Abstract
Background Limited walking ability is an important problem for patients with multiple sclerosis. A better understanding of how gait impairments lead to limited walking ability may help to develop more targeted interventions. Although gait classifications are available in cerebral palsy and stroke, relevant knowledge in MS is scarce.
Objective The aims of this study were: (1) to identify distinctive gait patterns in patients with MS based on a combined evaluation of kinematics, gait features, and muscle activity during walking and (2) to determine the clinical relevance of these gait patterns.
Design This was a cross-sectional study of 81 patients with MS of mild-to-moderate severity (Expanded Disability Status Scale [EDSS] median score=3.0, range=1.0–7.0) and an age range of 28 to 69 years.
Method The patients participated in 2-dimensional video gait analysis, with concurrent measurement of surface electromyography and ground reaction forces. A score chart of 73 gait items was used to rate each gait analysis. A single rater performed the scoring. Latent class analysis was used to identify gait classes.
Results Analysis of the 73 gait variables revealed that 9 variables could distinguish 3 clinically meaningful gait classes. The 9 variables were: (1) heel-rise in terminal stance, (2) push-off, (3) clearance in initial swing, (4) plantar-flexion position in mid-swing, (5) pelvic rotation, (6) arm-trunk movement, (7) activity of the gastrocnemius muscle in pre-swing, (8) M-wave, and (9) propulsive force. The EDSS score and gait speed worsened in ascending classes.
Limitations Most participants had mild-to-moderate limitations in walking ability based on their EDSS scores, and the number of walkers who were severely limited was small.
Conclusions Based on a small set of 9 variables measured with 2-dimensional clinical gait analysis, patients with MS could be divided into 3 different gait classes. The gait variables are suggestive of insufficient ankle push-off.
In patients with multiple sclerosis (MS), motor deficits most commonly affect the lower extremities, leading to limitations in walking ability and daily activity.1–5 About 85% of patients with MS indicate gait disturbances as their main complaint,6,7 which occur early in the MS disease course.3 Within 15 years of diagnosis, up to 50% of the patients require the assistance of a walking aid, and 10% will be wheelchair dependent.2
The course and progressive nature of MS imply that there will be changes in gait abnormalities over time. As MS can affect diverse neurological systems, neurodegeneration might result in loss of muscle force, spasticity, impaired coordination, and sensory impairment. Impairment in a single neurological system or in combination will contribute to a patient's gait deviations and may result in a specific pattern of gait.8 A better understanding of the gait changes associated with disease progression is necessary to implement more targeted interventions. Categorizing gait patterns could contribute to the development of such targeted treatments and may facilitate clinical communication between physical therapists and other health care providers.9,10
In cerebral palsy and stroke, gait analysis has been used to plan and evaluate the treatment of gait disorders.9–15 According to Toro et al,9 a gait “type” can be defined when a group of patients displays gait characteristics that are clinically identical in terms of etiology, effect on gait, and management of treatment. Patients with cerebral palsy are assigned to 1 of 5 gait patterns, mainly based on heel contact and knee position in mid-stance.16 Patients with stroke also have been assigned to gait patterns based on kinematic variables during walking.17 Classifications in spastic hemiplegia and diplegia had direct relevance in understanding the gait pattern and the treatment opportunities.10 It is plausible, therefore, that the classification of gait patterns in patients with MS would have considerable clinical utility.
A small number of studies have previously described gait patterns in patients with MS, with a focus on spatiotemporal parameters and differences compared with healthy controls.1,4,18 Patients with MS walk with a lower gait speed, shorter stride length, and more prolonged double support phase than healthy controls.6,19 However, these global spatiotemporal parameters do not reflect potential changes at the hip, knee, and ankle joints or describe the quality of gait.6,19 Moreover, spatiotemporal gait characteristics do not immediately link to treatment opportunities. A more complete gait assessment, therefore, will be achieved by including the kinematics of the lower extremities.6 Several studies have measured both spatiotemporal parameters and kinematics,2,6,20,21 and they all showed significantly smaller ranges of motion in the lower limb joints in patients with MS compared with healthy controls. Kelleher et al2 divided patients with MS into good ambulatory and reduced ambulatory groups, but no significant differences in either spatiotemporal parameters or kinematics were found. A limitation, in regard to the interpretation of gait analysis, is that small differences in joint angles do not reveal the quality of walking.
