Introduction

Tardieu has postulated that the major factor limiting active movement in spastic paresis is the antagonist (Tardieu 1966). It has been suggested in particular that the responsiveness of agonist motoneurones to descending commands may be impacted by the stretch imposed on the antagonist (Gracies 2005). In healthy subjects, motoneurone output is known to be sensitive to agonist stretch, depending on whether contraction is eccentric or concentric (Abbruzzese et al. 1994; Sekiguchi et al. 2001, 2003). The primary question addressed here was thus whether antagonist stretch might impede the ability to activate the agonist motoneurones in hemiparesis.

The second objective of this study was to determine whether effort perception (not the sense of the force achieved) (McCloskey et al. 1983) during agonist recruitment might also depend on antagonist stretch. Beyond feedforward information on the centrally generated command (McCloskey et al. 1974; Gandevia and McCloskey 1977a, b; Jones and Hunter 1983; Jones 1986; Cafarelli and Bigland-Ritchie 1979), feedback from peripheral receptors (Roland and Ladegaard-Pedersen 1977; Jones and Hunter 1985) may also contribute to effort perception, which is typically increased in disorders, of central origin (Gandevia and McCloskey 1977b; Jones and Hunter 1983; Simon et al. 2009; Holmes 1917; Solomon and Robin 2005). Whether a specific peripheral situation such as antagonist stretch might also impact on effort perception during agonist recruitment is unknown in spastic paresis. Finally, in the light of the impact of spastic co-contraction on motor weakness particularly when the co-contracting muscle is stretched (Gracies et al. 1997; Gracies 2005; Bourbonnais and Vanden Noven 1989; el-Abd et al. 1993; Dewald et al. 1995; Ikeda et al. 1998; Kamper and Rymer 2001; Vinti et al. 2013), we also aimed to explore the specific role of co-contraction in the effort perception in hemiparesis.

We have used a model of graded dorsiflexion efforts with or without gastrocnemius stretch (knee extended or flexed), positions that did not alter the length of the agonist dorsiflexors. Our primary hypothesis was that stretch in a contractured antagonist (i.e. a muscle with reduced length and reduced passive extensibility) such as the gastrocnemius might diminish the ability of the motor command to recruit agonist dorsiflexor motoneurones in hemiparesis. Further, we hypothesized an influence of spastic co-contraction on the perceived effort, suggesting that, in addition to limiting active movements, spastic co-contraction might also participate in the perception of weakness. A first part of the data using this paradigm but focusing on torques and co-contraction has previously been published (Vinti et al. 2013).

Methods

Subjects

This study was approved by the local ethics committee—Comité de Protection des Personnes Ile-de-France IX. Eighteen subjects with hemiparesis (8 women, age 54 ± 12, mean ± SD) from the Neurorehabilitation clinic of Henri Mondor University Hospitals (Créteil, France) and 18 healthy subjects (9 women, age 41 ± 13) underwent isometric dynamometric evaluation. Inclusion criteria for hemiparetic subjects were (1) vascular or traumatic brain injury-induced hemiparesis at least 6 months prior, and (2) a range of passive ankle dorsiflexion of at least 90° (Tardieu Scale, Gracies et al. 2010) with knee flexed or extended. Exclusion criteria consisted of intercurrent disease affecting ability to participate, cognitive dysfunction—in particular neglect—and major proprioceptive impairment, the latter two based on clinical examination including line bisection tests and determination of the joint perception thresholds using a goniometer. Inclusion criteria for healthy subjects were (1) absence of neurological disorder and (2) age <75.

Experimental procedures

Clinical examination in hemiparetic subjects included a 10-m ambulation speed and plantar flexor spasticity angle and grade using the Tardieu Scale (Gracies et al. 2010).

