Individuals With Multiple Sclerosis Exhibit More Regular Center of Mass Accelerations After Physical Therapy

Abstract

Objective

The purpose of this investigation was to explore if a physical therapy program involving strength, flexibility, balance, and walking can improve the uncharacteristic gait variability and overall mobility of persons living with multiple sclerosis (pwMS).

Design

Pre-post design to evaluate the mobility improvements after undergoing 6 weeks of a gait and balance physical therapy intervention.

Setting

The initial 2 weeks were conducted at a medical center under close supervision of a physical therapist. The remaining 4 weeks were performed by the patient at their home and monitored via teleconferences.

Participants

Fifteen pwMS with relapsing-remitting (N=11) or secondary progressive multiple sclerosis (N=4) were enrolled in this study (7 women; mean age: 54.8±9 years; Kurtzke Expanded Disability Status Score range: 3.0-6.5). A group of healthy age-matched controls (N=15) were used for comparisons.

Interventions

The 6-week physical therapy intervention included exercises that targeted strength, flexibility, balance, and walking. The initial 2 weeks of the intervention were performed on-site with the remaining 4 weeks home-based. The therapy was performed twice-a-day for 5 consecutive days each week. Each session was 45 minutes in length.

Main Outcome Measures

Preferred walking speed, spatiotemporal gait kinematics, and a 6-minute walk test were completed before and after therapy. The standard deviation (SD) and sample entropy were used to evaluate the amount of variability and the regularity of the time-dependent variations in the center of mass (COM) accelerations during the 6-minute walk test.

Results

Before the intervention, the SD of the COM was reduced, and the time-dependent variations were less regular in the pwMS than the control group. After therapy, the SD was 12% larger, and the time-dependent variations were more 7% regular in the pwMS. The effect size for these changes were large (0.91 and 0.94, respectively), suggesting these changes were meaningful. The changes in the regularity of the COM were related to the mobility improvements in the preferred walking velocity and 6-minute walk test.

Conclusions

The results suggest that pwMS have altered COM variability during gait, which can be improved with a similar physical therapy program. These changes appear to be linked with the extent of the mobility improvements.

Multiple sclerosis is a progressive, demyelinating disease that is typically diagnosed in adults who are between the ages of 20 and 40 years old.¹ This demyelination affects both the sensory and motor pathways, which can disrupt an individual's balance and overall mobility. Sustaining mobility has been reported as being the most important functional ability for persons living with multiple sclerosis (pwMS), with a vast majority relying on 1 or more assistive devices within their lifetime.² Additionally, about 52% of pwMS report falling within the past 6 months.³ pwMS often adopt compensatory strategies during walking to address their instabilities; these strategies include slower walking velocities, shorter step lengths, reduced cadence, and broader step widths.4, 5, 6, 7

The prevailing consensus is that deviations from neurologic health are associated with alterations in the amount of gait variability.8, 9, 10 Over the past decade, many studies have used the standard deviation (SD) or coefficient of variation to identify that pwMS deviate from the optimal amount of healthy variability. For example, several investigations have noted that pwMS tend to have a larger SD for the spatiotemporal kinematics, trunk kinematics, and lower extremity joint kinematics compared with healthy age-matched individuals.11, 12, 13, 14, 15, 16 This heightened amount of variability has emerged as a possible biomarker for identifying the deteriorations in the balance and gait in pwMS.^17,18 More recent trends have employed nonlinear analysis techniques to quantify the time-dependent changes in the variations that occur in the gait kinematics from 1 footfall to the next. These analysis techniques have been argued to provide more meaningful clinical insight because they account for the changes that occur over time, while the more traditional statistical methods for assessing overall variability (eg, SD) do not capture such information.⁸ However, very few investigations have used such techniques to characterize the gait patterns of pwMS. Outcomes from a study conducted by Kaipust et al¹⁹ have shown that the time-dependent variations within the gait of pwMS are altered when compared with healthy age-matched controls. In addition, outcomes from other studies suggest that the extent of these deviations might be linked to the severity of the multiple sclerosis symptoms.^20,21

