Abstract
Background Commonly used spasticity scales assess the resistance felt by the evaluator during passive stretching. These scales, however, have questionable validity and reliability. The tonic stretch reflex threshold (TSRT), or the angle at which motoneuronal recruitment begins in the resting state, is a promising alternative for spasticity measurement. Previous studies showed that spasticity and voluntary motor deficits after stroke may be characterized by a limitation in the ability of the central nervous system to regulate the range of the TSRT.
Objective The study objective was to assess interevaluator reliability for TSRT plantar-flexor spasticity measurement.
Design This was an interevaluator reliability study.
Methods In 28 people after stroke, plantar-flexor spasticity was evaluated twice on the same day. Plantar-flexor muscles were stretched 20 times at different velocities assigned by a portable device. Plantar-flexor electromyographic signals and ankle angles were used to determine dynamic velocity-dependent thresholds. The TSRT was computed by extrapolating a regression line through dynamic velocity-dependent thresholds to the angular axis.
Results Mean TSRTs in evaluations 1 and 2 were 66.0 degrees (SD=13.1°) and 65.8 degrees (SD=14.1°), respectively, with no significant difference between them. The intraclass correlation coefficient (2,1) was .851 (95% confidence interval=.703, .928).
Limitations The notion of dynamic stretch reflex threshold does not exclude the possibility that spasticity is dependent on acceleration, as well as on velocity; future work will study both possibilities.
Conclusions Tonic stretch reflex threshold interevaluator reliability for evaluating stroke-related plantar-flexor spasticity was very good. The TSRT is a reliable measure of spasticity. More information may be gained by combining the TSRT measurement with a measure of velocity-dependent resistance.
One of the common impairments observed after neurological lesions, such as a stroke, is spasticity. According to Lance, spasticity is a motor disorder characterized by a velocity-dependent increase in tonic stretch reflexes (“muscle tone”) with exaggerated tendon jerks, resulting from hyperexcitability of the stretch reflex as one component of the upper motor neuron syndrome.1 This motor disorder may lead to secondary complications, such as pain and muscle contractures. People with stroke-related spasticity, therefore, may experience difficulties in performing activities of daily living and a reduced health-related quality of life.2 The prevalence of spasticity in people after stroke is highly variable, ranging from 17.0% to 42.6% of people in the chronic phase of recovery.3 The use of an appropriate clinical measurement for spasticity is important for differentiating between types of hypertonia,4 such as spasticity1 and rigidity,5 evaluating the progression of spasticity over time, and determining the effectiveness of therapeutic interventions in recovery after stroke.
Ashworth Scales are the clinical scales most commonly used to estimate spasticity. Some features make these scales attractive for clinicians, such as their ease of use in clinical settings (because they do not require any sophisticated equipment) and the short time needed to administer them. The original Ashworth Scale6 is a subjective 5-point ordinal scale for grading resistance felt by the evaluator during passive stretching. To render the Ashworth Scale more sensitive, modifications were proposed by Bohannon and Smith7 (Modified Ashworth Scale) and then by Ansari et al8 (Modified Modified Ashworth Scale).
