Real-time mechanical responses to overload and fatigue using a flywheel training device

1 Original article 1 REAL-TIME MECHANICAL RESPONSES TO OVERLOAD AND FATIGUE 2 USING A FLYWHEEL TRAINING DEVICE 3 4 Authors: Alejandro Muñoz-López1, Marco Pozzo2, Pablo Floria3 5 6 Affiliations: 7 1Departamento de Motricidad Humana y Rendimiento Deportivo, University of Seville, 8 Seville, Spain 9 2SmartCoach Technologies, CA, USA 10 3Physical Performance & Sports Research, Universidad Pablo de Olavide, Seville, Spain 11 12 Corresponding author: 13 14 Alejandro Muñoz-López. Departamento de Motricidad Humana y Rendimiento 15 Deportivo, Education Sciences Faculty, University of Seville, Spain. E-mail: 16 [email protected] 17 18 Preferred running head: mechanical responses using flywheel 19 Abstract Word count: 247 20 Text-only Word count: 3779 21 Number of figures: 6 22 Number of tables: 2 23 24 25 26 27 28 29 30 2 Abstract 31 Purpose: This study aimed to analyze mechanical performance and mechanical losses 32 across different sets and repetitions of unilateral leg extensions under flywheel loading 33 conditions. Methods: In a cross-over design, 23 physically active participants executed 34 3 sets of 30 repetitions using two different loads. Angular speed, angular acceleration 35 and power were monitored, and mean and peak values of these variables were 36 calculated for each set of repetitions. We analyzed time-course differences every five 37 repetitions (described as ‘clusters’). Differences in the mechanical profile for Load by 38 Set and in the mechanical losses for Load by Set by Cluster repetitions were analyzed 39 using an ANOVA repeated measures test. A Principal Components (PC) analysis was 40 also conducted. Results: Only peak acceleration showed a significant Set by Load 41 interaction (p=0.006) in mechanical performance. At the intra-set level, all consecutive 42 clusters showed a significant decrement in all variables, but with different magnitudes 43 (speed < acceleration < power). Although all variables showed significant Set by 44 Cluster and Load by Cluster interactions, only the acceleration variables showed a 45 significant Set by Load interaction (p<0.05). All the variables showed a single PC. 46 Higher fatigue was found around the concentric peak for each variable. Conclusions: 47 The most sensitive variable for detecting fatigue using real-time feedback seems to be 48 the angular acceleration of the flywheel. Peak acceleration can be used to monitor 49 training workload and training volume in real time in leg extension exercises using 50 flywheel training devices. 51 52 Keywords 53 Feedback, level of effort, workload, volume, strength training 54 3 Introduction 55 Flywheel rotary training devices (FRTDs) have been used over the last 25 years in 56 resistance training programs to enhance performance via the induction of mechanical 57 fatigue (Berg and Tesch, 1994; Núñez et al., 2020). Recently, Sabido et al. (2018) 58 highlighted the lack of information on basic training using FRTD, especially with 59 regard to intensity and volume training variables. 60 Training intensity (workload) is typically managed using external loads (i.e., weights) in 61 resistance training (RT) programs, and is often defined and prescribed as a percentage 62 of one-repetition maximum (Kraemer and Ratamess, 2004). Some studies have 63 analyzed the overload characteristics of FRTDs in RT (Carroll et al., 2019; Castillo et 64 al., 2019; Martinez-Aranda and Fernandez-Gonzalo, 2017; Sabido et al., 2018; Spudić 65 et al., 2020; Vázquez-Guerrero et al., 2016). Studies have shown a relationship between 66 the moment of inertia used and the mechanical outcome. Mechanical variables such as 67 peak force (Carroll et al., 2019), mean force (Spudić et al., 2020), peak linear velocity 68 (Carroll et al., 2019; Spudić et al., 2020), mean linear velocity (Carroll et al., 2019), 69 mean power (Castillo et al., 2019) and peak power (Martinez-Aranda and Fernandez- 70 Gonzalo, 2017) have been used for this purpose. 