Outcomes from our prior clinical trial indicated that a physical therapy program involving strength, flexibility, balance, and walking (twice-a-day, 5 days-a-week, and over a 6-week duration) can result in clinically relevant mobility improvements in pwMS.²² Although these results are promising, it is currently unknown if the mobility changes are coupled with improvements to the variability seen in gait patterns of pwMS. The primary purpose of this investigation was to evaluate if the variability of the center of mass (COM) accelerations during walking improves after pwMS undergo a high-dose physical therapy protocol. We hypothesized that compared with a healthy cohort of adult controls, the amount of variability and regularity of the time-dependent fluctuations in the COM accelerations would be uncharacteristic for pwMS. However, after physical therapy, the variability would trend toward what is seen in the controls. Secondarily, we hypothesized that the walking velocity, spatiotemporal kinematics, and walking endurance would be improved after the pwMS undergo physical therapy and that these improvements would be related to the improvements in the variability of the COM accelerations.

Methods

Fifteen pwMS (7 women; mean age: 54.8±9 years; mean disease duration: 11.9±5 years) that had either relapsing-remitting (N=11) or secondary progressive multiple sclerosis (N=4) were enrolled in this study. This sample size was selected based on our previous investigation that identified that 12 pwMS would provide greater than 80% power in the selected outcome measures; therefore, accounting for a 20% dropout rate, we enrolled at least 14 persons.²² The participants were recruited based on the following inclusion criteria: between 30 and 70 years old, a Kurtzke Expanded Disability Status Score of 3.0-6.5, a definitive diagnosis by a neurologist, able to walk on a treadmill at a minimum speed of 0.5 miles per hour while holding onto handrails, cognitively competent, and a Mini-Mental State Examination score of >21. The exclusion criteria were as follows: documented multiple sclerosis-related relapse in the previous 6 months, major multiple sclerosis-specific medication changes in the previous 3 months, and the presence of another major co-morbidity such as neurologic disorders or uncontrolled pain, hypertension, and diabetes. Fifteen healthy, age-matched adults (10 women; mean age: 53.5±7 years) were also enrolled and acted as a control comparison group for the gait variables. The control group participants were free from any known orthopedic or neurologic impairments. All experimental procedures were reviewed and approved by the Institutional Review Board (IRB #0467-14-EP) and this study was registered on clinicaltrials.gov (NCT02524483). All participants provided written informed consent before participating in the experimental procedures. The enrolled pwMS completed all outcome measures before and after the therapeutic intervention and the controls completed all outcome measures once.

The total intervention period was 6 weeks. The therapy was performed twice-a-day for 5 consecutive days each week. This therapeutic dose and protocol were based on the successful outcomes of our prior investigation with pwMS.²² The initial 2 weeks were conducted at the medical center campus under the close supervision of a physical therapist (H.R., K.V.). The remaining 4 weeks were performed by the patient at their home and were monitored weekly via teleconferences with the therapists. Participants completed the same activities at home as they did during the initial 2 weeks and kept a home exercise program logbook to track their activity.

Each therapy session consisted of 15 minutes of strength and flexibility exercises, 15 minutes of postural balance exercises, and 15 minutes of treadmill walking. The activities selected for the therapeutic program were similar to those that would be performed in a group exercise program. Participants were instructed to complete each activity at their own pace for 3 minutes. Strength exercises included forward/backward lunges, stepping up/down a step, and squats. Flexibility training was completed both standing up and laying down on a mat. Participants were shown how to stretch the lower extremity muscles; especially any muscle that was specifically problematic to a participant. Both static and dynamic balance exercises were completed in each session. Static balance exercises included standing on a piece of foam with eyes open and feet wide apart or standing on 1 leg as long as possible with support. Dynamic balance exercises included stepping over small obstacles, walking sideways, or walking heel to toe. While walking on the treadmill, the participants were encouraged to remove 1 or both hands from the handrails if possible. The participants were allowed to increase and decrease their speed as needed to accomplish the total time. All participants reported their rate of perceived exertion based on the Borg scale and were instructed to attempt to work at a score of 12-13, which suggests that the exercise was somewhat hard. Rest was given as needed throughout the entire training session. For additional information about the therapeutic program, see an example home program in the supplemental material.

