THE UPPER MOTOR NEURON syndrome that follows a

Transcription

1 1016 Inhibition of the Triceps Surae Stretch Reflex by Stimulation of the Deep Peroneal Nerve in Persons With Spastic Stroke Peter H. Veltink, PhD, Michel Ladouceur, PhD, Thomas Sinkjær, MD, PhD ABSTRACT. Veltink PH, Ladouceur M, Sinkjær T. Inhibition of the triceps surae stretch reflex by stimulation of the deep peroneal nerve in persons with spastic stroke. Arch Phys Med Rehabil 2000;81: Objective: To reduce the triceps surae stretch reflex by electrical stimulation of the deep peroneal nerve. Design: Intervention study. Setting: Research institution. Participants: Sample of convenience of 10 spastic stroke individuals. Intervention: After the deep peroneal nerve was stimulated between 0.9 and 4 times tibialis anterior motor threshold, the triceps surae was stretched to elicit a reflex. Main Outcome Measure: The triceps surae stretch reflex was quantified by the amplitude of the reflex electromyography (EMG) in soleus and medial gastrocnemius muscles and mean ankle moment. Paired t test and the Wilcoxon signed rank test ( p.05) were used to evaluate the effect of conditioning stimulation. Results: The soleus stretch reflex EMG was reduced significantly ( p.001) by stimulating the deep peroneal nerve to 25% 6% (standard error) of the unconditioned value (relaxed triceps surae). The optimal interval between stimulation and stretch was msec. The velocity threshold increased significantly ( p.006) from a median value of 8 per second to 33 per second and the area under the stretch velocity/stretch reflex relation decreased significantly ( p.001) (soleus EMG). Conclusions: The stretch reflex of relaxed triceps surae in persons with spastic stroke can be extensively reduced by stimulating the deep peroneal nerve at several times motor threshold of the tibialis anterior. Key Words: Reflex, stretch; Inhibitional; Triceps surae; Electric stimulation; Stroke; Rehabilitation by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation THE UPPER MOTOR NEURON syndrome that follows a stroke, spinal cord injury, or multiple sclerosis is characterized by several symptoms, among others an increased joint stiffness (spastic hypertonia) caused by changes in the passive, intrinsic, and reflex components of stiffness. Spasticity, defined From the Institute for Biomedical Technology (BMTI) Faculty of Electrical Engineering, University of Twente, Enschede, the Netherlands (Veltink) and the Center for Sensory-Motor Interaction, Department of Medical Informatics and Image Analysis, University of Aalborg, Aalborg, Denmark (Ladoucern, Sinkjaer). Submitted May 20, Accepted in revised form December 22, Supported by the Training and Mobility of Researchers Program (project NEUROS) of the European Union, the Danish National Research Foundation, the Danish National Research Council, and the Danish Medical Research Council. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors or upon any organization with which the authors are associated. Reprint requests to Peter H. Veltink, Faculty of Electrical Engineering, University of Twente, PO Box 217, 7500 AE Enschede, The Netherlands /00/ $3.00/0 doi: /apmr as a velocity-dependent increase of the stretch reflexes, 1 has been linked to a decreased threshold of the stretch reflex, 2 which could be from increased motorneuronal excitability. 3-8 The origin of this increased motorneuronal excitability is unknown, but it appears to be related to a decreased presynaptic inhibition 9,10 and absent reciprocal inhibition. 11. Spastic patients may benefit if the motorneuronal excitability can be reduced when the muscle is relaxed. This reduced excitability, which will prevent unwanted reflex contractions, may be achieved by pharmacologic means. 12,13 An alternative way to reduce the stretch reflex is to electrically stimulate neural pathways to inhibit the excitability of the motorneuron pool. Apkarian and Naumann 14 found that the soleus stretch reflex could be inhibited in healthy subjects by a conditioning stimulation applied to the deep peroneal nerve at a level that just caused a small twitch in the tibialis anterior. This inhibition was not consistently observed in 6 spastic patients with varying neuromuscular disorders. The optimal conditioning test interval was found to be 160msec on average, which is much larger than found for disynaptic reciprocal inhibition in H-reflex studies (2msec), 11,15 as well as for presynaptic inhibition (25 to 60msec). By stimulating the deep peroneal nerve just below the motor threshold of the tibialis anterior, Crone and coworkers 11 found that disynaptic reciprocal inhibition depressed the conditioned H-reflex with 15%. However, this disynaptic reciprocal inhibition was not found in a group of 39 spastic multiple sclerosis patients, except in 4 patients who were using a foot-drop stimulator daily. In contrast, a facilitation of the reflex was seen at conditioning test intervals between 4 and 8msec. Capaday and colleagues 16 showed reciprocal inhibition by presynaptic inhibitory mechanisms of the soleus motor output in healthy subjects when stimulating the common peroneal nerve. The depression of soleus EMG as a response to the conditioning stimulation had a latency of approximately 40msec. The inhibition increased with contraction level in the same way for standing and the stance phase of gait. Inhibiting the soleus H-reflex is not only possible by stimulation of Ia afferents, but can also be obtained when stimulating cutaneous nerves. Fung and Barbeau, 17 reporting on healthy subjects, found that the soleus H-reflex was significantly inhibited in all phases of the gait cycle when the ipsilateral medial plantar arch was stimulated at 2.5 to 3 times sensory threshold and the conditioning test delay was approximately 45msec. At this stimulation site, mainly sensory nerve fibers are activated from cutaneous and mechanoreceptors of the sole. In moderately and severely impaired spastic paretic patients this conditioning stimulation restored phasic modulation of the H-reflex to nearly normal levels. Stimulating the sural, posterial tibial, and superficial peroneal nerves at the ankle during gait produced a reflex response in muscles of the ipsilateral leg. The responses were dependent on the phase of the gait cycle and on the nerve that was stimulated. 18,19 The above studies did not address how conditioning stimulation influenced the relation between stretch velocity and stretch reflex magnitude a relation believed to be of functional

