Content uploaded by Cédric YM Morio
Author content
All content in this area was uploaded by Cédric YM Morio on Oct 09, 2015
Content may be subject to copyright.
- 1 -
Original Article submitted to
Journal of Sports Medicine and Physical Fitness
INFLUENCE OF EXERCISE TYPE ON METABOLIC COST AND
GROSS EFFICIENCY: ELLIPTICAL TRAINER VERSUS
CYCLING TRAINER.
MORIO Cédric, PhD1*, HADDOUM Mohand, MSc1, FOURNET Damien, PhD2,
GUEGUEN Nils, PhD1.
Authors Affiliation:
1 Decathlon Research, Department of Movement Sciences, 4 boulevard de Mons, Villeneuve
d’Ascq, 59665 France
2 Decathlon Research, Department of Thermal Comfort Sciences, 4 boulevard de Mons,
Villeneuve d’Ascq, 59665 France
* Corresponding Author:
Mr MORIO Cédric
Decathlon Campus, R&D department
4 boulevard de Mons, BP299
59650 Villeneuve d’Ascq, France
Tel: 03.20.19.83.36
Email: cedric.morio@decathlon.com
Abstract
Elliptical trainers are known as a good mean to develop physical fitness. However, the
pedalling efficiency on an elliptical trainer has not been reported in the literature. The aim of
the present study is to compare metabolic cost and gross efficiency for two different trainers,
elliptical (ET) and cycling (CT). Fourteen participants were tested on ET and CT during two
exercise sessions. Participants pedalled at 9 different power outputs for 3 minutes each.
Oxygen consumption (VO2) and heart rate (HR) were recorded. Gross efficiency (GE) was
calculated during the last 30s of each 3min period. Maximal aerobic power (MAP) was
estimated for each participant for each condition. MAP was found to be significantly greater
in CT (237 ± 88W) compared to ET (151 ± 51W). Significant positive correlations were
found between power output and VO2 in both CT (r = 0.93) and ET conditions (r = 0.97).
Regarding the inter-individual variability in MAP, GE was significantly correlated to the
relative power output (%MAP) (r = 0.75 in CT and r = 0.69 in ET). The aim of the present
study was to investigate metabolic demand of different exercise type using %MAP in each
condition. The results confirmed that metabolic cost of ET was greater than CT at
similar %MAP. Gross efficiency was lowered in ET condition compared to CT. This could be
explained through the additional use of arms and the standing position during ET.
Keywords
Elliptical Trainer; Cycling; Gross Efficiency; Power Output; Oxygen Consumption.
- 2 -
INTRODUCTION
The elliptical trainer (ET) is known as a good mean to develop physical fitness. The aerobic
exercise associated with ET involve both lower and upper limbs. Moreover, compared to
classical training programs, ET has several advantages like a reduction of impact (running) or
vibration (cycling). Although ET is commonly practised in fitness centres as well as personal
use, there is a limited number of studies which investigate the use of ET.
Most of the previous work investigated the use of ET as either a cardiovascular fitness test
(1,2) or for specific metabolic prediction equations (2-6). As expected, previous work has
identified that the physiological demand of ET increased with the pedalling cadence and the
resistance level (3-5). When leg-only mode was used it induced less energy expenditure than
the use of both arms and legs simultaneously (4). Thus, ET has been used for physical fitness
assessment as well as physical fitness training. Research investigating the use of ET has
identified that if used a prolonged period (12 weeks) a significant improvement of the
physical fitness can be reached (7). As well as being a good method for increasing physical
fitness, ET has also been shown to be an effective method of maintaining physical fitness
level, during a period of three weeks of reduced activity in recently-trained runners (4 weeks
training) (8).
Literature has also compared the physiological demand of ET to other stationary exercise
such as treadmill running or cycle-ergometer training (9-12). It was shown that similar
maximal VO2peak and HRpeak could be achieved in either ET or treadmill running (10). Same
authors have also identified an equivalent linear relationship between HR, RPE and VO2 in
both exercises. However, other studies comparing the two methods have found there to be a
relative reduction in RPE. Although some authors did not find any differences in relative RPE
between treadmill running and ET (10), other studies presented either a relative reduction of
RPE in ET compared to treadmill running (9,11,12), stationary cycling (12) or leg-only ET
exercise (4) at a similar oxygen consumption or energy expenditure. In other words, an
increase in energy expenditure for similar RPE was assumed to be due to the addition of arms
movement to the overall action (4). Furthermore, arms were known to be less efficient than
legs in cyclical movement (13).
