When Antagonist Muscles at the Ankle Interfere with Maximal Voluntary Contraction under Isometric and Anisometric Conditions

Purpose: While resultant maximal voluntary contraction (MVC) is commonly used to assess muscular performance, the simultaneous activation of antagonist muscles could dramatically underestimate the strength of the agonist muscles. While quantification of antagonist torque has been performed in plantar- (PF) and dorsi-flexion (DF) joint in isometric conditions, it has yet to be determined in anisometric (concentric and eccentric) conditions. Methods: The experiment was performed in 9 participants through 2 sessions (reliability). The MVCs in DF and PF were measured in isometric, concentric and eccentric conditions (10°.s -1 ). Electromyographic (EMG) activities from the soleus, gastrocnemius medialis and lateralis, and tibialis anterior muscles were simultaneously recorded. The EMG biofeedback method was used to quantify antagonist torque, where participants were asked to maintain a level of EMG activity, corresponding to antagonist EMG activity and related to the muscle contraction type, according to a visual EMG bio-feedback displayed on a screen. Results: Resultant torque significantly underestimated agonist torque in DF MVC (30-65%) and to a lesser extent in PF MVC (3%). Triceps surae antagonist torque was significantly modified with muscle contraction type, showing higher antagonist torque in isometric (29 Nm) than eccentric (23 Nm, p < 0.001) and concentric (14 Nm, p < 0.001) conditions and resulting in modification of the DF MVC torque-velocity shape. The difference between DF eccentric and concentric MVC was attenuated when considered agonist torque (12%) rather than resultant torque (45%). Conclusion: Estimation of the antagonist torque in isometric or anisometric condition brings new insights to assessment of muscular performance and could result in costly misinterpretation in strength training and/or rehabilitation programs. muscle contraction type, Pearson correlation between agonist MVC and antagonist torque was performed. A Student T-test was performed to assess the eccentric/concentric MVC ratio between resultant and agonist MVC. When a significant main effect was found, Tukey post hoc test was performed to identify the significant differences between factors. The level of significance was set at p<0.05.


Introduction
Muscular performance is classically assessed through analysis of the mono-articular resultant torque developed during maximal voluntary contraction (MVC) under isometric or anisometric conditions. While the resultant torque assessment provides one measurement, it reflects the activation of the multiple synergist muscles around a joint (1), thereby highlighting the coactivation phenomenon (1)(2)(3)(4)(5). While it has been clearly demonstrated that antagonist muscles are electrically active during a high level of performance, antagonist assessments and their mechanical interpretation and impact on resultant torque are still unclear in the literature.
Coactivation phenomenon can be quantified by either electromyographic (EMG) activity (2,3,6) or by their corresponding mechanical contributions (7)(8)(9)(10)(11). EMG activity analysis of antagonist muscle contribution, named coactivation level, is related to the percentage of muscle activation acting as antagonist (2), and can be interpreted as the reflection of corticospinal neural drive (7). On the other hand, the mechanical contribution of antagonist muscle, named mechanical ratio, is related to the percentage of antagonist torque over the agonist torque (5), and provides measurement of mechanical output. Although coactivation level and mechanical ratio could be considered as complementary indicators of motor control output, previous studies have reported that coactivation level does not necessarily reflect the mechanical ratio depending on the muscles involved (5,(7)(8)(9)(10)(11), angle joint (7), age (8,11) and sex (9,10). Estimation of antagonist torque based on coactivation level can result in misinterpretation of about 5-15% compared to the mechanical ratio (5,7). More specifically, we previously demonstrated that plantar-flexors muscles acting as antagonist exerted a torque of about 30 N.m during dorsiflexion (DF) MVC caused an underestimation of ~70% of the maximal torque capacity of the dorsal flexors (5,7). On the other hand, with a similar coactivation level, dorsi-flexor muscles exerted a torque of about 3-6 N.m and underestimated only ~3% of plantar-flexion (PF) MVC (5,7). These findings in isometric conditions are difficult to transpose to ecological movements, and extension of our knowledge to anisometric conditions remains to be carried out. Even though coactivation level apparently does not differ between muscle contraction types in anisometric conditions (12)(13)(14)(15)(16), it has been clearly established that similar EMG activity results in greater torque in eccentric than in concentric sub-maximal contraction (17)(18)(19)(20). The only measure available in anisometric condition is the coactivation level, which does not reflect the mechanical impact of antagonist muscles.
To bridge this gap, we aimed to compare the mechanical impact of antagonist torque between eccentric, isometric and concentric in PF and DF MVCs. Based on isometric studies (5, 7), we hypothesized that resultant MVC strongly underestimates agonist MVC in DF and to a lesser extent in PF. In addition, since a similar EMG activity resulted in greater torque in eccentric than concentric conditions (17)(18)(19)(20), we expected that antagonist torque had greater mechanical impact during concentric MVC (antagonist acting in eccentric) than in eccentric MVC (antagonist acting in concentric) conditions, resulting in a modification of the torque-velocity relationship. Our results should provide substantial clues to identify muscular deficit in highlevel sport practice and rehabilitation programs.