Measurements of torque and effort perception

The precise methodology has been previously described (Vinti et al. 2013). In brief, isometric evaluation of ankle torque was performed using a dynamometric apparatus (Contrex™, CMV AG, Switzerland). Subjects were comfortably semi-seated with the tested ankle strapped to a rotating footplate, fixed at 90° of dorsiflexion, in one of two knee positions: extended and flexed at 90°. The hip was kept around 80°–100° flexion throughout the procedures. The axis between the two malleoli was used as the landmark for matching the ankle joint with the axis of rotation of the resistance adapter. The knee-extended position was arranged to place the trans-joint gastrocnemius muscle in maximally stretched position. The level of torque at each effort was recorded; considering measurement errors at low torques, only dorsiflexor torque levels greater than 2 Nm were considered “positive” in terms of capacity to help initiate foot dorsiflexion. The same investigator carried out all procedures.

After a few slow passive movements at low speed (10°/sec) from minimally to maximally stretched position of the plantar flexors to make the subject familiar with the apparatus, the ankle was positioned back at 90° dorsiflexion and the subject was asked to perform one maximal isometric dorsiflexion (“lift up your foot as hard as possible”) and one maximal isometric plantar flexion (“push down your foot as hard as possible”) to become familiarized with the feeling associated with a maximal contraction.

Then the subject was asked to produce three dorsiflexion efforts, one at light intensity (approximately 10–20 % maximum), one at medium intensity (approximately 50–60 %) and again one at maximal intensity, in that order. While different subjects could interpret “light” and “medium” efforts differently, producing some variability in the actual percent of maximum, each subject attempted to reproduce those same effort levels with the knee flexed and extended. Each effort was to be sustained for 5 s, and this was verified using EMG monitoring and encouraged verbally.

Immediately after each effort, subjects were asked to quantify the perception of the effort produced using a Visual Analog Scale of Perceived Effort (VASPE), much as in pain measurements (Huskisson 1974). At one end of a straight 100-mm line the number zero indicated a sensation of no effort, and at the other end the number 10 indicated a sensation of maximal effort. Subjects placed a check on the line at the point that would best reflect the intensity of their effort. In addition to quantifying subjective impression of the effort achieved, this served as a quality control of the three prescribed intensity levels.

EMG measurements

Muscle activity was assessed by surface EMG simultaneously from ankle dorsiflexors (tibialis anterior, TA) and plantar flexors (soleus, SO and medial gastrocnemius, MG) with pairs of surface electrodes (ARBO H135TSG, device ME 6000 from MEGA Electronics system, Cambridge, UK). The skin was cleansed and abraded with alcohol before electrode application. Electrode positioning followed the recommendations of Basmajian and Blumenstein (1980). The EMG signal was sampled at 2000 Hz, amplified (gain = 1000), filtered using a bank of anticausal notch filters to remove the power line interference (50 Hz) and its harmonics. Each notch filter was centered at the frequency of 50 × i Hz, with i ranging from 1 to 9, and had a bandwidth of 4 Hz. Filtered signals were rectified. All computations were performed by a custom-written Matlab program (version 7.1, Natick, Massachusetts, USA). From these processed surface EMG measurements, we derived in each knee position: (1) The agonist Root Mean Square (RMS) of ankle dorsiflexor/plantar flexor EMG during maximal isometric ankle dorsiflexion/plantar flexion, measured over the 500 ms around the recruitment peak. The timing of the agonist peak recruitment within the 5 s was not specifically monitored. (2) The Agonist Recruitment Index (ARI), i.e. the ratio of the RMS of a muscle acting as an agonist during a given period to its largest RMS value observed during the 500 ms around its agonist recruitment peak. (3) The Co-contraction Index (CCI), defined as the ratio of the RMS of the EMG of the muscle when it was antagonist to the intended effort (i.e. opposing the intended direction of the effort) to the RMS of its EMG when acting as an agonist during the 500 ms around its recruitment peak.

Statistical analysis

After a descriptive analysis deriving average values and standard deviations of all continuous variables, we performed two-factor (group × position) and (group × effort) ANOVAs with repeated measures to detect changes in RMSTA, ARITA and the perceived effort (VASPE) across the three-graded efforts (light, medium and maximal) in both populations, first whichever the position and second whichever the effort. We then compared the relative influence of potential predictors (ARITA, CCI and Torque) on the dependent variable VASPE in each population, using standardized regression coefficients obtained from multiple regression analysis. Significance was set at p < 0.05.

