Beruflich Dokumente
Kultur Dokumente
Fo
Comparison of different regression models to fit the force-
velocity relationship of a knee extension exercise
rP
Manuscript ID RSPB-2017-0247.R2
Complete List of Authors: Iglesias-Soler, Eliseo; University of A Coruna. Performance and Health
ev
Oviedo, Spain
Jaric, Slobodan; University of Delaware, Kinesiology and Applied Physiology
Sports Biomechanics
Submission checklist
A. For new submission and resubmission. (Please only check YES when these items are met, or
else the manuscript will be returned to you or immediately rejected.)
Yes
1. Have you added continuous line numbers from abstract to appendices?
2. Have you used the following sections in the manuscript: Abstract, Introduction,
Methods, Results, Discussion and Implications, Conclusion, References? (Note that
we use ‘Discussion and Implications’ but not only ‘Discussion’ as we request authors
to highlight the practical implications in sports biomechanics.)
3. Have you provided a testable hypothesis at the end of Introduction, i.e. ‘We
hypothesise that…’ or ‘It is hypothesised that…’. (Please answer the hypothesis in
Fo
5. Have you placed your documents in the following order: submission checklist,
response to reviewer’s comments (resubmission only), title page, abstract, keywords,
main text, acknowledgements, references, appendices, table(s) with caption(s), figure
ee
Yes
6. Have you prepared your point-by-point response to the Editor-in-Chief, Associate
ev
Editor, and each reviewer’s comments? Do not group the responses. Place this
document as the first document preceding the main text.
7. Have you provided a clean version of the revision with the changes highlighted by
iew
coloured text or highlighter? Do not use track changes function as it makes the
manuscript messy.
8. Have you used British English spelling? (e.g. analyse but not analyze, analysing but
not analyzing, normalise but not normalize, normalising but not normalizing, etc.
Search for ‘ize’, ‘yze’, ‘zing’ and make appropriate edits.)
On
9. Have you used ‘single quotation marks’ instead of “double quotation marks”?
10. Is the abstract in a single unstructured paragraph and within 200 words?
11. Have you provided 3-5 keywords which are not included in the title?
12. Have you used ‘participants’ instead of ‘subjects’ to refer to the human participants in
ly
the study?
13. Have you checked the preferred units, symbols and numbers? (e.g. SI unit preferred,
m/s but not ms-1, avoid too many significant figures beyond the accuracy of your tool,
use the same significant figures for mean and standard deviation, etc.)
14. Have you make sure your illustrations are readable with adequate font size and clear
graph line, and also when printed black-and-white? Have you provided the
illustrations in the highest quality, placed them as separated files in the image format
(i.e. JPG, TIFF)?
15. Are there no more than 8 figures?
Editor's Comments to Author: I will accept this one if you can well
address the last final comment from a reviewer. In addition, please make
sure all items on both Part A and B on the submission checklist are met
Dear Editor, thank you very much for your comments and consideration. We
Fo
believe that we were able to address all your suggestions and concerns. That
rP
also applies to the reviewer 2 comment that you also emphasized. We are
anyway ready to make new changes if the editor or reviewer consider them
ee
necessary.
rR
Country.
On
2. Line 11: If you have set the p <= 0.05 as the threshold for significance
then here please denote as p < 0.05. Why is it p <= 0.015? On Line 210 why
We actually set alpha at p < 0.05. We regret the inconsistent text that has been
because this was the maximum value observed for post-hoc comparisons of
linear and exponential with polynomial model. To avoid confusion we wrote p <
0.05 in the abstract as suggested. Finally, we have amended the mistake of line
210 as follows:
“Significantly lower Adj. R2 values were detected by post hoc Wilcoxon signed
rank tests for both LM and EM compared with PM for every limb (p ≤ 0.015 in all
Fo
cases)”.
rP
3. Line 83: Use kg/m2. Also on Line 532 and Table 2. This is one of the
ee
manuscript.
iew
4. Line 87: Give the full name of the ethics committee, e.g. The University
5. Line 159: Is that gain 1,000 (one thousand) or 1.000 (one)? I saw a
6. Reference list: give page numbers of the part you have cited for books
(Line 432).
7. Table 3 (Line 570): If you use p <= 0.05 as the threshold then here it
should not be p < 0.05, but p <= 0.05. Please change either one.
Fo
As previously said we actually set alpha at p < 0.05. We have corrected this
rP
mistake in the Statistical Analysis section
ee
8. Figure 4 caption: Please use open and close bracket to make them (A),
rR
(B), etc.
Reviewer: 2
clear that the comparison of MVC with Fmax is an invalid one because
average over range of motion vs. fixed angle, and different lever arms
Fo
from the point of rotation on the machine which would change force at the
rP
ankle for the same force at the stack. Thus, the authors should remove
manuscript.
rR
is fine and can remain as there is a sensible hypothesis that the two
iew
We are grateful to the reviewer for the comments and suggestions. We agree
On
with the reviewer that F0 and MVC don´t evaluate the same force modality.