All participants completed the standardized 6-minute walking endurance test in a hallway ∼40 meters long.²³ The pwMS were allowed to use their regular assistive devices during the test. Accelerations of the COM were measured during the 6-minute walk test using a tri-axial accelerometer^a that was positioned over the L2 vertebra and attached to a neoprene belt which was tightly wrapped around the participant's lower trunk. This experimental setup was selected because a prior investigation has shown that the measurement of the accelerations at L2 accurately captures the acceleration of the COM.²⁴ The accelerometer signal was streamed wirelessly to a computer through a custom LabView program^b (1000 Hz).

The unfiltered resultant accelerometer signal was analyzed using a custom Matlab program.^c The resultant accelerometer signal was evaluated in this study to account for any differences in the placement of the accelerometer at the different time points. The first and last 30 seconds of the 6-minute walk test were removed from the analysis to ensure the signal only included steady-state walking. SD was used to quantify the amount of variability present in the acceleration signal. A higher SD suggests the signal has a greater amount of overall variability. Sample Entropy (SampEn) is a mathematical algorithm that measures the repeatability within a time series and was used to quantify the regularity of the time-dependent changes in the resultant accelerometer signal time series.²⁵ We selected to use SampEn to evaluate the COM acceleration data because it has been shown to be more consistent than other non-linear techniques and is not influenced by the length of data or walking speed.^25,26 A more regular time-dependent change in the time series (eg, sinewave) will have a SampEn closer to zero, while a completely irregular time series will have a SampEn that approaches infinity (eg, white noise). SampEn is the negative natural logarithm of the conditional probability that 2 sequences that are similar at a certain point (m) remains similar at m+1 and can be expressed as

$$SampEn=-ln\left(\frac{A}{B}\right)$$ Equation 1

where A is the number of self-similar vectors at length m and B is the number of self-similar vectors at length m+1. The input parameters for the algorithm include the length of data that will be compared (m), the similarity criterion (r), the length of data (N), and the time lag (tau). For this study m was defined as 2 and r was defined as 20% of the SD of the signal; therefore, for our analysis, we compared 2 points (m=2) and selected the similarity criterion to be 20% of the SD of the signal (r=0.2). In principle, selecting an r that is too small will limit the number of data points that are similar, while selecting an r that is too large can result in too many self-similar matches. Multiple methods have been proposed for selecting a similarity criterion.27, 28, 29, 30, 31 Similar to prior investigations, the similarity criterion was determined through a parameter search until consistent and stable SampEN values were achieved.^27,31 Lastly, we selected our time lag to be 102 (tau=102) after using the Average Mutual Information (AMI) algorithm.³²

All participants also completed 2 self-paced walking trials where the spatiotemporal kinematics of gait were measured with a digital mat.^d These 2 walking trials were averaged together for the final statistical analyses. The variables of interest were gait velocity (meters/second), step width (meters), step length (meters), and cadence (steps/minute).

All statistical analyses were performed using IBM SPSS Statistics 23 statistical software.^e The pre-post changes in the respective outcome measures were evaluated using paired t tests, while independent t test were used for comparisons with the control group. Given the number of t tests conducted, the .05 alpha level was adjusted based on the false discovery rate (FDR) algorithm to .033 for the respective statistical comparisons.³³ Furthermore, we calculated the Cohen's d values to gauge the clinical relevance of the pre to post therapeutic change.³⁴ The calculated Cohen's d values were adjusted to account for the correlation between the pre and post measurements due to repeated measures. A Cohen's d value less than .3 represented a weak effect size, .5 was a medium effect size, and greater than .8 was large effect size. Lastly, Pearson correlation coefficients were calculated between the percent change in the variability measures and the percent changes in the gait measures. The data in the text are reported as the mean ± standard error of the mean.