In clinical practice, 2-dimensional (2D) gait analysis, including gait features and qualitative assessment of kinematics, is easier to use than 3-dimensional (3D) gait analysis. Also, the interpretation of 2D gait analysis results is more straightforward compared with advanced 3D gait analysis.22 Herbert et al23 suggested including the observation of additional gait features, such as push-off and clearance, combining several biomechanical variables during gait. Furthermore, a more complete gait analysis could be achieved with additional muscle coordination parameters derived from electromyography (EMG) and ground reaction forces (GRFs).
However, the literature also shows that gait assessments, including 2D gait analysis, are not implemented very well in physical therapy practices. The most important barriers to the implementation of these measurements are inexperience, administration time, and costs.24–26 The development of clinically meaningful gait patterns in MS may help to overcome some of these barriers and thus facilitate implementation and use of 2D gait analysis.
The purpose of this study was to explore whether distinct gait patterns in patients with MS could be identified based on 2D gait analysis, with the combined assessment of kinematics, gait features, and muscle activity during walking. Furthermore, the clinical interpretability and relevance of distinct gait patterns were established.
Method
Participants and Design
Patients with definite MS according to the Poser criteria27 who participated in a long-term cohort study on outcome measurement and functional prognosis were invited to participate in this cross-sectional study on gait patterns.28 From 1997 to 2001, 156 patients (age 16–59 years) who visited the participating neurology outpatient clinics and were diagnosed with MS less than 6 months previously were recruited and prospectively followed from definite diagnosis onward. For the follow-up measurements at 10 years after diagnosis, 137 patients could be approached. For this gait analysis study, patients were included if they were able to walk 10 m without assistance from another person.
Measurements
Gait analysis.
Participants walked barefoot over a 10-m walkway at a comfortable walking speed and, when possible, without a walking aid. Two high-speed video cameras recorded in the sagittal and frontal planes (sample frequency=50 Hz; Basler Pilot piA640-210gc GigE, Basler AG, Ahrensburg, Germany), while synchronized EMG recorded signals from 5 muscles of each leg. Four 10-m trials were recorded. In the first 2 trials, the participants walked back and forth on the walkway while the sagittal camera followed them. As many strides as possible were recorded for the analysis. The other 2 trials were used to collect the forceplate data for the left and right legs. The participants still used the total walkway, but the sagittal camera was focused on the forceplates. Gait speed was monitored for each trial. Preparation of the participant and the gait analysis took 20 minutes. The digital calculations and storage of data took an additional 10 minutes per participant.
EMG.
Electromyographic activity was recorded (sample frequency=1,000 Hz, bandwidth=20–1,500 Hz; Aurion, Milan, Italy). After standard skin preparation, electrodes (Ag-AgCl, lead-off area=1 cm2, interelectrode distance=2.5 cm; Tyco Healthcare Nederland BV, Zaltbommel, the Netherlands) were placed longitudinally over the muscle bellies of the following muscles: rectus femoris, semitendinosus, tibialis anterior, medial and lateral gastrocnemius, and soleus.29 The EMG signals were processed offline (low-pass second-order Butterworth filter, unidirectional at 2 Hz) to obtain smoothed rectified EMG (SR-EMG) envelopes.
GRF.
The vertical and fore-aft components of the GRF were recorded with a built-in forceplate at a sample frequency of 1,000 Hz (AMTI, OR6-5-1000, Watertown, Massachusetts), synchronously with the video analysis. From the signals, the magnitude, direction, and point of application of the GRF were calculated. Trials were repeated until a successful forceplate strike for each foot was recorded.
Descriptive variables.
The Expanded Disability Status Scale (EDSS) was used to indicate the severity of MS.30 Scores on the EDSS range from 0 to 10, with higher scores indicating more severe MS.
Gait Analysis Score Chart
A gait analysis score chart was developed by a gait analyst (C.D., gait laboratory manager with 15 years of research experience) and a clinical expert (V.dG., rehabilitation physician with 10 years of clinical experience) and based on the prerequisites of normal gait.11,13 The selected items had to be able to be scored from the 2D video analysis with forceplate and EMG registrations. Most kinematic items correspond to items on the reliable and valid Edinburgh Gait Scale.31 The score chart consisted of 3 parts (ie, kinematics, EMG, and GRF) and included 73 items (Tab. 1). A single, unblinded rater (J.K., physical therapist and movement scientist with 4 years of experience in clinical gait analysis) performed the scoring, using video and data analysis in the MoXie Viewer (VU Medisch Centrum, Amsterdam, the Netherlands).32 The scoring took a maximum of 30 minutes per participant.
Kinematic, EMG, and Forceplate Variables Included in the Latent Class Analysesa
Kinematics.
