2.1 Participants

A convenience sample of twenty people with Parkinson’s Disease (13 males and 7 females), ages 48–79 years (63.78 ± 8.97 years) were recruited from the Ottawa-Gatineau community. Since two individuals had severe dyskinesia and one individual had missing data, only 17 participants were included in the final analysis. Participants were assessed with the original Unified Parkinson’s Disease Rating Scale Motor Examination (11± 6) and were between I-III on the Hoehn & Yahr scale [29]. Further, seven participants reported freezing of gait based on the original Freezing of Gait Questionnaire. Participants were tested on their optimally medicated state. Exclusion criteria included any physical discomfort using a virtual reality system, any injuries and/or orthopedic surgeries that interfered with gait, walking only with a walking aid, and additional illnesses other than Parkinson’s Disease. All participants provided written informed consent and the study was approved by the Ottawa Health Science Network Research Ethics Board and the University of Ottawa Research Ethics Board. The study was conducted in accordance with the Tri-Council Policy statement; Ethical Conduct for Research Involving Humans; The International Conference on Harmonization- Good Clinical practice: Consolidated Guideline; and the provisions of the Personal Health Information Protection Act 2004.

2.2 Procedure

A familiarization period was provided for participants prior to data collection to determine their preferred walking speed on the treadmill. Experimental setup is depicted in Fig 1. Participants walked at their preferred speed on a dual-belt treadmill with two arm swing conditions (restricted and unrestricted). During the restricted arm swing trials, participants inserted their arms inside the safety harness, which effectively prevented arm motion. The unrestricted arm swing trials were participants’ normal arm swing movements. Arm swing conditions were paired with a slip that was induced on either the least or the most affected side (one slip per trial). This resulted in a total of four trials per participant with each trial lasting 2 minutes. Slips were caused by accelerating the left/right treadmill belt at 1.7m/s² for 0.75 seconds before decelerating at the same rate for 0.75 seconds to return to the participants preferred walking speeds [30]. Slips were automatically triggered, at heel-strike, after the participants’ heels crossed each other during the swing phase in the anteroposterior direction [30]. Twenty-five seconds were provided at the start of each trial for participants to reach steady-state walking before data collection began. To minimize slip anticipation, trials were randomized using a random number generator in Excel 2016 (Microsoft, Seattle, WA, USA) and slip onset within each trial was manually determined in a pseudorandom manner by the CAREN system operator. To ensure safety, and prevent falls, participants wore a safety harness attached to an overhead structure at all times. Participants were encouraged to rest whenever necessary to minimize fatigue.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Experimental setup for the CAREN system virtual environment.

https://doi.org/10.1371/journal.pone.0249303.g001

Data collection was completed using the CAREN-Extended System (Motek Medical, Amsterdam NL) in a virtual park terrain environment. This system combines a six degrees of freedom motion platform with embedded dual-belt instrumented treadmill, 12 camera Vicon Motion Capture system, 180-degree projector screen, and a safety harness. A 57-marker set was used for tracking full body kinematics [15, 31, 32]. Kinematic data were collected at 100Hz.

2.3 Data analysis

Markers were synchronized and processed in Vicon Nexus (nexus 2.6, Oxford, UK), while 3D kinematics were calculated in OpenSim with a full-body model [33]. A fourth-order, low-pass Butterworth filter with a 12Hz cutoff frequency was used to filter marker data. Data were analyzed by custom Matlab scripts (MathWorks, Natick, MA) to calculate step time, length, and width, as well as instantaneous linear and angular velocities at heel-strike for both feet. Linear velocities were calculated as absolute values. Additionally, the Margin of Stability (MOS) was calculated bilaterally as a measure of dynamic stability in the anteroposterior and mediolateral directions [34–36]. In the anteroposterior direction, this was calculated as the distance of the extrapolated COM to the right/left heel maker and in the mediolateral direction to the right/left lateral heel marker at heel-strike [15, 31, 36]. Data were categorized by least and most affected legs which were then divided into slipped and recovery limbs. The slipped limb was defined as the leg which was perturbed upon heel-strike, while the recovery limb was defined as the contralateral limb that performed the first heel-strike following the slip.