71 When using free weights, there is an external load that is sufficiently large it can only be 72 lifted once (the one-repetition maximum, 1RM). Weight resistance has an inertial 73 component and a constant component (the weight force), which is preponderant and 74 constitutes the minimum force threshold that must be overcome to initiate movement, 75 and below which failure occurs. FRTD resistance is a function of the flywheel moment 76 of inertia (Carroll et al., 2019) and the shaft diameter (Núñez et al., 2020). Dependency 77 on the gravitational component has been shown in some exercises using FRTDs (more 78 4 commonly during vertical component exercises) (Sjöberg et al., 2020). However, the 79 contribution of gravity is low in FRTDs, and frictional components also have a low 80 effect on mechanical loading. Thus, as long as Force is higher than zero, movement will 81 always occur. Therefore, training workload cannot be programmed as a percentage of 82 1RM in FRTDs, as it is with free weights, making it difficult to establish a proper 83 relative workload to work with when using FRTDs. 84 A linear relationship between relative exercise workload (i.e., percentage of 1RM) and 85 concentric speed has been found in free weight exercises (González-Badillo and 86 Sánchez-Medina, 2010). Carroll et al. (2019) showed a negative linear relationship 87 between the peak or mean concentric linear velocity and the moment of inertia in 88 FRTDs, while Spudić et al. (2020) found a linear force-velocity profile, as occurs with 89 free weights (Rahmani et al., 2018). Worcester et al. (2020) showed that the vertical and 90 angular velocity of the knee joint was reduced when the moment of inertia was 91 increased. Therefore, while force is increased, the speed of the flywheel decreases. In 92 weight-lifting exercises, force is the product of the mass of the load and the linear 93 acceleration. In FRTDs, torque is computed instead of force (Sabido et al., 2018), as the 94 product of the moment of inertia and the angular acceleration. Acceleration can thus be 95 computed in both cases. Consequently, characterizing the mechanical output over sets 96 and loads can provide interesting information in relation to training workload. To our 97 knowledge, this has not yet been studied in FRTDs using either angular velocity or 98 angular acceleration. 99 In addition to training workload, the training volume is also programmed during RT 100 exercises. For weightlifting exercises, volume has been traditionally programmed using 101 a fixed number of sets and repetitions. However, recent work supports the use of 102 5 velocity-based training to determine numbers of repetitions, suggesting that a higher 103 velocity loss within the set will result in higher acute fatigue (Sanchez-Medina et al., 104 2010; Sánchez-Medina and González-Badillo, 2011). The level of effort in this regard 105 will determine the chronic adaptations in RT (Pareja-Blanco et al., 2017). Therefore, 106 real-time measures of mechanical outcomes can guide decisions about the appropriate 107 volume of exercise (i.e., sets and/or repetitions) to be used. To date, only Sabido et al. 108 (2018) have studied this in FRTDs, showing a decrement in power with increased 109 repetitions. They also showed that the number of repetitions required to show a 110 statistical drop compared to the optimal repetitions (e.g., the highest power output) 111 decreased with the inertial load used. Understanding how other outcomes, such as 112 angular speed or angular acceleration, respond to different numbers of sets and 113 repetitions, may help in the development of more effective real-time monitoring of the 114 training volume in FRTDs. 115 In RT, the training workload (i.e., mechanical output) and the training volume (i.e., 116 mechanical loss) determine the effects of the program (González-Badillo and Sánchez- 117 Medina, 2010; Pareja-Blanco et al., 2017). Hence, it is important to be able to monitor 118 both training workload and training volume, in real time, when using FRTDs. The main 119 purpose of this study was to analyze the mechanical concentric output across several 120 sets and repetitions, with different external loads, during a single resistance training 121 session using an FRTD. A second objective was to analyze mechanical fatigue (i.e., 122 reduction in mechanical outcome in response to increasing numbers of repetitions or 123 sets) in the concentric phase of FRTD exercise. 