As ET was created to simulate running kinematics whilst eluding ground impact (12),
most of the previous studies focused on comparison with treadmill running. However,
pedalling on an ET is also a cyclic exercise that could be assimilated to stationary cycling. To
our knowledge, the efficiency of pedalling on ET has not been reported in the literature.
Therefore, the aim of the present study was firstly to give quantitative data on physiological
- 3 -
demands of elliptical cycling, and secondly to compare metabolic cost and gross efficiency
(GE is defined as the ratio of mechanical energy output divided by the total energy
expenditure) for two different types of trainer, elliptical and cycling. The main hypothesis was
that gross efficiency would be impaired while pedalling on an elliptical trainer compared to
cycling. Thus in order to compare these different activities, dependant variables (VO2, GE,
HR, RPE) were studied regarding relative power output during either elliptical trainer exercise
or cycling.
MATERIALS AND METHODS
Participants
Fourteen participants, 7 males (27.4 ± 3.6 years, 84.3 ± 13.4 kg, 182 ± 5 cm, 19 ± 2 % body
fat) and 7 females (27.4 ± 4.0 years, 64.1 ± 15.4 kg, 165 ± 6 cm, 29 ± 6 % body fat)
participated in this study. All participants were healthy, free from injury and were physically
active, practicing sport 2 to 3 hours per week. Participants gave informed written consent
before participating. This study was made in accordance with institutional and national ethical
rules.
Procedures
Each participant complete two sessions, each session was used to complete one of the
exercise types. The two sessions were performed one week apart and the order of testing was
randomised. The two exercise types were classical seated leg cycling (Domyos VM510,
thereafter called CT) and elliptical trainer (Domyos VE710, thereafter called ET) where
participants were required to combine both arm and leg movement to drive the elliptical
wheel.
For all the resting and exercise session, participants were equipped with a Fitmate Pro
(Cosmed) system which allowed the recording of relative oxygen consumption (VO2)
expressed in (mL.min-1.kg-1) and heart rate (HR) expressed in percentage of heart rate reserve
(%HRres). Respiratory exchange ratio (RER) was estimated according to Nieman et al. (14),
further validated by Louis et al. (15). Ratings of perceived exertion (RPE) were asked at the
end of the 3 min of every different power outputs through the Borg’s 6-20 RPE scale (16).
For each exercise session, an initial measurement of 5 min (HR and VO2) was taken at rest
then a warm-up protocol of 5 min was followed prior to exercise. During the rest period,
participants were asked to lie down on a foam mat in order to measure the proper resting
value of HR and VO2. During the exercise, participants pedalled at 9 different power outputs
- 4 -
(3 resistance levels x 3 cadences) for 3 min each (detailed resistance levels and cadences are
presented in Table 1). Although a 3 min period was short, this was enough to reach the
metabolic plateau for the low power output used in this study thus preventing subjects from
fatigue. The resistance levels and cadences were randomly proposed to limit the order and
rank effects. The exercise was split into 3 bouts of 9 min with 3 min rest periods, to limit the
fatigue effect.
In order to quantify the mechanical power output developed during exercise sessions, both
CT and ET were calibrated on a specific torque/cadence device provided by the manufacturer.
Table 1: The 3 resistance levels x 3 cadences were presented in the upper panel (A) for the
ET and CT conditions as well as the female vs male protocol adaptation. The lower panel
(B) presented the corresponding power outputs measured mechanically in specific testing
facility.