ly
relevant since the latter is a standard test typically used to assess the maximum
differences, our approach can provide the data that depict the maximum force
standard isometric MVC test. In other words, we believe that the discussed
the reviewer´s concern we have added the next paragraph in the discussion
section:
“Finally, the divergences between F0 and MVC can also be attributed to the fact
that the two variables do not exactly evaluate the same force production
modality since the former is extrapolated from dynamic mean values whereas
TITLE PAGE
Title:
1
Department of Physical Education and Sport, Faculty of Sports Sciences and
rP
2
Active and Healthy Lifestyle Observatory, Centre for Sport Studies, King Juan
rR
3
University School of Sports Medicine, University of Oviedo, Oviedo, Spain
ev
4
Department of Kinesiology and Applied Physiology & Biomechanics and Movement
iew
Corresponding author:
On
Eliseo Iglesias-Soler
ly
Fax: +34981167048
e-mail: eliseo@udc.es
Acknowledgments
Disclosure Statement:
1 Abstract
2 The aims of this study were to compare the goodness of fit and the concurrent
5 maximum load in the dominant and non-dominant leg were obtained in 24 male
11 significantly higher than the rest of the models (p < 0.05), while the standard error of
rR
12 estimate was slightly higher for the linear than for polynomial model (p = 0.001). MVC
ev
13 was underestimated by F0 calculated from the linear and polynomial models, while
14 the maximum power was accurately estimated by the linear model. In summary,
iew
15 while the polynomial model revealed somewhat better fit, the linear model more
16 accurately estimates the maximum power and provides the parameters of apparent
On
21 regression.
22
23 Introduction
24 The inverse relationship between muscular force and shortening velocity has been
25 consistently established both for isolated animal muscles and humans muscle groups
26 since the first half of 20-th century. Specifically, the seminal studies showed that this
28 functions that can be expressed by either exponential (Fenn & Marsh, 1935) or
Fo
29 hyperbolic equations (Hill, 1938; Wilkie, 1949). This type of upward concave curves
31 exercises, such as the knee extension or elbow flexion (MacIntosh, Herzog, Suter,
ee
32 Wiley, & Sokolosky, 1993; Tihanyi, Apor, & Fekete, 1982; Toji, Suei, & Kaneko, 1997;
33 Valour, Ochala, Ballay, & Pousson, 2003). When multijoint and monoarticular
rR
34 isoinertial exercises were analysed, the force-velocity relationship has been well
ev
35 fitted by linear models (Iglesias-Soler et al., 2016; Jaric, 2015; Rahmani et al., 1999;
36 Samozino, Rejc, Di Prampero, Belli, & Morin, 2012) although the comparison
iew
37 between linear and nonlinear models was not performed. On the other hand no
39 models have been observed for both vertical jump (Cuk et al., 2014) and bench press
40 throws (Sreckovic et al., 2015) but, to the best of our knowledge, no previous study
ly
41 has compared the goodness of fit of linear and nonlinear models for monoarticular
44 performance (Samozino et al., 2012). However, the analysis of the validity of the
45 parameters obtained from linear vs. nonlinear models has not been previously
46 reported.
49 evaluated and compared between contralateral limbs. In this regard, previous studies
50 (Samozino, Rejc, di Prampero, Belli, & Morin, 2014; Yamauchi, Mishima, Nakayama,
51 & Ishii, 2009) have used the linear model to explore the phenomenon of bilateral
52 deficit and asymmetries in the performance of the lower limb, but these asymmetries
53 were neither explored with nonlinear models nor the individual´s laterality was
Fo
54 considered (i.e. dominant vs. non-dominant limb). Differences between limbs are of
rP
55 particular importance in clinical assessments and monitoring post-operative
56 procedures, therefore the differences between the dominant and non-dominant need
ee
58 performance of the dominant and non-dominant limb are equally detected from the
60 Thus, the aims of this study were (a) to analyse the goodness of fit of linear and
iew
62 the unilateral isoinertial knee extension exercise; (b) to contrast the force-velocity
On
63 relationship obtained for dominant and non-dominant limbs with linear and nonlinear
65 parameters estimated from different models with respect to those directly measured
66 (i.e. the maximum strength and power); (d) to evaluate the relationship between
69 would be similarly fitted by linear and the nonlinear quadratic polynomial and
70 exponential models, so the simplest model (i.e. linear) would be a more efficient tool
71 for obtaining the mechanical profile for the knee extension exercise.
72 Methods
73 Experimental design
75 unilateral knee extension exercise in both dominant and non-dominant leg. Following
77 used for obtaining for each leg the force-velocity relationship and the one-repetition
Fo
78 maximum load (1RM). Finally, MVC of the knee extensor muscles was obtained in
81 Participants
rR
82 Twenty four male sports sciences students participated in this study (age, 22 ± 2
ev
83 years; height: 176 ± 8 cm; weight, 72.8 ± 9.0 kg; BMI, 23.31 ± 1.84 kg/m2). All of
84 them were physically active through their standard curriculum that included between
iew
85 6 and 8 hours a week of moderate physical activity and were free of musculoskeletal
86 pain or injury during the study. Participants signed a written informed consent before
On
87 participating in the study which was approved by the University of A Coruna Ethics
88 Committee.
ly
91 performed to allow the participants to acquire experience with the exercise technique
92 and to record the individual references to standardise the position for performing the
93 tested task.