Results

Patient characteristics

As reported in a previous paper using the same experimental paradigm (Vinti et al. 2013), the 18 patients included 9 cases of left hemiparesis. Ten-meter ambulation velocity was 0.64 ± 0.17 m/s (mean ± SEM) at comfortable speed and 0.84 ± 0.24 at maximal safe speed. In soleus, spasticity angle was 12.7° ± 6.8° and spasticity grade was 2.4 ± 0.8. In gastrocnemius, spasticity angle was 7.5° ± 4.5° and spasticity grade was 2.1 ± 0.8 (Gracies et al. 2010). During torque and EMG measurements all subjects were able to maintain effort (agonist EMG activation) for the full 5 s upon verbal encouragement. Torques for all contractions were close to three times higher in healthy subjects than in subjects with hemiparesis, as has been reported previously (Vinti et al. 2013).

Agonist tibialis anterior recruitment

Illustrations of typical EMG responses during the different levels of voluntary effort in each group have been previously provided (Vinti et al. 2013).

Effect of position

Whatever the effort, there was a major interaction between group (hemiparetic or healthy) and position (knee flexed or extended) for the raw agonist RMS of tibialis anterior (RMSTA, F 2,206 = 32.85, p < 0.0001) with a 25 ± 7 % (up to 98 %) reduction of RMSTA from knee flexed to extended in hemiparetic subjects, across all effort grades (p = 0.007), not seen in healthy subjects (p = 0.95; Fig. 1a). Averaging each effort grade separately in hemiparetic subjects, the difference knee flexed–knee extended for RMSTA was significant during the maximal efforts only (mean reduction 28 ± 8 %; p = 0.002; Fig. 1b, c). In individual patients, however, RMSTA reduction from knee flexed to knee extended was greater than 80 % in three subjects for light efforts, one subject for medium efforts and one subject for maximal efforts. These subjects with major inhibitory impact of the knee-extended position on agonist TA recruitment seemed more spastic in the plantar flexors (mean spasticity angle 19.3° ± 5.1° in soleus and 9.5° ± 1.5° in gastrocnemius) and slower walkers (mean 10-m ambulation velocity 0.60 ± 0.05 m/s at comfortable speed and 0.72 ± 0.07 m/s at maximal safe speed) than the rest of the group. The interaction between group and position for raw agonist recruitment (RMSTA) was lost when considering agonist recruitment indices only (ARITA; Fig. 2a, b).

Fig. 1
figure 1

Raw agonist recruitment of tibialis anterior. a Mean RMS EMG of all efforts combined in both knee positions; b values of each graded effort in healthy subjects and c values in hemiparetic subjects. **p < 0.01

Full size image
Fig. 2
figure 2

Agonist Recruitment Indices in tibialis anterior. Agonist Recruitment Index (ARI), i.e. ratio of the RMS of a muscle acting as an agonist during a given period to its largest RMS value during the 500 ms around agonist recruitment peak. a Healthy subjects; b hemiparetic subjects. ***p < 0.001

Full size image

Effect of effort

RMSTA increased from light to maximal efforts in healthy (F 2,102 = 40.87, p < 0.0001) and hemiparetic subjects (F 2,96 = 13.13, p < 0.0001) whatever the knee position (Fig. 1b, c). In parallel, ARITA also increased across the graded efforts in both groups (p < 0.0001; Fig. 2a, b). In both joint positions there was a trend for an interaction between group and effort for ARITA (F 2,201 = 2.58, p = 0.08). This interaction was highly significant for maximal efforts, where ARITA was greater in healthy subjects than in hemiparetic subjects (0.90 ± 0.01 vs 0.82 ± 0.02, p < 0.0001, ANOVA).