2 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink 1017 importance at the ankle joint. 20,21 Increased stretch reflexes in the ankle extensor may limit the gait performance in spastic subjects, during the gait phases in which the triceps surae is stretched. Changes in threshold and slope of this relation may also elucidate the inhibitory mechanism. 22 In the present study, we investigated the influence of deep peroneal nerve stimulation on the stretch velocity dependency of the stretch reflex in relaxed and voluntarily activated soleus and medial gastrocnemius muscles in 10 seated, spastic, stroke patients. This influence was investigated at several levels of the conditioning stimulation, varying between 0.9 and 4 times the motor threshold of tibialis anterior, and at the optimal interval between deep peroneal nerve stimulation and triceps surae stretch. METHODS Subjects A convenience sample of 10 stroke patients with varying levels of spasticity participated in the experiments. The main requirement for selection was that the subjects had had a stroke, that their Ashworth scale score was above 1, and they were able to walk unassisted for distances greater than 10 meters. No clinical spasticity tests were performed during the experiment, since the objective of this study was not to relate inhibition to clinical spasticity scores, but merely to test the feasibility of stretch reflex inhibition by stimulation in a broad selection of stroke patients with varying levels of spasticity. All subjects were more than 2 years poststroke and used their regular medication without intervention (patient characteristics, table 1). All subjects signed an informed consent approved by the local ethics committee. Experimental Setup The subject was seated in an adjustable chair with the foot of the affected leg firmly strapped to a foot plate. Knee and ankle joints were extended to approximately 100. The ankle was rotated by a DC motor a connected to the foot plate. The ankle joint s axis of the rotation was aligned with the axis of rotation of the foot plate. The imposed movements were 4 dorsiflexion ramps with a subsequent hold phase that lasted until 500msec after ramp onset. The stretch velocity of the ankle extensor (triceps surae) was varied by changing the duration of the reference ramp signal. The ankle moment was measured by means of strain gauges mounted on a beam connecting the foot plate with the motor. The angular position of the foot plate was measured by a potentiometer. The DC motor was powered by a DC amplifier b and controlled by a position servo system. The onset of movement was delayed 4msec with respect to the reference position signal supplied to the servo controller. The room s temperature was kept constant during the experiments in order to exclude any influence of temperature changes on stretch reflex characteristics. The reflex latencies for the same perturbation conditions and no stimulation varied less than 2% during the course of an experiment, indicating no effect of body temperature changes on nerve fiber conduction velocity. Further details about this setup have been described by Sinkjær and associates. 23 The electromyogram (EMG) signals of the soleus, medial gastrocnemius, and tibial anterior muscles were recorded by means of bipolar surface electrodes. A ground electrode was placed above the knee. The electromyogram signals were amplified c and filtered with a first-order band-pass filter (20Hz to 1kHz). Ankle angle, ankle moment, and amplified and filtered soleus and medial gastrocnemius signals were sampled at 2kHz and stored for further analysis. The deep peroneal nerve was stimulated d with a round 3-cm-diameter self-adhesive cathode placed on the skin of the shank, just distal to and approximately 5cm anterior to the head of the fibula. An oval anode electrode (4 by 6cm) was placed midway on the shank on the skin above the tibia. In all subjects, palpation of the tendons at the ankle ensured that the peroneus muscles, innervated by the superficial branch of the peroneus nerve, were not activated even at the highest stimulation levels used. These muscles are not pure antagonists of the triceps surae. 11,24 Stimulation of their afferents were therefore expected to have a facilitory effect on the triceps surae reflex. The deep peroneal nerve was stimulated with trains of 5 pulses at 200Hz, with a pulse width of 1msec. This stimulation frequency is markedly above the fusion frequency of the tibialis anterior, maximizing the activation of afferent nerve fibers, while obtaining a fused tibialis anterior contraction by the activation of motor fibers in the deep peroneal nerve branch. The stimulation strength was controlled by the stimulation current amplitude, which was manually adjustable on the current controlled stimulator. A personal computer compatible main computer controlled the experiment. It provided an ankle angular position reference signal, triggered the stimulator, and sampled the signals. Subject Sex Age (yrs) Time Since Injury Table 1: Patient Characteristics Body Side Affected Antispastic Medication Velocity Threshold MVC Soleus EMG/Ankle Moment Soleus EMG Level during Precontraction (µv) 1 Male 54 5yrs, 10mo Right No Male 53 6yrs, 6mo Left No Male 54 3yrs, 3mo Left No Male 61 6yrs, 6mo Left Yes Male 58 4yrs Left Yes Male 64 2yrs, 10mo Left No µV/35Nm 25 7 Male 52 5yrs Left Yes µV/46Nm 16 8 Female 63 13yrs, 4mo Right No µV/4Nm 6 9 Female 72 6yrs Left No µV/19Nm Male 52 3yrs, 5mo Left No µV/35Nm 16 Maximal voluntary contraction was measured in only 5 subjects. Velocity thresholds were determined from the trials in the velocity tests with no stimulation. For each subject, three measurements from different velocity tests were averaged. Velocity data are reported as degrees per second standard error. Abbreviations: MVC, maximal voluntary contraction; EMG, electromyography.

