Dynamic Balance in the Gait Cycle Prior to a 90° Turn in Individuals with Parkinson’s Disease

Abstract

Parkinson’s disease (PD), a prevalent neurodegenerative condition, is associated with fall-related injuries. Falls often occur during mobility tasks such as turning while walking. There is a paucity of research on the biomechanical etiology of falls, specifically, the control of dynamic balance during turns. The purpose of this study was to analyze dynamic stability, as measured by the margin of stability (MOS), during the gait cycle preceding a 90-degree turn during walking in persons with PD. Thirteen individuals with mild to moderate idiopathic PD and 10 healthy matched controls (CON) participated. Instrumented gait analysis was conducted during walking while performing 90-degree turns using the Plug-in Gait model and Vicon Nexus motion capture software. MOS variables at first double support, midstance, and second double support of the gait cycle preceding the turn were examined. The MOS variables and spatiotemporal gait parameters were compared between PD and CON using a multilevel mixed model ANOVA; post hoc analyses were conducted using two-sample t-tests. There were no differences in spatiotemporal gait parameters between groups. The PD group demonstrated significantly greater medio-lateral (M/L) MOS compared to CON for most variables. The changes seen in the M/L MOS in the PD group may reflect compensatory changes to increase dynamic stability during the gait cycle preceding a turn.

1. Introduction

Parkinson’s disease (PD) is the second most common neurological condition in the world, affecting more than 10 million people [1]. This disorder results from the degeneration of dopaminergic neurons within the substantial nigra causing changes in nigrostriatal pathways in the basal ganglia [2]. The basal ganglia are involved with the control of voluntary movement, postural tone, the automaticity of postural control strategies, and efficient gait function. Parkinson’s disease results in progressive gait and balance decline and difficulty adapting walking to varied task and environmental demands such as managing turns and obstacles [3, 4, 5]. Due to the effects of PD on the gait and postural control systems, all dynamic balance and walking activities require increased attention, which increases the cognitive load of locomotion [3]. Individuals with PD demonstrate increased gait variability and difficulty performing a second task while walking, which is also reflective of impaired balance and gait automaticity [3, 5]. Falls among the elderly population lead to the hospitalization of more than 800,00 Americans each year [6], and persons with PD are at even greater fall risk [7]. Falls during turns are very common in persons with PD and are eight times more likely to result in a hip fracture than other mechanisms of injury [8, 9, 10]. These injurious falls can force individuals with PD to become homebound, which reduces independence in their activities of daily living and participation in other meaningful activities.

Changing direction or turning while walking requires more proactive balance strategies and increased inter-limb coordination as compared to continuous forward walking [11]. Neural systems related to turning may be more vulnerable to functional impairments in individuals with PD [12]. The natural aging process results in gait changes including decreased gait velocity, stride length, single limb support time, and altered step width [13, 14]. In addition to these age-related changes, persons with PD also demonstrate PD-specific gait impairments including gait hypokinesia, shuffling gait with increased cadence and decreased stride length, reduced gait speed, and freezing of gait [5]. During walking and turning, persons with PD have impaired movement strategies [12], compromised head and neck stabilization, and abnormal muscle activation patterns [15]. Axial coordination is impaired during turns while walking as persons with PD display “en-bloc turning”, observed as a simultaneous rotation of their head and upper trunk instead of a cranial to caudal sequential rotation that is observed in healthy individuals [12]. Previous research provides evidence that persons with PD approach turns more slowly, take longer to turn, take more steps, and perform turns that are wider and less accurate than healthy age-matched adults [8, 12, 16]. During a turn, a person’s center of mass (COM) may be outside their base of support (BOS). Compared to healthy individuals during a walking turn, those with PD display a narrower BOS, which may lead to decreased dynamic stability and increased fall risk [8, 9]. Previous studies have examined the quality and sequencing of turning in persons with PD [8, 12, 17, 18], but research examining the biomechanical variables of dynamic balance during walking with turns is limited. Anticipatory postural strategies support dynamic stability during turns [12, 14]. At the time of this paper, no studies have analyzed the dynamic balance variables of the gait cycle just prior to a 90-degree turn. Given the motor planning impairments seen in persons with PD, these findings emphasize the necessity of continued research into the specific challenges related to the preparatory gait cycle directly before a turn.