Plantar flexor co-contraction

Plantar flexor co-contraction and torque data have been reported in a previous article (Vinti et al. 2013). In brief: (1) CCIs were abnormally high for the three dorsiflexion effort levels in hemiparetic subjects; (2) whatever the dorsiflexion effort, there was a strong interaction between position and group for the soleus (F 2,196 = 9.37, p < 0.001) and for medial gastrocnemius (F 2,196 = 21.91, p < 0.0001): the knee-extended position was associated with an increase in soleus CCI in hemiparetic subjects (p < 0.001) but not in healthy subjects and with reduced medial gastrocnemius co-contraction in healthy subjects but maintained levels in hemiparetic subjects; (3),in hemiparesis the increased co-contraction associated with gastrocnemius stretch was sufficient to reverse or cancel the intended dorsiflexor torque in 26 % of cases (Vinti et al. 2013).

Perception of dorsiflexion effort

Effect of effort

From light to medium and maximal efforts, perception increased in both healthy (F 2,202 = 354, p < 0.0001; Fig. 3a) and hemiparetic subjects (F 2,202 = 236, p < 0.0001; Fig. 3b) for both knee positions. The three-way ANOVA (group × position × effort) revealed a significant interaction, group × effort (p = 0.002) across both knee positions, whereby light efforts were perceived as lighter by healthy subjects (95 % CI VASPE [1.76–2.60]) than by hemiparetic subjects (95 % CI [2.67–3.49]), and maximal efforts were perceived as stronger by healthy subjects (95 % CI VASPE [8.52–9.36]) than by hemiparetic subjects (95 % CI [7.89–8.71]; Fig. 3a, b).

Fig. 3
figure 3

Dorsiflexor effort perception. Effort perception quantified by the Visual Analogical Scale of Perceived Effort (VASPE) in healthy (a) and hemiparetic (b) subjects in the two knee positions (flexed and extended)

Full size image

Effect of position

Whatever the effort, knee position had an impact on the dorsiflexion effort perception in hemiparetic subjects, which was on average 7 % greater in the extended than in the flexed position (5.6 vs 5.2, p = 0.03, ANCOVA; Fig. 3b). This effect was greatest for light efforts, where effort perception increased by 20 % when the knee was extended (2.8 vs 3.4, p < 0.05). In contrast, healthy subjects displayed no difference in the effort perception between the two knee positions, whatever the effort.

Effect of agonist contraction and antagonist co-contraction

When the knee was flexed, in healthy subjects the perception of dorsiflexion efforts (VASPE) was related to tibialis anterior activation only (standardized regression coefficient βARITA = 0.67, p < 0.0001), with no impact of medial gastrocnemius co-contraction or the torque produced (βCCIMG = 0.14, NS; βtorque = 0.11, NS). In hemiparetic subjects, however, effort perception was related to both tibialis anterior activation and gastrocnemius antagonist co-contraction (βARITA = 0.42, p < 0.002; βCCIMG = 0.39, p < 0.001; Fig. 4a, b).

Fig. 4
figure 4

Determinants of dorsiflexor effort perception. Standardized regression coefficients (β) of three potential predictors of effort perception in healthy (a) and hemiparetic (b) subjects knee flexed (left) and extended (right). Bars 95 % confidence intervals. In hemiparetic subjects knee extended, plantar flexor co-contraction becomes the strongest predictor

Full size image

When the knee was extended, the perception of effort by healthy subjects was related to tibialis anterior activation (βARITA = 0.54, p < 0.01) and the resulting torque (p < 0.01), with again no impact of antagonist co-contraction (βCCIMG = 0.12, p = 0.40; Fig. 4a). In hemiparesis, however, plantar flexor co-contraction was the strongest predictor of effort perception (βCCIMG = 0.45, p < 0.001), together with tibialis anterior activation (βARITA = 0.40, p < 0.001) and the torque (p < 0.05; Fig. 4b).

Discussion

This study demonstrates an increase in dorsiflexor paresis (reduced output from dorsiflexor motoneurones—sometimes virtually cancelled) with antagonist muscle stretch, a phenomenon defined here as stretch-sensitive paresis. The study also reports an increase in dorsiflexion effort perception with gastrocnemius stretch in spastic paresis, probably related to both antagonist stretch itself and the associated spastic co-contraction increase.

