Reliability of Force-Velocity Tests in Cycling and Cranking Exercises in Men and Women

The present study examined the reliability of the force-velocity relationship during cycling and arm cranking exercises in active males and females. Twenty male and seventeen female physical education students performed three-session tests with legs and three-session tests with arms on a friction-loaded ergometer on six different sessions in a randomized order. The reliability of maximal power (P max), maximal pedal rate (V 0), and maximal force (F 0) were studied using the coefficient of variation (CV), the intraclass correlation coefficient (ICC) and the test-retest correlation coefficient (r). Reliability indices were better for men (1.74 ≤ CV ≤ 4.36, 0.82 ≤ ICC ≤ 0.97, and 0.81 ≤ r ≤ 0.97) compared with women (2.34 ≤ CV ≤ 7.04, 0.44 ≤ ICC ≤ 0.98, and 0.44 ≤ r ≤ 0.98) and in cycling exercise (1.74 ≤ CV ≤ 3.85, 0.88 ≤ ICC ≤ 0.98, and 0.90 ≤ r ≤ 0.98) compared with arm exercise (2.37 ≤ CV ≤ 7.04, 0.44 ≤ ICC ≤ 0.95, and 0.44 ≤ r ≤ 0.95). Furthermore, the reliability indices were high for P max and F 0 whatever the expression of the results (raw data or data related to body dimensions). P max and F 0 could be used in longitudinal physical fitness investigations. However, further studies are needed to judge V 0 reliability.


Introduction
Maximal anaerobic power can be measured on frictionloaded cycle ergometers or isokinetic ergometers. Many protocols have been proposed for maximal power measurement: all-out tests against a single load (e.g., the Wingate test) [1,2], relationship between torque and pedal rate on an isokinetic ergometer [3,4], relationship between load and peak velocity [5], and force-velocity relationship during a single all-out test against a pure inertial load [6] or inertial + braking load [7][8][9].
On friction-loaded ergometer, maximal power corresponds to power at peak velocity or is computed during the acceleration phase taking into account the power necessary to increase the flywheel kinetic energy [10]. The relationship between pedal rate ( ) and braking force ( ) or torque ( ) can be described by a linear relationship [3,[5][6][7][8][9]11]. Linear force-velocity relationships have been described for all-out exercises performed on a cycle ergometer not only with the legs (i.e., cycling exercise) but also with the arms (i.e., cranking exercise). The individual characteristics of the force-velocity or torque-velocity relationship can be defined by two parameters: 0 (the intercept with the pedal rate axis which has the dimension of a maximal pedal rate) and 0 or 0 (the intercepts with the force or torque axis, which have the dimension of a maximal force or a maximal torque). Maximal power ( max ) corresponds to an optimal pedal rate ( opt ) equal to 0.5 0 and an optimal load or torque equal to 0.5 0 or 0.5 0 .
Previous studies reported that max [8] or peak power during a Wingate test [12][13][14][15] are significantly correlated with the percentage of the fast muscle fibers in the vastus lateralis. Furthermore, a significant positive correlation was observed between max and triceps surae musculotendinous stiffness at relative peak torque corresponding to the optimal cycling rate [16]. On the other hand, the value of opt during sprint cycling was significantly correlated with vastus lateralis myosin heavy chain II composition in a study comparing old 2 BioMed Research International and young participants [17]. The proportion of fast twitch fibres expressed in terms of cross-sectional area was highly correlated with opt ( = 0.88, < 0.001) [18], and the authors of this latter study suggested that opt would be the most accurate parameter to explore the fibre composition of the knee extensor muscle from cycling tests. The value of 0 in cycling depends on the strength and the rate of force development of muscle knee extensors [19]. The Wingate optimal braking force can also be determined from the result of a cycling force-velocity test as this braking force is close to 0.5 0 [5,20].
Therefore, it could be interesting to determine the parameters of the force-velocity relationships ( 0 , 0 , or 0 ) in addition to max on a cycle ergometer. Furthermore, the study of the changes in power-velocity relationship during an annual training cycle has been proposed in volleyball players [21], which assumes that the results of the force-velocity tests on cycle ergometers are reliable. The reliability of the cycling all-out tests has mainly been investigated by studying either the test-retest correlation coefficients ( test-retest ) or the intraclass correlation coefficient (ICC) or the standard errors of estimations (SEE) or the coefficients of variation (CV) for the indices of maximal power (Wingate peak power or max ) with the different protocols [1-4, 6, 9, 22-27]. In contrast, the reliability of the parameters of the force-velocity relationship (slope, 0 , 0 , and 0 ) has been investigated in a few studies, only [4,6,26]. Moreover, the validity of the statistical tests in these studies on reliability was probably questionable [28].
In a review on the reliability of power in physical performance tests, Hopkins et al. [29] suggested that nonathletic females might be less reliable than nonathletic males, probably because the nonathletic females may be less physically active than the nonathletic males. Similarly, cranking exercises are probably less familiar than cycling exercises and the effect of familiarisation sessions might be more important for force-velocity tests with the arms.
Thus, the aim of the present study was to examine the reliability of max , 0 , and 0 during force-velocity tests. In light of the literature observations, we hypothesized that reliability is lower in women than in men and for cranking force-velocity tests than for cycling tests.