Despite all of these attempts to improve Ashworth Scales, their validity is still questionable.9,10 Indeed, these scales assess the perceived resistance to passive movement, which is influenced not only by reflex responses in the stretched muscle but also by changes in passive resistance of noncontractile and contractile properties, as well as changes in the resistance of the shortening antagonist muscle.11,12 The fact that nonreflex components are not discriminated from reflex components compromises the construct validity of Ashworth Scales. More critically, scores on Ashworth Scales do not take into account the velocity dependence of the response. Because Ashworth Scales measure the perceived resistance to passive movement, which varies with the velocity of the stretch, the failure to control stretch velocity negatively affects the reliability of these scales.7
The Tardieu Scale has been suggested as a more valid alternative for spasticity assessment.13 Unlike Ashworth Scales, the Tardieu Scale takes into consideration the velocity-dependent aspect of the stretch reflex response. This scale was originally developed by Tardieu et al14 and has gone through many revisions.15 The most recent versions of the Tardieu Scale compare the passive resistance when the muscle is stretched at a slow velocity with that when the muscle is stretched at a fast velocity. The maximal angle reached during a slow stretch and the angle at which a catch or clonus is felt during a fast stretch are determined. The latter angle is then subtracted from the former angle, and the resulting value reflects the dynamic, or velocity-dependent, component of spasticity.16
The resistance perceived during stretching also is characterized on a subjective 6-point ordinal scale. Although the construct assessed by the Tardieu Scale is more closely related to Lance's definition of spasticity,1 this scale also focuses on the perceived resistance to passive movement, which is a consequence of spasticity and may not reflect the neurological origin of spasticity. More information may be gained by combining a resistance measurement with an electromyographic (EMG) response analysis. Indeed, there is consensus among researchers that neurophysiological measures should be at least part of a spasticity evaluation given that they provide information about the pathways that are altered in spasticity.17–22
The stretch reflex threshold (SRT),23–25 which indicates the joint angle at which motoneuronal recruitment begins for a specific velocity of stretch, has been suggested as a functionally relevant measure of spasticity. Previous studies demonstrated that spasticity and voluntary motor deficits in the upper limb after stroke may both be characterized by a limitation in the ability of the central nervous system to regulate the SRT.26–28 The SRT depends on stretch velocity (dynamic stretch reflex threshold [DSRT]) and reflects motoneuronal excitability.25,29 As for the other measurements of spasticity, the reliability of the SRT depends on how well the stretch velocity can be reproduced from one stretch to another; meeting this goal is difficult when the stretch is performed manually by a clinician.
A more reliable measure of the intrinsic state of the neuromuscular system can be provided by identifying the tonic stretch reflex threshold (TSRT). In other words, the TSRT identifies the minimal joint angle at which abnormal motoneuronal recruitment begins when the muscle is at rest and there is no motion. The theoretical construct of the TSRT has been validated as a measure of spasticity in elbow flexor28,30 and extensor30 muscles of patients with chronic stroke. In recognition of the clinical reality, TSRT testing has been implemented in a portable device requiring a minimal amount of equipment (Montreal Spasticity Measure).21,28
Some psychometric properties of TSRT estimation were evaluated previously; moderately good intra- and interevaluator reliability for the measurement of elbow flexor spasticity in participants after stroke was reported.28 Moreover, the TSRT can discriminate between spasticity and rigidity.5 The aim of the present study was to quantify the interevaluator reliability of the TSRT for measurement of the spasticity of the ankle plantar-flexor muscles, another muscle group frequently affected by stroke-related spasticity.31,32 Preliminary results appeared in abstract form.33
Method
Participants
People who sustained a stroke were recruited from institutions within the Center for Interdisciplinary Research in Rehabilitation of Greater Montreal (CRIR; Montreal, Quebec, Canada). People were included if they had a history of spasticity in one or both ankle plantar-flexor muscles, as estimated with the Ashworth Scale (≥1), and a passive ankle range of motion of at least 15 degrees, so that they would have adequate range of motion for stretching and for reaching high stretch velocities. People were excluded if they had pain in the evaluated ankle. Any medication use was documented.
Twenty-eight people (7 women) who had sustained a stroke and whose mean age was 57.4 years (SD=9.7, range=40–72) were included. All participants were in the subacute or chronic phase of recovery (time since injury ranging from 2 to 160 months). Most participants had sustained a stroke on the right side of the brain (n=18), resulting in left-side hemiparesis. The demographic characteristics of the participants and clinical data are summarized in Table 1. All participants read and signed a consent form approved by the CRIR Ethics Review Board.
Demographic Characteristics of Participants and Clinical Dataa
Experimental Protocol
In this interevaluator reliability study, participants came to the laboratory for a single 2-hour visit. For each participant, plantar-flexor spasticity in the most affected ankle was evaluated with the Montreal Spasticity Measure twice (evaluation 1 and evaluation 2) on the same day by 2 different evaluators.
