Vol. 92, Issue 2, 572-580, February 2002
Effects of training frequency on the dynamics of performance
response to a single training bout
Thierry
Busso1,
Henri
Benoit1,
Régis
Bonnefoy1,
Léonard
Feasson1, and
Jean-René
Lacour2
1 Groupe Physiologie et Physiopathologie de l'Exercice et
Handicap, Groupement d'Intérêt Public "Exercice,"
Université Saint-Etienne, 42023 Saint-Etienne cedex 2; and
2 Laboratoire de Physiologie, Faculté de Médecine
Lyon Sud, Université Lyon 1, 69921 Oullins cedex, France
 |
ABSTRACT |
The aim of this study was to analyze
the effect of an increase in training frequency on exercise-induced
fatigue by using a systems model with parameters free to vary over
time. Six previously untrained subjects undertook a 15-wk
training experiment composed of 1) an 8-wk training period
with three sessions per week (low-frequency training), 2)
1 wk without training, 3) a 4-wk training period with
five sessions per week [high frequency training (HFT)], and 4) 2 wk without training. The systems input
ascribed to training loads was computed from interval exercises and
expressed in arbitrary units. The systems output ascribed to
performance was evaluated three times each week using maximal power
sustained over 5 min. The time-varying parameters of the model were
estimated by fitting modeled performances to the measured ones using a
recursive least squares method. The variations over time in the model
parameters showed an increase in magnitude and duration of fatigue
induced by a single training bout. The time needed to recover
performance after a training session increased from 0.9 ± 2.1 days at the end of low-frequency training to 3.6 ± 2.0 days at
the end of HFT. The maximal gain in performance for a given training
load decreased during HFT. This study showed that shortening recovery time between training sessions progressively yielded a more persistent fatigue induced by each training.
mathematical model; overtraining; fatigue; adaptation
| |
INTRODUCTION |
IT IS GENERALLY ASSUMED
THAT training adaptations occur subsequent to exercise-induced
modifications of cellular homeostasis. These exercise-induced changes
would be the main stimulus driving the physiological responses leading
to the body's adaptations (25, 26). Nevertheless,
overloading for a long period might produce the persistent fatigue
generally associated with so-called overtraining (10, 15, 16,
18). Short-term overtraining or overreaching should be
distinguished from long-term overtraining, also called overtraining
syndrome or staleness. Overreaching is characterized by a transient
decline in performance that, after few days or few weeks of less
intensive training, can be recovered, sometimes at a level even higher
than that attained before the overload. Inversely, overtraining
syndrome or staleness is characterized in athletes by an incapacity to
train and perform over a longer period. A few weeks or months without
intensive training would be necessary to recover normal training
capacity. Although very little is known about the physiological
mechanisms of overreaching and staleness, there is evidence that an
imbalance between training loads and recovery time could be a major
cause of the temporary incapacity to perform.
To analyze the adaptations occurring during physical training and
exercise-induced fatigue, several studies have used various systems
modeling the effects of training on performance (1-8, 20,
22). These studies have considered performance as a systems output varying over time according to the training loads (i.e., systems
input). According to systems modeling, performance is mathematically
related to quantified training via a transfer function, including two
first-order filters. The first filter with a positive gain is ascribed
to training adaptations. The second filter with a negative gain is
ascribed to the fatiguing effects of exercise. The shorter time
constant of the negative function, compared with the positive function,
accounts for the transient decline in performance after exercise
completion. The model parameters (gain terms and time constants) are
determined by fitting model performances to experimental data.
Comparison of the model parameters has shown differences that would
arise from differences in the training regimen (6).
Differences in the time needed to recover performance after training
completion were reported with values ranging from 1-3 days for
subjects training four times a week (5) to 23 days for an
elite athlete training once or twice a day (4). These
differences could arise from the frequency and the size of the training
loads. The repetition of training loads could alter the response to a
given training dose. If exerc
ise occurs before complete recovery from previous exercise, the time
necessary to recover could be longer than if there were a longer gap
between the two training loads. Exercise-induced fatigue would thus be amplified by the repetition of stressful training bouts. To analyze more precisely the influence of training intensity on the time needed
to recover, the systems model described above was modified in earlier
work (6). A recursive least squares algorithm, used to
introduce time-dependent variations into model parameters, was applied
to data of strenuous training performed by two volunteers. This study
showed that time-dependent variations in model parameters generally
would not arise from noise in data. As suggested above, the observed
changes in model parameters pointed out a possible increase in the
magnitude and time course of long-term fatigue with repeated training
loads (6).
