase of exercise
duration up to twice the baseline (Kaikkonen,
Nummela, & Rusko, 2007; Kaikkonen et al., 2010;
S. Seiler et al., 2007). Moreover, the time course of
HRV markers during recovery has been studied by
several authors to quantify vagal reactivation after
exercise (Kaikkonen et al., 2010,2012; S. Seiler
et al., 2007). Recent studies showed that training
above the lactate threshold (LT) intensity delayed
HRV recovery compared with training below LT
(Plews, Laursen, Kilding, & Buchheit, 2014). Other
studies presented different post-exercise HRV kinet-
ics between continuous and intermittent running
sessions (Kaikkonen et al., 2008). In summary,
the duration for post-exercise HRV to return to
baseline values seems longer after exercise inducing
a greater metabolic demand (Stanley, Peake, &
Buchheit, 2013). Thereby, authors suggested that
post-exercise HRV may enable an objective TL
evaluation (Kaikkonen et al., 2010,2012). However,
most studies in line with post-exercise HRV recovery
and TL have been performed on subjects moderately
trained, in laboratory conditions (i.e. on treadmill or
ergocycle) and during exercise protocols far removed
from usual TSs (Kaikkonen et al., 2007,2010;
Martinmäki & Rusko, 2008). In addition, to our
knowledge, there exists no method or tool based on
HRV measurements for the rational evaluation of
TL in field conditions (Kaikkonen et al., 2012).
The aim of the present study was therefore to
propose a new HRV-based method for quantifying
TL. This method was tested in highly trained
athletes in field conditions during their usual TSs.
Finally, this new method was compared to two
previous methods commonly used by coaches and
athletes (i.e. Banister and Foster).
Methods
Subjects
Eleven well-trained long-distance male runners were
recruited from local running teams [competition
level: seven regional, three national and one interna-
tional; age = 32 ± 6 year; height = 182 ± 5 cm;
body mass = 76.3 ± 10.2 kg; maximal aerobic
speed (MAS) = 18.9 ± 1.2 km/h; resting HR = 44 ±
4 bpm; maximal HR = 187 ± 8 bpm; training = 450 ±
139 h/year]. Subjects receiving medical treatment,
or with asthma or cardiovascular disorders, were
excluded. The subjects, volunteers, gave written
informed consent to participate in this study. In
addition, throughout the experiment, the subjects
agreed not to change their living routine including
sleep duration, diet and professional occupation. The
protocol was approved by the ethical committee of
Lyon Sud-Est II and was in accordance with the
guidelines set by the Declaration of Helsinki.
Experimental design
The total duration of the study was two weeks. MAS
was first measured in a preliminary session using a
validated continuous multi‐stage track test: the Uni-
versité de Montréal track test (Leger & Boucher,
1980). Resting HR (HR
rest
) was measured with the
subject in a sitting position before the MAS test. HR
was recorded during the test (Suunto T6d heart rate
monitor, Suunto Oy, Finland) and the maximal HR
(5-s average) was considered to be the participant’s
HR
max
(Buchheit et al., 2010).
The subjects performed four different TSs
throughout the experiment. Exercises were per-
formed at the same time of day in a random order
2D. Saboul et al.
Downloaded by [UQ Library] at 07:32 08 February 2015
on four different days, separated by at least three
days. The subjects were asked to refrain from intense
physical exercise for two days and from alcohol and
caffeine consumption for one day prior to any
experimental session.
HRV measurements
For each session, HRV measurements were per-
formed in three different five-minute periods: [−5;
0] minutes before warm-up for all TS (Pre5), [5; 10]
minutes after TS (Post5) and [30; 35] minutes after
TS (Post30), respectively. HRV measurement con-
sisted of a five-minute R–R interval recording in
supine position in a quiet environment (Buchheit
et al., 2010; Plews, Laursen, Stanley, Kilding, &
Buchheit, 2013). Data were collected and recorded
using a validated HR monitor (Suunto T6d –1000
Hz). Data analyses were restricted to time domain
indices and the only HRV marker used was the root
mean square difference of successive normal R–R
intervals (RMSSD; Buchheit et al., 2010; Plews,
Laursen, Stanley, et al., 2013). RMSSD was chosen
because it represents short-term HRV variability and
especially vagal modulation (Buchheit et al., 2007,
2010; Task-Force, 1996). Consequently, this HRV
marker is widely used in the field of exercise physiology
(Buchheitetal.,2010; Plews, Laursen, Kilding, &
Buchheit, 2012). On the contrary, the spectral HRV
markerslike LF, HF and LF/HF ratio are less reliable in
the context of well-trained athletes (Saboul, Pialoux, &
Hautier, 2013,2014). Between Post5 and Post30 HRV
measurements (i.e. recovery phase), subjects had to
stay seated alone, in a locker room, in quiet and
comfortable environment (18.5–21°C). They were
asked to drink up to a maximum of 500 ml of water
but no food.