Participants
Nine (26.1 ± 2.7 years; 1.78 ± 0.05 m; 73.4 ± 6.5 kg) patients volunteered to participate in the 2 experimental sessions of the study. Subjects had no history of ankle surgery or other orthopaedic or neurological abnormalities of the lower limb during the two years preceding the study. Written informed consent was obtained, and the experimental design was approved by the local ethics committee (CPP Est: A00064-49) and was in accordance with the principles of the Declaration of Helsinki for human experimentation.

Experimental protocol
After a short warm-up that consisted of two series of 5 sub-maximal voluntary contractions in each muscle contraction type, the subjects randomly performed two MVCs in isometric, eccentric and concentric conditions in PF and DF. If a variation of more than 5% occurred between the first and the second MVC, participants were asked to perform a third MVC. The trial resulting in the maximal torque developed was used for further analysis. During each maximal contraction, the subject was strongly encouraged by the experimenters and visual force feedback was provided. A 2-min rest period was observed between each trial to avoid any fatiguing effect on the measurements.
Thereafter, participants were asked to randomly perform 2 sub-maximal isometric contractions in DF and PF, and 10 eccentric and sub-maximal eccentric contractions corresponding to the target of an EMG activity displayed on a screen (please see 'antagonist torque estimation' section for details).

Mechanical recording
Measurements were carried out in 2 experimental sessions by participant in order to examine reproducibility of antagonist torque measurement in isometric and anisometric conditions. Torque was measured using a dynamometer (System 3; Biodex Medical Systems, Shirley, NY).
Subjects were seated with a knee angle of 120° (180° corresponds to full extension) and hip angle at 100°. The MVCs in DF and PF were measured in isometric condition at 0° (perpendicular angle between the foot and the tibia), and in concentric (+10°.s -1 ) and eccentric (-10°.s -1 ) conditions with a range of motion of 30° (i.e., -15° DF and +15° PF). To minimize trunk and hip movement during contraction, the waist was stabilized by means of a belt; arms were positioned across the chest. The right foot was attached to the dynamometer by means of the Biodex ankle attachment, which was customized with a shoe bolted to the foot plate at the heel. Standard toe straps were used over the shoe. The muscle contraction type (i.e., isometric, concentric or eccentric) and action (i.e., PF or DF) were randomized to avoid any systematic effects. In anisometric (i.e., concentric or eccentric) condition, the subjects were asked to preactivate the agonist muscle by an isometric MVC 1-s before the ergometer movement.

Electromyographic recording
EMG activity was concurrently measured with torque during maximal PF and DF efforts. EMG activity of the soleus, gastrocnemius medialis, gastrocnemius lateralis and Tibialis Anterior (TA) muscles was recorded by means of two silver-chloride surface electrodes of 10-mm diameter (Controle Graphique Medical, Brie-Comte-Robert, France), with an inter-electrode (center-to-center) distance of 25 mm. For the soleus, the recording electrodes were placed along the mid-dorsal line of the leg, a few centimeters distal from the two heads of the gastrocnemius joining the Achilles tendon. Then, following the European recommendations regarding surface electromyography and the gastrocnemii medialis and lateralis, electrodes were placed on the 7 most prominent bulge of the muscle. The soleus and the gastrocnemii medialis and lateralis constitute the Triceps Surae (TS). In addition, electrodes were set at one third of the line between the tip of the fibula and the tip of the medial malleolus for the TA muscle. The ground electrode was attached on the patella of the other leg. Low impedance (<5 kΩ) at the skinelectrode interface was obtained by shaving, abrading and cleaning the skin with an alcoholether-acetone mixture.