95 velocity decrement with increasing load. This evaluation was conducted in the third
96 visit to the laboratory at least 48 hours after finishing the second familiarisation
97 session. The participant´s dominant leg was identified as that usually used to kick a
99 bicycle ergonometer (Monark 828E; Monark Exercise AB, Vansbro, Sweeden) with a
100 resistance of 1.25 W/kg of body mass, each leg was evaluated in a randomised order
Fo
101 using a knee extension cam-system machine (Technogym, Gambettola, Italy) with
rP
102 hip and knee angle set at 90º. The lever arm was individually adjusted by placing the
103 padded surface for the application of force at the level of the medial malleolus. Knees
ee
104 were placed at the point of the axis of rotation of the machine by adjusting the
rR
105 position of the seat and seat back. These adjustments were recorded to standardise
106 the individual exercise position throughout the experiment. Minimum movement
ev
107 range was set at 80º (i.e. from 90º to 10º of knee flexion) visually controlled using a
iew
108 graded scale placed at the axis of rotation. Participants´ hip and trunk were
109 immobilised by straps while the arms had to be maintained crossed at the chest
111
ly
113 The protocol started with the participants performing three knee extension repetitions
114 with a load of 15 kg. After a 1-min rest, the second trial of three repetitions was
115 performed with a load increment of 10 kg. This procedure was repeated until mean
116 velocity (MV) decreased by 25% compared with the best repetition corresponding to
117 the first load. In some cases, limitations in a participant´s range of movement (e.g.
118 shortening of ischiotibial muscles) caused the leg to lose contact with the lever arm at
119 the end of the movement, which may have resulted in an underestimation of the
120 velocity and force applied by the muscular action. Thus, only the portion of the
121 concentric period during which the acceleration was greater than acceleration due to
122 gravity was considered to obtain MV and the rest of the mechanical variables (see
123 below). Thereafter, successive sets of two repetitions with a 2-min rest interval
124 between sets were performed with load increments of 5 kg until MV decreased by
Fo
125 50% with respect to the initial performance. At this point, load increments were
rP
126 limited to 1-2.5 kg, with only one repetition required in each trial, with a 3-min rest
127 interval. This procedure was repeated until the participant was not able to complete
ee
128 at least 80º of knee extension. 1RM was defined as the maximum load properly lifted.
rR
129 After a 5-min period of rest, the procedure was repeated with the contralateral limb.
ev
130 To record the MV, mean force (MF), and mean power during the concentric phase of
131 each repetition, a dynamic measuring device (T-Force System, Ergotech, Murcia,
iew
132 Spain) was connected to the knee extensor machine to record the vertical
133 displacement of the load. Validity and reliability of this device have been previously
On
138 subsequently smoothed with a 4th order low-pass Butterworth filter with a cut-off
140 differentiation of V(t) with respect to time, and instantaneous force (F(t)) was
141 obtained as F(t) = m·(A(t) + g) where m is the mass of the weight stack plates and g
142 is the acceleration caused by the gravity. Instantaneous power (P(t)) was calculated
144 Participants were strongly encouraged during the test to perform each repetition with
145 the maximal velocity. When a load required performing more than one repetition, the
148 The maximum isometric voluntary contraction (MVC) was assessed in a modified
rP
149 knee extension machine (BF100, Biotech Bioiso, Brazil) with the hips and knees
150 flexed 90º. This position was selected in order to simulate the starting position in the
ee
152 Furthermore, this position has been frequently used to evaluate MVC of the knee
153 extensors (Wilson & Murphy, 1996). Participants had to perform an MVC with each
ev
154 limb and in a randomised order “as fast and as forcefully as possible” maintaining it
iew
155 for 4 s. A 3-minute rest period was provided between the assessments of both limbs.
156 In the case that the attempt was not considered maximal, another trial was performed
157 3 min later. Data were recorded using a force cell (sensitivity: 2 mV/V and 0.0028
On
158 VN; NL63-200, Digitimer Ltd, Welwyn Garden City, UK) that was attached to the
ly
159 knee extension machine, at a sample frequency of 5,000 Hz and filtered with a
160 bandwidth frequency ranging from 10 Hz to 1 kHz (gain = 1,000). MVC in N was
163 The relationship between values of velocity and force was obtained for each
164 participant using values of MV and MF recorded for every load of the 1RM test.
166 (i.e. ordinates) and MF as the predictor variable (i.e. abscissae). Additionally the
167 maximum mean power (MP) recorded throughout the progressive test was
169 Three models were considered for obtaining each individual regression: linear (LM),
170 quadratic polynomial (PM), and exponential model (EM). A detailed explanation of
171 these models is reported in the Appendix. The parameters obtained from each model
Fo
172 were: force intercept (F0), velocity intercept (V0), slope (s), and maximum power
173 (Pmax) for LM; F0, V0, Pmax and the concavity of the curve (a) for PM; and V0, Pmax and
rP
174 the force constant at which Pmax is obtained (k) for EM.
ee
176 Statistical analysis was performed using SPSS 20 (IBM, Armonk, NY, USA) and
177 GraphPad Prism software package for Windows (v5.03 GraphPad Prism Software
ev
178 Inc., San Diego, CA, USA). Normal distribution of data was explored by using
iew
179 Shapiro-Wilk test. Descriptive results are presented as means ± standard deviation
180 except in the case of the goodness of fit parameters that are presented as medians
181 and ranges. When normal distribution could not be assumed, nonparametric tests
On
182 were used. Adjusted coefficients of determination (Adj. R2) of individual regressions
ly
183 were used to compare the goodness of fit between LM, PM, and EM. Friedman´s test
184 was used in this case and post hoc paired comparisons were performed with
185 Wilcoxon signed-rank test with Bonferroni adjustment. Likewise, standard errors of
186 estimate (SEE) were contrasted between LM and PM by using paired t-test and
187 Wilcoxon signed-rank test if normality could be assumed or not respectively. SEE of
188 EM was not contrasted between PL and LM because it does not have an additive
189 scale (Glantz & Slinker, 1990). Similarly, a comparison between the dominant and
190 non-dominant limb was performed by paired t-test and Wilcoxon signed-rank test. To
191 analyse the association between the parameters of the force-velocity relationship
192 with the measurements of performance (1RM, MVC, and MP), Pearson´s correlation
193 coefficient or Spearman´s Rho were used if normality could be assumed or not,
194 respectively.