Kinematics were evaluated according to the list of items shown in Table 1. Each item was scored on a 2- or 3-point scale as normal or by its deviation from normal.11,13 For most items, the criteria to distinguish between score categories could be clearly formulated. Some items (ie, fast knee/hip flexion in initial swing, active push-off and movement of arms) inevitably required a more qualitative judgment.
EMG.
For each gait phase, timing of muscle activity was scored as normal or aberrant compared with the muscle activity patterns of healthy adult individuals (Tab. 1).11,13 This reference EMG value was available in the software. The scoring was done by visual inspection of the SR-EMG signals.
GRF.
The peak of the fore-aft component of the GRF at the end of terminal stance was scored as the amount of push-off force (in newtons). The typical M-wave, the vertical component of the GRF, was scored as present (normal) or absent (abnormal) (Tab. 1).
Data Analysis
The demographics of the study population were analyzed with descriptive statistics using the Statistical Package for the Social Sciences (version 18.0 for Windows, SPSS Inc, Chicago, Illinois). Mplus (version 6.1, Muthén & Muthén, Los Angeles, California) was used for latent class analysis (LCA).9,15,33 The latent class model consists of 54 dichotomous, 17 categorical, and 2 continuous variables (Tab. 1). The latent classes are defined by the criterion of conditional independence. This criterion means that within each latent class, each variable is statistically independent of every other variable. Maximum likelihood estimation was used to fit the latent class model to the data set and was repeated with 1,500 different starting values to achieve a global maximum.34
In order to determine the number of latent gait classes, we started by assessing one class and added an extra class stepwise until no further improvement in fit occurred. The model fit is estimated by fit indexes, that is, Akaike information criterion (AIC) and Bayesian information criterion (BIC). Smaller values represent a more optimal balance of model fit and parsimony. Both the AIC and BIC correct for the number of parameters in relation to the maximum possible number of parameters.34 Besides these 2 criteria, the model interpretability is an important factor in evaluating the classes.
With the Bayes theorem, patients are assigned to one of the latent classes by calculating the posterior probability of patient membership in each class (person orientation), which tends to be the largest when the homogeneity within a latent class and the latent class separation are both strong. The class membership of patients gives an impression whether the formed classes are clinically useful. A gait variable was defined as distinctive between the classes when the probability of the same score (normal or aberrant) within a class was 60% (homogeneity) and when the other classes had a higher probability of the contrary score (heterogeneity).
To determine the clinical relevance of the gait classes, the severity of MS (measured with the EDSS) and gait speed were used as indicative variables. A one-way analysis of variance was performed to determine whether the means were significantly different among the gait classes. The Bonferroni post hoc test was performed in case of a significant difference. The significance level was set at .05.
Role of the Funding Source
This study was supported by a grant of the Dutch MS Research Foundation (grant: 05-570 MS) and Fonds NutsOhra (grant: SNO-T-0601-68).
Results
Participants
For the measurement at 10 years after diagnosis, 137 patients with MS from the original inception cohort of recently diagnosed patients (within the previous 6 months) were approached, of whom 109 agreed to participate. A total of 81 participants performed the 2D gait analysis. Fifteen patients were not able to complete the 10-Meter Walk Test without help and were excluded from this part of the study, and 13 patients decided not to travel to the gait laboratory at the outpatient clinic.
The 81 participants (51 women, 30 men) had a mean age of 47.1 years (range=28–69), a median EDSS score of 3.0 (range=1.0–7.0) (Fig. 1), and a mean gait speed of 1.15 m/s (range=0.26–1.66).
Histogram of Expanded Disability Status Scale (EDSS) scores of the 81 participants.
During the gait analysis, 2 participants used a single cane, and 4 participants needed a walker. Forceplate data were missing for the 4 participants who required a walker. Electromyographic data were missing for 1 participant.
Gait Patterns
The results of the iterative process to find the best fitting latent class model are shown in Table 2. Based on the model fit parameters, in combination with the model interpretability, it was decided that 3 was the optimal number of gait classes.
Model Fit Indexes for Latent Class Analysesa
With a small set of gait variables (N=9), we were able to define the gait patterns of patients with MS. These 9 gait variables (of the total of 73 variables) satisfied the conditions of within-class homogeneity and between-class heterogeneity: heel-rise in terminal stance, push-off, clearance in initial swing, plantar-flexion position in mid-swing, pelvic rotation, arm-trunk movement, activity of the gastrocnemius muscle in pre-swing, M-wave, and propulsive force. Table 3 shows the probability of an aberrant score for these variables for each class. Figure 2 shows, for each class, video stills of a patient who satisfied the criteria. Terminal stance, initial swing, and mid-swing are presented, as these are the gait phases where, according to the set of gait variables, the classes are most distinctive from each other (Tab. 3). Patients in class 1 show an almost normal gait pattern. Heel-rise in terminal stance, clearance in initial swing, and ankle angle in mid-swing have a probability of 1.000. Patients in class 2 also show heel-rise, clearance, and a sufficient ankle angle, but with less convincing results compared with class 1 (probabilities of 1.000, .750, and .583, respectively). The bottom row in Figure 2 shows characteristics of a class 3 patient, with no heel-rise in terminal stance (probability of .714), no clearance in initial swing (probability of .786), and impaired plantar flexion of the ankle in mid-swing (probability of .857).