The inverted pendulum model has been used to examine static human balance (Figure 1) [19]. It models the human body as a COM, hinged at the ankle joint, moving around a static BOS. In this model, the location of the center of pressure (COP) can be defined as the reaction to changes in the movements of the COM. The model has since been adapted to allow the representation of dynamic movement, like walking and turning, by utilizing additional variables associated with dynamic balance [19]. Among these variables is the COM-COP inclination angle, i.e. the spatial relationship between the COM and COP, which has been used to analyze gait stability during sharp turns [21]. The extrapolated center of mass (XCOM) is a measure that accounts for the horizontal velocity of the COM during walking, and the margin of stability (MOS) is the minimum distance from XCOM to the boundaries of the BOS [21]. During gait, the COM naturally falls outside of the BOS. Dynamic measures of balance, such as the MOS, have been validated and are used often as part of gait analysis to describe both static and dynamic stability during ambulation [22]. The larger the MOS, the better the dynamic balance is within an individual. Maintaining a constant MOS may also reflect better dynamic control [23]. The medio-lateral (M/L) MOS increases as step width and single limb support time increase and is further dictated by the relationship between both limbs during the gait cycle [23, 24, 25, 26, 27].

Figure 1.

Schematic of the inverted pendulum model [19]. Represented by a single mass on top of a stick, the body’s center (CoM) is located at a distance l from the ankle joint center. The center of pressure (CoP, u) identifies the location of the ground reaction force, which is placed relative to the vertical projection of the center of mass, x. the boundaries of the base of support (BOS), U_MIN and U_MAX, demonstrate the potential range of the CoP. The sagittal plane of the laboratory coordinate system, y – z, indicates that the line of walking progression is along the y-axis. Note this figure originated in [20].

Turning requires interlimb coordination, modification of movement patterns, coordination of posture and gait, and an increased level of executive function, all of which are impaired in persons with PD [8]. Falls during turns present a significant threat to the health and wellness of individuals with PD, adversely affecting an individual’s independence and quality of life. Research regarding the control of dynamic balance during turns [20] may provide insight into where interventions may be targeted during therapeutic rehabilitation. Individuals with PD have impaired dynamic stability when completing turns while walking and may employ anticipatory adjustments in the gait and standing posture to maintain stability. Analysis of MOS variables in the gait cycle preceding a turn may provide information regarding biomechanical mechanisms underlying dynamic balance control in persons with PD. The purpose of this study was to analyze dynamic stability, as measured by the margin of stability, of the gait cycle just preceding a 90-degree turn during walking in persons with PD as compared to age-matched healthy adults. The results of this project may advance research into the biomechanical analysis of turning strategies and balance control in individuals with PD. This research may provide clinicians insight regarding balance strategies in persons with PD underlying their increased fall risk, as well as provide insight into possible fall prevention interventions.

2. Methods and materials

2.1 Recruitment and participants

This exploratory study was approved by the University Institutional Review Board (IRB reference#: 16–183-H). This project was part of a larger study that assessed a variety of tasks during walking that challenged dynamic balance in persons with PD. Since we collected data in a single session description of the data collection and reduction methods will be similar to previously published work [20].

Participants were recruited for this study from a local metropolitan area. Recruitment methods included flyers, presentations at PD support and exercise groups, speaking with medical professionals, and contacting participants from prior research projects. This study was advertised at local retirement communities, PD support groups, and through the local Parkinson’s Association. An age- and gender-matched control group was formed by a sample of convenience including acquaintances, relatives, and friends of researchers.

Thirteen individuals with mild to moderate idiopathic PD and 10 controls (CON) were recruited for this study (Figure 2). Individuals at a high risk of falling and those with an indication of moderate cognitive impairment were excluded. The Berg Balance Scale was utilized to determine the risk of falling [28, 29]. A score of <36 and/or > 2 falls per month was considered a high fall risk. A score of <21 on the Montreal Cognitive Assessment was indicative of cognitive impairment among persons with PD [30]. Refer to the previous work by Alderink et al. [20] for a complete list of inclusion and exclusion criteria for this study. The decade of life, as stated in the criteria, was defined as 50–59, 60–69, etc.