Torque estimation
Antagonist torque estimation. Antagonist torque was estimated by an EMG biofeedback method specifically developed to assess antagonist torque (5,11). Participants were asked to maintain a level of EMG activity according to the visual bio-feedback displayed on a screen: the current agonist EMG activity of the muscle had to correspond to its previously recorded antagonist EMG activity during the MVC. For anisometric condition, participants had to maintain EMG bio-feedback in isometric condition during 2-3-s at -15° or +15°, depending on the action and muscle contraction type. Thereafter, the movement was started and the subjects were asked to maintain the targeted EMG activity level throughout the duration of the movement. For the EMG biofeedback task, the RMS value of the EMG signal was provided by an integrated circuit, true RMS-to-DC converter (model AD536A, Analog Devices, USA; characteristics: maximal error for true RMS-to-DC conversion = 0.5 %, bandwidth > 450 kHz).
For the TS muscles, this circuit instantaneously computed the true RMS level of the amplified EMG signal for the 3 channels separately with an integration time of 375 ms. In this way, the RMS of soleus, gastrocnemii lateralis and medialis muscles were calculated and summed up by the circuit. The latter value was then displayed on an oscilloscope as visual EMG bio-feedback.
Two isometric sub-maximal contractions were performed in isometric condition with the target of the antagonist EMG activity recorded during isometric MVC; 10 sub-maximal contractions were performed in concentric condition corresponding to the antagonist EMG activity of the muscles acting in concentric condition during the eccentric MVC; and 10 sub-maximal contractions were performed in eccentric condition corresponding to the antagonist EMG activity of the muscles acting in eccentric condition during the concentric MVC. Two contractions were sufficient to reach the EMG target in isometric condition, while 10 were performed to ensure maintaining the EMG activity on the target during the movement in anisometric condition.
Agonist torque. Agonist torque is deduced by the addition of the estimated antagonist torque to the recorded resultant torque depending on the muscle contraction type: (1) For isometric contraction type: Agonist isometric MVC = Measured resultant isometric MVC + Antagonist isometric torque

Data analysis
Torque and EMG signals were acquired with a sampling frequency of 2 kHz and processed with a multi-channel analogue-digital converter (Biopac Systems Inc., USA). The EMG signal was filtered with a bandwidth frequency ranging from 10 Hz to 5 kHz, (gain of 1000). EMG-RMS was measured over a 0.5-s period after the torque had reached a plateau for isometric condition and a 0.5-s period divided into two periods of 0.25-s on both sides of the 90° position time in anisometric condition (i.e., constant angle of 2.5° under and over the 90° position) via Matlab software. For the anisometric condition, trials within 10% of the targeted EMG were kept for analysis. 91% and 71% of the trials were kept for analysis in PF and DF submaximal contraction, respectively. The ratio between eccentric and concentric MVC was calculated to compare MVC in anisometric muscle contraction type for the resultant and agonist MVC. The level of coactivation corresponds to the normalization of the EMG of a muscle acting as antagonist by the EMG activity of the same muscle maximally acting as agonist in a specific muscle contraction type (5,20). In addition, we calculated a mechanical ratio corresponding to the antagonist torque normalized by the maximal agonist torque of the same muscle in a specific muscle contraction type (5).