Probability of the 9 Gait Variables Defining Gait Patterns in Patients With Multiple Sclerosis
Typical examples of the 3 gait classes. The video stills show for each class (rows) a patient from the position of the left leg. Each column corresponds to a variable of the core set (ie, heel-rise in terminal stance, clearance in initial swing, and ankle angle in mid-swing). IC=initial contact.
From the EMG variables (Tab. 1), only the nonactivity of the gastrocnemius muscle in pre-swing fulfilled the conditions of within-class homogeneity and between-class heterogeneity. In all classes, the participants had a high probability for preactivity of the soleus and gastrocnemius muscles in mid-stance (data not shown).
The forceplate data showed the same course as the kinematic data. The probability of a normal M-wave decreased with each successive class. The same applies for the propulsive force, which significantly declined from class 1 to class 3.
Clinical Interpretation
In Table 4, the median EDSS and mean gait speed are shown to confirm the clinical relevance of the distinctive gait patterns. The median EDSS increased, whereas gait speed decreased, in the classes with impaired gait patterns. The Bonferroni post hoc test showed a significant difference among the 3 classes for both the EDSS and gait speed (P<.01).
Participant Characteristics for Each Gait Classa
Discussion
The aim of this study was to identify distinct gait patterns in patients with MS in order to facilitate targeted treatment for patients with declined walking ability. The results showed that 3 gait patterns could be defined based on a small set of only 9 gait characteristics. The patterns identify patients with MS as good walkers (class 1), minimally impaired walkers (class 2), and moderately impaired walkers (class 3).
The 3 consecutive gait patterns seem to be induced by increasing insufficiency of the ankle push-off that may result from increasing disability. The mean propulsive force component shows a reduced push-off. Three other variables in the small set of the 9 gait characteristics are linked to a reduced push-off: absent heel-rise in terminal stance, no clearance in initial swing, and a lack of properly timed gastrocnemius muscle activity in pre-swing. In clinical practice, reduced push-off is commonly associated with kinematic changes of the knee and hip joints (ie, decreased knee extension in terminal swing and increased hip flexion in pre-swing).35 A small increase in hip flexion is sufficient to achieve greater power in the hip joint, and this is a common strategy to compensate for reduced push-off ability.2,21 In the present study, such small differences in joint angles were not measured, as a 3D gait analysis would have been required. Normally, the push-off generates the energy required to move the limbs forward.36,37 The observed insufficient knee extension in terminal swing could be a result of the reduced acceleration of the leg in swing.38 Although knee extension in terminal swing did not fully meet the criteria for inclusion in the core set, the LCA showed an increased probability of an aberrant score of knee flexion score in terminal swing (class 1=16%, class 2=58%, class 3=57%).
The deterioration in kinematic gait variables were not all related to deviations in EMG variables. For example, the persistent plantar flexion of the ankle during swing was inconsistent with the normal pattern of the tibialis anterior muscle, which could be due to weakness of this muscle. However, in the present study, EMG signals could not estimate individual muscle forces. The persistent plantar flexion also could be caused by spasticity of the gastrocnemius muscle during swing. However, the proportion of patients with EMG activity of the gastrocnemius and soleus muscles in swing, which were variables on the score form, was almost zero for all 3 classes (data not shown). Furthermore, despite early activity of the gastrocnemius and soleus muscles in mid-stance, a late heel-rise was found in the walkers with minimal and moderate impairment. Benedetti et al21 found similar discrepancies between EMG abnormalities and the expected alterations of gait kinematics. They concluded that this early activation of the calf muscle might represent a primary gait disturbance, rather than a compensatory gait disturbance, because early activation of the calf muscles is observed with higher speeds during normal walking.39
The latent LCA was performed, starting with a total of 73 gait variables (39 kinematics, 30 EMG, and 4 forceplate). A sensitivity analysis, with fewer variables in the LCA, confirmed that the set of 9 gait characteristics was sufficient to achieve the same distinctive gait patterns (results not shown). These 9 gait variables can be easily observed with 2D video analysis with forceplate and EMG registrations and enables a clinically relevant classification of patients with MS, which seems directly related to the increasing disability.