Figure 2.

Participant recruitment and participation flow diagram.

Initial screening was conducted via phone interview. Upon meeting inclusion criteria, individuals were screened in person at the Biomechanics and Motor Performance Laboratory. Testing included completion of the Freezing of Gait Questionnaire (FOG-Q), Berg Balance Scale, and Montreal Cognitive Assessment. All participants were non-freezers for the FOG-Q, defined by responses to item 3, “Do you feel that your feet get glued to the floor while walking, making a turn or when trying to initiate walking (freezing)?”, which has been found to be a good screen for FOG frequency [31]. Additionally, the functional walking criteria were screened to ensure participants were able to complete the 300 m walk test unassisted. All participants provided written informed consent prior to inclusion in the study. Once consent was obtained, participants provided medical information including the history of their PD diagnosis, fall history, medical history, demographic information, and current medications. Participant demographic information is provided in Table 1.

	CON (n = 10)	PD (n = 11)*
Age (yrs.)	65.1 ± 7.4	67.5 ± 10.2
Gender (M:F)	9:1	10:1
Height (m)	1.8 ± 0.1	1.8 ± 0.1
Mass (kg)	86.4 ± 12.6	90.5 ± 17.5
BBS	55.3 ± 1.6	53.4 ± 3.3
MoCA	27.2 ± 2.4	25.9 ± 2.8
FOG-Q	Not Tested	2.7 ± 3.6

Table 1.

Demographic means ± standard deviation for control (CON) and Parkinson’s (PD) groups.

*Thirteen participated in this project but data from two participants did not meet the analysis criteria.

BBS = Berg Balance Scale where the highest score of 56 indicates no balance deficits. MoCA = Montreal Cognitive Assessment where the highest score of 30 indicates no cognitive deficits. FOG-Q = Freezing Gait Questionnaire where the highest score of 24 indicates the most severe freezing of gait. Only the PD group was tested for FOG-Q.

2.2 Procedures

Walking and turning test procedures were completed during the “on” period of the participants’ PD medication schedule as this reflects participants’ usual functioning in a medicated state. Participants were asked to wear their normal walking shoes and clothing, including tight-fitting shorts, and sports bras for women, to enhance the exposure of anatomical landmarks and placement of motion capture markers. To prepare for optical motion capture trials, the following anthropomorphic measurements were taken: ankle, knee, elbow, and wrist width, hand thickness, leg length, and inter-ASIS (anterior-superior iliac spine) width; a shoulder offset was set at 3.5 cm for all participants. Lower extremity range of motion and strength assessment were completed using goniometric and myotome grading measures, respectively. Following the physical examination, 40 spherical markers (14 mm) were placed on landmarks on the body according to a modified version of the full-body Plug-in Gait (PIG) model (Appendix A). Modifications to the PIG model included additional markers on the heads of the 5th metatarsals and placement of the thigh markers on the mid-lateral thigh and tibial markers on the mid-lateral tibiae. Markers were placed by a single examiner and checked between walking sessions for any displacement. All examiners were trained by the lab director, but only a single examiner placed anatomical markers to eliminate inter-examiner variability [20].

Testing procedures included the following walking conditions 1) self-selected pace (which we defined, in this study, as Walking), 2) walking with 90-degree turns at a normal pace, 3) termination of gait and 4) walking with an obstacle crossing task. Dynamic balance variables were analyzed for the 90-degree turn and obstacle-crossing tasks in prior studies. This study focused its analysis of dynamic balance on gait cycles (referred to as Pre-turn walking) preceding the 90-degree turn.

Each participant completed a static standing calibration trial prior to beginning the walking trials to ensure correct marker positioning, create a subject-specific skeletal model, and identify where markers are in relation to joint centers or other bony landmarks. One walking trial was also completed and processed to examine joint kinematics prior to further data collection. Joint kinematics were graphed against a normative database to check for errors in the frontal and transverse knee and hip transverse plane kinematics that are indicative of marker placement error. If these kinematic data were abnormal, marker placement was revised and the static calibration and walking trial were repeated [20].