Participants.
Twenty healthy males (24.20 ± 2.69 years, 1.80 ± 0.06 m, and 76.48 ± 8.93 kg) and seventeen healthy females (23.53 ± 2.12 years, 1.68 ± 0.06 m, and 61.18 ± 9.58 kg) volunteered to participate in this study. The participants were all active physical education students but none of them were familiarized with sprint cycling or arm cranking before participation in the study. Before any data collection, all participants were fully informed of the possible risk and discomfort associated with the experimental procedures and gave written informed consent. The experimental protocol was approved by the Institutional Review Board of the University and carried out according to the guidelines of the Declaration of Helsinki.

Procedures.
The participants performed three session tests with the legs and three session tests with the arms on six different sessions in random order. All the tests were performed within a period of four weeks with at least 48 hours between the sessions. Participants were instructed to avoid any strenuous activity between sessions and to follow their usual diet throughout the experimental period. All tests were performed at the same time of day to minimize the effects of circadian rhythms [30] and with similar standard environmental conditions for all participants (mean temperature and humidity: 22 ± 0.1 ∘ C and 35 ± 0.4%, resp.). Body mass and height measures of all subjects were examined before each testing session.
The participants performed a standard warm-up consisting of 5 min cycling (80 W and 50 W for men and women, resp.) before the leg tests or arm cranking (50 W and 20 W for men and women, resp.) for the arm tests, with two short accelerations (3-s) at the end of the third min and the fifth min. After 5 minutes of passive recovery, participants performed the force-velocity test which consisted of repetitive short maximal sprints of 6-s against increasing braking forces. The braking forces administrated at the beginning of the sprints cycling were 2 kg and 1.5 kg for men and women, respectively, while during arm cranking the loads were equal to 1.5 kg and 1 kg for men and women, respectively. Then, the braking force was increased after 5 min of passive recovery (sprints cycling: 2 and 1.5 kg for men and women, resp.; arm cranking: 1.5 and 1 kg for men and women, resp.) until the participant was unable to reach a peak velocity higher than 100 rpm. The same order of braking force application was respected across session tests.
All force-velocity tests were performed on a frictionloaded cycle ergometer with weights (Monark 864, Monark Exercise AB, Vansbro, Sweden) adjustable for both leg and arm exercises [31,32]. During sprint cycling exercises, participants were seated on the cycle ergometer equipped with toe clips and well-fastened straps to avoid losing the pedals. The same riding position was used throughout the study. Participants were instructed to cycle in seated position to avoid the effect of postural changes [33][34][35]. During arm cranking exercises, the pedals were replaced with handles and the cycle ergometer was fixed on a metal frame. The participants were standing on their feet in front of the ergometer during the exercises. The center of the pedal axis was approximately 20 cm lower than the level of the shoulder axis. All sprints were performed from the same initial pedal position. Participants were encouraged by the same investigator to reach the maximal velocity rate as quickly as possible. Instantaneous pedal rate in cycling or cranking was monitored throughout a PC computer by means of an encoder placed on the cycle ergometer flywheel. Then, the velocity was averaged over 1-s intervals.
The peak velocity ( ) was measured for each braking force ( ) and was used to calculate the linear force-velocity relationship for cycling exercises according to the least squares method: BioMed Research International 3 The above relationship was transformed as follows [33]: In this equation, 0 and 0 corresponded to the intercepts with the velocity axis and force axis, respectively ( 0 = and 0 = / ). Since a linear relationship between and was assumed, max corresponded to an optimal velocity and an optimal braking force equal to 0.5 0 and 0.5 0 , respectively. Hence, max was calculated as follows [5,33]: The performance variables were expressed in absolute units and according to dimensional scaling. 0 was expressed in absolute unit (rpm) and relative to body height (rpm ⋅ BH −1 ). 0 was expressed in absolute unit (kg) and relative to body mass raised to the power of 0.67 (kg ⋅ BM −0.67 ). max was expressed in absolute unit (W) and relative to body mass (W ⋅ BM −1 ).