Training sessions
The experimental TSs were designed first to repres-
ent the usual TSs regularly undertaken during the
season by these well-trained athletes (Stepto, Haw-
ley, Dennis, & Hopkins, 1999) and second to cover a
representative range of track sessions in terms of
intensity and duration. Consequently by design the
workloads were not matched for intensity × volume.
Each session was performed on an athletic track for a
similar duration for each subject. To control exercise
intensity, the speeds were individualised as a per-
centage of each athlete’s MAS.
TS 1 (S
70%
) consisted of a 10-minute warm-up
run followed by active endurance running at 70% of
MAS for 34 minutes. The session ended with
10 minutes of cool down at low speed (50% MAS)
for a total training duration of 54 minutes (total
volume = 3580 a.u.).
TS 2 (S
85%
) consisted of a 20-minute warm-up
run followed by three 10-minute bouts at 85% of
MAS with three minutes of passive recovery. The
session ended with 10 minutes of cool down at low
speed (50% MAS) for a total training duration of 69
minutes (total volume = 4350 a.u.).
TS 3 (S
95%
) consisted of a 20-minute warm-up
run followed by eight 2-minute bouts at 95% of
MAS with 1 minute of active recovery at 60% of
MAS. The session ended with 10 minutes of cool
down at low speed (50% MAS) for a total training
duration of 54 minutes (total volume = 3800 a.u.).
TS 4 (S
100%
) consisted of a 25-minute warm-up
run followed by one bout of 6-minute duration at
100% of MAS. The session ended with 10 minutes
of cool down at low speed (50% MAS) for a total
training duration of 41 minutes (total volume =
2700 a.u.).
TL estimation
Three methods were used in order to quantify the
TL of each session. First, we used the HR recorded
during exercise. TRIMP was calculated according to
equation 1 (Banister et al., 1986).
TRIMP ¼THRexe HRrest
HRmax HRrest
0:64
e1:92HRexeHRrest
HRmaxHRrest ð1Þ
TRIMP: training impulse
T: duration of TS (min)
HR
exe
: mean HR of the TS (bpm)
HR
rest
: HR at rest (bpm)
HR
max
: maximal HR (bpm)
e: Naperian logarithm of 2.712
Second, RPE (scale 0–10) was obtained 30 min-
utes after the exercise and multiplied by the duration
of the TS (min) according to Foster method (Borre-
sen & Lambert, 2009; Foster, 1998). Third, we
defined a new index (TL
HRV
) for quantifying the TL
from the change from pre- to post-exercise RMSSD.
TL
HRV
index was calculated according to
equation 2.
TLHRV ¼In TPre5 Post5
Post30 Post5
ð2Þ
TL
HRV
: TL index
T: duration of the TS (min)
Pre5: RMSSD value before TS (ms)
Post5: RMSSD value five minutes after TS (ms)
Post30: RMSSD value 30 minutes after TS (ms)
Training load quantification with HRV 3
Downloaded by [UQ Library] at 07:32 08 February 2015
This equation was designed to reflect the kinetics
of HRV recovery with both the disturbance (i.e.
Pre5–Post5) and the return to homeostatic state of
rest (i.e. Post30–Post5). The first part of the calcu-
lation was designed to take into account training
intensity through RMSSD decrease (Pre5–Post5)
since previous results demonstrated that HRV mod-
ifications were more sensitive to exercise intensity
than exercise duration (Kaikkonen et al., 2007).
Moreover, post-exercise RMSSD increase (Post30–
Post5) was included in the formula to evaluate
exercise effects on vagal reactivation (Buchheit et al.,
2007). We chose to assess the second measurement
30 minutes after the end of exercise: first, since
previous results reported that 30 minutes of HRV
recovery is a “mid-point”and a good compromise to
investigate recovery processes (Kaikkonen et al.,
2007,2008; S. Seiler et al., 2007) and second, in
comparison with the Foster method that also uses
feedback recorded after 30 minutes (Borresen &
Lambert, 2009; Foster, 1998). Finally, the method
included a ratio between RMSSD decrease (Pre5–
Post5) and post-exercise increase (Post30–Post5)
which normalised HRV changes, lowering the influ-
ence of day-to-day baseline HRV fluctuations linked
to sleep, diet or stress. Post-HRV measure being not
significantly influenced by training duration (Kaik-
konen et al., 2010), according to the Banister and
Foster methods, our (Pre5–Post5)/(Post30–Post5)
ratio was also multiplied by the TS duration (i.e.
T). Finally, due to the skewed nature of HRV
recordings, these data were log transformed by
taking the natural logarithm (ln; equation 2;
Buchheit et al., 2010; Plews et al., 2012).
Statistical analysis
All values were expressed as means ± 90% confid-
ence limits. The normality of data was tested with
the Shapiro–Wilk test. Data were not normally
distributed. Thus, the Friedman test is used for
one-way repeated measures analysis of variance in
order (1) to examine the difference in HRV value
(Pre5 vs. Post5 vs. Post30) for each TS, (2) to
compare the difference between all sessions (S
70%
vs.