The purpose of the present investigation was to analyze the dynamic
variations of performance response to exercise with stepwise changes in
training frequency. Changes over time in the responses to training
loads were assessed using the systems model with time-varying parameters. More precisely, the aim of this study was to determine whether an increase in training frequency and thus a decrease in
recovery time between training sessions would induce a progressive increase in the magnitude and duration of long-term fatigue induced by
an identical training load.
| |
METHODS |
Subjects and experimental methods.
Informed consent to participate in the study was obtained from six
healthy men. All were sedentary or involved in recreational activities,
and none was engaged in physical activity on a regular basis for at
least 6 mo before the experiment. The protocol was approved by the
local ethics committee (Conseil Consultatif de la Protection des
Personnes dans la Recherche Biomédicale de la Loire). Mean age,
weight, and height were, respectively, 32.7 ± 5.0 (SD) yr,
83.5 ± 12.6 kg, and 182 ± 8 cm.
Throughout the experiment, the subjects performed incremental tests
until exhaustion to measure their maximal oxygen consumption (
O2 max) and trials to determine the
maximal power that they could sustain for 5 min (Plim 5').
O2 max was measured during incremental
exercise on a cycle ergometer (model 818, Monark, Stockholm, Sweden).
Subjects warmed up for 5 min at a work rate ranging from 100 to 150 W,
according to their fitness. The work rate was then increased every 2 min by 30 W until exhaustion. A 20-W increment was used for the
preexperiment measurement in one subject who had a maximal aerobic
power (MAP) equal to 200 W. The subjects breathed through a two-way
non-rebreathing valve (model 2700, Hans Rudolph, Kansas City, MO). The
expiratory gases were collected in a polyethylene bag (HP Production,
Saint-Etienne, France) during the last 30 s of each 2-min bout.
Gas composition was analyzed using a paramagnetic analyzer for
O2 (Servomex Serie 1440, Crowborough, UK) and an infrared
analyzer for CO2 (Datex Normocap, Helsinki, Finland).
The gases in the bag were emptied in a Tissot spirometer
(Techmachine Gymrol, Andrézieux, France) to measure minute
ventilation. The external MAP was computed from power output and
duration of the last increment as proposed by Kuipers et al.
(17). Three minutes after cessation of exercise, a
fingertip blood sample was taken to be analyzed for lactate concentration (YSI 2300, Yellow Springs Instruments, Yellow Springs, OH).
The trial to measure Plim 5' consisted of a 10-min warm-up
and then an all-out exercise over 5 min on a cycle ergometer (model
829E, Monark). Breaking force was predetermined, and the pedaling
frequency was adapted by the subject according to his own
possibilities. The breaking force during the warm-up was equal to 50%
of the breaking force during the subsequent 5-min test. To obtain a
pedaling frequency around 70-80 rpm during the 5-min test, the
breaking force was estimating using either the limit power obtained
from the previous test or the MAP from the first test. The power output
developed by the subject during the trial was registered throughout via
an interface between the cycle ergometer and the computer.
Plim 5' was determined as the average power output
sustained over the test.
During the 2 wk preceding the training intervention,
O2 max was measured, and the subjects
performed three trials to measure Plim 5'. The training
intervention was then composed of four periods: 1) an 8-wk
period with three training sessions per week [low-frequency training
(LFT), weeks 1-8]; 2) a 1-wk period without
training (week 9); 3) a 4-wk period with five
training sessions per week [high-frequency training (HFT), weeks
10-13]; and 4) a 2-wk period without training
(weeks 14 and 15). During LFT, the three training
sessions were generally separated by 2 days without training. On each
day of training, the subjects performed first one test to measure
Plim 5', and, after 15 min of rest, they trained on a
cycle ergometer using intermittent exercise with 5 min of work,
interspersed with 3 min of active recovery, repeated four times.
Exercise intensity was prescribed to 85% of the last measured
Plim 5'. During HFT, the subjects trained for 5 days
consecutively per week. On Monday, Wednesday, and Friday, they
performed the same training session as during LFT. On Tuesday and
Thursday, the subjects did not perform the test to measure
Plim 5' but repeated the training sequence five times
instead of four. During the week between LTF and HFT, the subjects
performed two tests to measure Plim 5' and one test to
measure O2 max. During the period
subsequent to HFT, Plim 5' was measured three times during
the first week and twice the second week.
O2 max was measured at the end of the
last week of the experiment. These two
O2 max tests were done 2 days after the
last Plim 5' measurement. Therefore, Plim 5'
was measured two or three times for each week of the experiment, and
O2 max was measured on three separated
occasions: before the experiment, after LFT, and after HFT.