Participants practiced self-paced walking trials and were instructed to cleanly strike the force plates without targeting (i.e., looking at) them. All walking trials were completed during one session per participant. A gait belt was available for safety if needed per the examiner’s discretion. Participants walked down a 10-meter walkway with three force plates aligned straight down the walkway. After sufficient practice, a minimum of five left and five right gait cycles with clean force plate strikes were collected [20].

For the 90-degree turn condition, a modified procedure based on Yang et al. was used [32]. Participants were allowed to look at the ground before the turn and target the force plate for this condition since we believed this simulated actual turning in normal ambulation of people with PD. To mark where the turn should occur, a cone was placed at the edge of the second force plate. Participants received standardized instructions for the trial. “For this task, we will be examining how you complete a 90-degree turn. Please walk at a comfortable speed along the walkway and turn to your (left/right) between the two pieces of tape placed next to the walkway. You will complete five trials of this task on each leg. I will demonstrate the task for you.” For our study, we analyzed the kinematics at the force plate foot strike and gait of the step preceding the turn (Figure 3).

Figure 3.

Laboratory coordinate system, force plate set-up, and turning procedure (see [20]). For 90-degree turns to the right, participants walked in the negative y-direction and pushed off with their left foot in the negative x-direction. For a 90-degree turns to the left participants walked in the positive y-direction and pushed off with their right foot in the negative x-direction. The orange triangles represent the cones set up at the corner of the force plate.

2.3 Instrumentation

Vicon Nexus v2.6.1 motion capture software was used to synchronize 16 T40S/MXF40/MX40 (120 Hz) cameras (Vicon Motion Systems LTD, Oxford, UK) and three AMTI (Advanced Mechanical Technology Inc., Watertown, MA) floor-embedded force plates (1200 Hz) and collect marker trajectories and ground reaction forces, respectively.

2.4 Data processing and reduction

Marker and ground reaction force data were post-processed and trials were trimmed to a single gait cycle in Vicon Nexus 2.6.1 (refer to [20] for details). Six representative gait cycles (three right and three left) were selected and exported to Visual3D (C-Motion, Inc., Germantown, MD) for the dynamic stability analysis. The MOS variables considered in this study were COM - COP inclination angle (Figure 4), COP - COM (Figure 4), COP - XCOM (Figure 5), and U_MAX - XCOM (Figure 5). The COM – COP inclination angle was defined relative to the laboratory coordinate system (Figure 3) as the spatial angle between the vertical projection of the COM and a line connecting the COM to the COP. The other variables were determined relative to the virtual foot coordinate system (Figure 6) in both the A/P and M/L directions.

Figure 4.

Center of mass (COM) – Center of pressure (COP) inclination angle ($\theta )$ is the spatial angle between the vertical projection on the ground of the COM and a line connecting the COM to the most lateral position of the COP. A) Posterior view of the COM – COP inclination angle and the linear medio-lateral (M/L) distance between the vertical projection on the ground of the COM and location of the COP, i.e., COP – COM M/L, at midstance on the right B) lateral view of the COM – COP inclination angle and the linear antero-posterior (a/P) distance between the projection on the ground of the COM and location of the COP, i.e., COP – COM a/P, at first double-support, and C) lateral view of the COM – COP inclination angle and COP – COM a/P at second double-support (see [20]).

Figure 5.

Schematic illustrating the determination of the linear distance between the extrapolated center of mass (XCOM) and the center of pressure (COP), or U_MAX. A) the linear distance between the COP and XCOM in medio-lateral (M/L) and antero-posterior (a/P) directions, i.e., COP – XCOM M/L and COP - XCOM a/P, respectively, B) the linear distance between U_MAX (i.e., head of the 5th metatarsal) and XCOM in the medio-lateral direction, i.e., U_MAX – XCOM M/L, and C) the linear distance between U_MAX (i.e., between the heads of the 1st and 2nd metatarsals) and XCOM in the antero-posterior direction, i.e., U_MAX – XCOM a/P (see [20]).