SRT Testing With the Montreal Spasticity Measure
Participants adopted a comfortable position, lying supine, with the knees slightly flexed (∼30°). Before testing was done, participants performed an isometric maximal voluntary contraction of the plantar-flexor muscles to adjust the gain of the EMG recordings. With the participants completely relaxed, the baseline EMG signal and the initial ankle angle, which corresponded to the neutral position in the frontal plane and full plantar flexion in the sagittal plane, were then determined and recorded. The plantar-flexor muscles were then stretched manually by moving the ankle from the initial position toward full dorsiflexion at different velocities. Before each stretch, an auditory cue emitted by the Montreal Spasticity Measure indicated to the evaluator approximately which velocity to use (slow, moderate, or fast velocity). Auditory signals consisted of 7 consecutive tones decreasing in pitch. They were randomly generated and equally distributed among slow, moderate, and fast velocities. Each evaluator performed a minimum of 20 stretches. During the stretches, participants were instructed to relax completely and not assist or resist the passive angular displacement. Each evaluation lasted about 20 minutes, with a 10-minute rest period between evaluation 1 and evaluation 2.
Nine different evaluators participated in this experiment. Over the course of the data collection process (∼2 years), unexpected staff movements caused modifications in pairs of evaluators. There was no systematic pairing of evaluators with participants. The number of assessments per evaluator ranged from 3 to 15. To standardize the TSRT measurement, all evaluators read documentation and participated in two 1-hour training sessions with the Montreal Spasticity Measure portable device.
Data Recording
Responses to stretching in the ankle plantar-flexor muscles of the paretic lower limb were recorded. The ankle angular position was recorded by use of an electrogoniometer (servo-type rotational-position potentiometer P2200, Novotechnik US Inc, Southborough, Massachusetts) with the axis of rotation positioned over the lateral malleolus. The proximal and distal arms of the electrogoniometer were aligned with the fibula and the fifth metatarsal bone, respectively, and were attached with self-adhesive Velcro (Velcro USA Inc, Manchester, New Hampshire) straps. Ankle passive range of motion was evaluated manually with the electrogoniometer. For EMG recordings, after cleaning of the skin with alcohol, silver-silver chloride disposable surface electrodes (Ambu BlueSensor P, Ambu A/S, Ballerup, Denmark) were placed over the lateral gastrocnemius muscle or the soleus muscle (for participants 20, 24, and 28). A reference electrode was positioned over the head of the fibula. For electrode placement and locations, recommendations from the Surface ElectroMyoGraphy for the Non-Invasive Assessment of Muscles (SENIAM) were followed.34 In brief, anatomical landmarks were used to determine muscle position. Bipolar electrodes were then placed over the belly of the muscle parallel to the muscle fibers. Interelectrode distance was about 2 cm. Plantar-flexor EMG signals were first preamplified (by ∼700). Depending on the gain adjustment executed before the testing, signals may have been amplified with another factor, ranging from 1 to 10. Plantar-flexor EMG signals were sampled at 1,000 Hz.
Data Analysis
The analysis was performed automatically by the Montreal Spasticity Measure software. Electromyographic signals were digitally filtered with a zero-lag fourth-order Butterworth filter (band-pass filter, 50–350 Hz), rectified, and then smoothed with a 21-millisecond moving average to obtain the envelope. Angular position signals were processed with a zero-lag fourth-order Butterworth filter (low-pass filter, 20 Hz) to remove noise introduced by analog-to-digital conversion. Angular velocities were obtained by differentiation of the angular position signals.
Before each stretch, the Montreal Spasticity Measure software ensured that 3 criteria were met: (1) the EMG activity level was below the mean baseline plus the standard deviation, (2) the starting joint angle was within ±10 degrees of the initial ankle angular position, and (3) the velocity was less than 10 degrees per second. For each trial, a computer algorithm was used to determine the stretch response onset on the basis of the plantar-flexor EMG envelope through identification of the point in time at which the envelope rose and remained above 3 standard deviations of the baseline for a minimum of 25 milliseconds, whether or not clonus was present. This time was used to identify the corresponding angle and velocity from respective time plots. A combination of threshold angle and velocity, therefore, was recorded (DSRTs) (Fig. 1) for each identified EMG onset.
Examples of dynamic stretch reflex threshold (DSRT) identification in spastic ankle plantar-flexor muscles at 3 different velocities (slow, moderate, and fast) for participant 16. For each velocity of stretch, ankle angular position (top), velocity (middle), and rectified plantar-flexor electromyographic (EMG) activity (bottom) are shown. Vertical dashed lines indicate DSRTs (combinations of threshold position and velocity) for each trial. DF=dorsiflexion.