Criterion performance and quantification of training.
Plim 5' was used as a criterion of performance for
modeling. The amount of training was quantified from work done during
training and trials. The daily quantity of training was computed as a
function of the exercise duration and the intensity referred to
Plim 5'. The work done during warm-up and recovery was not
considered in the computation. The test to measure Plim 5'
(i.e., 100% intensity) was ascribed to 100 training units. For
training sessions, each 5-min bout of exercise was weighted by
intensity of the effort (power output/Plim 5' × 100). For example, the work done during a training session composed of
four repetitions of a 5-min effort at 85% of Plim 5' would be 4 × 85 = 340 training units. For the test to
measure O2 max, the amount of training
was set at 100 units as for the trial to measure Plim 5'.
Modeling training effects on performance.
The model used in this study was entirely described in a previous study
(6). The model is based on a systems model initially proposed by Banister et al. (1). The subject is
represented by a system with a daily amount of training as input and
performance as output. The system operates in accordance with a
transfer function resulting from the sum of two first-order filters
whose impulse response is given by
|
(1)
|
where g(t) is the impulse response of the transfer
function, k1 and k2 are
gain terms, t is time, and 
1 and
2 are time constants. Figure
1 shows that the impulse response of
performance in case k2 is greater than that in
case k1 and that 1 is greater than 2. After training completion, the decline in
performance is given by k2 
k1. Afterward, recovery allows the performance to
return to its initial value (time noted tn) and
then to peak at a maximal value (time noted tg).
The maximal gain in performance tg days after a
training load of 1 training unit is noted pg. These
notations were chosen according to previous studies (6, 9).
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 1.
Response of performance to a training dose of 1 unit.
pg, Maximal gain in performance; k1
k2, difference between gain terms for
positive component (k1) and negative component
(k2) of the model; tn,
time needed to recover performance after training completion;
tg, time needed to reach maximal performance
after training completion.
|
|
The time functions of performance [p(t)] and training
[w(t)] are mathematically related as
|
(2)
|
where p* is an additive term that depends on the initial
training status of the subject and * denotes the product of convolution.
The definition of the convolution product leads to
|
(3)
|
Discretization of Eq. 3 results in an estimation of
the model performance on day n
(
n) from the successive training loads
wi, with i varying from 1 to
n 1
|
(4)
|
For the time-varying model, the parameters were fitted by using
a recursive least squares algorithm with an exponential window (6). The model parameters were evaluated on each day that
performance was measured. The parameters at a given time were estimated
from the previous and present data. On day n, the parameters
are obtained by minimizing the following recursive function
|
(5)
|
where 
is a constant with a value ranging between 0 and 1.
Sn was minimized for each day when actual
performance was measured. The model parameters for the day n
were estimated with successive minimization of
Sn using a grid of values for the time
constants: from 30 to 60 days for 1 and from 1 to 20 days for 2. The model parameters
(k1, k2,
1, 2) were initialized with the six
values for performance measured during the preexperiment and the first
week of training using the least squares method. The value of for
the set of subjects was chosen so that the mean SE of the fit remained
higher than the mean intraindividual variability of
Plim 5' assessed from the SD values computed before the
training experiment.
The tn was defined as the time needed after an
impulse training stimulus for the effects of fatigue to be dissipated
sufficiently to allow the effects of training to return performance to
the pretraining level. Thereafter, the performance exceeds its
pretraining level. The tn was estimated by
|
(6)
|
The time needed to reach maximal performance after an impulse
training stimulus (tg) was estimated as follows
|
(7)
|
The maximal gain in performance due to 1 unit of training
(pg) was estimated as follows
|
(8)
|
An additional analysis was undertaken to assess the specific
effect of variations in time constants on results. For this, variations
in performance were modeled with fixed time constants: 30 days for
1 and 10 days for 2. In this computation,
only the gain terms were free to vary over time, using the same
procedure as described above.
Statistics.