Statistical Analyses.
Statistical procedures were carried out using Statistica 7.1 Software (StatSoft, France). Data of 0 , 0 , and max are presented as mean and standard deviation (mean ± SD). Before statistical analysis, each performance variable was tested for normality with the Shapiro-Wilk test. With the assumption of normality confirmed, systematic change in performance from trials 1 to 3 was examined using one-way ANOVA with repeated measures and a Tukey's post hoc test. All significance thresholds were set at < 0.05.
Absolute reliability, which concerns the consistency of individual's scores [36], was determined using the standard error of measurement SEM and the coefficient of variation (CV) using the following formulas [37]: where SD diff was the standard deviation of the differences between consecutive session tests (i.e., sessions 1 and 2 and sessions 2 and 3). Relative reliability, which concerns the consistency of individual's position in the group relative to others [36], was assessed using the intraclass correlation coefficient of twoway random effects model with single measure for each pair of consecutive session tests (i.e., sessions 1 and 2 and sessions 2 and 3) as follows: In this formula MS represents the participant mean square, MS represents the error mean square, is the number of trials, MS represents the trials mean square, and is the number of participants. The ICC is considered as high for values above 0.90, moderate for values between 0.80 and 0.90, and low for values below 0.80 [38].
In addition, the test-retest correlation coefficient ( test-retest ) was calculated for each pair of consecutive session tests in order to compare the results of the present study to the data in the literature [29]. The Bland-Altman plots were used to check for heteroscedasticity [28].
All the differences between men and women were highly significant ( < 0.001) even when the data were related to body mass ( max ⋅ BM −1 , 0 ⋅ BM −1 , and 0 ⋅ BM −0.67 ). The significance level of the difference in 0 ⋅ BH −1 between men and women was equal to < 0.05, only.  and cranking, resp.) and women ( (2,32) = 4.55, < 0.05 and (2,32) = 6.10, < 0.01, for cycling and cranking, resp.). Tukey's post hoc tests revealed that 0 at session 1 was significantly lower by comparison to sessions 2 and 3. In contrast, there was no significant main effect of sessions on 0 and max for arms and legs in men and women ( > 0.05).
The CV (%) of 0 , 0 , and max are presented in Tables  3 and 4. The highest CV values were obtained for 0 by comparison with 0 and max . The greatest CV values were observed for cranking exercises in female participants.
The values of test-retest are presented in Tables 3 and 4. The values of test-retest increased for the correlations between sessions 2 and 3 when compared with the correlations between sessions 1 and 2. Except 0 with the arms in women, the lowest test-retest were observed for 0 .
For the correlations between the results of the first and second sessions, the values of test-retest for 0 were significantly different between cycling and cranking but in the female group, only ( = 0.030 for 0 ; = 0.036 for 0 related to BM −0.67 ). Similarly, the values of test-retest between the first and second sessions were significantly different between male and female groups for 0 and max ( = 0.007 for 0 , = 0.005 for 0 related to BM −0.67 , and = 0.047 for max in watts). For the correlations between the results of the second and third sessions, the values of test-retest for 0 and max were significantly different between cycling and cranking but in the female group, only ( = 0.01 for 0 ; = 0.006 for 0 related to BM −0.67 and = 0.023 for max in watts). All the other comparisons of test-retest between men and women or cycling and cranking were not significantly different.   The ICC of each performance variable across sessions 1 and 2 and sessions 2 and 3 in male and female participants are presented in Tables 3 and 4. The values of ICC improved for sessions 2 and 3 by comparison with sessions 1 and 2. Excepting 0 with the arms in female participants, the lowest ICC values were observed for 0 .