124 Methods 125 Subjects 126 6 Twenty-three healthy men with resistance training experience (men, n= 15, 21.40 ± 0.99 127 years, 74.85 ± 6.95 kg, 1.79 ± 0.07 m, Leg extension isometric peak torque= 273.33 ± 128 63.28 and women, n= 8, 21.50 ± 1.38 years, 62.05 ± 5.79 kg, 1.62 ± 0.07 m, Leg 129 extension isometric peak torque= 196.85 ± 49.70) participated in this study. The sample 130 size was determined using G*Power software (v.3.1.9.2, Universität Kiel, Germany). 131 For an alpha error of 0.05, effect size f equal to 0.40 and a power of 80%, we obtained a 132 recommended sample size of twenty-three. We provided all subjects with extensive 133 information on the purposes and risks associated with the study, and all of them gave 134 written consent. This study was approved by a local Ethics Committee, following the 135 Declaration of Helsinki standard guidelines. 136 Design 137 This study followed a cross-over design, where two different training loads were 138 randomized and crossed for each participant. A week before the intervention, the 139 participants undertook two familiarisation sessions with the exercise protocol. For each 140 session, the protocol consisted of three sets of unilateral leg extension exercises on an 141 FRTD with the dominant limb. A gap of a week between training sessions was used as a 142 wash-out period for the cross-over. Before each measurement session, the participants 143 performed a warm-up consisting of 5 minutes exercise on a cyclo-ergometer (Ergoselect 144 200, Ergoline, Germany) at 100W power and a cadence of 75rpm. They then performed 145 two sets of 8 isotonic leg extensions at submaximal load (40 and 80N·m-1) on an 146 isokinetic dynamometer (Biodex System 4, Biodex Medical Systems, Shirley, NY, 147 USA). In addition, prior to and immediately after each exercise set, the participants 148 performed three 5s-long maximal voluntary contractions with 30s of rest between each 149 attempt. 150 7 Procedures 151 The main exercise consisted of unilateral knee extension bouts of 3 sets of 30 152 repetitions each, using an FRTD with a cylindrical flywheel shaft (Kbox 3, Exxentric, 153 Sweden). We chose this number to ensure the onset of fatigue. The values used as 154 flywheel moments of inertia were 0.025 and 0.050 kg·m2 (Martinez-Aranda and 155 Fernandez-Gonzalo, 2017). The first two repetitions were used to get the flywheel 156 spinning and were not considered in the analysis (Sabido et al., 2018). Participants 157 performed maximal effort throughout the entire concentric phase. In the subsequent 158 coupled eccentric phase, we instructed them to commence braking in the last third of the 159 range of motion. Verbal encouragement was given throughout the exercise. Body 160 position and range of motion (from 90 to 180deg – full extension - knee angle) were 161 standardized with participants strapped to the seat of the isokinetic dynamometer. 162 We fastened the flywheel device to the back of the dynamometer, in a vertical position. 163 An ankle support was attached to the participant's ankle around both ankle malleoli. We 164 horizontally adjusted the flywheel pulling strap with the leg flexed at 90deg for each 165 participant (Figure 1). A 5 minutes rest time was given between sets, during which the 166 participants remained strapped in position. We continuously measured the raw angular 167 speed of the flywheel using a quadrature, incremental rotary encoder (EMS22Q, 168 Bourns), with a resolution of 256 pulses per revolution, fastened to the flywheel shaft. 