195 To assess the concurrent validity of the parameters F0 and Pmax estimated from three
Fo
196 regression models with the directly measured performance (MVC and MP), Pearson
198 Wilcoxon signed-rank test were used. Additionally, concurrent validity was graphically
ee
199 analysed by Bland-Altman plots, obtaining biases and 95% limits of agreement (LoA).
200 When a significant positive correlation was detected between the absolute
rR
201 differences and mean values, the natural logarithms of both measurements were
ev
202 used as previously recommended (Nevill & Atkinson, 1997). In these cases, by taking
203 antilogs of the bias, the dimensionless ratio between geometric means is calculated
iew
204 while antilogs of LoA result from bias multiplied or divided by a second ratio that
205 indicates the level of agreement. Thus, values close to 1 represent concordance
On
206 between methods (Nevill & Atkinson, 1997). Alpha was set at p < 0.05.
207 Results
ly
208 Regarding Adj. R2, Friedman test was significant for both the dominant (p = 0.002)
209 and non-dominant limb (p = 0.004). Significantly lower Adj. R2 values were detected
210 by post hoc Wilcoxon signed rank tests for both LM and EM compared with PM for
211 every limb (p ≤ 0.015 in all cases). SEE was significantly higher for LM in comparison
212 with PM in the dominant and in the non-dominant limb (p = 0.001 in both legs; Table
213 1). As previously indicated, SEE of EM was not compared with the rest of models
216 No significant differences were detected between the dominant and non-dominant
217 leg for both 1RM (70 ± 13 and 71 ±12 kg in the dominant and non-dominant limb,
218 respectively; p = 0.489) and MVC (1321 ± 295 and 1276 ± 268 N for the dominant
Fo
219 and non- dominant leg, respectively; p = 0.319). Likewise, MP of the dominant and
221 0.976).
ee
222 Wilcoxon test did not detect significant differences between limbs for s (p = 0.189), a
rR
224 paired-samples t-test did not reveal significant differences between the limbs for F0
ev
225 estimated from LM (p = 0.193), F0 obtained from PM (p = 0.745), V0 of LM, PM, and
iew
228 MVC values were significantly higher than those estimated by LM and PM (p < 0.001
229 for all the cases). F0 calculated from LM correlated significantly with MVC (r = 0.872
ly
230 and 0.855 for dominant and non-dominant limbs, respectively; p < 0.001 in both
231 cases). F0 estimated from PM, was also significantly associated with MVC in
232 dominant limbs (r = 0.545; p = 0.006) but only a tendency was observed for the non-
234 Pmax estimated from LM was 264 ± 52 and 263 ± 55 W for the dominant and non-
235 dominant limbs, respectively. Comparisons of these variables with MP were non-
236 significant (p = 0.174 and 0.620, respectively). Pmax calculated from PM was 255 ± 48
237 and 254 ± 49 W for dominant and non-dominant legs, respectively. These values
238 were significantly lower than MP (p = 0.001 in both cases). Mean values of Pmax
239 calculated from EM were 247 ± 50 and 247 ±51 W for dominant and non-dominant
241 Pmax of EM and MP in both limbs (both p < 0.001). All the correlations between Pmax
242 and MP were statistically significant (p < 0.001): LM (r = 0.987 and 0.976 for
Fo
243 dominant and non-dominant limb, respectively), PM (r = 0.987 and 0.983 for
rP
244 dominant and non-dominant legs, respectively) and EM (r = 0.988 and 0.981 for
246 In the Bland-Altman plots for comparisons between MVC and F0, significant
rR
247 correlations were detected between the absolute differences and mean values, thus
ev
248 natural logarithms of measurements were used to calculate the bias and LoA (Figure
249 1). Antilogs of bias (i.e. the ratio between geometric mean of MVC and F0) in the
iew
250 dominant limb were 1.46 (LoA = [1.06, 1.99]) and 1.37 (LoA = [0.93, 2.03]) for LM
251 and PM , respectively. In the non-dominant leg, the same values were 1.37 (LoA =
On
253
ly
255 Antilogs of bias for comparison between MP and Pmax for the dominant limb were
256 0.99 (LoA = [0.92, 1.06]), 1.03 (LoA = [0.96, 1.09]) and 1.06 (LoA = [1.00, 1.13]) for
257 LM, PM, and EM, respectively. The same values for the non-dominant leg were 0.99
258 (LoA = [0.91, 1.09]), 1.03 (LoA = [0.96, 1.11]) and 1.06 (LoA = [0.98, 1.15]) (Figure
259 2).