Relation between the Variabilities of
in men:

Discussion
In the present investigation, we studied the reliability of max , 0 , and 0 during cycling and arm cranking exercises in active men and women. In order to study the reliability Table 4: Differences between sessions 2 and 3; coefficients of variation (CV), intraclass correlation coefficients (ICC), and test-retest correlation coefficients ( test-retest ) for 0 , 0 , and max for the leg or arm force-velocity tests in men and women. of these parameters, force-velocity tests on cycle ergometer were separately repeated three times in different sessions for each exercise. It was assumed that reliability was lower (1) in women than in men and (2) for cranking force-velocity tests than for cycling tests. The results of the present study were in agreement with this hypothesis: the reliability indices were better for the men and the leg indices when compared with the women and arm indices (Tables 3 and 4). Whatever the force-velocity parameter ( 0 , 0 , and max ), familiarisation sessions might be more important for women and arm tests as indicated by the lower values of CV in men and leg tests when the results of the first and second sessions were compared ( Table 3). The reliability of max was similar to the reliability of the different indices of maximal power in previous studies. For example, the reliability of the results of the Wingate is good for the peak power ( test-retest > 0.90) and the mean power ( test-retest between 0.91 and 0.93) [1,2,22], in contrast with the reliability of the fatigue index ( test-retest = 0.43). Similarly, the reliability of the power indices measured with the different force-velocity protocols was high when measured with isokinetic cycle ergometers [3,4,9], frictionloaded ergometers [23,24,26], or the inertial load method [6,25]. In a study by Winter et al. [23], the maximal power computed during the acceleration phase ( corr ) estimated according to Lakomy [10] was 10% higher than max but the reliability of corr was lower ( test-retest : 0.530 for corr versus 0.972 for max in men, and 0.922 for corr versus 0.952 for max in women). In the same study of Winter et al. [23], the CV values of corr were higher in men (6.9% for corr versus 2.7% for max ) but not in women (3.7% for corr versus 4.2% for max ). Furthemore, according to Winter et al. [23], these results of optimization procedures (i.e., the method of Vandewalle et al. [5]) add further support and have securer fundations than those enjoyed by correction procedures [10]. For arm exercises, Smith et al. [39] reported CV values of 4.5% for corr and 2.8% for max . It is likely that the lower reliability of corr is explained by oscillations of corr (product of and corr that takes into account not only the braking force but also the force necessary for the flywheel acceleration). On isokinetic cycle ergometers, the coefficients of variation of the slope and intercept of   Figure 3: Relationships between intersession differences in 0 (ordinates) and intersession differences in 0 (abscissae) for the leg and arm force-velocity tests in men and women. Continuous lines and black points: differences between the first and second sessions. Dashed line and empty circles: differences between the second and third sessions.
the regression between torque and pedal rate were 13.7 and 10.5%, respectively [4]. The values of CV of 0 , 0 , and max in the present study were similar to the values of CV for the different parameters measured with the inertial method (4 trials on the same day): 3.3% for corr , 2.7% for 0 , and 4.4% for 0 [6]. For friction-loaded ergometers, the reliability of the force-velocity parameters in cycling has been tested in male physical education students [26]. For 0 and max , SEE was lower than 5% and test-retest or ICC were higher than 0.90 as in the present study for the cycling force-velocity test in the male participants. However, the comparison and the validity of the reliability indices must take into account the characteristics of the data [28,37]. The data are said to be homoscedastic when the random error does not depend on the size of the measured value. Homoscedastic errors are generally expressed in the same units as those of their measurements and they can be analysed with conventional parametric analyses. SEM is valid when the data are homoscedastic. The data are said to be heteroscedastic when the random error increases as the measured values increase. Heteroscedastic data should be measured on a ratio scale (e.g., percentage) and be investigated with an analysis based on nonparametric analyses (i.e., rank tests). CV is valid even when the data are heteroscedastic. The heterogeneity of values between participants influences the results of the reliability tests.
(1) The coefficient of correlation of test-retest ( test-retest ) is sensitive to the heterogeneity of data between participants.
(2) The effect of heteroscedascity on the observed "errors" in a test-retest is low when the data range is narrow.