A":"Resistance"levels"x"cadences"protcols"
""
""
cadences"(rpm)"
""
""
30"
40"
50"
60"
70"
80"
ET"
Female"
1"
3"
5"
1"
3"
5"
1"
3"
5"
"
"
"
"
"
"
"
"
"
""
Male"
""
""
""
2"
5"
8"
2"
5"
8"
2"
5"
8"
""
""
""
""
""
""
CT"
Female"
""
""
""
""
""
""
1"
3"
5"
1"
3"
5"
1"
3"
5"
""
""
""
""
Male"
""
""
""
""
""
""
""
""
""
1"
3"
5"
1"
3"
5"
1"
3"
5"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
"
B":"Corresponding"power"outputs"(W)"
""
""
cadences"(rpm)"
""
""
30"
40"
50"
60"
70"
80"
ET"
Female"
22"
32"
39"
38"
54"
65"
56"
80"
98"
"
"
"
"
"
"
"
"
"
"
Male"
"
"
"
50"
65"
80"
71"
98"
118"
97"
135"
165"
"
"
"
"
"
"
CT"
Female"
""
""
""
""
""
""
28"
53"
76"
38"
73"
105"
49"
95"
137"
""
""
""
""
Male"
""
""
""
""
""
""
""
""
""
38"
72"
105"
49"
95"
137"
61"
118"
171"
Data Analysis
In addition to VO2, HR and RPE, other metabolic parameters were calculated during the last
30 seconds of each 3 min period at each power output. First, the gross efficiency (GE) was
calculated as the ratio between the mechanical energy output and the human energy
expenditure. The mechanical energy output in Joules was obtained by integrating the power
output along time. The human energy expenditure in Joules was obtained through the VO2
measurement and the RER estimation (14). Then, maximal oxygen consumption (VO2max) and
maximal aerobic power (MAP) were estimated for each participant for each exercise type.
These variables were estimated as the projection of the theoretical maximal HR on the linear
- 5 -
regression between the HR and the VO2 or mechanical power output (17,18). Thus, it was
possible to express the mechanical power output in relative percentage of MAP (%MAP).
Linear regressions between %HRres, VO2 and absolute power output in Watt as well
as %MAP were calculated on global results for each of the nine power outputs, for every
participant in both exercise types. %HRres was also compared to the RPE scale in order to
determine whether an exercise was perceived as more difficult than the other. Global
logarithmic regression was calculated between GE and %MAP for each exercise type,
wherein the logarithmic coefficient represented the increase slope of GE regarding %MAP
and the intercept coefficient corresponded in GE at MAP. Individual logarithmic regressions
between GE and %MAP were also calculated for each participant, in order to statistically
compare the two exercise types.
Statistical Analyses
VO2max and MAP estimated from ET were compared to VO2max and MAP estimated from CT
using a paired sample t-tests. As the population used for comparison was mixed (male &
female), the Wilcoxon comparison test was preferred. In case there was no difference between
ET and CT exercise type either for VO2max or MAP, an intraclass correlation coefficient
ICC(2,1) would be applied to test the exercise type accordance between each other (19).
Pearson correlation coefficients were calculated between %HRres, VO2 and power outputs
(absolute and relative), as well as between %HRres and RPE scores. Additional paired
comparison tests (Wilcoxon) between trainer types were performed on individual coefficients
obtained through regression calculation for each participant, to compare GE slope and
maximal GE at MAP. Statistical α-level of significance was set at 0.05 for all statistics.
RESULTS
The estimations of VO2max and the associated MAP are presented in the Table 2. The VO2max
estimated did not significantly differ between exercise types (p = 0.24). Moreover, there was a
good accordance between the two groups of results for each individuals between estimations
of ET VO2max and CT VO2max (ICC(2,1) = 0.83). The absolute MAP estimated for the CT
exercise (237 ± 88 W) was significantly greater (p = 0.001) than the ET MAP (151 ± 51 W).
- 6 -
Table 2: Estimated VO2max (mL.min-1.kg-1) and MAP (W) for cycling and elliptical trainers
in males and females. Mean (±sd) are presented. * significantly different between CT and
ET (p < 0.05).
Cycling
Trainer
Elliptical
Trainer
VO2max (mL.min-1.kg-1)
39.0
40.9
8.3
7.4
MAP (W)
237
151*
88
51
%HRres showed a significant positive relationship with absolute power output in both CT
and ET conditions (r = 0.65; p < 0.001, r = 0.69; p < 0.001, respectively). The representation
of these results in Figure 1A also showed higher %HRres (p < 0.001) during ET condition
compared to CT condition at similar absolute power output. %HRres also showed a
significant relationship with the relative power output in CT and ET conditions (r = 0.90;
p < 0.001, r = 0.96; p < 0.001, respectively). The representation of these results in Figure 1B
showed similar %HRres regarding power output in %MAP either during CT or ET.