The algorithm was then used to compute a regression line based on a first-order linear equation through DSRTs on a velocity/angle plot. The coefficient of determination (measure of goodness of fit), slope, and intercept with the x-axis were extracted from the equation. The intersection of the regression line and the x-axis corresponded to the joint angle at which the TSRT was located.35 A low TSRT corresponded to a high level of spasticity.
A time-dimensional parameter, μ, that defined the sensitivity of the DSRT to stretch velocity was calculated as the inverse of the slope (α) of the velocity/angle curve (μ=−1/α).
The term dynamic SRT was used to distinguish the SRTs evoked by stretches at different velocities from tonic SRT, which was computed from the data. The word dynamic in reference to the SRT is a general physical term and should not be identified with the dynamic component of muscle spindle afferents; it also depends on properties of other afferents, interneurons, and alpha motoneurons.
For each participant, clonus was identified qualitatively on the basis of plantar-flexor EMG signals. Clonus was considered to be present when a phasic response (multiple bursts) was observed in plantar-flexor activity after a fast velocity stretch, but it was not quantified because it occurred after the threshold was reached.
Statistical Analysis
For the assessment of interevaluator reliability, 2 complementary approaches were used: paired-sample t test and 2-way random-effects intraclass correlation coefficient model. The paired-sample t test was used to compare the parameters describing the regression line—that is, intercept with x-axis (TSRT), velocity sensitivity (μ), and coefficient of determination—derived from evaluations 1 and 2. An additional statistical analysis, intraclass correlation coefficient (ICC [2,1]),36 was used to calculate the absolute agreement between the TSRT estimations made by the different evaluators. Statistical analyses were performed with SPSS Statistics 17.0 software (SPSS Inc, Chicago, Illinois).
Role of the Funding Source
This work was supported by the Canadian Physiotherapy Foundation, Univalor, and Collaborative Health Research Projects.
Results
Computation of the Regression Line Through DSRTs
Stretch responses were not evoked in 17.3% (SD=16.5%) of the trials, and DSRTs were identified and then rejected because of false-positive results due to EMG artifacts in 9.6% (SD=8.5%) of the trials. Electromyographic increases observed after stretching of the muscle through the entire range of motion were considered to be EMG artifacts. A mean of 17.3 (SD=9.8) DSRTs (a minimum of 9 DSRTs) executed at different velocities was used to compute the regression lines for all 28 participants. The goodness of fit (coefficient of determination) for each regression line was acceptable, ranging from .712 to .999. Examples of regression lines computed to estimate TSRTs in representative participants are shown in Figure 2.
Examples of tonic stretch reflex threshold (TSRT) estimations in 2 representative participants (participant 5 [left panel] and participant 27 [right panel]). Dynamic stretch reflex thresholds (DSRTs) in plantar-flexor muscles at different velocities in evaluation 1 (EVAL 1) (gray diamonds) and evaluation 2 (EVAL 2) (black diamonds) are shown. Regression line equations, coefficients of determination (R2), inverse of the slope of the velocity/angle curve (μ), and TSRTs are shown.
Interevaluator Reliability
Mean TSRTs in evaluations 1 and 2 were 66.0 degrees (SD=13.1°) and 65.8 degrees (SD=14.1°), respectively (Tab. 2). The absolute difference between evaluations was 6.1 degrees (SD=4.3°) (95% confidence interval=4.43263, 7.76737). The mean μ value measured in evaluation 1 was 0.008 second (SD=0.026), whereas a mean μ value of 0.042 second (SD=0.149) was obtained in evaluation 2. The paired-sample t test revealed no significant difference in either TSRTs (t=0.176, P=.86) or μ values (t=−1.144, P=.26) between evaluations.
Interevaluator Reliability for Tonic Stretch Reflex Threshold (TSRT) Plantar-Flexor Spasticity Measurementa
The ICC for the absolute agreement between the TSRT estimations made by the different evaluators for all participants was .851 (95% confidence interval=.703, .928; P<.001), indicating very good interevaluator reliability. Figure 3 shows the agreement between TSRT estimations.