169 Digital encoder pulses were fed into a multipurpose acquisition system (SmartCoach 170 MultiChannel, SmartCoach Technologies Inc., CA, USA). Angular speed was acquired 171 at a sample rate of 100 samples/s using ancillary computer software (SmartPlot V4.7.0, 172 SmartCoach Technologies Inc., CA, USA). Then, we imported these data (one raw data 173 file for each participant, repetition, set, and load) into Matlab R2019a (The MathWorks 174 8 Inc., Natic, MA, USA) for processing and analysis. We smoothed the angular speed 175 (Speed) using a Gaussian-weighted moving average filter. Then, we calculated the 176 angular acceleration (acceleration) as the time derivative of filtered angular speed, and 177 torque as the derivative of angular momentum (moment of inertia multiplied by angular 178 acceleration). Finally, we calculated power as the product of torque and speed. The 179 intraclass correlation coefficients (3, k) of the angular speed for the first three 180 repetitions of the first set were 0.97 (0.95 – 0.99 95% confidence interval) and 0.96 181 (0.92 – 0.98 95% confidence interval) for the 0.025 kg·m2 and 0.050 kg·m2 loads, 182 respectively. 183 <<<INSERT FIGURE 1 ABOUT HERE>>> 184 For time-course analyses, we segmented all the variable curves into repetitions. For 185 each repetition, we defined the beginning of the concentric phase as the point at which 186 the flywheel acceleration changed from negative to positive values, and the beginning 187 of the eccentric phase as the point of change from positive to negative values. We only 188 used data from the concentric phase in this study. We calculated mean and peak 189 concentric values for speed, acceleration, and power for each repetition. For more 190 straightforward time-course interpretation and analyses, we only considered the values 191 for every five repetitions (clusters). The first cluster was represented as the highest 192 mechanical outcome for each variable. 193 Statistical analysis 194 Data are shown as mean ± standard deviation (SD). First, we tested Gaussian 195 distribution using the Shapiro-Wilk test of normality. Then, we calculated the R- 196 squared (r2) measure of goodness of fit for mechanical fatigue using all the repetitions 197 for each Set and Load for each variable, using a Microsoft Excel 2016 spreadsheet 198 9 (Microsoft Corporation). We calculated the mechanical profile as the highest repetition 199 magnitude within the first three repetitions. Then, we tested the possible differences for 200 each variable in Load by Set using a two-way ANOVA repeated measures test. In 201 addition, we analyzed mechanical fatigue using the same analyses, but following a 202 three-way design (Load by Set by Cluster). In both cases, we used a post-hoc test with 203 Bonferroni adjustment for pairwise comparisons. In addition to these analyses, we 204 calculated the variation in performance between sets, calculating the typical error, 205 expressed as coefficient of variation (CV) using a customized spreadsheet (Hopkins, 206 2015). 207 As a secondary analysis, we carried out a Principal Component Analysis (PCA) in 208 Matlab (Deluzio et al., 2014). First, we resampled the dataset for each parameter and 209 each repetition on a 0 - 100 relative time scale (101 points = 100 intervals) using a 210 linear length normalization procedure. For each variable, we created a matrix (630 x 211 101) containing the time series data of all participants’ clusters (23 participants x 2 212 loads x 3 sets x 6 clusters x 101 relative time points per cluster). The PCA of these 213 matrices returned eigenvector components, eigenvalues, and scores. The eigenvalues 214 indicated the amount of variation for a given principal component. A criterion of 90% 215 of the variance was used to determine the number of principal components (PC) 216 extracted for analysis (Deluzio et al., 2014). 217 Results 218 Mechanical performance 219 Data from the best-averaged repetition (mechanical performance) are shown in Table 1. 220 Only peak acceleration showed a significant Set by Load interaction (p=0.006). As a 221 main effect, Set and Load showed significant changes in all the variables. Post-hoc 222 16 The authors would like to thank the Sports Science degree students at the University of 365 Seville; AR, JC, JP and GR, who were actively involved in data acquisition and 366 participant recruitment for this study. The authors declare that no financial assistance 367 was received for this project. 368 Conflict of interest 369 MP is the owner of Smartcoach Europe AB, but was not involved in the data acquisition 370 or the statistical analyses. 