S
85%
vs. S
95%
vs. S
100%
) at each HRV recovery time
and (3) to compare TL of all TSs (TL-S
70%
vs.TL-
S
85%
vs.TL-S
95%
vs.TL-S
100%
) calculated with the
three TL methods. Post hoc analyses were per-
formed with the Wilcoxon signed rank test. In
addition, the magnitudes of change between HRV
values (Pre5 vs. Post5 vs. Post30) for each TS and
between all TSs calculated with the three TL
methods where expressed as the standardised mean
differences [effect size (ES)], using Hopkins’spread-
sheet. The following threshold values for effects size
statistics were adopted: ≤0.1 (trivial), ≥0.2 (small),
≥0.6 (moderate), ≥1.2 (large) and ≥2.0 (very large).
Spearman’s correlation coefficient was used to
study the relationships between exercise intensity
vs.normalised Post5 HRV values (i.e. relative to
Pre5 HRV values), total exercise volume vs. normal-
ised Post5 HRV values and exercise duration vs.
normalised Post5 HRV values. In addition, Spear-
man’s correlation coefficient was used to study the
relationships between the three TL indexes (Foster
vs. Banister vs. TL
HRV
). The R
2
values were
converted to Rvalues in order to use the following
adapted criteria to interpret the magnitude of the
relationship, where ≥0.1–0.3 is small, ≥0.3–0.5 is
moderate, ≥0.5–0.7 is large, ≥0.7–0.9 is very large
and ≥0.9–1.0 is almost perfect.
Agreement between the three methods was exam-
ined by Bland and Altman plots. Since the three
methods do not have the same unit, the individual
TL calculated by each method was expressed as a
percentage of the total TL of the four TSs in order to
construct Bland and Alman plots; for example, %
TL-S
70%
= 100 × [TL-S
70%
/(TL-S
70%
+ TL-S
85%
+
TL-S
95%
+ TL-S
100%
)]. The differences between the
measurements of TL performed with the three
methods (expressed as a percentage) were devised
in relation to the mean values; 95% of the differences
were expected to lie between the two “limits of
agreement”that were the mean difference ± 1.96 SD
of the differences, expressed as bias ± random error.
In addition, heteroscedasticity was tested. Because
all data have been normalised (i.e. expressed as a
percentage), all bias of Bland and Altman plots are
equal to zero. The data were analysed using StatSoft
software (Statistica 7.1, StatSoft, Inc., USA) and the
statistical significance was set at p< 0.05.
Results
The RMSSD values were significantly different
between Pre5, Post5 and Post30 within each session
(p< 0.05). As shown in Table I, Post5 RMSSD
values were largely lower than Pre5 values in all ses-
sions (ES ± 90% confidence limits: −1.26 ± 0.10).
Conversely, Post30 RMSSD values were moderately
greater than at Post5 values (ES ± 90% CL: 0.96 ±
0.11).
RMSSD values between the four sessions were not
significantly different (p= 0.16) in baseline (Pre5)
where as they were significantly different in Post5
(p= 0.0004) and Post30 (p= 0.00006; Table I).
As presented in the last column of Table I, there
was a significant difference (p= 0.0001) between the
TL
HRV
index calculated from the three RMSSD
values of each TS. A significant correlation [R=
−0.70 (very large); p< 0.000001] was observed
between exercise intensity and normalised Post5
4D. Saboul et al.
Downloaded by [UQ Library] at 07:32 08 February 2015
HRV values (expressed as a percentage of Pre5;
Figure 1). By contrast, no correlation was observed
between exercise duration and normalised Post5
HRV values [R= 0.22 (small); p= 0.15] and
between total training volume and Post5 HRV values
[R = 0.20 (small); p= 0.19]. In addition, no
significant correlations were found between Post30
HRV values and exercise intensity, exercise duration
or training volume [respectively, R=−0.27 (small),
R=−0.04 (small) and R=−0.08 (small) with p>
0.05 for all correlations].
TL of each session was evaluated by the three
different methods. Results are presented in Figure 2
for Banister (top), Foster (middle) and TL
HRV
(bottom) methods. Banister’s method revealed sig-
nificant TL differences for all TS except between
S
70%
and S
95%
. ES is large (−1.25) for S
70%
vs. S
85%
,
small (−0.25) for S
70%
vs. S
95%
, large (1.49) for S
70%
vs. S
100%
, moderate (−1.17) for S
85%
vs. S
95%
, large
(−1.84) for S
85%
vs. S
100%
and large (1.61) for S
95%
vs. S
100%
.Using Foster’s method, we observed sig-
nificant differences for TL of all TS except between
S
85%
and S
95%
. ES is large (−1.51) for S
70%
vs. S
85%
,
large (−1.36) for S
70%
vs. S
95%
, moderate (−1.00)
for S
70%
vs. S
100%
, small (−0.52) for S
85%
vs. S
95%
,
large (−1.25) for S
85%
vs. S
100%
and moderate (0.93)
for S
95%
vs. S
100%
. Finally, TL
HRV
provided signi-
ficant differences for TL of all TS except between
S
85%
and S
95%
and between S
70%
and S
100%
.ESis
large (−1.28) for S
70%
vs. S
85%
, large (−1.28) for
S
70%
vs. S
95%
, small (−0.48) for S
70%
vs. S
100%
,
small (−0.42) for S
85%
vs. S
95%
, moderate (−1.14)
for S
85%
vs. S
100%
and moderate (1.06) for S
95%
vs. S
100%
.