Figure 6.

A virtual foot coordinate system (CS) is defined by markers (light blue) over the medial and lateral malleoli and the toe (between the first and second metatarsal heads). The markers were projected onto the floor (dark blue) of the lab to align the CS parallel to the floor for estimation of the position of the center of pressure relative to the foot. The origin of the virtual foot CS is located at the ankle joint center (midway between the medial and lateral malleoli) projected to the floor. For the right foot, the right-handed CS is positive-x (red, medio-lateral), positive-y (green, antero-posterior), and positive-z (blue, superior-inferior). FP1 = force plate one; FP2 = force plate two; FP3 = force plate three.

The extrapolated center of mass ($\mathit{\text{XCOM}}$) was found using:

$$\mathit{\text{XCOM}}=\mathit{COM}+\left(V_{\mathit{COM}}/{\omega }_0\right)$$E1

where $\mathit{COM}$ and $V_{\mathit{COM}}$ are the instantaneous position and velocity of the total body $\mathit{COM}$ and ${\omega }_0$ is the angular eigenfrequency of the pendulum,

$${\omega }_0=\sqrt{g/l}$$E2

where $g$ = 9.81 m/s² and $l$ is the leg length (the distance from the $\mathit{COM}$ to the ankle joint center in meters) [21].

Dynamic stability was calculated as the distance between the $\mathit{\text{XCOM}}$ and the limits of the BOS:

$$\mathit{MOS}=U_{\mathit{MAX}}-\mathit{\text{XCOM}}$$E3

where $U_{\mathit{MAX}}$ in the anterior direction is defined by the toe marker and $U_{\mathit{MAX}}$ in the lateral direction is defined by the 5th metatarsal marker.

The MOS variables were analyzed at three points in the gait cycle: first double-support (FDS) at 6% of stance, midstance (MS) at 50% of stance, and second double-support (SDS) at 94% of stance [20]. Spatiotemporal parameters (ST) of velocity, stride width, stride length, step length, cadence, stance time, and swing time were also calculated. All study variables were exported to spreadsheets and R (version 4.2.2 (2022-2110-31)) [33] and RStudio (Version 2023.03.0 + 386) [34] statistical software for descriptive and statistical analyses.

2.5 Statistical analysis

Preliminary statistical analysis demonstrated no significant differences in dynamic balance measures between the right and left limbs, therefore, we pooled the right and left gait cycle data. Statistical analyses for primary and secondary variables were performed using nlme [35] and emmeans [36] packages in R/RStudio. MOS variables were analyzed with a three-level mixed model ANOVA. The fixed factors were: phase (FDS, MS, and SDS), condition (CON and PD), and their interaction. Phase was modeled as a repeated measure within gait cycle and subject. Spatiotemporal variables were analyzed with a two-level mixed model ANOVA. The fixed factors were: task (Pre-Turn and Walking), condition (CON and PD), and their interaction. Both models used subject as a random intercept and the gait cycle as the observational unit. Primary and secondary variables were also analyzed with post hoc analyses using two-sample t-tests (α = 0.05). The effect size was reported as the estimated difference between the means (Appendix B).

3. Results

3.1 Spatiotemporal gait parameters

Since we suspected that participants with PD may have altered their gait when anticipating a turn, we analyzed ST parameters between conditions (i.e., between PD and CON groups), as well as between tasks: Pre-turn, i.e., defined as the gait cycle immediately preceding the turn, and Walking, i.e., defined as a gait cycle during self-selected normal walking. When comparing Pre-turn to Walking within the CON group, all variables were significantly different (Table 2). Likewise, within the PD group, all ST parameters were significantly different between the two tasks (p < 0.05) (Table 3). For the Pre-turn task, there were no significant differences between CON and PD for any of the ST parameters (Table 4). All ST parameters were significant for task, demonstrating that these parameters were different for Pre-turn and Walking (Table 5); however, there were no differences for condition. Double limb support, swing time, cadence, and velocity were significant for task and condition, representing a difference between PD and CON within Pre-turn and Walking (Table 5).