3 1018 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink Protocol The effect of the conditioning stimulation on the stretch reflex inhibition in the soleus and medial gastrocnemius was tested at several conditioning stimulation levels applied to the deep peroneal nerve. At each stimulation level, the interval between the conditioning stimulus burst and the onset of the triceps surae stretch was varied. At the optimal interval, the stretch velocity was varied. The tests were performed in relaxed and precontracted triceps surae. The deep peroneal nerve was stimulated at 0.9, 1.5, 2.0, 3.0, and 4.0 times the motor threshold of the tibialis anterior. Stimulation at 3 or 4 times motor threshold was only applied if 2 times motor threshold stimulation did not result in marked soleus stretch reflex inhibition. Stimulation of up to 4 times motor threshold was justified by the study of Gracies and coworkers, 25 who showed that additional recruitment of Ia afferent nerve fibers can take place up to this stimulation level, when using skin electrodes. For each stimulation level, the interval between the beginning of the conditioning stimulus burst and the onset of the stretch was varied between 6msec and 204msec in 10 steps ( 6, 14, 29, 54, 79, 104, 129, 154, 179, 204msec). Additionally, stimulation without subsequent stretch and stretches without conditioning stimulation were applied. All 12 conditions (10 intervals, only stretch, only stimulation) were applied approximately 8 times in random order. A 40-msec stretch rise time was used for the test in which the interval between deep peroneal nerve stimulation and triceps surae stretch was varied, resulting in a maximal stretch velocity of 120 /sec. Subsequently, the conditioning stimulation s effect on the relation between stretch reflex amplitude and stretch velocity was determined for the optimal conditioning-stretch interval. Stretch rise times between 20 and 400msec were imposed, resulting in maximal stretch velocities between 15 /sec and 190 /sec. For each subject the stretch velocity threshold was first determined. Stretches were then performed for 7 stretch velocities above this threshold, with and without conditioning stimulation. An extra condition was also performed: stimulation only and no stretch. All 15 conditions were applied approximately 8 times in random order. To reduce the effects of possible time dependencies of the stretch reflex and conditioning processes, we compared in the analysis only those stretch reflexes acquired in random order in the same test. In summary, the protocol consisted of the following parts: 1. Variation of the interval between deep peroneal nerve conditioning stimulation and triceps surae stretch with relaxed triceps surae at 0.9, 1.5, 2.0, and, if required, 3.0 or 4.0 times motor threshold of tibialis anterior. 2. Velocity tests with relaxed triceps surae at the same conditioning stimulation levels and at the optimal interval between deep peroneal nerve stimulation and triceps surae stretch, as determined in part Variation of the interval between deep peroneal nerve stimulation and triceps surea stretch with precontracted triceps surae at the highest stimulation level used in the relaxed tests. 4. Velocity test with precontracted triceps surae at the same stimulation level and at the optimal interval between deep peroneal nerve stimulation and triceps surae stretch, as determined in part 3. Two or three stretch velocities above threshold were used. All 10 subjects participated in parts 1 and 2 of the protocol and 7 of the 10 participated in parts 3 and 4 of the protocol. They were asked to contract the triceps surae at a level they could hold for at least 1.5 minutes. These levels are specified in table 1. For comparison, the soleus electromyography levels during maximal voluntary contraction were specified for some subjects. During the precontraction tests, the high-pass filtered, rectified, and low-pass filtered electromyogram of soleus was presented to the subject as visual feedback for the level of contraction. The subjects were asked to keep the contraction at the reference level without taking notice of the stimulation and stretch disturbances. Precontraction tests were interrupted every 1.5 minutes for a 1-minute break. Data analysis We inspected the soleus electromyogram for all trials of protocol parts 1 and 2 to check for relaxed muscles. Skipping nonrelaxed trials was only necessary in exceptional cases for 9 of 10 subjects. Only subject 6 did not relax continuously and had to be reminded regularly. Therefore, approximately 30% of his trials had to be omitted during the analysis. The data were analyzed with MATLAB e software. In order to remove remaining movement artifacts the electromyograms were digitally high pass-filtered using a second-order filter with a cut-off frequency of 20Hz. The electromyograms were rectified and low-pass filtered with a second-order filter with a cut-off frequency of 20Hz. Both filters were composed of a first-order Butterworth filter, which was applied twice to the signal, once in normal time order and once in reverse time order, resulting in zero phase shift (effectively second order). The peak of the stretch reflex electromyogram was determined by taking the maximum value of the rectified and low-pass filtered electromyograms after averaging all responses for each condition. These peak electromyogram stretch responses with conditioning and without conditioning are referenced in the present report, respectively, as peak conditioned stretch reflex EMG and peak unconditioned stretch reflex EMG. We determined background EMG levels by taking the mean value of the EMG signals of all trials without stimulation in the same test from 204msec before stretch onset to 4msec before stretch onset. For all conditions the background EMG was subtracted from the peak unconditioned and conditioned stretch reflex EMG. The mean values of the ankle moment signals were determined between 204msec before to 496msec after stretch onset. This interval encompassed the tibialis anterior response to the stimulation burst and the triceps surae reflex response to the stretch and ended just before the ankle angle was reset after the stretch. Note that the inertial moments