Relative VO2 (mL.min-1.kg-1) also showed significant relationship with the absolute power
output in both CT and ET conditions (r = 0.80; p < 0.001, r = 0.86; p < 0.001, respectively).
The Figure 1C presents higher VO2 (p < 0.001) in ET compared to CT condition for similar
absolute power output. VO2 measurements were still higher (p < 0.001) in ET compared to
CT condition while comparing VO2 with relative power output expressed in percentage of
estimated MAP (Figure 1D). The relationships between these parameters were also
significant for both CT and ET conditions (r = 0.77; p < 0.001, r = 0.78; p < 0.001,
respectively).
- 7 -
Figure 1: Heart rate reserve (%HRres) (Fig. 1A, 1B) and VO2 (mL.min-1.kg-1) (Fig. 1C,
1D) in relation to absolute power output (W) and relative power output (%MAP). Each data
point corresponds to individual measurements at each power output (14*9 = 126
observations).
The rating of perceived exertion (RPE) results are presented in the Figure 2 where RPE
scores increased significantly with the %HRres measurements during CT condition (r = 0.86;
p < 0.001) as well as during ET condition (r = 0.85; p < 0.001). The RPE scores tended to
increase more rapidly during CT (slope coefficient = 12.3) than during the ET condition
(slope coefficient = 10.9). This increase was not confirmed by the individual slopes of
the %HRres vs. RPE linear regression (p > 0.05).
- 8 -
Figure 2: Rate of perceived exertion (RPE) in relation to heart rate reserve (%HRres).
Each data point corresponds to individual measurements at each power output (14*9 = 126
observations).
GE increased as the power output increased. This relationship was not linear but can be
approximated by a logarithmic regression between GE and absolute power output (Figure
3A) during CT (r2 = 0.55) and ET conditions (r2 = 0.55). Logarithmic regressions were
improved with the relative power output (Figure 3B) for the CT condition (r2 = 0.61) but not
for the ET condition (r2 = 0.54). Through studying GE in regard with the relative power
output in %MAP, the maximal aerobic GE could be observed on the regression curve for
100% of MAP. This value corresponded to the intercept coefficient of the logarithmic
regression equation (24.5% in CT and 14.4% in ET). These GE slope and intercept
coefficients were reported and compared in Table 3, where GE slope and intercept
coefficients were significantly lower for the ET condition compared to the CT condition.
Table 3: Logarithm coefficients and Intercept coefficients (based on logarithmic regression
between GE and %MAP) for cycling and elliptical trainers in males and females. Mean (±
SD) are presented. * significantly different between CT and ET (p < 0.05).
Cycling
Trainer
Elliptical
Trainer
Logarithm Coefficient
6.7
4.5 *
1.3
0.7
Intercept Coefficient
25,4
14.9 *
1,5
1.5
- 9 -
Figure 3: Relationship between gross efficiency and absolute power output (A) or relative
power output expressed in percentage of maximal aerobic power (B). Estimated gross
efficiency at maximal aerobic power presented in the figure with black arrows
corresponded to the intercept coefficient of the logarithmic regression (B). Each data point
corresponds to individual measurements at each power output (14*9 = 126 observations).
DISCUSSION
The main finding of the present study was that VO2 was found to be higher in elliptical trainer
(ET) compared to cycling trainer (CT) at both similar absolute and relative power outputs.
Thus, the hypothesis of a lower gross efficiency (GE) in ET was confirmed.
One point of interest for the current study was to illustrate results in percentage of
maximal aerobic power in addition to absolute power output. As the maximal aerobic power
was clearly shown to be task dependant, the comparison of ET and CT exercises at same
absolute power output would be erroneous. Previous studies tried to compare ET with other
- 10 -
exercise types like walking, running, cycling or stepping (9-12). They all compared different
level of power output which might induce bias in the interpretations of the observed
differences. These different activities could have been compared at similar cadence or speed,
as previously done for electromyographic or kinematic analyses (20,21). But, the use of
identical cadence during walking, cycling and pedalling on an ET did not imply the same
energy consumption and did not produce the same power output. It was possible to compare
ET and CT condition at similar %MAP, with the individual- and task-dependant estimations
of the MAP. Thus, individual slope and intercept coefficients were used to compare ET and
CT conditions (10). In the present study, the relatively high ICC on estimated VO2max between
the two conditions confirmed the possibility to compare the two exercises.