Means, SD, and SE were computed for the selected variables, and the
coefficients of variation (%CV) were computed for each subject for
Plim 5' measured before the beginning of training. One-way
ANOVA was used to test differences in the studied variables occurring
during the experiment. The parameters measured during the
O2 max test and the closest measurement of Plim 5' (preexperiment, after LFT and after HFT) were
compared with ANOVA and the Scheffé post hoc test. The
differences in model parameters (k1,
k2, 1, 2,
tn, tg,
k2 k1, and
pg) were examined after averaging the values over each week
of the experiment. To discard the variability during the first weeks due to the initialization of the model parameters, ANOVA was applied to
the values from the second week of the experiment. The variations in
model parameters were examined first over LFT (weeks
2-8) and then over the entire experiment (weeks
2-15). The variations over time were examined with contrast
analysis (Scheffé's method) to compare each value for
weeks 9-15 with the mean value over weeks
2-8.
| |
RESULTS |
Measurement variability was estimated from the
Plim 5' tests run during the 2 wk preceding the experiment
and the test done on the first day of LFT (i.e., before any training). The first measurement was discarded to take learning into account. The
variability for the three remaining values ascribed to the CV
(SD/mean × 100) was 1.58 ± 0.81% (SD = 4.28 ± 1.94 W).
Table 1 shows the gain in MAP and
Plim 5' after LFT and HFT.
O2 max increased by 20.5 ± 7.0%
after LFT (P < 0.001) and by 22.6 ± 5.0% after
HFT (P < 0.001), compared with preexperiment levels.
MAP increased by 25.7 ± 12.1 and 28.6 ± 13.7% after LFT
and HFT, respectively (P < 0.001). The time course of
Plim 5' for each subject is given in Fig. 2. When the closest measurements to each
O2 max test are analyzed, the
statistical analysis showed a significant increase in
Plim 5' compared with preintervention values: 26.9 ± 6.7% after LFT and 29.5 ± 5.3% after HFT (P < 0.001). The postexercise blood lactate concentration also increased
significantly after LFT (P < 0.001) and after HFT
(P < 0.01). There was no statistical difference
between the values obtained after LFT and HFT in any of the above
variables. However, peak heart rate decreased significantly after HFT
compared with the preintervention value (P < 0.05).
View larger version (36K):
[in this window]
[in a new window]
|
Fig. 2.
Performance fit for the model with gain terms and time constants
free to vary over time. A-F: subjects
1-6, respectively. Plim 5', maximal power
sustained for 5 min; Pre, before training; LFT, low-frequency training;
HFT, high-frequency training.
|
|
Fit of performances measured throughout the study for each subject to
the model with time-varying parameters is illustrated in Fig. 2. For
all of the subjects, the fit was better during LFT than during HFT when
the day-to-day variations of Plim 5' were greater. Table
2 gives the statistics of the fit of the model with time-varying parameters using a value of 0.95 for the weighting factor . The coefficient of determination ranged from 0.957 to 0.982, and the SE was 4.76 ± 0.84 W. This 0.95 value for
was retained because the SE of the fit remained higher than the
average SD of Plim 5' measurements done before training.
Figure 3 shows the variations over time
of the model parameters (k1,
k2, 1, and 2) and
their derived variables (tn,
tg, k2 k1, and pg). When only LFT (weeks
2-8) is considered, no statistical difference was observed
for any variable. When the values from weeks 2-15 were
analyzed using ANOVA, statistical differences were observed for
tn, k2 k1, and pg. The difference between the two
gains k1 and k2 was
greater than during LFT in week 11 (P < 0.05), week 12 (P < 0.001), and week
13 (P < 0.05). The time necessary to recover
performance after training completion, tn,
increased significantly during HFT. With a mean value lower than 1 day
during LFT (0.9 ± 2.1 days in week 8),
tn increased to 2.3 ± 1.8 days in
week 11 (P < 0.05), 3.5 ± 1.6 days in
week 12 (P < 0.001), and 3.6 ± 2.0 days in week 13 (P < 0.001). The maximal
gain in performance after a training dose of 1 unit, pg, decreased progressively during HFT. However, the decrease was statistically significant only for weeks 14 and
15 after HFT (P < 0.01). The increase in
tg, the time necessary for performance to reach
its maximal level after training completion, was not statistically
significant. To illustrate these variations in model parameters, the
impulse responses of performance (i.e., variations over time of
performance after a training dose of 1 unit) using mean model
parameters for the first week of HFT (week 10) and the third
and fourth weeks of HFT (weeks 12 and 13) are
compared in Fig. 4. The comparison of
these responses showed 1) an increase in the decrement in
performance after training completion for weeks 12 and
13 compared with week 10, 2) a
progressive increase in the time needed to recover the initial level of
performance from week 10 to 13, and 3)
a decrease in the maximal gain in performance after recovery in
week 13, although this decrease appeared to be statistically
significant only 1 wk later.
View larger version (35K):
[in this window]
[in a new window]
|
Fig. 3.