from accelerating and decelerating the foot and foot plate should cancel each other in these mean moments because the interval included both the acceleration and deceleration phases. The mean moments of the stretch trials with and without conditioning stimulation are referenced in the present report as mean conditioned and mean unconditioned stretch reflex moment, respectively. The mean moments of the trials with only stimulation, but no stretch, quantify how much the tibialis anterior contraction contributed to the mean ankle moments. The mean offset ankle moments were determined by taking the mean of the moment signals of all trials without stimulation in the same test from 204msec before to the onset of the stretch. For all conditions the mean offset ankle moment was subtracted from the mean unconditioned and conditioned stretch reflex moment. The mean stimulation response without stretch was subtracted from the mean conditioned stretch reflex moment, to exclude the tibialis anterior motor response from the analysis. The ankle angular velocity signals were determined by low-pass filtering (cut-off frequency: 50Hz, second-order zero phase shift Butterworth filter, see above) and subsequently

4 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink 1019 digitally differentiating the ankle angle signals. The peak velocity was taken for evaluation. Variation of the interval between stimulation and stretch. The effect of the conditioning stimulus on the normalized inhibition of the triceps surae stretch reflex EMG was determined for each conditioning test delay by normalizing the peak conditioned stretch reflex EMG on the peak unconditioned stretch reflex EMG. The normalized inhibition of the stretch reflex moment was determined by dividing the mean conditioned stretch reflex moments by the mean unconditioned stretch reflex moments. The optimum conditioning test delay and the associated minimal value of the stretch reflex were determined separately for EMG and moment. Velocity tests. The thresholds and slopes of the velocity dependencies of the peak conditioned and unconditioned stretch reflex EMG were determined from linear fits, omitting points below threshold and above saturation (generally below 10% and above 80% of maximal unconditioned stretch reflex). Saturation of the unconditioned reflex at high stretch velocities was observed in 6 subjects. The conditioning stimulation s effect on velocity relations was quantified by the area under the relations between stretch velocity and conditioned reflex magnitudes (peak stretch reflex EMG and mean stretch reflex moment), normalized on the same area for unconditioned reflexes. Statistics. Different values of relevant quantities with and without stimulation were tested for statistical significance using the paired t test if the data passed the normality test (Kolmogorov-Smirnov test), or, alternatively, the Wilcoxon signed rank test. The influence of stimulation level (categories 0.9 and 1.5 times tibialis anterior motor threshold and maximum stimulation level) was tested using the one-way analysis of variance (ANOVA) test if the data were normal, or, alternatively, the Kruskal-Wallis one-way ANOVA on ranks test. Subsequent multiple comparisons of the stimulation levels were performed with the Tukey test (in case of regular ANOVA) or Dunnett s method (in case of Kruskal-Wallis ANOVA). The p values.05 were considered statistically significant. The statistical tests were performed with the SIGMASTAT f statistical software package. RESULTS Variation of the Interval Between Stimulation and Stretch In order to give an impression of the measured signals, figure 1 shows sample registrations of the averaged EMG recordings, ankle joint moments, and angular velocities for varying intervals between conditioning stimulation of the deep peroneal nerve and triceps surae stretch. The varying timing of the conditioning stimulation is apparent in these recordings by the shifting stimulus artifacts in soleus and medial gastrocnemius EMG. In the example, the triceps surae was relaxed and the deep peroneal nerve was stimulated at 3 times the motor threshold of the tibialis anterior. A marked inhibition is apparent in the EMG for both soleus (fig 1A) and medial gastrocnemius (fig 1B). The direct motor response of the tibialis anterior is evident in the recordings of the ankle joint moments (fig 1C). For comparison, the bottom traces show the response to a triceps surae stretch without stimulation and the tibialis anterior motor response to deep peroneal nerve stimulation without a triceps surae stretch. The ankle dorsiflexion that stretches the triceps surae is apparent as a pulse in the ankle angular velocity recording (fig. 1D). Inertial contributions to the joint moments (fig 1C) are apparent at the beginning and end of the stretch. The differences between the stretch reflexes with and without stimulation are difficult to distinguish from the joint moment recordings because they also contain the tibialis anterior motor response. In the further analysis, this motor response was subtracted from the moment recordings with stimulation. In precontracted muscle, 2 to 3 reflex peaks with short to medium latencies can be distinguished in the soleus and medial gastrocnemius EMGs. 26 In the present analysis the maximum peak was considered as a measure for the size of the reflex. The average dependencies of the stretch reflexes on the interval between conditioning stimulation and the onset of triceps surae stretch are shown in figure 2 for the highest levels of conditioning stimulation. These average dependencies show a clear inhibition when the triceps surae is relaxed (fig 2A,B,C, averaged over 10 subjects), but not when it is precontracted (fig 2D,E,F, averaged over 7 subjects). Since these average dependencies do not represent the maximum inhibitions if the Fig 1. Measured stretch reflexes elicited in the relaxed triceps surae for varying intervals between conditioning stimulation of the deep peroneal nerve and triceps surae stretch (subject 10; stimulation level: 3 times motor threshold of the tibialis anterior). Shown are the averaged recordings of (A) soleus EMG, (B) medial gastrocnemius EMG, (C) ankle joint moment, and (D) ankle angular velocity.