260
262 Correlations between the parameters of mechanical profile (F0, V0, k, and a) and
263 variables of performance (1RM, MVC, and MP) were obtained from Pearson´s
264 correlation coefficients for V0 in both limbs and F0 in the non-dominant leg, and
265 Spearman´s Rho in the remaining variables (Table 3). Coefficient a from PM was
266 negative (i.e. downward concavity) in 3 and 4 participants for the dominant and non-
Fo
267 dominant legs, respectively (Figure 4). Therefore, we decided to report both the
268 results for the whole sample and exclusively for participants with upward concavity
rP
271
273 The main findings of this study were: (a) the goodness of fit of the force-velocity
iew
274 relationship in an isoinertial knee extension exercise was high for the three models
275 although it was slightly higher for PM in comparison with LM and EM; (b) in
276 concordance with the lack of differences in performance, there were no significant
On
277 differences between limbs regarding the parameters obtained from each model; (c)
ly
278 the magnitudes of F0 and MVC were markedly different despite being highly
279 correlated, while Pmax appeared to be a valid estimation of MP, especially in the case
280 of the LM; and (d) excluding the quadratic coefficient of PM, and V0 of LM and EM,
281 most of parameters calculated from each model positively correlated with the
282 muscular performance, although these associations were particularly high for F0
284 Our results showed that PM fits the force-velocity relationship better than LM and
285 EM. These results are partially coincident with those previously reported showing that
286 when mean values of force and velocity from a bench press exercise were used and
287 averaged through the sample, the correlation coefficients of the polynomial
288 regressions were higher than that obtained by linear regressions (Sreckovic et al.,
289 2015). On the contrary, some studies reported no differences in the coefficient of
290 correlation of linear and polynomial regressions for the force-velocity relationship
Fo
291 obtained from vertical jumps (Cuk et al., 2014; Feeney, Stanhope, Kaminski, Machi,
rP
292 & Jaric, 2016). However, in these studies Adj. R2 and SEE were not compared
293 between models. In this regard, Adj. R2 is appropriate to contrast the goodness of fit
ee
294 between models since including another independent variable, as in the case of PM
rR
295 with respect to LM, increases the coefficient of determination and therefore the
297 freedom. Adj. R2 takes this into account allowing the comparison of models with
iew
298 different numbers of variables (Glantz & Slinker, 1990). On the other hand, SEE
299 expresses the variability of the obtained regression lines and represents a proper
300 criterion for selecting the best regression model (Glantz & Slinker, 1990). A quadratic
On
301 polynomial regression has been previously used to approach the force-velocity
ly
302 relationship, as a way to reflect the upward concavity of this relationship for several
303 exercises such as leg press (Allison, Brooke-Wavell, & Folland, 2013), a multijoint
304 lower limbs exercise (Rejc, Lazzer, Antonutto, Isola, & di Prampero, 2010), or a
305 isokinetic knee extension (Callahan & Kent-Braun, 2011). Nonetheless, this model
306 was not compared with others, which represents an apparent novelty of our study.
307 Nevertheless, it must be pointed out that although lower than in PM, LM goodness of
308 fit was exceptionally high, and confirms the findings of a recent study showing a
309 linear force-velocity relationship in unilateral isokinetic knee extension within a wide
310 range of contraction velocities when the isometric force is excluded (Grbic et al.,
311 2017). Therefore, the obtained results suggest an advantage of the linear model in
312 this type of exercise considering its simplicity and the physiological meaning of its
313 parameters.
314 Regarding the comparison between the dominant and non-dominant legs, the
Fo
315 parameters derived from the models proved to be similar between the limbs. The
316 same applies regarding the measured performance. This result contrasts the
rP
317 asymmetries observed in young football players (Daneshjoo, Rahnama, Mokhtar, &
ee
318 Yusof, 2013) and suggests a practical application in the context of injuries and
319 surgeries in general physical active population, since we can presume that non-
rR
320 injured leg can be used as a reference for the mechanical profile of the injured limb.
ev
321 Results of the present study showed a low concurrent validity of F0 to estimate the
MVC of knee extension at 90º of knee flexion. Although there was a significant
iew
322
323 correlation between F0 and MVC in both limbs and models (i.e. LM and PM),
324 significant differences between assessments, values of bias, and wide LoA revealed
On
325 that the magnitudes of two variables do not correspond. Previous studies reported
326 the validity of F0 obtained from a LM with respect to directly recorded 1RM (Garcia-
ly
327 Ramos, Jaric, Padial, & Feriche, 2016; Sreckovic et al., 2015), MVC (Cuk et al.,
328 2014), and peak torque of isokinetic tests (Feeney et al., 2016), but in all these cases
329 validity was analysed simply by Pearson´s correlation, which is not appropriate to
330 assess the agreement between two tests or evaluations (Nevill, 1996). In this regard,
331 our results expand those previously reported, adding a more comprehensive
333 regression with respect to MVC. It must be pointed out that F0 obtained from LM and
334 PM were significantly lower than the directly recorded MVC, which contrasts with
335 other studies in which F0 overestimated MVC (Rahmani, Viale, Dalleau, & Lacour,
336 2001). Differences in the type of exercise and isometric test used could explain these
337 discrepancies. It must be considered that for the knee extension exercise the last
338 load (i.e. 1RM) was usually moved at velocities higher than 0.20 m/s and, therefore,
339 F0 was a result of a relatively remote extrapolation of force associated with null
340 velocity. Furthermore, the applied force could be underestimated since the mass of
Fo
341 the lever, the leg inertia and the possible role of friction were not taken into account
rP
342 (Rahmani et al., 2001; Rahmani et al., 1999; Rambaud, Rahmani, Moyen, & Bourdin,
343 2008). Finally, the divergences between F0 and MVC can also be attributed to the
ee
344 fact that the two variables do not exactly evaluate the same force production modality
rR
345 since the former is extrapolated from dynamic mean values whereas the latter is a
347 extrapolating average values obtained over the entire range of motion which entails
iew
348 variable lever arms, whereas MVC is performed at a fixed angle and therefore at a
351 validity regarding the measured MP, since the correlation between them was high, no
ly
352 significant differences between them were detected, the bias was close to 1, and LoA
353 were quite narrow. This was also the case for Pmax obtained from PM and EM, except
354 that the estimated values were significantly lower than the values directly recorded.