Variations in parameter estimates and derived variables (mean value
per week ± SE) when a model with time-varying time constants and
gain terms is used. A: time constant 1.
B: time constant 2. C:
k1. D: k2.
E: tg. F:
tn. G: pg. H:
k2 k1. au,
Arbitrary units. Significant difference: * P < 0.05; ** P < 0.01; *** P < 0.001.
|
|
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 4.
Response of performance to a single training dose of 1 unit computed from mean model parameters estimated in weeks
10, 12, and 13 using a model with
time-varying gain terms and time constants.
|
|
As a final study, the time-varying model was applied to the data with
only the gain terms free to vary over time and the time constants fixed
to constant values: 30 days for 1 and 10 days for
2. The value for was set at 0.94 to obtain a SE
close to the results of the model with time-varying gains and time
constants. Indeed, the coefficient of determination ranged from 0.958 to 0.980 and the SE was 4.81 ± 0.68 W (Table 2). The variations over time of the gains and the derived variables are depicted in Fig.
5. Despite the lack of variation in time
constants, the gain terms and the derived variables
(tn, tg,
k2 k1, and
pg) showed patterns similar to those with the free-varying
time constant model. Additionally, a relationship between gain terms
was observed for each subject when the values averaged over each week
from week 2 were used. The negative gain
k2 was correlated with the positive gain
k1 in each of six subjects, with
r2 ranging from 0.66 to 0.97 (P < 0.001). The mean linear regression coefficients were 1.837 ± 0.363 for the slope and 0.023 ± 0.012 for the intercept.
View larger version (32K):
[in this window]
[in a new window]
|
Fig. 5.
Variations in parameter estimates and derived variables (mean value
per week ± SE) when a model with time-varying gain terms and time
constants set at 30 days for 1 and 10 days for
2 is used. A: k1.
B: k2. C:
tg. D:
tn. E: pg. and
F: k2 k1. Significant difference:
* P < 0.05; *** P < 0.001.
|
|
| |
DISCUSSION |
The major finding of this study was that an increase in training
frequency induced changes in the dynamics of performance response to a
single training bout. There was 1) an increase in the
magnitude and duration of the fatiguing effect for a single training
session and 2) a decrease in exercise-induced adaptation assessed by the maximal gain in performance for a given amount of training.
The time course of performance during the experiment assessed by the
limit power over 5 min showed a regular increase during the 8 wk of
LFT. During HFT, in which recovery time between training sessions was
diminished, day-to-day variations were clearly observed. The 26.9 ± 6.7% increase in Plim 5' after LFT over the
preexperiment value was comparable to the 25.7 ± 12.1% increase for MAP. However, the increase in the time trial result was slightly greater than the 20.5 ± 7.0% gain in
O2 max. Although O2 max appeared to be the main
determinant, other factors played a role in the gain in
Plim 5'. This greater improvement in Plim 5'
than in O2 max could be explained by an
amelioration of the net efficiency or a greater contribution with
training of the anaerobic metabolic pathway. An increase in the
anaerobic contribution would be in accordance with the increase in peak
blood lactate after the O2 max trial
(15.5 ± 1.4 mM after LFT vs. 11.5 ± 1.3 mM during
preexperiment; P < 0.001). These results were in line
with those obtained after HFT, although no statistically significant
further improvement was observed.
The systems model with time-varying parameters was applied with a
recursive least squares algorithm (6). The 0-1 range assigned to the recursive algorithm's parameter enabled changes in
model parameters. Setting too small would allow rapid changes in
model parameters, making them overly sensitive to noise in performance
evaluation. Inversely, if is too large, the ability of the process
to follow time variations in the parameters would be limited. To
optimize the process of following parameter variations and avoiding
oversensitivity to noise, the value of was chosen to correspond to
the variability of Plim 5' measured during the first runs
before training was begun. Preexperiment measurements of
Plim 5' showed a 4.28 ± 1.94 W intraindividual variability after the first value was discarded to take learning into
account. The 1.6 ± 0.8% CV is in keeping with the 1-3.5% range reported in the literature for time trials lasting 15-60 min
(11, 13, 24). This kind of trial (maximal power over a
given time) should give more reproducible results than expected for the
measurement of the time that a subject can sustain a given power. Poor
reproducibility with CV ranging from 17 to 27% has been reported for
such constant-load trials (13, 14, 19). To limit the
influence of noise on variations over time in model parameters, the
value of 0.95 was finally retained for because it yielded a mean SE
for performance (4.76 ± 0.84 W) slightly greater than its
variability without training.