2.4 Statistical analysis

Data were analyzed using SPSS 23.0 (SPSS, IBM, Chicago IL), and p < 0.05 was considered statistically significant. An alpha = 0.05 was set a prioi for statistical significance. The normality of variables was verified using Shapiro-Wilk’s test. A two-way repeated measures ANOVA for arm (unrestricted and restricted) and leg (least and most affected leg) was performed to find main effects and interactions for the slipped limb. Additionally, an identical but separate two-way repeated measures ANOVA assessed main effects and interactions for the recovery limb. If statistical significance was achieved, then pairwise comparisons with a Sidak-Bonferroni adjustment for multiple comparisons were used for post-hoc analyses.

3.1 Arm swing

As no statistical significance (p > 0.05) in range of motion (ROM) existed between the least and most affected arms, the two arms were averaged for further analysis. The paired samples t-test revealed that the restricted arm swing condition had a reduced arm ROM than the unrestricted arm swing condition (t(16) = -6.13, p < 0.001). The ANOVA revealed that participants had a faster step time (F(1,16), p = 0.006, η_p² = 0.389) during the restricted arm swing condition compared to unrestricted arm swing during heel-strike of the slipped limb. No further arm swing main effects or interactions were found.

3.2 Least and most affected legs

Our results demonstrated that when the least affected leg was the slipped limb, it had a wider step than the most affected (F(1,16) = 10.788, p = 0.005, η_p² = 0.403). Further, when the least affected leg was the recovery limb, it had a larger anteroposterior MOS (F(1,16) = 9.654, p = 0.007, η_p² = 0.376), a longer step time (F(1,16) = 6.825, p = 0.019, η_p² = 0.299), and had a trend for a longer step length (F(1,16) = 3.955, p = 0.064, η_p² = 0.198) than the most affected. No additional main effects or interactions were found.

4.1 Arm swing

Contrary to our hypothesis, there were no differences between the unrestricted and restricted arm swing conditions neither for the MOS nor linear and angular velocities. This occurred despite differences in arm ROM between our arm swing conditions. These findings were unexpected due to the functional role arm swing has in the recovery strategy to slips [16–18]. Marigold et al. [16, 17] demonstrated that in response to a slip, healthy individuals execute a rapid arm elevation response where the arms are flexed forward and upward. The authors discussed that the arms flex specifically in these directions to move the COM further anteriorly and in the opposite direction of the trunk’s initial backward displacement caused by the slip [16]. An anterior shift in the COM’s position would theoretically be reflected as an increase in the anteroposterior MOS as it measures the distance of the COM’s dynamical state to the BOS [10, 36]. This in turn would reflect that the arm elevation strategy is effective in reducing dynamic stability loss in the backward direction [16–18]. However, when our MOS results are examined alongside our linear and angular velocities, the lack of findings indicate that our participants’ arm movements did not influence their trunk kinematics. One potential explanation for this is that our participants may have been ineffective in executing a targeted arm elevation response to the slip. Carpenter et al. [6] demonstrated that the arm responses in people with Parkinson’s Disease during platform perturbations were more variable in their trajectory and had a shorter response time compared to controls. Further, the arms returned to a position near their trunk relatively quickly in people with Parkinson’s Disease [6]. The authors further found that the recovery arm responses in people with Parkinson’s Disease were particularly ineffective and abnormal when perturbations were induced in the backward direction (the same destabilizing direction for slips) [6]. This could explain the lack of findings in our study between our arm swing conditions for the slipped and recovery limb analyses. However, further research is needed to discriminate the impact of age and Parkinson’s disease on arm strategies in response to slips. Carpenter et al. [6] found that the arm response to perturbations was ineffective despite the deltoid muscles activating earlier in people with Parkinson’s Disease than controls. Therefore, future research should examine the EMG profile of the deltoid muscles between these two groups.