355 This is also a novel finding of this study that suggests that LM better predicts MP
356 than the other 2 models. This result seems to be contradictory to the fact that SEE
357 was higher for LM than for PM, but it could be explained by the mathematical
358 characteristics of both models. The simplicity of LM entails that Pmax is calculated
359 only using F0 and V0 while in the case of PM the calculation of maximum power is
360 based on three estimated coefficients (see Appendix). Thus, it can be hypothesized
361 that these three coefficients could cause a higher accumulated deviation from Pmax
363 Except for the parameter a of the quadratic polynomial, and V0 estimated from LM
364 and EM, the parameters calculated from each model correlated positively with 1RM,
Fo
365 MVC, and MP, although the associations of V0 obtained from PM were weaker and
366 significant only in 3 out of 6 cases (Table 3). The lower correlations observed for V0
rP
367 could be caused by the characteristics of the sample comprised of young participants
ee
368 since it has been shown that the improvement in power after resistance training is
369 dependent on both the force and velocity for older individuals but only on the force for
rR
370 younger participants (Petrella, Kim, Tuggle, & Bamman, 2007). On the other hand, a
ev
371 remarkable result was that the correlations of F0 obtained from LM with 1RM, MVC,
372 and MP were consistently higher than those corresponding to PM (Table 3), which
iew
373 suggests a better physiological meaning of this variable when it is obtained from a
374 LM. Additionally, correlations of a with 1RM, MVC and MP were not significant even
On
375 when only positive values of this coefficient were considered. A negative value of a
376 means that the force-velocity profile has a downward concavity contrary to classic
ly
377 models. This kind of profile was observed for 3 and 4 participants in the dominant
378 and non-dominant leg, respectively; and corresponds to profiles previously reported
379 and attributed to inhibitory mechanisms (Perrine & Edgerton, 1978; Wickiewicz, Roy,
380 Powell, Perrine, & Edgerton, 1984). Less curvature has been observed in individuals
381 with higher percentages of fast-twitch fibres (Tihanyi et al., 1982) whereas fatigue
382 shifts the force-velocity curve downward (De Ruiter, Jones, Sargeant, & De Haan,
383 1999). However, our results suggest that curvature is not indicative of muscular
384 performance in a unilateral knee extension exercise. Further studies are warranted in
385 order to study the meaning of this coefficient in other types of exercises.
386 There are some potential limitations in this study that should be noted. Firstly, the
387 exercise was performed using a cam-system machine that entailed variable moment
388 arm. Secondly, friction, mass of lever arm, and leg inertia were not considered when
389 calculating the force. Thirdly, we used mean values of force and velocity although it
Fo
390 has been previously suggested that the maximum values could provide a better
391 reliability (Garcia-Ramos et al., 2016). However, other studies revealed a similar
rP
392 reliability for the two types of variables (Cuk et al., 2014; Sreckovic et al., 2015) while
ee
393 the mechanical output averaged over the entire movement was suggested to be
394 more representative of the muscular effort (Samozino et al., 2014). Fourthly, MVC
rR
395 was obtained only for 90º of knee flexion and, therefore, the question remains
ev
396 whether the validity differs at other knee angles. Lastly, a number of other major
398 Conclusion
399 Although the goodness of fit of the force-velocity relationship for the isoinertial knee
On
400 extension exercise was somewhat better with PM, LM was more accurate when
ly
401 estimating MP. Considering comprehensively both the simplicity and utility of each
402 model, in addition to their goodness of fit and the physiological meaning of its
403 parameters, it seems that LM should be recommended for research and routine
404 testing of the isoinertial knee extension exercise. In that case, the testing procedure
405 could be simplified by using only two distinctive loads (i.e. a light and a heavy load)
406 as previously suggested (Jaric, 2016; Zivkovic, Djuric, Cuk, Suzovic, & Jaric, 2016).
407 References
408 Allison, S. J., Brooke-Wavell, K., & Folland, J. P. (2013). Multiple joint muscle
409 function with ageing: The force-velocity and power-velocity relationships in young
410 and older men. Aging Clinical and Experimental Research, 25, 159-166.
411 doi:10.1007/s40520-013-0024-y
412 Callahan, D. M., & Kent-Braun, J. A. (2011). Effect of old age on human skeletal
413 muscle force-velocity and fatigue properties. Journal of Applied Physiology, 111,
Fo
414 1345-1352. doi:10.1152/japplphysiol.00367.2011
rP
415 Cuk, I., Markovic, M., Nedeljkovic, A., Ugarkovic, D., Kukolj, M., & Jaric, S. (2014).
416 Force-velocity relationship of leg extensors obtained from loaded and unloaded
ee
418 doi:10.1007/s00421-014-2901-2
ev
419 Daneshjoo, A., Rahnama, N., Mokhtar, A. H., & Yusof, A. (2013). Bilateral and
422 doi:10.2478/hukin-2013-0005
On
423 De Ruiter, C. J., Jones, D. A., Sargeant, A. J., & De Haan, A. (1999). The
ly
427 Feeney, D., Stanhope, S. J., Kaminski, T. W., Machi, A., & Jaric, S. (2016). Loaded
428 vertical jumping: Force-velocity relationship, work, and power. Journal of Applied
430 Fenn, W. O., & Marsh, B. S. (1935). Muscular force at different speeds of shortening.
432 Garcia-Ramos, A., Jaric, S., Padial, P., & Feriche, B. (2016). Force-velocity
433 relationship of upper body muscles: Traditional versus ballistic bench press.