More generally, TL
HRV
and Foster values were
significantly correlated [R= 0.61 (large); p=
0.00001]. In addition, correlation between TL
HRV
and Banister values and between Foster and Banister
values was also significant [respectively, R= 0.57
(large); p= 0.00006 and R= 0.43 (moderate);
p= 0.004].
The Bland and Altman plots presented in Figure 3
showed that x-axis values of all methods are hetero-
geneously distributed and the differences (i.e. y-axis
values) are normally distributed. In the middle graph
(i.e. Foster vs. TL
HRV
), all the differences are
comprised between the 95% limits of agreement
(mean ± 1.96 SD). In the top and bottom graphs
(i.e. Banister vs. TL
HRV
and Foster vs. Banister),
only one point is not included between the 95%
limits of agreement (less than 5%). According to
heteroscedasticity results, there is a positive relation-
ship between the mean values and difference values
of TL
HRV
vs. Foster methods (R= 0.45; p< 0.01).
Discussion
The aim of the present study was to examine the
relationship between TL and HRV variations
induced by aerobic exercise in field conditions on
highly trained athletes. The results of this work can
be summarised by two main findings. First, as
Table I. Mean ± 90% confidence limits of RMSSD (ms), HR
exe
(bpm) and RPE score for each TSs
Pre5 RMSSD Post5 RMSSD Post30 RMSSD Pre5–Post5 ES Post5–Post30 ES TL
HRV
HR
exe
RPE score
S
70%
78 ± 25 33 ± 12* 119 ± 59
#
Moderate ↘Moderate ↗3.7 ± 0.3 146.7 ± 5.5 4.2 ± 0.8
S
85%
89 ± 27 20 ± 9* 65 ± 42
#a
Large ↘Moderate ↗5.5 ± 0.7
a
147.8 ± 3.6 6.3 ± 0.7
a
S
95%
73 ± 28 11 ± 3*
ab
36 ± 15
#
*
ab
Large ↘Moderate ↗5.0 ± 0.5
a
149.4 ± 5.3 7.2 ± 0.8
a
S
100%
83±23 9±2*
ab
76 ± 30
#c
Large ↘Large ↗4.0 ± 0.3
bc
141.0 ± 2.1
bc
7.5 ± 0.7
a
S
70%
/S
85%
/S
95%
/S
100%
: TS number 1, 2, 3 or 4.
Pre5: RMSSD before TS (ms).
Post5: RMSSD five minutes after TS (ms).
Post30: RMSSD 30 minutes after TS (ms).
RMSSD: root mean-square difference of successive normal R–R intervals (ms).
TL
HRV
: HRV coefficient for TL estimation (a.u.).
HRexe: mean HR of the TS (bpm).
RPE score: RPE for TS (1–10).
*p< 0.05: different from Pre5;
#
p< 0.05: different from Post5;
a
p< 0.05: different from S
70%
;
b
p< 0.05: different from S
85%
;
c
p< 0.05:
different from S
95%
.
Figure 1. Linear regression between exercise intensity and nor-
malised Post5 HRV values.
o: Individual values of each subject.
▬: Mean values of each training intensity.
Curving lines represent 90% of confidence interval.
Training load quantification with HRV 5
Downloaded by [UQ Library] at 07:32 08 February 2015
reported by previous studies in laboratory condi-
tions, HRV decrease immediately after the exercise
session (Post5) performed in field conditions is
closely related to, and enables evaluation of, exercise
intensity. Second, the present TL
HRV
index can
reflect the TL of aerobic exercise performed in field
conditions similarly to previous validated methods.
Immediate post-exercise HRV is related to exercise
intensity
During all TSs, from Pre5 to Post5, we observed a
significant decrease in RMSSD values. As observed
Figure 2. TL quantification of each TS calculated with the
methods of Banister, Foster and HRV (data are presented as
means and 90% confidence intervals).
S
70%
,S
85%
,S
95%
and S
100%
: TSs.
a
p< 0.05: different from S
70%
;
b
p< 0.05: different from S
85%
;
c
p< 0.05: different from S
95%
.
*(stars) represents a “large ES”of this significant difference.
□(square) represents a “moderate ES”of this significant
difference.
Figure 3. Bland and Altman plots for assessing agreement
between the three methods of TL quantification.
Since the three methods do not provide the same unit of TL, the
individual TL calculated by each method was expressed as a
percentage of the total TL corresponding to the some of the four
TSs; for example, % TL-S
70%
= 100 × [TL-S
70%
/(TL-S
70%
+
TL-S
85%
+ TL-S
95%
+ TL-S
100%
)].