371 Bibliography 372 Bagley, J.R., McLeland, K.A., Arevalo, J.A., Brown, L.E., Coburn, J.W., Galpin, A.J., 373 2017. Skeletal Muscle Fatigability and Myosin Heavy Chain Fiber Type in 374 Resistance Trained Men. J. Strength Cond. Res. 31, 602–607. 375 https://doi.org/10.1519/JSC.0000000000001759 376 Berg, H.E., Tesch, P.A., 1994. A gravity-independent ergometer to be used for 377 resistance training in space. Aviat. Sp. Environ. Med. 65, 752–756. 378 Carroll, K.M., Wagle, J.P., Sato, K., Taber, C.B., Yoshida, N., Bingham, G.E., Stone, 379 M.H., 2019. Characterising overload in inertial flywheel devices for use in exercise 380 training. Sport. Biomech. 18, 390–401. 381 https://doi.org/10.1080/14763141.2018.1433715 382 Castillo, D., Domínguez, R., Rodríguez-Fernández, A., Raya-González, J., 2019. 383 Effects of caffeine supplementation on power performance in a flywheel device: a 384 randomised, double-blind cross-over study. Nutrients 11, 255. 385 https://doi.org/10.3390/nu11020255 386 Chorley, A., Lamb, K.L., 2020. The Application of Critical Power, the Work Capacity 387 17 above Critical Power (W′), and Its Reconstitution: A Narrative Review of Current 388 Evidence and Implications for Cycling Training Prescription. Sports. 389 https://doi.org/10.3390/sports8090123 390 Deluzio, K.J., Harrison, A.J., Coffey, N., Caldwell, G.E., 2014. The analysis of 391 biomechanical waveform data, in: Robertson, D.G.E., Caldwell, G.E., Hamill, J., 392 Kamen, G., Whittlesey, S.N. (Eds.), Research Methods in Biomechanics. Human 393 Kinetics, pp. 312–337. 394 González-Badillo, J.J., Sánchez-Medina, L., 2010. Movement velocity as a measure of 395 loading intensity in resistance training. Int. J. Sports Med. 31, 347–52. 396 https://doi.org/10.1055/s-0030-1248333 397 González-Badillo, J.J., Yañez-García, J.M., Mora-Custodio, R., Rodríguez-Rosell, D., 398 2017. Velocity loss as a variable for monitoring resistance exercise. Int. J. Sports 399 Med. 38, 217–225. https://doi.org/10.1055/s-0042-120324 400 Hopkins, W.G., 2015. Spreadsheets for Analysis of Validity and Reliability. 401 Sportscience 39, 16–42. 402 Kraemer, W.J., Ratamess, N.A., 2004. Fundamentals of Resistance Training: 403 Progression and Exercise Prescription. Med. Sci. Sports Exerc. 36, 674–688. 404 https://doi.org/10.1249/01.MSS.0000121945.36635.61 405 Martinez-Aranda, L.M.M., Fernandez-Gonzalo, R., 2017. Effects of inertial setting on 406 power, force, work, and eccentric overload during flywheel resistance exercise in 407 women and men. J. Strength Cond. Res. 31, 1653–1661. 408 https://doi.org/10.1519/JSC.0000000000001635 409 Núñez, F.J., Galiano, C., Muñoz-López, A., Floria, P., 2020. Is possible an eccentric 410 18 overload in a rotary inertia device? Comparison of force profile in a cylinder- 411 shaped and a cone-shaped axis devices. J. Sports Sci. 38, 1624–1628. 412 https://doi.org/10.1080/02640414.2020.1754111 413 Pareja-Blanco, F., Rodríguez-Rosell, D., Sánchez-Medina, L., Sanchis-Moysi, J., 414 Dorado, C., Mora-Custodio, R., Yáñez-García, J.M., Morales-Alamo, D., Pérez- 415 Suárez, I., Calbet, J.A.L.L., González-Badillo, J.J., 2017. Effects of velocity loss 416 during resistance training on athletic performance, strength gains and muscle 417 adaptations. Scand. J. Med. Sci. Sports 27, 724–735. 418 https://doi.org/10.1111/sms.12678 419 Rahmani, A., Samozino, P., Morin, J.B., Morel, B., 2018. A simple method for 420 assessing upper-limb force–velocity profile in bench press. Int. J. Sports Physiol. 421 Perform. 13, 200–207. https://doi.org/10.1123/ijspp.2016-0814 422 Sabido, R., Hernández-Davó, J.L., García-Valverde, A., Marco, P., Asencio, P., 2020. 423 Influence of the Strap Rewind Height during a Conical Pulley Exercise. J. Hum. 424 Kinet. 74, 109–118. https://doi.org/10.2478/hukin-2020-0018 425 Sabido, R., Hernández-Davó, J.L., Pereyra-Gerber, G.T., 2018. Influence of different 426 inertial loads on basic training variables during the flywheel squat exercise. Int. J. 427 Sports Physiol. Perform. 13, 482–489. https://doi.org/10.1123/ijspp.2017-0282 428 Sánchez-Medina, L., González-Badillo, J.J., 2011. Velocity loss as an indicator of 429 neuromuscular fatigue during resistance training. Med. Sci. Sports Exerc. 