435 Glantz, S. A., & Slinker, B. K. (1990). Primer of applied regression and analysis of
Fo
436 variance (pp. 245-255). New York: McGraw-Hill, Health Professions Division.
rP
437 Grbic, V., Djuric, S., Knezevic, O. M., Mirkov, D. M., Nedeljkovic, A., & Jaric, S.
ee
438 (2017). A novel two-velocity method for elaborate isokinetic testing of knee
440 0043-113043
ev
441 Hill, A. V. (1938). The heat of shortening and the dynamic constants of muscle.
iew
442 Proceedings of the Royal Society of London. Series B, Biological Sciences, 126,
443 136-195.
On
444 Iglesias-Soler, E., Fernandez-Del-Olmo, M., Mayo, X., Farinas, J., Rio-Rodriguez, D.,
446 profile after short resistance training programmes differing in set configurations.
449 maximum performance tasks. International Journal of Sports Medicine, 36, 699-
451 Jaric, S. (2016). Two-load method for distinguishing between muscle force, velocity,
453 doi:10.1007/s40279-016-0531-z
454 MacIntosh, B. R., Herzog, W., Suter, E., Wiley, J. P., & Sokolosky, J. (1993). Human
455 skeletal muscle fibre types and force: Velocity properties. European Journal of
459 Nevill, A. M., & Atkinson, G. (1997). Assessing agreement between measurements
460 recorded on a ratio scale in sports medicine and sports science. British Journal of
rR
462 Perrine, J. J., & Edgerton, V. R. (1978). Muscle force-velocity and power-velocity
iew
463 relationships under isokinetic loading. Medicine and Science in Sports, 10, 159-
464 166.
On
465 Petrella, J. K., Kim, J. S., Tuggle, S. C., & Bamman, M. M. (2007). Contributions of
466 force and velocity to improved power with progressive resistance training in
ly
467 young and older adults. European Journal of Applied Physiology, 99, 343-351.
468 doi:10.1007/s00421-006-0353-z
469 Rahmani, A., Viale, F., Dalleau, G., & Lacour, J. R. (2001). Force/velocity and
472 Rahmani, A., Belli, A., Kostka, T., Dalleau, G., Bonnefoy, M., & Lacour, J. (1999).
475 doi:10.1123/jab.15.3.337
476 Rambaud, O., Rahmani, A., Moyen, B., & Bourdin, M. (2008). Importance of upper-
477 limb inertia in calculating concentric bench press force. Journal of Strength and
Fo
478 Conditioning Research, 22, 383-389. doi:10.1519/JSC.0b013e31816193e7
rP
479 Rejc, E., Lazzer, S., Antonutto, G., Isola, M., & di Prampero, P. E. (2010). Bilateral
480 deficit and EMG activity during explosive lower limb contractions against different
ee
482 doi:10.1007/s00421-009-1199-y
ev
483 Samozino, P., Edouard, P., Sangnier, S., Brughelli, M., Gimenez, P., & Morin, J. B.
484 (2014). Force-velocity profile: Imbalance determination and effect on lower limb
iew
486 doi:10.1055/s-0033-1354382
On
487 Samozino, P., Rejc, E., Di Prampero, P. E., Belli, A., & Morin, J. B. (2012). Optimal
ly
490 doi:10.1249/MSS.0b013e31822d757a
491 Samozino, P., Rejc, E., di Prampero, P. E., Belli, A., & Morin, J. B. (2014). Force-
494 doi:10.1249/MSS.0b013e3182a124fb
496 neuromuscular fatigue during resistance training. Medicine and Science in Sports
498 Sreckovic, S., Cuk, I., Djuric, S., Nedeljkovic, A., Mirkov, D., & Jaric, S. (2015).
503 fiber composition in human knee extensor muscles. European Journal of Applied
ee
505 Toji, H., Suei, K., & Kaneko, M. (1997). Effects of combined training loads on
ev
506 relations among force, velocity, and power development. Canadian Journal of
508 Valour, D., Ochala, J., Ballay, Y., & Pousson, M. (2003). The influence of ageing on
511 Wickiewicz, T. L., Roy, R. R., Powell, P. L., Perrine, J. J., & Edgerton, V. R. (1984).
514 435-443.
515 Wilson, G. J., & Murphy, A. J. (1996). The use of isometric tests of muscular function
517 Wilkie, D. R. (1949). The relation between force and velocity in human muscle. The
519 Yamauchi, J., Mishima, C., Nakayama, S., & Ishii, N. (2009). Force-velocity, force-
520 power relationships of bilateral and unilateral leg multi-joint movements in young
522 doi:10.1016/j.jbiomech.2009.05.032
Fo
523 Zivkovic, M. Z., Djuric, S., Cuk, I., Suzovic, D., & Jaric, S. (2016). A simple method
rP
524 for assessment of muscle force, velocity, and power producing capacities from
526 doi:10.1080/02640414.2016.1221521
rR
527
ev
528
iew
On
ly
529 Appendix
531 Force-Velocity linear model: Considering velocity (V) as the dependent variable, the
535 force) and s is the slope of the linear regression. Additionally, from this equation is
rP
536 obtained the theoretical maximum value of force (F0) when V equals zero (i.e. force
ee
537 axis intercept). Thus s can be also obtained by the following equation:
rR
542 Force-Velocity quadratic polynomial model: The quadratic polynomial model of the
On
545 where V0 has the same meaning as for the linear approach and a is the value of the
546 second derivative and therefore it is a measurement of the concavity of the curve.