6D. Saboul et al.
Downloaded by [UQ Library] at 07:32 08 February 2015
by other authors, each exercise induced a disturb-
ance of the homeostatic balance with an alteration of
autonomic cardiac control (Buchheit et al., 2007).
We did not find correlation between immediate post-
exercise HRV and exercise volume or between
immediate post-exercise HRV and exercise durations.
Although this study was not designed to test the
influence of exercise duration on HRV modifications,
previous works reported that increases exercise dura-
tion did not affect immediate or acute HRV recovery
(Stanley et al., 2013). Thus with respect to our study
conditions including TSs lasting less than 70 min-
utes, we can assume that exercise duration impacted
less Post5 HRV than intensity. Conversely, whatever
exercise type (i.e. continuous or intermittent), exer-
cise intensity was the main factor of the HRV
decrease observed between baseline and immediate
post-exercise values (Figure 1). As reported by several
laboratory studies, Post5 HRV value may reflect the
blood lactate concentration and thus exercise intens-
ity (Kaikkonen et al., 2010,2012; S. Seiler et al.,
2007). Unfortunately, blood lactate was not meas-
ured in the present study; however, it can be assumed
that HRV measurement performed immediately after
exercise could be a relevant and objective tool to
assess training intensity in field conditions.
Post-training HRV increase
Between Post5 and Post30 (acute recovery phase),
the significant RMSSD increase observed in all four
TSs may be explained by a reduction in cardiac
sympathetic activity with a simultaneous increase
in vagal nerve activation (Buchheit et al., 2007;
Kaikkonen et al., 2010). It seems that RMSSD
reactivation reflects the individual subject training
response in relation to the specificities and contents
of the exercise performed (Kaikkonen et al., 2012;S.
Seiler et al., 2007). Indeed, Post30 in S
100%
did not
present significant difference with Post30 in S
85%
while it was significantly lower in Post5. In addition,
Post30 in S
100%
was significantly greater than Post30
in S
95%
. These results suggest that the reactivation of
vagal modulation is a complex process and does not
only depend on exercise intensity. Indeed, despite
that Post5 was strongly linked to exercise intensity
it is important to note that Post30 HRV is not
correlated to either intensity or duration and training
volume (p> 0.05 for all). These findings are in
accordance with recent study who reported signific-
ant differences between Post30 HRV values of two
TSs performed at the same intensity (85% of MAS)
but with different methods (continuous vs.inter-
mittent; Kaikkonen et al., 2008). Finally, the post-
exercise RMSSD increase (i.e. from Post5 to
Post30) can be delayed depending on several
parameters, such as intensity, muscular and cardio-
vascular demands of exercise, as well as the fatigue
status or even the training level of the subject
(Aubertetal.,2003; Kaikkonen et al., 2012;
S. Seiler et al., 2007).
TL
HRV
calculation
Several parameters have to be taken into account
when assessing TL (Borresen & Lambert, 2009). As
described above, the different information pro-
vided by the RMSSD measures appears to be closely
related to the nature of the exercise. Therefore, we
propose a new TL
HRV
formula (equation 2) that
includes pre- and post-exercise HRV data. Interest-
ingly, none of the three HRV measurements of our
TL
HRV
marker taken individually is correlated with
either Banister or Foster TL. This suggests that the
inclusion of the three measures is important for the
validity of TL
HRV
. In addition, the ratio between
RMSSD decrease and post-exercise RMSSD
increase provides normalised values with respect to
inter-individual differences. The main interest of our
method includes the three-point HRV measures
which provide a wide range of information. First,
Pre5 HRV measurement performed at rest before
exercise represents baseline states that reflect current
or short-term fitness/fatigue status of athlete and
was generally use during athlete monitoring
(i.e. day-to-day HRV follow-up; Bosquet, Merkari,
Arvisais, & Aubert, 2008; Kiviniemi et al., 2010;
Plews, Laursen, Kilding, & Buchheit, 2013). Sec-
ond, Post5 HRV measurement which is strongly
linked to exercise intensity (Kaikkonen et al., 2010;
S. Seiler et al., 2007). Finally, Post30 HRV meas-
urement potentially reflects the acute athlete’s
recovery ability of ANS (S. Seiler et al., 2007).
TL
HRV
index vs. other TL methods
As shown in Figure 2, the new TL
HRV
provided TL
repartition between the four TS similar and the two
other methods. This visual observation is corrobo-
rated with the significant correlations obtained
between TL
HRV
indices and the two other methods.