The variations in model parameters over the experiment showed similar
patterns in all six subjects. No statistical variations were observed
during LFT when 2 or 3 days of recovery separated two successive
training sessions. Changes in model parameters appeared when recovery
time decreased to 1 day between the exercise stimuli (HFT). The changes
in time constants did not reach the level of statistical significance,
despite large variations. The 1 decreased systematically
in each subject over the second part of the experiment until the limit
value of 30 days. The 2 increased over HFT with an
important reduction of interindividual variability. It is difficult to
address adequately the possible implications of the variations in time
constants. To assess the importance of these changes, the results were
compared with those obtained using a model with only gain terms free to
vary over time and fixed time constants. Their values were kept
constant over the experiment to arbitrary values of 30 days for
1 and 10 days for 2. The variations in
gain terms and the derived variables exhibited very similar patterns
when time constants were used that were free to vary over time or when
they were kept constant. Using the model with only time-varying gain
terms would thus be helpful in further investigations. However, it is
difficult to interpret the correlation observed between the two gain
terms. The structure of the model being the sum of two exponentials,
variations in k1 might compensate in part for
variations in k2. The more composite variables
derived from model parameters would provide a clearer picture of the
response to exercise. With the use of time-varying gain terms and time
constants, statistical differences were observed between LFT and HFT
for 1) the difference between the two gains k2 and k1, which reflects
the magnitude of the decrement in performance after exercise
completion, and 2) tn, which is the
time needed to recover performance after exercise completion. These
differences were statistically significant for the second to fourth
weeks of HFT compared with mean values for LFT. The pg
decreased progressively during HFT. However, this decrease became
statistically significant only during the subsequent 2 wk without training.
Our observation that more frequent training yields a greater fatiguing
effect is in accordance with data in the literature about modeling of
training effects on performance. As noted in a previous study
(6), differences in time-invariant parameters of the
systems model would arise from differences in training intensity. The
time to recover performance after exercise, assessed from model
fitting, ranged from 1-3 days for subjects performing endurance
training four times a week (5) to 23 days for an elite
explosive athlete who trained once or twice a day (4). Values of 12.2 ± 5.7 days have been reported for elite swimmers (22) and 8 and 11 days for two subjects training once a
day for 28 days (20). These differences in time to recover
performance after training completion could be related to workload and
training frequency (6). The results of this study are in
line with these conclusions. The increase in training frequency yielded
a significant increase in the time necessary to recover performance
after a single training bout. The values obtained for
tn, when the subjects trained 5 days a week,
were lower than data in the literature for the time-invariant model.
The lower workloads compared with the previous experiments could
explain this gap. The tn values increased up to
10 and 14 days for the two subjects studied in our laboratory's
previous study (6). Although these two subjects also
trained 5 consecutive days per week, their daily training doses were
about twice those in the present investigation. High-dose training
added to HFT could thus explain the high values for
tn in athletes or during intensive training
reported in the above-cited studies. Data reported for
tn computed with time-varying or time-invariant models are in accordance with short-term overtraining lasting a few
days to a few weeks (10, 15, 16, 18). However, considering their initial fitness, the subjects of this study could behave differently from well-trained elite athletes. Higher fit subjects could
better tolerate the training sequences used in this study. The data of
this study could not thus be directly translated to elite athletes.
Nevertheless, the high tn values reported above in elite athletes would not be simply the time necessary to recover after a single training bout but the result of the repetitions of
training loads. The new insight arising from the results of this study
is that long-duration fatigue associated with short-term overtraining
would be a progressive process in which the magnitude and duration of
exercise-induced fatigue would increase with the repetition of the
training doses. The time needed to recover could thus progressively
increase up to 2-3 wk as reported for strenuous training. This
period, ranging from a few days to a few weeks, would be the ideal
period for tapering, which is the period of reduced training used to
optimize performance for a competition (12, 21, 23).
The decrease in the maximal gain in performance for a given amount of
training has been less well documented. This observation is in line
with the conclusions that Wenger and Bell (27) drew from
data pooled from training protocols, with training frequency ranging
from two to six sessions per week. These authors showed that the
optimal combination of intensity and frequency of training is
90-100% O2 max at a frequency of
four sessions per week and that training effects plateau with higher
frequencies. A possible limitation of the body to adapt to training
loads could yield an apparent decline in the benefit of a new training
stimulus. A progressive decrease in exercise-induced adaptation could
also be associated with overtraining. Long-term overtraining is
associated with a lack of adaptation to training loads, yielding
underperformance and necessitating reduced training for weeks or months
before recovering performance level. Although the underlying mechanisms of long-term overtraining are unclear, alterations of the body's response to disturbed homeostasis could affect the capacity of overtrained subjects to recover and adapt to exercise
(15).