5 1020 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink Fig 2. Average dependencies of the stretch reflexes on the interval between conditioning stimulation of the deep peroneal nerve and triceps surae stretch onset for highest levels of conditioning stimulation. (A,B,C) Relaxed triceps surae, averaged over 10 subjects. (D,E,F) Precontracted triceps surae, averaged over 7 subjects. Data are means and standard errors of the (A,D) normalized peak stretch reflex EMG for soleus, the (B,E) medial gastrocnemius, and of the (C,F) normalized mean moment. Normalized peak EMG responses are only shown for stimulation-to-stretch intervals at which the stimulus artifact did not interfere. conditioning stretch intervals at which they occur vary between subjects, we determined the minimum conditioned stretch reflexes, normalized on the unconditioned reflex magnitudes and the conditioning stretch intervals at which they occurred for each subject. Figure 3 shows the means and standard errors of the normalized peak EMG reflex responses for soleus (Sol) and medial gastrocnemius (MG) and of the normalized mean moments over all subjects for relaxed and precontracted triceps surae. For the relaxed condition (10 subjects), the results are shown for 3 levels of conditioning stimulation: 0.9 and 1.5 times tibialis anterior motor threshold and the highest stimulation levels used for each individual subject ( max in fig 3). The inhibition of the stretch reflex depended significantly on stimulation level (ANOVA, p.001 for Sol, MG, and moment) and was significantly different ( p.001) from one at maximum stimulation levels. Comparisons of the stimulation levels showed that higher stimulation levels yielded higher inhibition in all cases. We also tested the optimal intervals between deep peroneal nerve stimulation and triceps surae stretch for a dependency on stimulation level, while discarding the intervals when the maximum inhibition was smaller than 10% of the unconditioned reflex response. The dependency of optimal intervals on stimulation level was not significant for soleus EMG (ANOVA, p.77) and mean ankle moment ( p 0.6), but was marginally significant for medial gastrocnemius EMG ( p.03). At the highest levels of stimulation the normalized stretch reflexes were significantly smaller than 1.0 ( p.001 for sol, MG, and mean ankle moment), indicating that the stretch reflex was effectively reduced at sufficiently high stimulation level. The optimal interval between deep peroneal nerve stimulation and triceps surae stretch was approximately 150msec. For the precontracted condition (7 subjects), the tests were only performed at maximum stimulation levels. In contrast to relaxed triceps surae, 2 reflex peaks with short to medium latencies can be distinguished in the EMG of soleus and medial gastrocnemius. 26 In the current analysis, the maximum peak was considered as a measure for the size of the reflex. In general, the stretch reflex inhibition at the optimal interval between deep peroneal nerve stimulation and tricep surae stretch was less in precontracted than in relaxed triceps surae (fig 3), but still significant in soleus, for which the minimum normalized peak reflex EMG was significantly different from 1 ( p.02). However, this was not the case for medial gastrocnemius EMG ( p.07). The minimal normalized mean moment was marginally significantly different from 1 ( p.05). The optimal interval between deep peroneal nerve stimulation and triceps surae stretch, at which the reflexes were lowest, was between 100 to 150msec (fig 3). Velocity tests. The velocity dependence of the triceps surae stretch reflex is illustrated in figure 4, showing the peak unconditioned and conditioned stretch reflex EMG for soleus and medial gastrocnemius and the mean unconditioned and conditioned stretch reflex moment as a function of the ankle

6 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink 1021 Fig 3. Normalized minimum stretch reflexes after conditioning stimulation of the deep peroneal nerve (upper panel) and the intervals between conditioning stimulation and triceps surae stretch for which they occur (lower panel). For the relaxed condition (n 10), the average mean standard error are shown for 3 levels of conditioning stimulation: 0.9 and 1.5 times tibialis anterior motor threshold and max, the highest stimulation level used for each individual subject (2 to 4 times motor threshold). Results are shown for normalized peak EMG reflex responses for soleus ( ) and medial gastrocnemius ( ), and for normalized mean moment ( ). angular velocity during triceps surae stretch. Conditioning (by stimulating the deep peroneal nerve) clearly influences this function if the stimulation is performed at a sufficiently high level. The stimulation level in this example was 3 times the motor threshold of the tibialis anterior muscle. The values of the mean ankle moment with and without conditioning stimulation (fig 4C) are the same for small angular velocities below the velocity threshold of the stretch reflex, indicating that subtracting the tibialis anterior motor response from the recordings with conditioning stimulation is an accurate operation that does not influence the comparison of mean moments with and without stimulation. In order to characterize the velocity dependence of the stretch reflex EMG, we fitted linear relations to the velocity dependencies of the peak stretch reflex EMG with and without prior conditioning stimulation (fig 4A,B) and then analyzed what influence stimulation level had on thresholds and slopes. The relative slopes with and without stimulation and the difference of the thresholds of these fitted lines are shown in figures 4D,E. At the highest stimulation levels, the relatives slopes were significantly smaller than 1 for soleus ( p.02), but not for medial gastrocnemius ( p 0.1). The thresholds increased significantly in both muscles ( p.006 for soleus, p.02 for medial gastrocnemius). The median increase of threshold at the highest stimulation levels was 25 /sec for soleus and 31 /sec for medial gastrocnemius. As a reference, the median threshold without stimulation was 8 /sec for