547 The value of F0 can be obtained by calculating the root of the polynomial at which the
548 first derivative is negative (i.e. the function is decreasing at that point). In those cases
549 in which the polynomial does not have a real root, F0 can be approximated by
550 calculating the minimum of the function (i.e. the point at which the first derivative
551 equals 0). Finally, the expression for power of this model is:
553 Therefore, the value of force at which Pmax is performed (FP) can be estimated
554 by equating to 0 the first derivative of eq.5 and calculating the root of the
562 where V0 has the same meaning as previously and k is the exponential force
563 constant in N that corresponds to the force value at which the velocity is reduced to
On
564 0.37 times V0. However, with this model, F0 cannot be calculated since 0 is only
567 Obtaining the first derivative of eq.8 and equating to 0, it can be deduced that
570
Table 1. Goodness of fit of each regression model of the force velocity relationship.
Data are presented as medians and ranges in brackets
LM PM EM
Adj. R2 Dominant 0.964* 0.984 0.971*
(0.885-0.992) (0.906-0.995) (0.778-0.990)
Non- 0.965* 0.982 0.972*
dominant (0.901-0.991) (0.910-0.994) (0.887-0.989)
SEE (m/s) Dominant 0.037* 0.031
(0.022-0.078) (0.016-0.082)
Non- 0.039* 0.033
dominant (0.020-0.072) (0.017-0.045)
Adj. R2: adjusted coefficient of determination; SEE: standard error of estimate; LM:
Fo
Linear model; PM: quadratic polynomial model; EM: exponential model. *Significant
differences in comparison with PM.
rP
ee
rR
ev
iew
On
ly
Table 2. Mechanical profiles of the dominant and non-dominant leg (data averaged across the participants)
Fo
LM PM EM
F0 V0 s F0 V0 a V0 k
(N)
rP
(m/s) (1/103 m/s/N) (N) (m/s) (1/106 m/s/N2) (m/s) (N)
ee
Dominant 898 ± 163 1.18 ± 0.09 -1.36 ± 0.30 950 ± 170 1.32 ± 0.17 0.75 ± 0.90 1.57 ± 0.23 433 ± 97
rR
Non-dominant 924 ± 167 1.14 ± 0.10 -1.27 ± 0.25 939 ± 207 1.29 ± 0.19 0.86 ± 0.86 1.52 ± 0.24 451 ± 111
LM: linear model; PM: quadratic polynomial model; EM: exponential model; F0: force intercept; V0: velocity intercept; s: slope; a:
ev
quadratic parameter of the quadratic polynomial; k: exponential parameter.
iew
On
ly
571 Table 3. Correlations between parameters calculated from each model and performance assessments.
Fo
1RM MVC MP
rP
Dominant. Non-dominant Dominant Non-dominant Dominant Non-dominant
ee
s 0.786** 0.655** 0.501* 0.577* 0.690** 0.686**
rR
F0 LM 0.953** 0.907** 0.754** 0.685** 0.915** 0.936**
PM
EM
0.350
-0.115
0.439*
0.004
0.526**
-0.153* iew
0.192
-0.182
0.424*
0.054
0.451*
-0.027
ly
a>0 -0.136 -0.121 0.227 0.095 -0.077 -0.042
572 LM: linear model; PM: quadratic polynomial model; EM: exponential model; F0: force intercept; V0: velocity intercept; s: slope; a:
573 quadratic coefficient of the polynomial model; k: exponential force constant; 1RM: 1-repetition maximum; MVC: maximum voluntary
574 contraction. * p < 0.05; ** p < 0.001.
576 Figure 1. Example of the initial position in the knee extension machine.
577
578 Figure 2. Bland-Altman plots showing the agreement between the actual
579 measurement of maximum voluntary contraction (MVC) and the estimated maximum
580 force (F0). Ln: natural logarithm; LM: linear model; PM: quadratic polynomial model;
Fo
581 EM: exponential model; LoA: limits of agreement.
582
rP
583 Figure 3. Bland-Altman plots showing the agreement between the maximum mean
ee
584 power (MP) and the maximum estimated power (Pmax). Ln: natural logarithm; LM:
585 linear model; PM: quadratic polynomial model; EM: exponential model; LoA: limits of
rR
586 agreement.
ev
587
588 Figure 4. Force-Velocity profiles of two representative participants: (A) with positive
iew
589 a coefficient from the PM (i.e. upward concavity); (B) with negative a coefficient from
590 PM (i.e. downward concavity). The solid lines represent linear regression, the dotted
On
591 lines represent polynomial regression, and the dashed lines represent the
Fo
rP
ee
rR
ev
iew
On
ly
Fo
rP
ee
rR
ev
Figure 2. Bland-Altman plots showing the agreement between the actual measurement of maximum
iew
voluntary contraction (MVC) and the estimated maximum force (F0). Ln: natural logarithm; LM: linear
model; PM: quadratic polynomial model; EM: exponential model; LoA: limits of agreement.
Fo
rP
ee
rR
Figure 3. Bland-Altman plots showing the agreement between the maximum mean power (MP) and the
maximum estimated power (Pmax). Ln: natural logarithm; LM: linear model; PM: quadratic polynomial
model; EM: exponential model; LoA: limits of agreement.
ev
Fo
rP
ee
rR
ev
iew
On
ly
Figure 4. Force-Velocity profiles of two representative participants: A) with positive a coefficient from the PM
(i.e. upward concavity); B) with negative a coefficient from PM (i.e. downward concavity). The solid lines
represent linear regression, the dotted lines represent polynomial regression and the dashed lines represent
the exponential regression.