43, 430 1725–1734. https://doi.org/10.1249/MSS.0b013e318213f880 431 Sanchez-Medina, L., Perez, C.E., Gonzalez-Badillo, J.J., 2010. Importance of the 432 propulsive phase in strength assessment. Int. J. Sports Med. 31, 123–9. 433 19 https://doi.org/10.1055/s-0029-1242815 434 Sjöberg, M., Berg, H.E., Norrbrand, L., Andersen, M.S., Gutierrez-Farewik, E.M., 435 Sundblad, P., Eiken, O., 2020. Influence of gravity on biomechanics in flywheel 436 squat and leg press. Sport. Biomech. 1–17. 437 Spudić, D., Smajla, D., Šarabon, N., 2020. Validity and reliability of force–velocity 438 outcome parameters in flywheel squats. J. Biomech. 107, 109824. 439 https://doi.org/10.1016/j.jbiomech.2020.109824 440 Talbot, J., Maves, L., 2016. Skeletal muscle fiber type : using insights from muscle 441 developmental biology to dissect targets for susceptibility and resistance to muscle 442 disease 5. https://doi.org/10.1002/wdev.230 443 Tous-Fajardo, J., Maldonado, R.A., Quintana, J.M., Pozzo, M., Tesch, P.A., 2006. The 444 flywheel leg-curl machine: offering eccentric overload for hamstring development. 445 Int. J. Sports Physiol. Perform. 1, 293–298. https://doi.org/10.1123/ijspp.1.3.293 446 Vázquez-Guerrero, J., Moras, G., Baeza, J., Rodríguez-Jiménez, S., 2016. Force outputs 447 during squats performed using a rotational inertia device under stable versus 448 unstable conditions with different loads. PLoS One 11, 1–13. 449 https://doi.org/10.1371/journal.pone.0154346 450 Worcester, K.S., Baker, P.A., Bollinger, L.M., 2020. Effects of Inertial Load on Sagittal 451 Plane Kinematics of the Lower Extremity During Flywheel-Based Squats. J. 452 Strength Cond. Res. 1. https://doi.org/10.1519/jsc.0000000000003415 453 454 Figure Legends 455 20 Figure 1. Flywheel device setting used. The Kbox3 was vertically positioned (1) and 456 fastened to the back of the isokinetic dynamometer. The individual pulled the flywheel 457 strap (2) using a support attached to the ankle (3). 458 Figure 2. Representative random example of single set execution raw data for angular 459 speed, angular acceleration, and power. Time execution is represented on the horizontal 460 axis. A total of 30 waveforms can be observed in angular speed, representing 30 461 repetitions (one waveform equals one repetition). 462 Figure 3. Averaged clusters (black bars) time-course for Mean (A) and Peak (B) speed 463 over the three sets for each moment of inertia. The error bars represent the standard 464 deviation. Between-sets significant differences (p<0.05) for the same cluster are shown 465 as † (with Set 1), ‡ (with Set 2) or $ (with Set 3). All clusters were significantly 466 different. Averaged values for each repetition are also shown for representative 467 visualization as white bars. 468 Figure 4. Averaged clusters (black bars) time-course for Mean (A) and Peak (B) 469 acceleration over the three sets for each moment of inertia. The error bars represent the 470 standard deviation. Between-sets significant differences (p<0.05) with the same cluster 471 are shown as † (with Set 1), ‡ (with Set 2) or $ (with Set 3). All clusters were 472 significantly different. Averaged values for each repetition are also shown for 473 representative visualization as white bars. 474 Figure 5. Averaged clusters (black bars) time-course for Mean (A) and Peak (B) power 475 over the three sets for each moment of inertia. The error bars represent the standard 476 deviation. Between-sets significant differences (p<0.05) with the same cluster are 477 shown as † (with Set 1), ‡ (with Set 2) or $ (with Set 3). All clusters were significantly 478 21 different. Averaged values for each repetition are also shown for representative 479 visualization as white bars. 480 Figure 6. Angular speed, angular acceleration and power averaged clusters waveforms 481 for each moment of inertia during the normalized concentric phase. Darker curves show 482 first clusters. The dotted grey lines represent the load vector for each principal 483 component found. The vertical green dotted lines represent the part of the load vector, 484 which represents the 90 to 100% of the load vector values (area with higher waveform 485 differences). 486