The spread of the data between participants is different for 0 , 0 , and max expressed in percentage of the group averages even when they are related to body dimension (Table 5). Heteroscedasticity was expected for 0 , 0 , and max raw data. However, this expectation was not confirmed with Bland-Altman plots of these data, especially in men ( Figure 4). The data ranges of parameters 0 , 0 , and max were lower than 62% in men (Table 5), which could partly explain that heteroscedascity was not suggested by the Bland-Altman plots of 0 , 0 , and max raw data (Figure 4). In Table 5: Ranges of parameters 0 , 0 , and max expressed in percentage of the means of the male or female groups. women, the data ranges were larger than in men when the ranges were expressed as percentages of the means (Table 5) but the correlations of the absolute values of the differences versus the means of the results in the first and second sessions ( Figure 5) were not significant. All other things being equal, the differences between sessions are probably lower in well-motivated individuals and experts in cycling and the average of their performances in sessions 1 and 2 should be higher (and inversely for the nonexperts and not motivated individuals). Therefore, the effects of motivation and expertise can alter the results of the Bland-Altman plot in this kind of physical tests. As in the study by Attiogbé et al. [26], the values of test-retest and ICC were lower for 0 than for 0 and max , which can be partly explained by the smaller variance of this parameter. Indeed, the range of 0 is smaller (Table 5) than the range of 0 and max . The small variance of 0 in the present study is probably an expression of the small variance of 0 when compared with the variances of 0 and max in a general athletic population [35]. The small range of 0 also probably explains that the values of CV in men and women were lower for 0 than for 0 and max in the cycling as well as the cranking force-velocity tests. Excepting the study by Buśko [21], there is no data about the changes in 0 during an annual training cycle and, therefore, it is difficult to know whether its reliability is good enough for the estimation of the training effect on this parameter. The ranges of 0 and max were similar but the values of test-retest or ICC were higher for max than for 0 (and 0 ). It is likely that the variations in 0 and 0 between sessions are not totally independent (Figure 3). Indeed, the values of 0 and 0 are extrapolated from the relationship between braking force and peak velocity. An underestimation of the peak velocity corresponding to the highest braking force induces a rotation of the F-V regression line (i.e., a more negative slope) and, consequently, an overestimation of 0 in addition to an underestimation of 0 . Inversely, an underestimation of the peak velocity corresponding to the lowest braking force induces a less negative slope of the -regression line and, consequently, and overestimation of 0 in addition to an underestimation of 0 . The value of max depends on 0 and 0 and the effect of an underestimation of 0 on max should be compensated by the effect of an overestimation of 0 , and vice versa. This could partly explain why the values of test-retest , ICC, or CV were better for max than for 0 .
The values of 0 , 0 , and max were lower in women than in men. The differences in BH and BM were not the only explanations of the lower values of 0 , 0 , and max in women. Indeed, these differences were still significant when forcevelocity parameters were related to BH or BM ( 0 ⋅ BH −1 , 0 ⋅ BM −0.67 , and max ⋅ BM −1 ). This gender effect could partly be explained by a difference in muscle fiber composition as, for example, the higher percentage of the cross-sectional area that corresponds to the slow fibers in women [40][41][42]. The lower values of 0 ⋅ BM −0.67 , 0 ⋅ BM −1 , and max ⋅ BM −1 might partly be explained by a lower percentage of lean body mass in women. The lower values of test-retest in women cannot be explained by a lower range of the individual data ( Table 5). The lower reliability in women might partly be explained by the effect of menstrual cycle, but it is possible that this effect is less important in trained women because training might reduce the cyclical hormonal fluctuations [29].
The variability of 0 and max depends on the variability of BM when these data are related to body mass ( 0 ⋅ BM −1 , 0 ⋅BM −0.67 , and max ⋅BM −1 ). In spite of the instructions about diet, hydration, and training, the standard deviations of the differences in BM between the sessions were not negligible (<1.25% of BM).

Methodological Considerations
To the best of our knowledge, this is the first study examining the reliability of force-velocity tests on cycle ergometer during sprint cycling and arm cranking exercises in active men and women. One of the limitations inherent to the experimental protocol in the present study is that the crank length was the same for all participants. The usual crank length is probably higher than the optimal length in small participants, which could partially explain the lower reliability in women. Therefore, familiarization sessions are required in small participants.

Conclusion
The present study showed high reliability of max and 0 , allowing the use of these parameters in longitudinal evaluations. Furthermore, the reliability of max was better than that of 0 whatever the expression of the results (expressed in absolute unit or data related to body dimension). The reliability indices were also better in men and cycling forcevelocity tests than in women and cranking force-velocity tests. Further studies are needed to judge the reliability of 0 .