Previous studies using the time-invariant model showed both lower
pg and higher tn with increasing
training load (6). However, differences in the initial
level of fitness could explain these differences. The lower
pg for the more fit subjects could be essentially due to
the need for greater training loads to obtain similar or even lower
gain in performance. The decrease in aerobic performance increment with
training when fitness level increases was well documented (see, for
example, Ref. 27). This phenomenon could obscure the
findings of this study. The results about HFT could be different if HFT
were done first before the increment in fitness. Because the subjects
undertook HTF after LFT, the much higher fitness could be responsible
for the observed decrease in pg during HFT. Nevertheless,
the higher fitness at the beginning of HFT could not alone explain the
decreased pg. This decrease appeared only at the beginning
of HFT and not at the end of LFT in which the level of performance was
close to the level during HFT. On the other hand, the increase in
tn with HFT could not be attributed to the order
of the training sequences. The reverse of the order of the two training
phases could eventually enhanced the difference in
tn between HFT and LFT.
In contrast to the findings of this study, the previous application of
the time-varying model showed an increase in pg with training for the two studied subjects (6). Furthermore,
for both subjects, variations in pg were correlated with
unsteady variations in performance interpreted as a change in the
subjects' adaptability to train (6). This contradiction
with the present results could arise from differences in the training
program. In the earlier experiment, higher training loads than the ones used in the present study were applied from the beginning, thus yielding a decrease in pg at the start of the training
program. The subsequent increase in pg partly due to 2 wk
of reduced training in the middle of the experiment would also be an
adaptation to heavy training. The decrease in exercise adaptation
assessed from pg in the present study would only be
temporary and restorable with adequate training. This points out the
importance of progressively increasing training load with an adequate
period of recovery to prevent staleness (16).
Nevertheless, the methodology employed in our previous study could also
have obscured the results. The training loads referred to MAP measured
every other week and were assumed to vary linearly between two
measurements. In contrast with the present study, day-to-day variations
in performance were not taken into account, possibly leading to an
underestimation of the signal input and thus an overestimation of
pg. Nevertheless, because of its potential importance in
training periodization, the possible increase in training adaptability
with adequate increase in training loads and regeneration periods
deserves further investigations.
In conclusion, this study showed that the reduction in recovery time
between training sessions yielded a progressive increase in the
magnitude and duration of fatigue induced by each bout of training
stimulus and also led to a decrease in the resulting adaptations.
Reduced adaptation to training loads could arise from lower tolerance
to exercise, from higher fitness level, or from a limitation on the
body's capacity to adapt to greater training loads.
| |
FOOTNOTES |
Address for reprint requests and other correspondence: T. Busso, Laboratoire de Physiologie, CHU de Saint-Etienne, Hôpital de Saint-Jean-Bonnefonds, Pavillon 12, 42055 Saint-Etienne Cedex 2, France (E-mail: busso{at}univ-st-etienne.fr).
The costs of publication of this
article were defrayed in part by the
payment of page charges. The article
must therefore be hereby marked
"advertisement"
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
10.1152/japplphysiol.00429.2001
Received 3 May 2001; accepted in final form 25 September 2001.
| |
REFERENCES |
1.
Banister, EW,
Calvert TW,
Savage MV,
and
Bach TM.
A systems model of training for athletic performance.
Aust J Sports Med
7:
57-61,
1975.
2.
Banister, EW,
Carter JB,
and
Zarkadas PC.
Training theory and taper: validation in triathlon athletes.
Eur J Appl Physiol
79:
182-191,
1999.
3.
Banister, EW,
and
Hamilton CL.
Variations in iron status with fatigue modelled from training in female distance runners.
Eur J Appl Physiol
54:
16-23,
1985.
4.
Busso, T,
Candau R,
and
Lacour JR.
Fatigue and fitness modelled from the effects of training on performance.
Eur J Appl Physiol
69:
50-54,
1994.
5.
Busso, T,
Carasso C,
and
Lacour JR.
Adequacy of a systems structure in the modeling of training effects on performance.
J Appl Physiol
71:
2044-2049,
1991[Abstract/Free Full Text].
6.
Busso, T,
Denis D,
Bonnefoy R,
Geyssant A,
and
Lacour JR.
Modeling of adaptations to physical training by using a recursive least squares algorithm.