Interevaluator agreement for tonic stretch reflex threshold (TSRT) estimations. The scatterplot illustrates TSRTs obtained in evaluation 1 (EVAL 1) and evaluation 2 (EVAL 2) for each participant. Each point represents one participant. Points falling on the dashed gray line would indicate perfect agreement between evaluators. The solid black line represents the best fit for all data. R2=coefficient of determination.
Discussion
The interevaluator reliability of the TSRT for the evaluation of ankle plantar-flexor spasticity after stroke was very good.
SRT and Gain
Different parameters, such as SRT and gain, have been used to characterize the stretch reflex response.25,29,37–39 Although both likely contribute to the clinical phenomenon of spasticity, the present study focused on the spatial threshold, defined as the joint angle at which motoneuronal recruitment begins and the gain relationship originates. Many studies have suggested that neurophysiological measures, such as EMG activity, either alone or in combination with biomechanical measurements, be used to evaluate spasticity.17–22 However, the quantification of response magnitude or gain may not be optimal given that the amplitude of EMG responses may vary because factors such as the placement of electrodes, skin resistance, subcutaneous fat, and muscle atrophy can lead to additional variability between and within individuals (from session to session).40 The use of TSRTs identified with EMG signals may therefore be a good alternative.
Interevaluator Reliability of TSRTs
Many factors may affect spasticity measurement. Reflex responses may be influenced by the initial tension in the muscle41 and by different biomechanical parameters, such as stretch velocity and joint position.27,42–44 To minimize the impact of differences in evaluator techniques on the reliability of the measurement, evaluators used a dedicated portable spasticity measurement device specifically designed to evoke DSRTs and compute TSRTs. In contrast to the commonly used clinical evaluation of spasticity, the Montreal Spasticity Measure software used to measure the TSRT takes these parameters into consideration by helping the evaluator reproduce similar conditions. Before stretching the muscle, the software ensured that the plantar-flexor EMG signal was below a certain level of activity (mean baseline + SD) and that the starting joint angle was within ±10 degrees of the preset initial angle. A cursor helped the evaluator target the same initial joint angle before each stretch and perform the stretch smoothly throughout the entire range of motion.
In addition, to minimize differences in evaluator techniques, the evaluators read the Montreal Spasticity Measure user guide, which provided them with specific instructions on how to perform muscle stretching; participated in two 1-hour training sessions to familiarize themselves with the equipment; and practiced muscle stretching at different velocities, with feedback from the instructors. Acquiring experience with the portable device and practicing a standardized technique to execute the spasticity evaluation probably contributed positively to the reliability score obtained in the present study.
However, despite the very good result in the present study, some factors may have negatively influenced reliability. The fact that 9 different evaluators participated in the present study may have been a limitation because having multiple pairs of evaluators represents a source of variability, especially in an interrater reliability study. If the same pair of evaluators had collected data throughout the entire study, then the absolute agreement between evaluators may have been better or, at least, not have been worse. Although the evaluator could not have directly or indirectly influenced the TSRT identification, the lack of masking of the evaluator could have introduced bias in the present study.
Although the reliability of resistance measurements is negatively affected by a lack of standardization in the stretch velocity used to evoke the response, TSRT reliability takes advantage of large variations in stretch velocities. Indeed, the precision of the intercept (TSRT) depends on the model used to estimate the relationship and the goodness of fit of the regression line. A wider range of stretch velocities, therefore, would result in a regression line with a better fit and—consequently—a more precise intercept with the x-axis, which corresponds to the TSRT.
Validity of TSRT Measurement
As mentioned in Lance's definition,1 spasticity is a velocity-dependent phenomenon. Therefore, spasticity should be evaluated at different stretch velocities. Contrary to the commonly used clinical evaluation of spasticity, our procedure measured SRTs for stretches applied at different velocities. Stretch reflex thresholds critically depend on the velocity of the stretch (DSRTs); the faster the velocity of the imposed stretch, the earlier the stretch response occurs. The TSRT, which is a unique measure (one TSRT per muscle group), is then based on these multiple DSRTs. The notion of the DSRT does not exclude the possibility that spasticity is dependent on acceleration as well as velocity; this possibility is a topic for future work.