However, Bland and Altman plots (Figure 3) showed
that the three methods have assessed different training
impacts for each TS. For example, Foster under-
estimated S
70%
compared to TL
HRV
and Banister
methods whereas Banister underestimated S
100%
compared to TL
HRV
and Foster methods. These
findings can be explained by the characteristics of
each method. Indeed, TRIMP may underestimate
the energetic and sympathetic stress of short high-
Training load quantification with HRV 7
Downloaded by [UQ Library] at 07:32 08 February 2015
intensity bouts (especially interval training; Borresen
& Lambert, 2009; K. S. Seiler & Kjerland, 2006)as
demonstrated by the similar TL given for S
70%
and
S
95%
. Indeed, these two sessions expressed the same
TRIMP since they have the same duration and HR
means (Borresen & Lambert, 2009; Lucía, Pardo,
Durántez, Hoyos, & Chicharro, 1998) whereas it is
obvious that S
95%
had a higher physiological impact
than S
70%
as reported by TL
HRV
and RPE methods.
TRIMP method is not valid for high-intensity interval
training (HIT) with short repetitions and high anaer-
obic contribution because of the shifted kinetic of HR
make totally ineffective (Buchheit & Laursen, 2013).
On the contrary, during such HIT, the acidosis
resulting from anaerobic metabolism solicitation
induces metaboreflex stimulation that can be quanti-
fied by HRV.
Using RPE scale, athletes may evaluate only the
difficulty of the body of the session while the Foster
method takes into account the total duration of
the TS (including warm-up and cool down at lower
intensities). This may lead to an overestimation of
high-intensity TSs assessed by Foster (Borresen &
Lambert, 2008,2009). In addition, because the
Foster method is subjective, it is also possible that
athletes may change their choice according to the
coach’s expectations (Borresen & Lambert, 2009;
Foster et al., 2001). Conversely, TL
HRV
was built to
assess TL with objective parameters like current
fitness/fatigue status (Pre5; Plews, Laursen, Kilding,
et al., 2013), exercise intensity (Post5; Kaikkonen
et al., 2010) and acute athlete’s recovery ability
(Post30; S. Seiler et al., 2007). The fact that TL
HRV
does not discriminate S
70%
vs.S
100%
and S
85%
vs.
S
95%
suggested that, despite different content, the
results of the homeostatic perturbations induced by
these TSs may be similar. Indeed, TL is modulated
by both intensity and duration of exercise and
despite different content (high intensity/short time
or low intensity/long time), TL of these sessions may
be identical.
Limitations, perspectives and practical applications
We should acknowledge that TL
HRV
method is more
complex than the Foster method for daily use.
Consequently, this study represents the first step
of a more global work aiming to demonstrate that
HRV could be used to quantify TL on the field.
Obviously, the future studies should rule out the
limitations that emerge from the present study more
particularly the limited sensitivity of intensity meas-
ured by Post5 HRV (see Figure 1). In addition, the
future experimentation should aim to test the
HRV measurement in standing position during an
active walking phase (Boullosa, Barros, Del Rosso,
Nakamura, & Leicht, 2014) in the context of
TL
HRV
. Then, other studies should shorten the
protocol to obtain a more simple method for daily
monitoring. In this context, the linearity of the post-
HRV reactivation observed during the first hour may
justify a post-exercise recording reduced to 10
minutes (Casties et al., 2006). Lastly, the TL
HRV
will have to be validated on a larger range of training
modalities [e.g. resistance training that may also be
assessed by post-HRV measurements where non-
linear indexes have been shown to be relevant
(Caruso et al., in press)].
From a practical point of view, this new TL
HRV
marker may also be used by elite athletes during
routine TSs (performed at the end of each month or
training cycle) to objectively and simply measure
their current fitness/fatigue status. Indeed, consider-
ing that the parasympathetic reactivation is closely
related to the current athlete level and/or fatigue
status (Buchheit et al., 2007; S. Seiler et al., 2007),
we can assume that TL
HRV
measurement performed
regularly with exactly the same TS may provide
information on the current fitness level of an athlete.
In this sense, future investigations will be conducted
to verify this relation.
Conclusion
The main purpose of the present study was to define
a new method for quantifying TL by using pre- and
post-exercise RMSSD measurements in field condi-
tions. TL
HRV
may provide objective and rational
information about the current intensity of the exer-
cise but also on the TL in line with the two main
validated methods (i.e. Foster and Banister). It is
also the first study to test HRV tools (i.e. with
formula that provide numeric and rational data) in
relation to TL.
Acknowledgement
The authors would like to thank the SUUNTO
company for its material support (Hervé Riffault).
References
Aubert, A. E., Seps, B., & Beckers, F. (2003). Heart rate variability
in athletes. Sports Medicine,33, 889–919. doi:10.2165/000072
56-200333120-00003
Banister, E. W., Good, P., Holman, G., & Hamilton, C. (1986).
Modeling the training response in athletes. Paper presented at the
The 1984 Olympic Scientific Congress Proceedings sport and
elite performers. Retrieved from http://www.ncbi.nlm.nih.gov/
entrez/query.fcgi?cmd=Retrieve& db=PubMed& dopt=Citation&
list_uids=6778623
Borresen, J., & Lambert, M. I. (2008). Quantifying training load:
A comparison of subjective and objective methods. International
Journal of Sports Physiology and Performance,3(1), 16–30.