J Appl Physiol
82:
1685-1693,
1997[Abstract/Free Full Text].
7.
Busso, T,
Häkkinen K,
Pakarinen A,
Carasso C,
Lacour JR,
Komi PV,
and
Kauhanen H.
A systems model of training responses and its relationships to hormonal responses in elite weight-lifters.
Eur J Appl Physiol
61:
48-54,
1990.
8.
Busso, T,
Häkkinen K,
Pakarinen A,
Kauhanen H,
Komi PV,
and
Lacour JR.
Hormonal adaptations and modelled responses in elite weightlifters during 6 weeks of training.
Eur J Appl Physiol
64:
381-386,
1992.
9.
Fitz-Clarke, JR,
Morton RH,
and
Banister EW.
Optimizing athletic performance by influence curves.
J Appl Physiol
71:
1151-1158,
1991[Abstract/Free Full Text].
10.
Fry, RW,
Morton AR,
and
Keast D.
Overtraining in athletes. An update.
Sports Med
12:
32-65,
1991[ISI][Medline].
11.
Hickey, MS,
Costill DL,
McConell GK,
Widrick JJ,
and
Tanaka H.
Day to day variation in time trial cycling performance.
Int J Sports Med
13:
467-470,
1992[ISI][Medline].
12.
Houmard, JA.
Impact of reduced training on performance in endurance athletes.
Sports Med
12:
380-393,
1991[Medline].
13.
Jeukendrup, A,
Saris WHM,
Brouns F,
and
Kester ADM
A new validated endurance performance test.
Med Sci Sports Exerc
28:
266-270,
1996[ISI][Medline].
14.
Krebs, PS,
and
Powers SK.
Reliability of laboratory endurance tests (Abstract).
Med Sci Sports Exerc
21:
S10,
1989.
15.
Kuipers, H.
Training and overtraining: an introduction.
Med Sci Sports Exerc
30:
1137-1139,
1998[ISI][Medline].
16.
Kuipers, H,
and
Keizer HA.
Overtraining in elite athletes. Review and directions for the future.
Sports Med
6:
79-92,
1988[ISI][Medline].
17.
Kuipers, H,
Verstappen FTJ,
Keizer HA,
Geurten P,
and
van Kranenburg G.
Variability of aerobic performance in the laboratory and its physiological correlates.
Int J Sports Med
6:
197-201,
1985[ISI][Medline].
18.
Lehmann, M,
Foster C,
and
Keul J.
Overtraining in endurance athletes: a brief review.
Med Sci Sports Exerc
25:
854-862,
1993[ISI][Medline].
19.
McLellan, TM,
Cheung SS,
and
Jacobs I.
Variability of time to exhaustion during submaximal exercise.
Can J Appl Physiol
20:
39-51,
1995[ISI][Medline].
20.
Morton, RH,
Fitz-Clarke JR,
and
Banister EW.
Modeling human performance in running.
J Appl Physiol
69:
1171-1177,
1990[Abstract/Free Full Text].
21.
Mujika, I.
The influence of training characteristics and tapering on the adaptation in highly trained individuals: a review.
Int J Sports Med
19:
439-446,
1998[Medline].
22.
Mujika, I,
Busso T,
Lacoste L,
Barale F,
Geyssant A,
and
Chatard JC.
Modeled responses to training and taper in competitive swimmers.
Med Sci Sports Exerc
28:
251-258,
1996[Medline].
23.
Neufer, PD.
The effect of detraining and reduced training on the physiological adaptations to aerobic exercise training.
Sports Med
8:
302-320,
1989[ISI][Medline].
24.
Palmer, GS,
Dennis SC,
Noakes TD,
and
Hawley JA.
Assessment of the reproducibility of performance testing on an air-braked cycle ergometer.
Int J Sports Med
17:
293-298,
1996[ISI][Medline].
25.
Viru, A.
The mechanism of training effects: a hypothesis.
Int J Sports Med
5:
219-227,
1984[Medline].
26.
Viru, A.
Postexercise recovery period: carbohydrate and protein metabolism.
Scand J Med Sci Sports
6:
2-14,
1996[Medline].
27.
Wenger, HA,
and
Bell GJ.
The interactions of intensity, frequency and duration of exercise training in altering cardiorespiratory fitness.
Sports Med
3:
346-356,
1986[ISI][Medline].
J APPL PHYSIOL 92(2):572-580
8750-7587/02 $5.00
Copyright © 2002 the American Physiological Society
TOP
ABSTRACT
INTRODUCTION