Results

Thirteen participants completed the entire 6 weeks of the therapeutic intervention. One individual discontinued because of a non-multiple sclerosis-related health condition and 1 individual discontinued because of a fall-related injury that occurred during the intervention period. No other adverse multiple sclerosis-related events occurred during the intervention period for the remaining participants who completed the program. All participants who completed the program are included in all analyses. See table 1 for demographic information about the participants.

Table 1.

Baseline demographic and clinical characteristics of the participants

Subject	Sex	Age (Years)	MS Type	MS Duration	Assistive Device
1	Man	69	RR	24	Cane/AFO
2	Man	48	SP	12	Cane
3	Woman	57	RR	12	Three foot cane
4	Woman	57	RR	12	None
5	Woman	36	SP	14	Walker
6	Man	55	SP	4	AFO
7	Man	59	RR	5	AFO
8	Man	65	SP	7	Walker
9	Woman	50	RR	15	Cane
10	Woman	50	RR	12	None
11	Woman	60	RR	15	Cane
12	Man	56	RR	19	Bioness & Cane
13	Man	64	RR	10	Cane
14	Man	40	RR	8	Cane/AFO
15	Woman	56	RR	9	None
Average	7 Women	54.8±9	11 RR	11.9±5

Abbreviations: AFO, ankle foot orthosis; MS, multiple sclerosis; RR, relapsing-remitting multiple sclerosis; SP, secondary progressive multiple sclerosis.

Table 2 displays the outcomes of the gait assessments performed in this investigation and the effect sizes. Before undergoing the physical therapy protocol, the pwMS had a lower SD than the control group indicating that there was less variability in the COM accelerations for the pwMS (P=.0004; fig 1A). However, the SD was increased 12% after the therapeutic program (P=.008; fig 1A) indicating that there was a greater amount of variability in the COM accelerations after the intervention. The effect size for the SD was .91 indicating that the change was large. Nevertheless, the pwMS still had a lower post-intervention SD than the control group (P=.03; fig 1A).

Table 2.

Respective gait assessments for the persons with multiple sclerosis (pwMS) and the control comparison group

Variable	pwMS Pre	pwMS Post	r	d	Control Group
6-Minute walk (m)	298.8±23*	318±24*	0.94	0.65	494.7±1.5
Standard deviation	0.75±0.07*	0.84±0.07*,†	0.92	0.91	1.09±0.08
Sample entropy	1.64±0.05*	1.50±0.06*,†	0.71	0.94	1.31±0.04
Velocity (m/s)	0.83±0.06*	0.90±0.06*	0.84	0.63	1.17±0.02
Step length (m)	0.51±0.02*	0.54±0.02†	0.87	0.71	0.58±0.01
Step width (m)	0.14±0.01	0.14±0.02	0.93	0.20	0.17±0.01
Cadence (steps/min)	96.9±4*	99.6±4*	0.90	0.39	117.6±3

NOTE. Gait assessments were performed at the pre and post physical therapy time points for the pwMS. The displayed Pearson correlations (r) and Cohen's d effect sizes were calculated between the pre and post physical therapy data points for the pwMS. The r values were used to adjust the Cohen's d values to account for repeated measures. A Cohen's d value of 0.3 represents a weak effect size, 0.5 was a medium effect size, and greater than 0.8 was large effect size. The presented data are reported as the mean ± standard error of the mean.

^⁎Significantly different from the control group at a 0.03 alpha level.

^†Significantly different from the pre therapy assessment time point at a 0.03 alpha level.

Fig 1.

Standard deviation and sample entropy results of COM acceleration. (A) The standard deviation of the resultant accelerations during the 6-minute walk test of the pwMS before and after the physical therapy intervention and the standard deviation of the control group. This graph shows that the pwMS significantly increased the amount of variability in their trunk accelerations after the physical therapy but still had a reduced amount of variability when compared with the control group. (B) The sample entropy of the resultant accelerations during the 6-minute walk test before and after the physical therapy by the pwMS and the control group. This graph displays that after the completion of the physical therapy, the pwMS had more regular accelerations but still displayed less regular accelerations than the control group. * indicates P<.05.