soleus and 14 /sec for medial gastrocnemius. The mean value of the relative slopes at the highest stimulation level was.66 for soleus and.71 for medial gastrocnemius. The threshold differences with and without stimulation were significantly dependent on the stimulation level (ANOVA, p.002 for Sol, p.007 for MG). When we compared stimulation levels we found that only the thresholds at maximum stimulation level differed significantly from those at a stimulation level of 0.9 times the motor threshold of tibialis anterior for both soleus and medial gastrocnemius EMG. The relative slopes with and without stimulation did not depend on stimulation level (one-way ANOVA, p.07 for Sol, p 0.3 for MG). The conditioning stimulation s overall effect on the velocity relations was quantified by the area under the velocity relations with conditioning stimulation, normalized on the area without stimulation. Figure 5 shows the means and standard errors of the normalized areas under the velocity relations over all subjects for relaxed and precontracted triceps surae. For the relaxed condition (10 subjects), the results are shown for 3 levels of conditioning stimulation: 0.9 and 1.5 times tibialis anterior motor threshold and the highest stimulation levels used for each individual subject ( max in figure 5). The normalized areas under the velocity relation depended significantly on stimulation level (ANOVA, p.001 for the peak stretch reflex EMG of Sol and MG, p.002 for the mean ankle moment). At the highest levels of stimulation the normalized areas under the velocity relations were significantly smaller than 1 ( p.001 for peak stretch reflex EMG of Sol, MG, and mean ankle moment), indicating that the stretch reflex was effectively reduced over the whole velocity range at sufficiently high stimulation levels. In precontracted triceps surae, the inhibiting effect of the conditioning stimulation on the triceps surae stretch reflex was less pronounced than for the relaxed situation (fig 5). The stretch reflex inhibition in precontracted muscle, as expressed by the normalized area difference under the velocity relations, was only significant in the mean stretch reflex moment (p.02), but not for peak stretch reflex EMG ( p 0.4 for Sol and p 0.7 for MG). DISCUSSION In the present study, we found large and significant inhibitions of the stretch reflex in relaxed triceps surae when the deep peroneal nerve was stimulated before the stretch. These significant inhibitions occurred if the interval between conditioning stimulation and triceps surae stretch was chosen optimally and the level of stimulation was sufficiently high, in most cases between 2 to 4 times motor threshold of the tibialis anterior. At the highest stimulation levels reflex reduction was clearly associated with significantly increased stretch velocity thresholds above which reflexes in the triceps surae occurred. Also, the normalized area under the relation between stretch velocity and stretch reflex decreased significantly when the deep peroneal nerve received conditioning stimulation. In precontracted muscles, this inhibition effect was much smaller, but still significant in the moment, although not in the peak stretch EMG of soleus and medial gastrocnemius. We have demonstrated that conditioning stimulation of the peroneal nerve in spastic stroke patients can prevent stretch reflexes in the triceps surae. Conditioning stimulation may counteract exaggerated triceps surae stretch reflexes during early stance phase and allow for increased dorsiflexion speeds during early swing phase of gait, when the triceps surae is stretched.

7 1022 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink Fig 4. The dependency of the stretch reflex in relaxed triceps surae on the ankle dorsiflexion velocity. (A,B,C) An example of the stretch velocity dependencies of the peak reflex EMG with and without conditioning stimulation for (A) soleus and (B) medial gastrocnemius and (C) of the mean moment with and without conditioning stimulation (subject 10; stimulation level: 3 times motor threshold of the tibialis anterior). The error bars indicate standard errors. Dashed lines in (A) and (B) indicate the stretch velocity dependencies of the peak reflex EMG, which were fitted with linear relations, only considering points above the velocity threshold and below reflex saturation. (D,E) The distribution of threshold differences with and without conditioning stimulation versus relative slopes of these linear fits for (D) soleus and (E) medial gastrocnemius EMG for all 10 subjects;, stimulation at 0.9 times motor threshold of the tibialis anterior;, stimulation at 1.5 times this threshold;, stimulation at the highest stimulation levels. Possible Inhibitory Mechanisms The inhibition of the stretch reflex in the triceps surae by conditioning stimulation of the deep peroneal nerve may be attributed to several mechanisms, of both central and peripheral nature. Crone and Nielsen 27 reported that subthreshold stretches of the relaxed triceps surae by an Achilles tendon tap may result in a long-lasting inhibition of the H-reflex, lasting over 8sec (postactivation depression). Hultborn and colleagues 28 suggested that the phenomenon is most likely caused by a reduced transmitter release from previously activated Ia afferent fibers. Although we tightly strapped the ankle, the contraction of tibialis anterior may have caused the triceps surae to stretch slightly, because the ankle is not a rigid structure. Subsequent stretch reflexes may be reduced by postactivation depression. However, this peripheral effect can only cause part of the stretch reflex inhibition found in our study because (1) about half of the inhibition is still present in precontracted muscle at optimal conditioning test interval (see fig 3), while postactivation depression has only been found in relaxed muscle; and (2) inhibition of the soleus stretch reflex still existed when we blocked the common peroneal nerve of a healthy subject distal to the site of stimulation in a preliminary experiment using lidocaine, effectively suppressing the contraction of the tibialis anterior. Evidently, direct