Impact of antagonist position on agonist recruitment: stretch-sensitive paresis

A previous paper already demonstrated and quantified the increase in plantar flexor co-contraction and the decrease in dorsiflexor torque at three dorsiflexion effort levels in hemiparetic subjects when the knee was in extended position (Vinti et al. 2013). Here, we demonstrate for the first time that the reduction in agonist torque with antagonist stretch is not only due to the increase in plantar flexor co-contraction (Levin et al. 2000; Knutsson and Mårtensson 1980) but also to a decrease in agonist motoneuroneal recruitment, which we were able to measure. The phenomenon of decreased agonist recruitment with antagonist stretch thus goes over and above the classically described spastic restraint, which merely corresponded to torque limitation due to passive and active antagonist resistance (Kamper et al. 2003; Dietz et al. 1981; Koo et al. 2003; Hu et al. 2006; Gracies et al. 1997; Tardieu 1966; Vinti et al. 2013; Ikeda et al. 1998; Kamper and Rymer 2001). This study helps explain torque reduction during antagonist stretch by a decrease in agonist EMG.

In spastic paresis, reduced EMG signals may reflect at least in part decreased excitatory drive to agonist motoneurones, contributing to reduced force or movement (Dietz et al. 1981; Koo et al. 2003; Hu et al. 2006; Sahrmann and Norton 1977; Knutsson and Richards 1979; Inman et al. 1952; Woods and Bigland-Ritchie 1983; Bourbonnais et al. 1989; Visser and Aanen 1981). In the face of triceps surae co-contraction, the net torque associated with dorsiflexion efforts could then become negligible or even be reversed into plantar flexion (Vinti et al. 2013; Levin et al. 2000). This effect was higher during strong efforts but was also major during light or medium efforts in individual cases.

The classically altered EMG–force relationship in spastic paresis (Dietz et al. 1981; Bourbonnais et al. 1989; Dewald et al. 1995; Canning et al. 2000) was not a critical issue here as our conclusions rely mostly on quantitative EMG measures, and because the observed decrease in agonist tibialis anterior EMG paralleled that in dorsiflexion torque (Vinti et al. 2013). These findings are also not due to muscle or skin conformation changes because tibialis anterior length and skin stretch on the anterior aspect of the leg are not affected by knee position changes. Accordingly, tibialis anterior EMG was not different in healthy subjects when knee position was changed (Fig. 1a), which confirmed previous evidence (Cafarelli and Bigland-Ritchie 1979; Komi and Buskirk 1972; Marsh et al. 1981; Newman et al. 2003).

Rather, the reduction of dorsiflexor recruitment by gastrocnemius stretch in hemiparesis might involve reciprocal inhibition from triceps surae (Cody et al. 1987; Simonetta-Moreau et al. 1999; Marque et al. 2001). Tonic gastrocnemius stretch could exert a post-synaptic inhibitory effect on tibialis anterior motoneurones, making them less able to respond to descending voluntary drives (Simonetta-Moreau et al. 1999; Marque et al. 2001). This effect would further diminish the ability of a weakened descending command to bring motoneurones to firing threshold, sufficiently to completely abolish voluntary dorsiflexors EMG in some patients. The larger effects in maximal efforts (Fig. 1) and the loss—or rather masking—of these inhibitory effects when looking only at agonist recruitment indices (Fig. 2) might involve preferential effects of such reciprocal inhibition on the latest recruited motoneurones therefore on the high-threshold motoneurones. However, while some studies mention various sensitivities of motoneurones of different sizes to Ia reciprocal inhibition or Group II inputs, no quantification is available to our knowledge (Henneman 1985; Marchand-Pauvert et al. 2000; Pierrot-Deseilligny and Burke 2005). Finally, Ia reciprocal inhibition effects would certainly increase from muscles with higher spindle sensitivity, which is the case of the most contractured muscles (Gioux and Petit 1993). Even though the present study was not designed to study reciprocity in stretch-sensitive paresis (the position knee extended or knee flexed altered only the length of gastrocnemius and not of tibialis anterior), we would speculate that this increased paresis might thus affect particularly the less contractured of two muscles around a joint when its more contractured antagonist is stretched, a phenomenon termed stretch-sensitive paresis (Gracies 2005).