	Pre-Turn	Walking	P-Value*	Effect size†	95% CI
Double limb support (s)	0.355 (0.01526)	0.299 (0.01526)	<0.001	0.0564	0.04640, 0.0665
Cycle time (s)	1.209 (0.02943)	1.133 (0.02958)	<0.001	0.0758	0.06232, 0.0893
Stance time (s)	0.774 (0.02154)	0.723 (0.02169)	<0.001	0.0513	0.04009, 0.0626
Swing time (s)	0.430 (0.00811)	0.417 (0.00828)	<0.001	0.0130	0.00612, 0.0199
Step length (m)	0.662 (0.02089)	0.711 (0.02108)	<0.001	−0.0490**	−0.06143, −0.0366
Stride length (m)	1.337 (0.04536)	1.434 (0.04553)	<0.001	−0.0965**	−0.11391, −0.0791
Cadence (steps/min)	100.680 (2.51985)	105.605 (2.54318)	<0.001	−4.9250**	−6.46580, −3.3841
Velocity (m/s)	1.122 (0.03215)	1.256 (0.03215)	<0.001	−0.1335**	−0.15141, −0.1156

Table 2.

Mean (standard deviation) for spatiotemporal (ST) gait parameters for the control group with task (1) pre-turn* and (2) walking*.

*P-values are for a t-test of the difference in means between Pre-Turn and Walking for the control group. Likewise, we are reporting the 95% confidence interval (CI) for the difference between the means.

**Negative effect size indicates the Pre-turn value was less than the Walking value.

†The effect size was determined relative to the differences between the means for the Pre-Turn and Walking tasks.

	Pre-Turn	Walking	P-Value*	Effect size†	95% CI
Double limb support (s)	0.358 (0.01477)	0.326 (0.01477)	<0.001	0.0324	0.0252, 0.0396
Cycle time (s)	1.216 (0.02952)	1.133 (0.02961)	<0.001	0.0828	0.0689, 0.0968
Stance time (s)	0.780 (0.02160)	0.730 (0.02167)	<0.001	0.0498	0.0392, 0.0603
Swing time (s)	0.437 (0.00863)	0.401 (0.00871)	<0.001	0.0360	0.0286, 0.0434
Step length (m)	0.663 (0.02089)	0.707 (0.02098)	<0.001	−0.0448**	−0.0557, −0.0339
Stride length (m)	1.338 (0.04533)	1.421 (0.04540)	<0.001	−0.0834**	−0.0987, −0.0682
Cadence (steps/min)	98.878 (2.53787)	106.755 (2.55478)	<0.001	−7.8766**	−9.5368, −6.2164
Velocity (m/s)	1.100 (0.03212)	1.266 (0.03209)	<0.001	−0.1662**	−0.1824, −0.1500

Table 3.

Mean (standard deviation) spatiotemporal (ST) gait parameters for the Parkinson’s group with task (1) pre-turn and (2) walking.

*P-values are for a t-test of the difference in means between Pre-Turn and Walking tasks. Likewise, we are reporting the 95% confidence interval (CI) for the difference between the means.

**Negative effect size indicates the Pre-turn value was less than the Walking value.

†The effect size was determined relative to the differences between the means for the Pre-Turn and Walking groups.

	CON	PD	P-Value*	Effect size†	95% CI
Double limb support (s)	0.355 (0.01526)	0.358 (0.01477)	0.892	−0.002921**	−0.0480, 0.0421
Cycle time (s)	1.209 (0.02943)	1.216 (0.02952)	0.869	−0.006984**	−0.0953, 0.0814
Stance time (s)	0.774 (0.02154)	0.780 (0.02160)	0.86	−0.005486**	−0.0702, 0.0592
Swing time (s)	0.430 (0.00811)	0.437 (0.00863)	0.593	−0.006455**	−0.0316, 0.0186
Step length (m)	0.662 (0.02089)	0.663 (0.02089)	0.986	−0.000537**	−0.0632, 0.0621
Stride length (m)	1.337 (0.04536)	1.338 (0.04533)	0.992	−0.000654**	−0.1366, 0.1353
Cadence (steps/min)	100.680 (2.51985)	98.878 (2.53787)	0.621	1.801392	−5.7802, 9.3830
Velocity (m/s)	1.122 (0.03215)	1.100 (0.03212)	0.633	0.022151	−0.0742, 0.1185

Table 4.