stimulation of the afferents in the peroneal nerve contributes significantly to the inhibition, and peripheral mechanisms like postactivation depression can only partly explain the inhibition. Stimulation of the deep peroneal nerve up to 4 times tibialis anterior motor threshold may cause several populations of afferent fibers to be activated. Questions remain about which afferents cause the inhibition and which central pathways are involved. No strong indication was found for different mechanisms of stretch reflex inhibition at different stimulation levels, since the optimal conditioning stretch interval was only marginally significantly dependent on stimulation level in medial gastrocnemius stretch reflex EMG, and not at all significant in soleus stretch reflex EMG and ankle moment. The inhibitory mechanisms may be of postsynaptic or presynaptic nature. Our study indicated that the inhibition of the triceps surae stretch reflex by stimulation of the deep peroneal nerve was mainly associated with a clear increase in velocity threshold, which depends strongly on stimulation level. The reduction of the slope of the velocity dependency of the stretch reflex was only significant for soleus stretch reflex EMG, not for medial gastrocnemius stretch reflex EMG, and appeared not to depend on stimulation level. The predominant effect of a change in threshold velocity indicates that postsynaptic inhibition mechanisms prevail over presynaptic mechanisms. The

8 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink 1023 blocked peroneal nerve distal to the stimulation site indicated that a large part of the inhibition still occurred without tibialis anterior contraction. Therefore, the inhibition at relatively long conditioning test intervals may also involve more complex spinal and supraspinal pathways, or may result from the direct stimulation of smaller, more slowly conducting afferents. Fig 5. Normalized areas under the relations between stretch velocity and conditioned reflex magnitude. For the relaxed condition (n 10), the results are shown for 3 levels of conditioning stimulation: 0.9 and 1.5 times tibialis anterior motor threshold and max, the highest stimulation level used for each individual subject (2 to 4 times motor threshold). Results are shown for normalized peak EMG reflex responses for soleus ( ) and medial gastrocnemius ( ), and for normalized mean moment ( ). The normalized areas under the velocity relation for relaxed triceps surae were significantly dependent on stimulation level (ANOVA, p F.001) for the peak stretch reflex EMG of both soleus and medial gastrocnemius and p.002 for the mean ankle moment) and significantly different from one at the maximum stimulation levels (p F.001). analysis of thresholds and slopes of the velocity dependencies of the stretch reflex should be interpreted cautiously, because the linearized stretch velocity/reflex relations yield a rather coarse simplification. 22 Disynaptic reciprocal inhibition by stimulation of Ia afferents was not tested systematically in the present study. The optimal conditioning test intervals for disynaptic reciprocal inhibition is approximately 2msec. 11 In the present study, no inhibition was observed in the moment at conditioning test intervals of 6msec and 14msec (the normalized mean ankle moment was not significantly different from 1 for these intervals). However, we did not check at intermediate intervals, which may have been necessary to observe the disynaptic inhibition effect. Crone 11 reported this inhibition to occur for a conditioning test interval range between 0 and 5msec. However, the average inhibition at a conditioning test interval of 2msec found by Crone in healthy subjects was only 15%, when applying a conditioning stimulation at the motor threshold of the tibialis anterior. This inhibition level is relatively small with respect to the inhibition levels we found at the highest stimulation levels between 2 and 4 times tibialis anterior motor threshold. Also, Crone 11 did not find disynaptic reciprocal inhibition in spastic multiple sclerosis patients. We found optimal conditioning test intervals of about 150msec at the highest conditioning stimulation levels (see fig 3). This finding is in agreement with the average optimal conditioning test interval of 160msec found by Apkarian and Naumann. 14 Also in H-reflex studies a large inhibition of 46% in the soleus H-reflex was reported by Capaday 16 at conditioning test intervals of 100 to 120msec, when stimulating the common peroneal nerve. Several mechanisms may play a role in the inhibition effects at the large optimal conditioning test intervals found. Apkarian and Naumann 14 suggest that the contraction of the tibialis anterior is important for the inhibition effect. Our preliminary experiment on a healthy subject with Functional Implications In spastic patients the stretch reflex may be affected. Sinkjær 29,30 found a reduced soleus stretch reflex modulation during gait in spastic multiple sclerosis patients, which was not found in spastic stroke patients. 20 However, Nielsen 20 found that the threshold velocity of the soleus stretch reflex was markedly lower in the early swing phase of gait in spastic stroke patients (108 /sec) than in healthy subjects (309 /sec). Evidently the soleus stretch reflex is elicited at a smaller dorsiflexion velocity in persons with stroke than in normal persons. It has been argued that an affected reflex modulation or decreased velocity thresholds of the triceps surae stretch reflex may limit the gait performance in spastic subjects, 20,21 specifically during the gait phases in which the triceps surae is stretched. Such dorsiflexion movements occur in the beginning of the stance and swing phases. Phasic stretch reflex activity in the ankle extensors may be seen during the beginning of the stance phase. 21,31 It is unclear how much this affects actual gait, but it may prevent spastic subjects to progress during stance. 