8D. Saboul et al.
Downloaded by [UQ Library] at 07:32 08 February 2015
Borresen, J., & Lambert, M. I. (2009). The quantification of
training load, the training response and the effect on perform-
ance. Sports Medicine,39, 779–795. doi:10.2165/11317780-
000000000-00000
Bosquet, L., Merkari, S., Arvisais, D., & Aubert, A. E. (2008). Is
heart rate a convenient tool to monitor over-reaching? A
systematic review of the literature. British Journal of Sports
Medicine,42, 709–714. doi:10.1136/bjsm.2007.042200
Boullosa, D. A., Barros, E. S., Del Rosso, S., Nakamura, F. Y., &
Leicht, A. S. (2014). Reliability of heart rate measures during
walking before and after running maximal efforts. International
Journal of Sports Medicine,35, 999–1005. doi:10.1055/s-0034-
1372637
Buchheit, M., Chivot, A., Parouty, J., Mercier, D., Al Haddad,
H., Laursen, P. B., & Ahmaidi, S. (2010). Monitoring
endurance running performance using cardiac parasympathetic
function. European Journal of Applied Physiology,108, 1153–
1167. doi:10.1007/s00421-009-1317-x
Buchheit, M., & Laursen, P. B. (2013). High-intensity interval
training, solutions to the programming puzzle Part I: Cardio-
pulmonary emphasis. Sports Medicine,43, 313–338.
doi:10.1007/s40279-013-0029-x
Buchheit, M., Laursen, P. B., & Ahmaidi, S. (2007). Parasym-
pathetic reactivation after repeated sprint exercise. AJP: Heart
and Circulatory Physiology,293(1), H133–H141. doi:10.1152/
ajpheart.00062.2007
Caruso, F. C., Arena, R., Phillips, S. A., Bonjorno Junior, J. C.,
Mendes, R. G., Arakelian, V. M., …Borghi-Silva, A. (in
press). Resistance exercise training improves heart rate variab-
ility and muscle performance: A randomized controlled trial in
coronary artery disease patients. European Journal of Physical
and Rehabilitation Medicine. Retrieved from http://www.ncbi.
nlm.nih.gov/pubmed/25384514
Casties, J.-F., Mottet, D., & Le Gallais, D. (2006). Non-linear
analyses of heart rate variability during heavy exercise and
recovery in cyclists. International Journal of Sports Medicine,27,
780–785. doi:10.1055/s-2005-872968
Chalencon, S., Busso, T., Lacour, J.-R., Garet, M., Pichot, V.,
Connes, P., …Barthélémy, J. C. (2012). A model for the
training effects in swimming demonstrates a strong relationship
between parasympathetic activity, performance and index
of fatigue. PLoS One,7(12), e52636. doi:10.1371/journal.
pone.0052636.s003
Chandola, T., Heraclides, A., & Kumari, M. (2010). Psychophy-
siological biomarkers of workplace stressors. Neuroscience and
Biobehavioral Reviews,35(1), 51–57. doi:10.1016/j.neubiorev.
2009.11.005
Foster, C. (1998). Monitoring training in athletes with reference to
overtraining syndrome. Medicine and Science in Sports and Exercise,
30, 1164–1168. doi:10.1097/00005768-199807000-00023
Foster, C., Heimann, K., Esten, P., Brice, G., & Porcari, J.
(2001). Differences in perceptions of training by coaches and
athletes. South African Journal of Sports Medicine,8(2), 3–7.
Hautala, A. J., Kiviniemi, A. M., & Tulppo, M. P. (2009).
Individual responses to aerobic exercise: The role of the
autonomic nervous system. Neuroscience and Biobehavioral
Reviews,33(2), 107–115. doi:10.1016/j.neubiorev.2008.04.009
Hellard, P., Avalos, M., Lacoste, L., Barale, F., Chatard, J.-C., &
Millet, G. P. (2006). Assessing the limitations of the Banister
model in monitoring training. Journal of Sports Sciences,24,
509–520. doi:10.1080/02640410500244697
Hopkins, W. G. (1991). Quantification of training in competitive
sports. Methods and applications. Sports Medicine,12, 161–183.
doi:10.2165/00007256-199112030-00003
Kaikkonen, P., Hynynen, E., Mann, T., Rusko, H., & Nummela, A.
(2010). Can HRV be used to evaluate training load in constant
load exercises? European Journal of Applied Physiology,108,435–
442. doi:10.1007/s00421-009-1240-1
Kaikkonen, P., Hynynen, E., Mann, T., Rusko, H., & Nummela, A.