TSRT Measurement for Ankle Plantar-Flexor Muscles
The TSRT was previously used to evaluate spasticity in elbow flexor and extensor muscle groups.5,26–28,35 However, to our knowledge, the present study is the first to show the feasibility of using the TSRT for the evaluation of ankle plantar-flexor spasticity. The evaluation of spasticity at the ankle differed from that at the elbow in the following ways. The ankle plantar-flexor range for the TSRT measurement was smaller than that used to determine the TSRT response in elbow muscles. In particular, the mean passive ankle range of motion was 27.3 degrees (SD=7.2°); in comparison, the value for the upper limb is more than 90 degrees.27 Despite this limitation, it was possible to identify plantar-flexor TSRTs in all participants.
One of the main differences between the results obtained at the elbow joint and those obtained at the ankle joint concerns the sensitivity of DSRTs to stretch velocity. The μ values suggested that ankle plantar-flexor muscle groups are less sensitive to stretch velocity than elbow muscle groups.5,28 This suggestion is consistent with the earlier observations of Given et al45 showing that passive stiffness was significantly higher at the ankle than at the elbow. This difference could be explained by the larger relative amount of intramuscular connective tissue in the ankle muscles or the larger cross-sectional area of the ankle plantar-flexor muscles than of either the elbow flexor or the elbow extensor muscles.
Another factor in the evaluation of plantar-flexor spasticity is the presence of clonus, which is defined as an involuntary rhythmic muscle contraction inducing distal joint oscillations.46–51 In the present study, clonus was elicited in about 75% of the participants (Fig. 1). Despite the fact that clonus can be observed in other upper and lower limb joints,49 the prevalence of clonus at the ankle is higher than that at the elbow. The relationship between the TSRT and clonus is still unknown. Therefore, the SRT onset determined in individual trials may have been under- or overestimated in the presence of clonus. Despite evidence of moderate to good intraevaluator reliability of the TSRT in elbow flexor muscles in people with stroke and cerebral palsy, future investigations will focus on ankle plantar-flexor muscles.
Clinical Relevance
Measures of spasticity commonly used in clinical settings, such as Ashworth Scales and the Tardieu Scale, focus on the perceived resistance to passive movement, which is a consequence of spasticity. Although commonly used, these scales have questionable validity and inconsistent reliability. The TSRT reflects one of the neurophysiological causes of spasticity, that is, limitations in the range of central regulation of the spatial threshold of the stretch reflex, which may better indicate changes in spasticity and motor recovery. Ongoing studies will investigate the responsiveness of this measurement. Any commercial system capable of measuring EMG signals, passive range of motion, and velocity can collect the data needed for TSRT estimation. However, software is necessary to make such a tool operable and interpretable by clinicians.
In conclusion, the TSRT has very good interevaluator reliably and is a physiologically valid measurement of spasticity. More information may be gained by combining the TSRT measurement with a measure of velocity-dependent resistance. The TSRT can be measured with a portable device, the Montreal Spasticity Measure, which can be easily used in clinical settings.
Footnotes
Ms Mullick and Dr Levin provided concept/idea/research design. Dr Blanchette and Dr Levin provided writing and project management. All authors provided data collection and reviewed the manuscript before submission. Dr Blanchette, Ms Moïn-Darbari, and Dr Levin provided data analysis. Dr Levin provided fund procurement, participants, facilities/equipment, institutional liaisons, administrative support, and consultation (including review of manuscript before submission). The authors thank Rhona Guberek, Sandeep Subramanian, Revital Hacmon, Vanessa Gatti, and Ruth Dannenbaum for their help in data collection. They also thank Christian Beaudoin and Valeri Goussev for their valuable contributions in software programming and Eric Johnstone for technical support.
This study was approved by the CRIR Ethics Review Board.
This work was supported by the Canadian Physiotherapy Foundation, Univalor, and Collaborative Health Research Projects.
Dr Levin owns part of the intellectual property rights for the spasticity measurement device used in the study.
- Received May 28, 2014.
- Accepted October 4, 2015.
- © 2016 American Physical Therapy Association