Before the therapeutic program, the pwMS had a higher SampEn than the control group indicating that the pwMS had less regular time-dependent changes in their COM accelerations (P<.0001; fig 1B). There was an ∼8% decrease in SampEn for the pwMS after the therapeutic program (P=.008; fig 1B) suggesting that the time-dependent changes became more regular after the intervention. The effect size for the change in the SamEn was 0.94 suggesting that the post-therapy change in the regularity of the COM fluctuation was large. However, the pwMS still had a higher post-intervention SampEn than the control group (P=.014; fig 1B).

Before undergoing physical therapy, the pwMS walked a shorter distance for the 6-minute walk test (P<.0001), had a slower preferred walking velocity (P<.0001), had a shorter step length (P=.005), and a slower cadence (P=.0004) when compared with the controls. The pwMS were able to walk 7% further for the 6-minute walk test (P=.042; d= .65), had a 10% improvement in their preferred walking speed (P=.049; d= .63), and a 6% improvement in their step length (P=.027; d= .71) after undergoing the physical therapy. The effect sizes of the respective gait changes were medium to large. However, the changes seen in the 6-minute walk and preferred walking speed were not considered to be significant based on the FDR adjusted 0.03 alpha level. No changes were also seen in step width (P=.52; d=.21) or cadence (P=.20; d=.39).

There was a strong negative correlation between the percent change in SampEn and the percent change in the distance walked during the 6-minute walk test (r=-0.64; P=.02; fig 2A). This suggests that the participants who had a greater improvement in the distance walked after the therapeutic intervention also tended to exhibit a greater improvement in the regularity of the COM accelerations. There also was a negative correlation between the percent change in SampEn and the percent change in the walking velocity (r=-0.56; P=.05; fig 2B) indicating that the participants who exhibited a greater improvement in walking speed after the therapeutic intervention also tended to have a greater improvement in the regularity of the COM accelerations. There were no correlations between the percent change in the SD and the improvements in any of the other gait measures (Ps>.05).

Fig 2.

Sample entropy, 6-minute walk test, and velocity correlations. (A) The correlation between the percent change in sample entropy of the resultant accelerations during the 6-minute walk test and the percent change in distance walked during the 6-minute walk test after therapy for the pwMS. (B) The correlation between the percent change in sample entropy of the resultant accelerations during the 6-minute walk test and the percent change in walking velocity after the physical therapy for the pwMS.

Discussion

The primary purpose of this study was to determine whether a physical therapy program involving strength, flexibility, balance, and walking has the potential to alter the variability of the COM accelerations of pwMS during gait. The outcomes of our investigation display that individuals with MS who have Kurtzke Expanded Disability Status Score scores between 3.0 and 6.5 have uncharacteristic variability in their COM accelerations and that this variability can potentially be improved after undergoing physical therapy. We suggest that the lower amount of variability and less regular pattern in the COM accelerations at baseline may reflect a disruption in the proper integration of the visual and somatosensory information, which are vital for maintaining the dynamic walking balance.35, 36, 37 Our results also showed that changes in the regularity of the COM accelerations were related to the degree of the improvements in walking endurance, and preferred walking speed. Altogether, these results suggest that physical therapy has the potential to enhance the control of the COM during gait in pwMS.

Prior investigations have demonstrated that the time-dependent gait variations are altered in pwMS and that these variations may depend on the severity of the symptoms.19, 20, 21 Our SampEn results were similar to those found by Shema-Shiratzky et al²¹ who evaluated changes in SampEn of a wearable sensor signal during the 6-minute walk test in individuals with mild to moderate MS. However, this previous investigation identified that the SampEn was lower in individuals who had higher levels of disability. Similarly, the results from Kaipust et al¹⁹ suggested that individuals with MS have more regular time-dependent changes in their spatiotemporal gait kinematics than healthy adults. We suspect that these conflicting results could be due to the differences in the methods used to evaluate gait variability in each study. The devices used by Shema-Shiratzky et al²¹ were placed on the lower back and ankle. Foot accelerations have been seen to display less complexity in individuals with MS,²¹ thus the less complexity seen in the data from Shema-Shiratzky et al's²¹ study could be related to the location of the sensors. Kaipust et al¹⁹ evaluated gait variability by investigating step width and stride length, while our investigation quantified gait variability by evaluating COM accelerations. Potentially, the amount and temporal structure of variability is different for the various components of gait. It is possible that the more regular time-dependent variability in the spatiotemporal kinematics of pwMS that were seen in Kaipust et al's¹⁹ investigation is a compensatory technique for the less regular variations in the COM accelerations so that individuals may maintain a steady gait. Additionally, the difference may be related to the use of the treadmill in Kaipust et al's investigation because it has been shown that the time-dependent variations seen in the gait kinematics are suppressed while walking on a treadmill.^38,39 Nonetheless, these prior studies and the results shown here demonstrate that the time-dependent gait variations are affected by multiple sclerosis.