The results of our study show that the gait patterns are associated with increasing disability. Because increasing disability, to a large extent, is based on disease progression, we may hypothesize that future longitudinal studies will show the gait patterns are based on a progressive deterioration of gait variables. Kelleher et al2 also expected to find degenerative gait patterns as a consequence of disease progression, but they could not confirm this finding. These authors clustered patients with MS a priori as good or impaired walkers based on the Hauser Ambulation Index, but no significant differences in kinematics were found between good and impaired walkers. Furthermore, a functional score such as the Hauser Ambulation Index is apparently too nonspecific and not directly related to the deviations in kinematics during walking.
An advantage of the scoring method used in our study is that gait features (eg, push-off, clearance), which are direct indicators for targeted treatment (eg, orthoses, physical therapy) also were scored. The outcome of our scoring method corresponds with the gait speeds in the observed classifications (ie, a higher walking class is related to lower gait speed). This finding agrees with findings often seen in patients with reduced walking ability.40 Furthermore, the differences in EDSS scores among the classes also confirmed the deterioration in walking ability with this scoring method.
With regard to implications for clinical practice, our study showed that the gait patterns of patients with MS can be classified by scoring only a few easily observable gait variables. This classification seems useful for monitoring and evaluating patients with MS, as it reflects the degenerative course of the disease. Furthermore, as reduced push-off ability seems to be one of the main causes of reduced walking ability, it will be a challenge to investigate whether treatment options aimed at improving ankle push-off, such as spring-like ankle-foot orthoses41 and functional electrical stimulation of the calf muscles, are effective in patients with MS.
Limitations
Although 2D gait analysis is beneficial in clinical practice, the reliability of this method is often discussed.19,42–44 As we could confirm the clinical relevance of the identified gait patterns, we believe that the benefits of 2D gait analysis in this study exceed the possible disadvantages. As described in the Method section, we tried to formulate the score categories as unambiguously as possible to maximize the reproducibility and reliability of this scoring method. In our study, one experienced physical therapist filled in the score form with gait variables. Nevertheless, a priority of future research should be to test the intrarater and interrater reliability and external validity of the score chart in other MS study populations.
Although 81 gait analyses to characterize gait patterns in patients with MS give a robust indication, several study limitations should be considered. Most of the participants (68%) were classified in class 1, showing minimal-to-moderate walking problems according to the EDSS (median=2.5, range=1.0–5.0). Therefore, the number of patients in classes 2 and 3, both impaired walkers, was relatively small and prevented the identification of further classes. Furthermore, most participants had impairments of several neurological systems (eg, pyramidal, sensory, cerebellar), which made it impossible to conduct detailed subgroup analyses at the level of neurological systems. As the clinical gait analyses were conducted in a safe gait laboratory environment, barefoot, and without a walking aid when possible, translation of the results to daily activities will require further monitoring and evaluation.
In conclusion, this study showed that with a small set of 9 2D gait variables, 3 clinically relevant gait patterns in patients with MS were identified. From the 9 gait variables, it can be concluded that a decline in ankle push-off seems to be the common factor to induce limited walking ability in patients with MS.
Footnotes
Dr Kempen, Dr Doorenbosch, Professor de Groot, and Dr Beckerman provided concept/idea/research design and project management. Dr Kempen, Professor de Groot, and Dr Beckerman provided writing, participants, and facilities/equipment. Dr Kempen and Dr Doorenbosch provided data collection. Dr Kempen, Dr Doorenbosch, Dr Knol, and Professor de Groot provided data analysis. Professor de Groot and Dr Beckerman provided fund procurement. Dr Kempen, Professor de Groot, and Dr Beckerman provided participants and facilities/equipment. Professor de Groot provided institutional liaisons. Dr Knol and Dr Beckerman provided consultation (including review of manuscript before submission). Professor de Groot and Dr Doorenbosch are the clinical expert and gait analysis expert, respectively. The authors thank Kim van Hutten, Tanneke Vogelaar, and Jannet Meester for their assistance during the clinical gait analysis.
The Medical Ethics Committee of VU University Medical Center granted approval for this study.
This study was supported by a grant of the Dutch MS Research Foundation (grant: 05-570 MS) and Fonds Nuts Ohra (grant: SNO-T-0601-68).
- Received September 23, 2015.
- Accepted May 1, 2016.
- © 2016 American Physical Therapy Association
References
Issue highlights