%HRres and VO2 expressed in regards to absolute power output presented significantly
higher regression values in ET compared to CT. %HRres did not present any differences
between conditions when expressed in percentage of MAP. VO2 presented similar increase
with percentage of MAP but higher intercept coefficient in ET indicating a constant higher
metabolic rate in ET exercise compared to CT.
Although a slight reduction could be observed in ET condition, RPE did not differ
significantly between the two exercise types. Previous studies demonstrated lower RPE in ET
exercise compared to treadmill running, reported to similar VO2 (9,11,12). Only the study of
Porcari et al (12) compared with CT and found an increased in VO2 at similar RPE. As the
present study compared exercise at the same absolute and relative power outputs, it could be
assumed that this difference in RPE results could be explained with an impairment of gross
efficiency in ET compared to CT condition. GE was indeed lower in ET condition compared
to CT either while looking at absolute power output or relative percentage of maximal aerobic
power.
The lower GE in ET condition has not been described before in the literature.
Nevertheless, increase VO2 at similar RPE was once reported in ET exercise compared to CT
(12). The impairment from 24.5% of maximal GE at 100% of MAP in CT to 14.4% of GE in
ET exercise could be explained through different underlying reasons. Firstly, the additional
use of arms in ET was demonstrated to increase VO2 for identical cadence and resistance level
while pedalling with both arms and legs compared to the leg-only exercise (4). It was also
known that cyclic arm movements were less efficient than leg ones (13). Significantly
increased activations in upper limb muscles were also previously reported in ET compared to
both treadmill running and CT (22). Secondly, the ET could also be seen as an unusual
exercise, with a difference in the lower limb muscles coordination and their implication (20-
- 11 -
22). ET might need of greater experience of participants to get used to such specific
coordination. For example, most of the participants reported the need of arm at higher
resistance levels, whereas the arm moved only to follow the elliptical device without helping
to produce power at lower resistance levels. Thirdly, the standing position induced by the
elliptical device compared to the seated cycling exercise might provoke greater energy
consumption only to support the whole body mass. Nevertheless, conflicting results were
reported in the literature whether standing cycling induced greater VO2, thus reduced cycling
GE (23,24) or not (25,26). Finally, all the forces applied on the elliptical pedal were not
effective like those observed in classical pedalling (27). Power output was measured directly
on the pedal connection with the resistance wheel for both ET and CT. The distance of the
lever arm and the pedal angle changes relative to the resistance wheel of the ET were not
taken into account in the calculation of the power output and might reduce the gap with
classical CT.
Another interesting perspective of the present work would be to compare the
physiological demand of elliptical cycling without using the arms with a stationary standing
cycling condition. Thus, the main independent variable would be the leg pedalling profile. A
longer habituation of the participant to the protocol would be interesting as well.
CONCLUSIONS
The present data provides a better understanding of the physiological demand of elliptical
trainer exercise. ET exercise was previously proposed as a good alternative for rehabilitation
or cross-training program (increased arm workload and lowered induced impact). However,
the low maximal GE of 14.4% observed in the present study implies that coaches should
beware of these results while they use ET instead of CT or treadmill exercise. Prolonged
training at a low GE might be detrimental for performance improvement, as it could accustom
to use ineffective gestures during athletes’ rehabilitation. ET training, however, might also
improve GE through better economical technique. Further investigations on ET training
programs are needed to clarify the benefits of such training exercise.
ACKNOWLEDGEMENTS
The authors would like to thank the Domyos™ brand for providing the trainers and the
power/cadence technical reports needed for the present study. Special thank to Dr James
Webster for the english improvements he has proposed us.
- 12 -
REFERENCES
1. Dalleck LC, Kravitz L, Robergs RA. Maximal exercise testing using the elliptical cross-trainer and
treadmill. JEPonline 7(3):94-101, 2004.
2. Parker SB, Griswold L, Vickers RR Jr. Development of an elliptical trainer physical fitness test. Naval
Health Research Center Report No. 06-06, 2006.
3. Dalleck LC, Kravitz L. Development of a metabolic equation for elliptical crosstrainer exercise. Percept
Mot Skills 104(3):725-732, 2007.