Statistical analysis
All statistical tests were performed with Sigma Stat software (Sigma Stat 3.5, SPSS Inc., USA).
Descriptive statistical methods, including means and their standard deviations (SDs), were calculated for each parameter. The data are presented as means ± SD in the text, the figures, and the table. Normality of the data was checked using the Kolmogorov-Smirnov test, and equality of variances was verified by the Levene test. In order to test the reliability of our measurement, the Intra-Class Correlation (ICC) was calculated between sessions 1 and 2. For this study, ICC values from 0.60 to 0.79 were considered as "good reliability", and those greater than 0.80 as "excellent reliability" (21,22). In case of excellent or good reliability (Reproducibility: Intra-Class Correlation results), data were pooled, analyzed and presented as a mean of the 2 sessions. The statistical analyses were performed separately for PF and DF MVC, or TS and TA muscles. A two-way ANOVA with repeated measures was performed to examine the effect of the MVC (resultant vs. agonist), and the muscle contraction type (eccentric, isometric, concentric) on PF and DF MVC. A one-way ANOVA with repeated measures was performed to assess the effect of muscle contraction type (eccentric, isometric, concentric) on TS and TA antagonist torque. A two-way ANOVA with repeated measures was performed to assess coactivation method (coactivation level vs mechanical ratio), and muscle contraction type (eccentric, isometric, concentric) on PF and DF MVC. Regardless of the muscle contraction type, Pearson correlation between agonist MVC and antagonist torque was performed. A Student T-test was performed to assess the eccentric/concentric MVC ratio between resultant and agonist MVC. When a significant main effect was found, Tukey post hoc test was performed to identify the significant differences between factors. The level of significance was set at p<0.05.

Reproducibility: Intra-Class Correlation
The ICC scores are presented in Table 1

TA co-activation level versus mechanical ratio.
Results showed significant main effect of muscle contraction type (p = 0.013, F = 5.71) and coactivation method (p < 0.001, F = 108.3) on coactivation quantification, whereas no significant interaction between muscle contraction type and coactivation method was found (p = 0.93, F = 0.07) ( Table 3). Post-hoc analysis showed that coactivation methods were significantly higher in isometric than in concentric (p = 0.004, t = 3.34), while no significant difference was found between isometric and eccentric (p = 0.054, t = 2.11), and between eccentric and concentric (p = 0.24, t = 1.23). In addition, coactivation level was significantly higher than mechanical ratio (p < 0.001, t = 10.41) regardless of muscle contraction type.
A significant positive linear correlation was found between DF agonist MVC and TS antagonist torque (r = .69, p < .001) (Figure 2, lower panel).

TS Co-activation level versus mechanical ratio.
Results showed no significant main effect of muscle contraction type (p = 0.54, F = 3.52) and coactivation method (p = 0.15, F = 2.55), whereas interaction between muscle contraction type and coactivation method was significant (p < 0.001, F = 16.21) ( Table 3). The coactivation level was not significantly different between isometric, concentric, and eccentric muscle contraction types (p > 0.10). However, the mechanical ratio was significantly higher in isometric than concentric (p < .001) and eccentric (p = .013) muscle contraction type. The mechanical ratio was significantly higher for eccentric compared to concentric muscle contraction type (p = .005). Post-hoc analysis showed that coactivation level was significantly lower than mechanical ratio in isometric condition (p < 0.001, t = 4.59), while coactivation level was significantly higher than mechanical ratio in eccentric condition (p = 0.42, t = 2.16). No significant difference between coactivation method was found in eccentric condition (p = 0.30, t = 1.07).   $ p < 0.001, significant difference between resultant and agonist MVC. ** p < .01, *** p < .001, significant difference between muscle contraction type for MVC and antagonist torque.