Interestingly, the leg undergoing the slip had a faster step during the restricted arm swing condition compared to the unrestricted. This response could reflect destabilization from the restricted arm swing condition in our participants as it may have impeded their movement timing [15]. Indeed, a faster step reduces the amount of time the contralateral leg (recovery limb) has to determine appropriate foot placement. Previous research suggests that walking with restricted arm swing disrupts the rhythmic temporal sequence of foot placement in this demographic [15]. One of the primary symptoms of Parkinson’s Disease is impaired internal movement timing [37]. Therefore, our finding supports the notion that when arm swing becomes completely absent (due to the disease), foot placement timing is further disrupted [15]. As such, disruption to temporal foot placement arises both mechanically from the absent arm swing and neurologically from the Basal Ganglia’s neurodegeneration [15].

4.2 Least and most affected leg

As hypothesized, our analysis revealed differences between the least and most affected leg. This potentially reflects the asymmetric neurodegeneration in people with Parkinson’s Disease. Indeed, our participants demonstrated asymmetric impairment to mediolateral foot placement as they had a wider step on their least affected leg, compared to the most affected, when the least affected leg was undergoing the slip. When stability is threatened, both healthy individuals and people with Parkinson’s Disease widen their steps to maintain the COM within the lateral boundaries of the base of support [15, 38]. However, during walking, the neuromuscular system determines foot placement by predicting the COM’s future position at the upcoming heel-strike [39–41]. In our study, slip occurrence was randomized between the legs and slip onset within each trial, was established in a pseudorandom manner. This minimized the possibility for our participants to predict the timing and laterality of the perturbation. Thus, the least affected leg’s larger step width (at slip onset) reflects an already existing step asymmetry prior to the perturbation. Since our study did not examine differences between the first and subsequent slips, step width differences potentially arose as participants attempted to widen their BOS after initial slip exposure. However, the results indicate that our participants were more effective in adjusting the BOS mediolaterally with their least affected leg.

In addition, asymmetric neurodegeneration would account for the differences between the legs in attenuating backward stability loss. Indeed, the longer step time and larger anteroposterior MOS when the least affected leg was the recovery limb would reflect appropriate and asymmetric slip responses. A longer step time provides more time for stable foot placement of the contralateral leg before the COM is transferred between limbs. Additionally, the larger anteroposterior MOS reduces the risk of backward stability loss by moving the COM’s dynamical state further anteriorly and in the opposite direction to the slip [10, 20, 21]. As step time and anteroposterior COM movement are regulated passively by subcortical structures, the differences between the legs may reflect the asymmetric neurodegeneration of dopamine within the Basal Ganglia [10, 24, 25]. These interlimb differences hold important implications for clinicians. Specifically, clinicians should consider the efficacy of targeting the least affected leg to promote gait adaptation and stability in slip recovery responses. This would be in line with initial evidence from Ricciardi et al. [42], who demonstrated in a pilot study that physical therapy targeting the least affected side improved more in the UPDRS-III and Tinetti scale subscores than standard therapies targeting both legs.

4.3 Limitations

Several limitations should be considered in the context of our study. For instance, participants were tested on their optimally medicated state which affects foot placement [37]. Additionally, differences between freezers and non-freezers were not examined. As freezers demonstrate greater postural instability than non-freezers, restricted arm swing may have a distinct effect on their dynamic stability [43, 44]. Future research should also consider examining differences between people with Parkinson’s Disease and healthy aged-matched controls to parse out differences due to age and those that arise due to Parkinson’s Disease. In addition, it has been shown that after perturbation exposure, healthy adults adopt a more cautious gait i.e. with shorter, wider, and slower steps as a means to enhance their dynamic stability to prevent potential future perturbations [10, 22]. It is possible that our participants used a similar strategy when walking after initial slip exposure. Therefore, this should be taken into account when interpreting the findings of this article. Similarly, individuals with a higher Hoehn & Yahr score may respond differently to slips than those with a lower score as these individuals have further reductions in their postural control. Research on perturbation recovery responses in people with Parkinson’s Disease would also benefit from an examination of the harness load data, an aspect unexamined in our study, to assess balance loss versus recovery. Further, future studies should examine differences in the recovery response between the first slip and following slips since differences may occur in the proactive and reactive recovery responses in restoring dynamic stability. Finally, this study did not examine differences between the dominant and non-dominant leg which may elicit further differences in response to slip recovery.