(2012). Heart rate variability is related to training load variables
in interval running exercises. European Journal of Applied Physi-
ology,112,829–838. doi:10.1007/s00421-011-2031-z
Kaikkonen, P., Nummela, A., & Rusko, H. (2007). Heart rate
variability dynamics during early recovery after different endur-
ance exercises. European Journal of Applied Physiology,102(1),
79–86. doi:10.1007/s00421-007-0559-8
Kaikkonen, P., Rusko, H., & Martinmäki, K. (2008). Post-
exercise heart rate variability of endurance athletes after
different high-intensity exercise interventions. Scandinavian
Journal of Medicine and Science in Sports,18, 511–519.
doi:10.1111/j.1600-0838.2007.00728.x
Karvonen, J., & Vuorimaa, T. (1988). Heart rate and exercise
intensity during sports activities. Practical application. Sports
Medicine,5, 303–311. doi:10.2165/00007256-198805050-
00002
Kiviniemi, A. M., Hautala, A. J., Kinnunen, H., Nissilä, J.,
Virtanen, P., Karjalainen, J., & Tulppo, M. P. (2010). Daily
exercise prescription on the basis of HR variability among men
and women. Medicine and Science in Sports and Exercise,42,
1355–1363. doi:10.1249/MSS.0b013e3181cd5f39
Leger, L., & Boucher, R. (1980). An indirect continuous running
multistage field test: The Universite de Montreal track test.
Canadian Journal of Applied Sport Sciences,5(2), 77–84.
Lucía, A., Pardo, J., Durántez, A., Hoyos, J., & Chicharro, J. L.
(1998). Physiological differences between professional and elite
road cyclists. International Journal of Sports Medicine,19, 342–348.
doi:10.1055/s-2007-971928
Martinmäki, K., & Rusko, H. (2008). Time-frequency analysis of
heart rate variability during immediate recovery from low and
high intensity exercise. European Journal of Applied Physiology,
102, 353–360. doi:10.1007/s00421-007-0594-5
Plews, D. J., Laursen, P. B., Kilding, A. E., & Buchheit, M.
(2012). Heart rate variability in elite triathletes, is variation in
variability the key to effective training? A case comparison.
European Journal of Applied Physiology,112, 3729–3741.
doi:10.1007/s00421-012-2354-4
Plews, D. J., Laursen, P. B., Kilding, A. E., & Buchheit, M.
(2013). Evaluating training adaptation with heart rate mea-
sures: A methodological comparison. International Journal of
Sports Physiology and Performance,8(6), 688–691. Retrieved
from http://www.ncbi.nlm.nih.gov/pubmed/23479420.[Epub
2013 Mar 8].
Plews, D. J., Laursen, P. B., Kilding, A. E., & Buchheit, M.
(2014). Heart rate variability and training intensity distribution
in elite rowers. International Journal of Sports Physiology and
Performance,9(6), 1026–1032. doi:10.1123/ijspp.2013-0497
[Epub 2014 Apr 3].
Plews, D. J., Laursen, P. B., Stanley, J., Kilding, A. E., & Buchheit,
M. (2013). Training adaptation and heart rate variability in
elite endurance athletes: Opening the door to effective monitoring.
Sports Medicine,43, 773–781. doi:10.1007/s40279-013-0071-8
Saboul, D., Pialoux, V., & Hautier, C. (2013). The impact of
breathing on HRV measurements: Implications for the lon-
gitudinal follow-up of athletes. European Journal of Sport
Science,13(5), 534–542. doi:10.1080/17461391.2013.767947
[Epub 2013 Feb 8].
Saboul, D., Pialoux, V., & Hautier, C. (2014). The breathing
effect of the LF/HF ratio in the heart rate variability measure-
ments of athletes. European Journal of Sport Science,14(Suppl. 1),
S282–S288. doi:10.1080/17461391.2012.691116 [Epub 2012
May 31].
Seiler, K. S., & Kjerland, G. O. (2006). Quantifying training
intensity distribution in elite endurance athletes: Is there
evidence for an “optimal”distribution? Scandinavian Journal
of Medicine and Science in Sports,16(1), 49–56. doi:10.1111/
j.1600-0838.2004.00418.x
Training load quantification with HRV 9
Downloaded by [UQ Library] at 07:32 08 February 2015
Seiler, S., Haugen, O., & Kuffel, E. (2007). Autonomic recovery
after exercise in trained athletes: Intensity and duration effects.
Medicine and Science in Sports and Exercise,39, 1366–1373.
doi:10.1249/mss.0b013e318060f17d
Stanley, J., Peake, J. M., & Buchheit, M. (2013). Cardiac
parasympathetic reactivation following exercise: Implications
for training prescription. Sports Medicine,43, 1259–1277.
doi:10.1007/s40279-013-0083-4
Stepto, N. K., Hawley, J. A., Dennis, S. C., & Hopkins, W. G.
(1999). Effects of different interval-training programs on
cycling time-trial performance. Medicine and Science in Sports
and Exercise,31, 736–741. doi:10.1097/00005768-199905000-
00018
Task-Force. (1996). Heart rate variability. Standards of measure-
ment, physiological interpretation, and clinical use. Task Force
of the European Society of Cardiology and the North American
Society of Pacing and Electrophysiology. European Heart
Journal,17, 354–381.
10 D. Saboul et al.
Downloaded by [UQ Library] at 07:32 08 February 2015