Discussion
The current study investigated the mechanical impact of antagonist muscles related to their muscle contraction type during PF and DF MVC. In accordance with our hypothesis, results showed that resultant torque significantly underestimated agonist torque with a major impact in DF MVC and a minor impact in PF MVC. In addition, TS antagonist torque depends on muscle contraction type resulting in DF MVC torque-velocity shape modification, while TA antagonist torque had no effect on PF MVC torque-velocity shape. The difference between DF eccentric and concentric MVC was attenuated considering agonist torque (12%) rather than resultant torque (45%).
In accordance with previous studies comparing resultant and agonist MVC in isometric condition (5,7,11), the current study showed the overwhelming impact of plantar-flexor muscles on DF resultant MVC (from 30% to 65%) and the minor impact of dorsi-flexor muscles on resultant PF MVC (~2%). In addition, similar results were found in concentric and eccentric muscle contraction type showing that the impact of antagonist torque on resultant MVC is highly specific as regards the muscles involved. It is safe to assume that the mechanical contribution of the dorsi-flexor muscles could be neglected when assessing PF performances (5,7,11), Focusing on plantar-flexors muscles, our results did not highlight any differences between the coactivation level in isometric, concentric and eccentric muscle contraction type. Although these results corroborate the literature (12)(13)(14)(15)(16), the antagonist torque was more significantly impacted by muscle contraction type. Our results indeed showed that the estimated TS antagonist torque was significantly greater in isometric than eccentric than concentric. As a consequence, we observed that isometric agonist MVC became higher than eccentric and concentric agonist MVC (+16.5% and +22%, respectively), while resultant torque showed greater eccentric MVC than isometric and concentric muscle contraction type. The shape of the 'DF MVC-angular velocity' relationship was thereby modified, highlighting lesser difference between eccentric and concentric MVC when considering agonist (12%) rather than the resultant (45%) MVC. In echo to the force-velocity relation of a muscle or a single fiber, in vitro studies reported higher force during lengthening (eccentric) than shortening (concentric) or static (isometric) muscle contraction type (24,25). While in vitro studies have reported that lengthening contraction force represents 1.4-1.9 times than isometric (25)(26)(27)(28), in situ studies have reported slight increases (15,16,20,(29)(30)(31)(32) or no change (13,17,(33)(34)(35)(36) in eccentric compared with isometric MVC. It can be admitted that the shape of the force-velocity relation depends only on muscular factors in isolated fibers, whereas both neural and muscular components contribute to MVC in vivo. In addition, the specific control of muscles during lengthening is still discussed in the literature (18,19). In a recent review, Duchateau and Enoka (19) identify three points characterizing neural specificities during eccentric contractions. The authors reported that (i) EMG activity is most of the time lower in eccentric than in concentric effort (37)(38)(39)(40)(41); (ii) the neural activation is most of the time lower in eccentric than in concentric contraction (13,33,42); (iii) peak discharge rate of motor units is lower in eccentric than I concentric contraction (43). All in all, studies have paradoxically reported an alteration of neural component and greater resultant MVC in eccentric compared to concentric contraction.
The mechanical titin's resistance mechanism described in the enhanced force in eccentric contractions may be still greater than the deficit of neural efficiency in eccentric compared to concentric contractions (44,45). It could therefore be suggested that the greater PF eccentric resultant MVC observed in the literature is partially due to the smaller antagonist torque in concentric in comparison with isometric and eccentric muscle contraction type.
Electromechanical efficacy, i.e. the ratio between EMG activity and the mechanical output, should be considered to evaluate antagonist coactivation.
The quantification of the mechanical contribution of each muscle group in torque production at a specific joint may highlight the deficits between agonist and antagonist torques in athletes or different pathologies. In a practical way, assessment of the mechanical contribution of antagonist torque in isometric and anisometric conditions provides new opportunities to investigate antagonist muscles in injured athletes, for example by monitoring the antagonist/agonist ratio in hamstring injury (46,47) or preventively. In addition, this approach could improve understanding of the mechanical output synergies in muscles of patients with neurological disorders such as post-stroke patients (23,48).

Methodological considerations
Isometric and anisometric antagonist torque estimations were performed under sub-maximal level corresponding to a targeted EMG activity (5,7,8,11). While this method was easy to perform in isometric conditions with only two sub-maximal contractions needed, it was more challenging in anisometric conditions, particularly when a low level of EMG activity was targeted. Given this context, 71% and 91% of data were retained for analysis in DF and PF submaximal contraction, respectively. Consequently, reliability measurement showed good to excellent reliability for TA antagonist torque (0.78 to 0.90), while excellent reliability was observed for TS antagonist torque (0.84 to 0.90). All in all, our study indicated that the quantification of antagonist torque with the EMG biofeedback method is easy and reliable in anisometric conditions when medium and high levels of EMG activity is targeted (>10%) and more challenging for a low level of EMG activity (<10%).

Conclusion
In addition to evaluation of the resultant MVC and the coactivation level, it matters to consider the mechanical role of the antagonist muscles when assessing muscular performance in isometric and anisometric contractions. The antagonist torque was substantially altered by muscle contraction type affecting the 'MVC-angular velocity' relationship in DF, reducing the gap between eccentric and concentric MVC. All in all, our findings provide new elements to assess performance in strength training and rehabilitation programs.