21 The markedly reduced stretch reflex threshold in early swing may prevent spastic stroke patients to dorsiflex their ankle at normal speed. The results of this study call for further investigations of stretch reflex inhibition during gait 32 to determine if an inhibition of the exaggerated triceps surae stretch reflex will improve walking in spastic patients. Acknowledgments: The research was done at the Center for Sensory-Motor Interaction (SMI) of the University of Aalborg (Denmark). We acknowledge Knud Larsen, MS, EE, for his support and assistance in the experiments and the subjects for their participation in the experiments. References 1. Lance JW. Symposium synopsis. In: Feldman RG, Young RR, Koella WP, editors. Spasticity: disordered motor control. Chicago: Yearbook; p Katz RT, Rymer WZ. Spastic hypertonia: mechanisms and measurement. Arch Phys Med Rehabil 1989;70: Garcia-Mullin R, Mayer RF. H reflexes in acute and chronic hemiplegia. Brain 1972;95: Magladery JW, Teasdall RD, Park AM, Languth HW. Electrophysiological studies of reflex activity in patients with lesions of the nervous system. I. A comparison of spinal motoneurone excitability following afferent nerve volleys in normal persons and patients with upper motor neurone lesions. Bull Johns Hopkins Hosp 1952;91: Olsen PZ, Diamantopoulos E. Excitability of spinal motor neurones in normal subjects and patients with spasticity, parkinsonian rigidity, and cerebellar hypotonia. J Neurol Neurosurg Psychiatry 1967;30: Sax DS, Johnson TL. Spinal reflex activity in man. Measurement in relation to spasticity. In: Feldman RG, Young RR, Loella WP, editors. Spasticity: disordered motor control. Chicago: Yearbook; 1980, p Takamori M. H reflex study in upper motoneuron diseases. Neurology 1967;17: Yap CB. Spinal segmental and long-loop reflexes on spinal motoneuron excitability in spasticity and rigidity. Brain 1967;90: Delwaide PJ. Human monosynaptic reflexes and presynaptic inhibition. An interpretation of spastic hyperreflexia. In: Desmedt

9 1024 STRETCH REFLEX INHIBITION IN SPASTIC STROKE, Veltink JE, editor. Human reflexes, pathophysiology of motor systems, methodology of human reflexes. Basel: Karger; p Faist M, Mazevet D, Dietz V, Pierrot-Deseilligny E. A quantitative assessment of presynaptic inhibition of Ia afferents in spastics. Differences in hemiplegics and paraplegics. Brain 1994;117: Crone C, Nielsen J, Petersen N, Ballegaard M, Hultborn H. Disynaptic reciprocal inhibition of ankle extensors in spastic patients. Brain 1994;117: Noth J. Trends in the pathophysiology and pharmacotherapy of spasticity. J Neurol 1991;238: Nielsen JF, Sinkjær T. Peripheral and central effect of baclofen on ankle joint stiffness in multiple sclerosis. Muscle Nerve In press. 14. Apkarian JA, Naumann S. Stretch reflex inhibition using electrical stimulation in normal subjects and subjects with spasticity. J Biomed Eng 1991;13: Crone C, Hultborn H, Jespersen B, Nielsen J. Reciprocal Ia inhibition between ankle flexors and extensors in man. J Physiol 1987;389: Capaday C, Lavoie BA, Comeau F. Differential effects of a flexor nerve input on the human soleus H-reflex during standing versus walking. Can J Physiol Pharmacol 1995;73: Fung J, Barbeau H. Effects of conditioning cutaneomuscular stimulation on the soleus H-Reflex in normal and spastic paretic subjects during walking and standing. J Neurophysiol 1994;72: Van Wezel BMH, Ottenhoff FAM, Duysens J. Dynamic control of location-specific information in tactile cutaneous reflexes from the foot during human walking. J Neurosci 1997;17: Zehr EP, Komiyama T, Stein RB. Cutaneous reflexes during human gait: electromyographic and kinematic responses to electrical stimulation. J Neurophysiol 1997;77: Nielsen JF, Andersen JB, Barbeau H, Sinkjær T. Input-output properties of the soleus stretch reflex in spastic stroke patients and healthy subjects during walking. NeuroRehabil 1998;10: Stein RB, Yang JF, Bélanger M, Pearson KG. Modification of reflexes in normal and abnormal movements. Prog Brain Res 1993;97: Kernell D, Hultborn H. Synaptic effects on recruitment gain: a mechanism of importance for the input-output relations of motoneurone pools. Brain Res 1990;507: Sinkjær T, Toft E, Andreassen S, Hornemann BC. Muscle stiffness in human ankle dorsiflexors: intrinsic and reflex components. J Neurophysiol 1988;60: Meunier S, Pierrot-Deseilligny E, Simonetta M. Pattern of monosynaptic heteronymous Ia connections in the human lower limb. Exp Brain Res 1993;96: Gracies JM, Pierrot-Deseilligny E, Robain G. Evidence for further recruitment of group I fibres with high stimulus intensities when using surface electrodes in man. Electroencephalogr Clin Neurophysiol 1994;93: Toft E, Sinkjær T, Andreassen S, Larsen K. Mechanical and electromyographic responses to stretch of the human ankle extensors. J Neurophysiol 1991;65: Crone C, Nielsen J. Methodological implications of the post activation depression of the soleus H-reflex in man. Exp Brain Res 1989;78: Hultborn H, Illert M, Nielsen J, Paul A, Ballegaard M, Wiese H. On the mechanism of the post-activation depression of the H-reflex in human subjects. Exp Brain Res 1996;108: Sinkjær T, Andersen JB, Nielsen JF. Impaired stretch reflex and joint torque modulation during spastic gait in multiple sclerosis patients. J Neurology 1995;243: Sinkjær T. Muscle, reflex and central components in the control of the ankle joint in healthy and spastic man. Acta Neurol Scand Suppl 1997;96: Veltink PH, Ladouceur M, Sinkjær T. Stretch reflex contribution to soleus activation during spastic gait. In: Proceedings of the 20th Annual International Conference of the IEEE-EMBS; 1998 Oct 29-Nov 1; Hong Kong. Piscataway (NJ): IEEE; p Merletti R, Andina A, Galante M, Furlan I. Clinical experience of electronic peroneal stimulators in 50 hemiparetic patients. Scand J Rehabil Med 1979;11: Suppliers a. DC motor, CEM model 26; ETS Parvex, 8 avenue du Lac, B.P. 249, Dijon Cedex, France. b. DC power amplifier, model 2708; Bruel and Kjær, Nærum, Denmark, DK c. EMG amplifier, DISA model 15C01; Dantec Measurement Technology A/S, Tonsbakken 16, Skovlunde, Denmark, DK d. Isolator 11 Stimulator; Axon Instruments Inc, Chess Dr, Foster City, CA e. MATLAB; MathWorks, Inc, 24 Prime Park Way, Natick, MA f. SIGMASTAT; SPSS Science Software GmbH, Postfach 4107, Erkrath, Germany.