The group of pwMS walked 7% further for the 6-minute walk test and had a 10% improvement in their preferred walking speed. However, the noted changes for the group did not exceed the minimal detectable change values previously reported for pwMS, suggesting that the mobility gains were not remarkable.40, 41, 42 Our evaluation of the individual data shows that some participants had notable improvements in the respective gait variables (ie, greater than 20%), while others had minimal change or no change at all. Clearly, there were responders and non-responders in the cohort of pwMS that underwent the physical therapy protocol employed in this investigation. Nevertheless, the outcomes from this investigation are novel in that they suggest that the degree of improvements seen in walking speed and endurance were linked with the extent of the improvements seen in controlling the COM variability during gait (ie, SampEn). This suggests that improvements in the step-to-step control of the COM might be important for the overall mobility improvements of pwMS after physical therapy. Based on our results, we contend that changes in the gait variations might provide a viable biomarker of the success of a physical therapy protocol. Potentially, pwMS that do not respond as well to the current therapeutic trends might not display changes in the variability of the COM during gait, while those that do respond display robust changes.

Limitations

This study did not have follow-up after the completion of the study and subsequent discontinuation of the physical therapy. Therefore, the duration of the resulting beneficial changes in the control of the COM during gait cannot be determined from the current investigation. Future investigations should follow participants after the intervention period has ended to evaluate the long-term benefits of the current therapeutic intervention. We also did not have a control group that consisted of pwMS who completed a more traditional therapeutic program. The extent of the improvements seen in the preferred walking velocity and 6-minute walk for the group of pwMS were also less than minimal detectable change values previously reported for pwMS, suggesting that the gains were minimal.40, 41, 42 Nevertheless, we did find strong correlations between the extent of the mobility improvements and improvements in the control of the COM variability during gait, suggesting that those that responded well to the physical therapy tended to have meaningful improvements in their gait performance also tended to improve their ability to control the COM during gait.

Additionally, this study was limited by the sample size and is partly the reason that the amount of change seen in the 6-minute walk and preferred walking velocity did not meet the FDR corrected alpha level. Based on the effect sizes seen for the 6-minute walk and preferred gait velocity in this investigation, 14 participants would have resulted in 80% power to find a similar outcome at the .03 alpha level using a 1-sided t test. Hence, the few dropouts we experienced likely influenced the statistical power that was necessary to detect clinically significant differences in the respective gait outcome measures at the corrected alpha level. This information should be taken into consideration when designing the next generation of clinical trials for pwMS.

Conclusions

In conclusion, the outcomes from our study indicate that after 6 weeks of a physical therapy program involving strength, flexibility, balance, and walking might result in pwMS having more normal amounts of COM variability and more regular time-dependent variations in the COM accelerations during walking. This extent of the improvements in the variability of the COM appears to be linked with the overall mobility improvements. Altogether, these results suggest that pwMS that respond well to physical therapy might improve the ability to control the COM for improved gait. Furthermore, therapeutic approaches that focus on learning to better control the COM during gait and balance exercises might result in beneficial gait improvements in pwMS.

Suppliers

• a.Trigno, Delsys Inc.
• b.DAQ, National Instruments Inc.
• c.Matlab, The Mathworks, Inc.
• d.GAITRite, CIR Systems Inc.
• e.SPSS, IBM Corp.