4. Mier CM, Feito Y. Metabolic cost of stride rate, resistance, and combined use of arms and legs on the
elliptical trainer. Res Q Exerc Sport 77(4):507-513, 2006.
5. Swier DP. Comparison of oxygen consumption on different brands of elliptical trainers in the
development of a metabolic prediction equation. MSc dissertation of the University of Texas at
Arlington, 2008.
6. Vickers RR Jr, Griswold L, Hodgon JA. A comparison of three models of elliptical trainer. Naval
Health Research Center Report No. 06-31, 2006.
7. Egana M, Donne B. Physiological changes following a 12 week gym based stair-climbing, elliptical
trainer and treadmill running program in females. J Sports Med Phys Fitness 44:141-6, 2004.
8. Joubert DP, Oden GL, Estes BC. The effect of elliptical cross training on VO2max in recently trained
runners. Int J Exerc Sci 4(1):4-12, 2011.
9. Brown GA, Cook CM, Krueger RD, Heelan KA. Comparison of energy expenditure on a treadmill vs.
an elliptical device at a self-selected exercise intensity. J Strength Cond Res 24(6):1643-1649, 2010.
10. Mercer JA, Dufek JS, Bates BT. Analysis of peak oxygen consumption and heart rate during elliptical
and treadmill exercise. J Sport Rehabil 10:48-56, 2001.
11. Pecchia K, Evans BW, Edwards JE, Bell F. Physiological responses to exercise on the elliptical trainer
compared to the treadmill. Med Sci Sports Exerc 31(5):S158, 1999.
12. Porcari JP, Zedaker JM, Naser L, Miller M. Evaluation of an elliptical exerciser in comparison to
treadmill walking and running, stationary cycling, and stepping. Med Sci Sports Exerc 30(5):S168,
1998.
13. Eston RG, Brodie DA. Responses to arm and leg ergometry. Br J Sports Med 20(1):4-6, 1986.
14. Nieman DC, LaSosso H, Austin MD, Pearce S, McInnis T, Unick J. Validation of Cosmed’s FitMate™
in measuring exercise metabolism. Res Sports Med 15: 67-75, 2007.
15. Louis J, Hausswirth C, Brisswalter J. Reproducibility of the Cosmed’s FitMate™ metabolic system in
measuring steady-state cycling energy expenditure. Sci Sports 27: 20-23, 2012.
16. Borg G. Borg’s perceived exertion and pain scales. Human Kinetics Publishers 1998.
17. Dalleck LC, Kravitz L. Relationship between %heart rate reserve and %VO2 reserve during elliptical
crosstrainer exercise. J Sports Sci Med 5:662-671, 2006.
18. Wyndham CH. Submaximal tests for estimating maximum oxygen intake. Canad Med Ass J 96:736-
742, 1967.
19. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 86(2): 420-
428, 1979.
20. Damiano DL, Norman T, Stanley CJ, Park HS. Comparison of elliptical training, stationary cycling,
treadmill walking and overground walking. Gait Posture 34(2):260-264, 2011.
21. Prosser LA, Stanley CJ, Norman TL, Park HS, Damiano DL. Comparison of elliptical training,
stationary cycling, treadmill walking and overground walking. Gait Posture 33(2):244-250, 2011.
22. Sozen H. Comparison of muscle activation during elliptical trainer, treadmill and bike exercise. Biol
Sport 27:203-206, 2010.
23. Ryschon TW, Stray-Gundersen J. The effect of body position on the energy cost of cycling. Med Sci
Sports Exerc 23(8):949-953, 1991.
24. Tanaka H, Bassett DR, Best SK, Baker KR. Seated versus standing cycling in competitive road cyclists:
uphill climbing and maximal oxygen uptake. Can J Appl Physiol 21(2):149-154, 1996.
25. Millet GP, Tronche C, Fuster N, Candau R. Level ground and uphill cycling efficiency in seated and
standing positions. Med Sci Sports Exerc 34(10):1645-1652, 2002.
26. Hansen EA, Waldeland H. Seated versus standing position for maximization of performance during
intense uphill cycling. J Sports Sci 26(9):977-984, 2008.
27. Sanderson DJ. The influence of cadence and power output on the biomechanics of force application
during steady rate cycling in competitive and recreational cyclists. J Sports Sci 9:191-203, 1991.