Effect of antagonist position—and of spastic co-contraction—on the perception of agonist effort

This study showed increased dorsiflexor effort perception with gastrocnemius stretch in hemiparetic subjects, but not in healthy subjects. It may appear paradoxical that effort perception would increase in the position knee extended, while agonist recruitment decreased in that condition, even though effort perception partially depended on agonist recruitment (Fig. 4). Such paradox could be explained by an increased magnitude of the dependence of effort perception on antagonist co-contraction, as shown in Fig. 4. Here, increased dorsiflexor effort perception with the knee extended has to do with the associated increased plantar flexor co-contraction in that situation (Vinti et al. 2013), rather than with plantar flexor stretch per se (the degree of gastrocnemius stretch was similar in all patients with the knee extended but plantar flexor co-contraction and sense of effort vastly differed). To further establish this hypothesis might require pure motor blocks (e.g. using curare) to filter out the effects of antagonist co-contraction from those of antagonist stretch on agonist effort perception. The study otherwise confirmed increased effort perception with the isometric force exerted (Stevens and Cain 1970; Eisler 1962; Cain and Stevens 1971) at least with the knee extended, for both healthy and hemiparetic subjects (Fig. 4).

Reduced range of effort perception in spastic paresis

The correspondence between the prescribed level of effort (Fig. 3, X axis) and the level of effort retrospectively estimated (Y axis) suggests gross capacity of subjects (healthy and hemiparetic) to appropriately follow the light, medium and maximal effort levels. However, the range of available effort perceptions appeared reduced in hemiparesis, raising the question of a dampened effort perception (Fig. 3). Alternatively, the striking parallelism to the reduced range of agonist recruitment indices (Fig. 2) suggests a correlation between effort perception and agonist recruitment index, demonstrated in Fig. 4, the effort perception itself being adequate in hemiparesis. On the high end of the range, reduction of maximal agonist recruitment index over 5 s efforts could reveal higher fatigability of large muscle fibers in paresis, explaining short tolerance to strong efforts (Dietz et al. 1981, 1986; Edström 1970 ). On the low end, changes in recruitment gain of motoneurone pools with non-uniform distribution of synaptic effects to low- and high-threshold motor units could explain previously reported sustained tonic motoneurone firing with little or no synaptic input in spastic paresis, especially in low-threshold units (Hultborn et al. 2004; Heckmann et al. 2005; Mottram et al. 2009). Yet, reduced proprioception in hemiparesis may also contribute to a reduced range of effort perceptions but patients with major sensory impairment were not included in this study (see “Methods”) (Carey et al. 1996; Tyson et al. 2013).

Study limitations

Exact knee angles were not verified with goniometry, thus the two knee positions might not have been comparable for the two groups, as hamstrings shortening might have limited ability to obtain full knee extension in the paretic group. However, significant hamstring shortening is rare with adult-acquired brain lesions, as opposed to infant-acquired lesions (Van Reeth et al. 2013a, b). More importantly, the exact knee angle in the situation “knee extended” was not relevant here, rather the fact that the knee was maximally extended in that situation, thus providing maximal gastrocnemius stretch. Another potential limitation is the slightly—but not significantly—different age between the two groups.

Conclusions/implications

This study first showed the existence of a direct link between the degree of paresis—measured by the agonist motoneurone output—and the antagonist in the spastic paresis context. As plantar flexors are almost invariably contractured in chronic adult hemiparesis (Kwah et al. 2012), decreased dorsiflexor recruitment when the antagonist plantar flexor is stretched suggests that this phenomenon may occur with any pair of muscles around a joint in spastic paresis, when one is far more contractured than the other (e.g. finger flexors and extensors, pronators and supinators, etc.). If such a generalization proves true in future studies, there is additional incentive for clinicians to ascribe higher priority to desensitizing spindles by lengthening contractured muscles. Antagonist lengthening might thus both decrease its spastic co-contraction and increase voluntary agonist recruitment.