Mean (standard deviation) spatiotemporal (ST) gait parameters for pre-turn for the control (CON) and Parkinson’s (PD) groups.

*P-values are for a t-test of the difference in means between CON and PD groups. Likewise, we report the 95% confidence interval for the difference between the means.

**Negative effect size indicates the CON value was less than the PD value.

†The effect size was determined relative to the differences between the means for the CON and PD groups.

	Task		Condition		Task: Condition
	F	P-value	F	P-value	F	P-value
Double limb support (s)	186.3	<0.001	0.491480	0.493	14.6754	<0.001
Cycle time (s)	258.9	<0.001	0.000351	0.985	0.5077	0.477
Stance time (s)	167.1	<0.001	0.018181	0.894	0.0396	0.842
Swing time (s)	86.3	<0.001	0.155876	0.698	20.0854	<0.001
Step length (m)	125.9	<0.001	0.000246	0.988	0.2498	0.618
Stride length (m)	235.0	<0.001	0.000362	0.985	1.2405	0.266
Cadence (steps/min)	120.4	<0.001	0.008546	0.927	6.5891	0.0109
Velocity (m/s)	618.6	<0.001	0.007701	0.931	7.1445	0.00816

Table 5.

Fixed effects ANOVA of condition (Parkinson’s, control), task (pre-turn, walking), and interaction for spatiotemporal (ST) gait parameters.

3.2 Measures of dynamic stability: a qualitative description of the shapes of the curves for the metrics of dynamic balance

Visual inspection of the graphical representations for the dynamic balance metrics revealed similar trends in the direction of change of the curves, but some differences in the magnitude of changes during the stance phase of gait for both the Parkinson’s and control groups (Figures 7 and 8). In general, it’s apparent that there was greater trial variation in COP – COM, COP - XCOM, and U_MAX – XCOM metrics in the medio-lateral direction for both PD and CON groups. To interpret these figures and to interpret the data presented in subsequent sections, the following guidelines will help:

• A larger magnitude of change in the distance between the center of mass and the base of support is consistent with a larger margin of stability.

• For the COP – COM A/P, COP – XCOM A/P, and U_MAX – XCOM A/P metrics a positive value indicates that the COP (and U_MAX) is anterior to the COM; conversely, a negative value indicates that the COP lies posterior to the COM. Thus, as participants moved from FDS through MS to SDS, the COP progressed from in front of (anterior) the COM to behind (posterior).

• For the COP – COM M/L, COP – XCOM M/L, and U_MAX – XCOM M/L metrics a positive value indicates that the COP (and U_MAX) is lateral to the COM, i.e., the COM lies within the base of support, whereas a negative value indicates that the COP lies medial to the COM.

• For the COP – COM inclination angle increasing positive values are consistent with a greater angle, thus a greater distance between the COP and COM. For example, as expected, the inclination angle was greater at first and second double support.

Figure 7.

Dynamic balance metrics for the right limb for all subjects during percent stance of the gait cycle for the a) CON and B) PD groups. For the a/P COP - COM, positive values indicate that the COP is anterior to the COM. In M/L COP - COM, increasing positive values indicate that the COP is more lateral to the COM. The same is true for the remaining pairs, with positive COP and positive U_MAX indicating that the toe marker is anterior to XCOM in the a/P direction, and increasing positive values indicating that the COP and U_MAX are lateral to XCOM in the M/L direction.

Figure 8.

Dynamic balance metrics for the right limb for all subjects during percent stance of the gait cycle for the a) CON and B) PD groups. For the a/P COP-COM, positive values indicate that the COP is anterior to the COM. In M/L COP - COM, increasing positive values indicate that the COP is more lateral to the COM. The same is true for the remaining pairs, with positive COP and positive U_MAX indicating that the toe marker is anterior to XCOM in the a/P direction, and increasing positive values indicating that the COP and U_MAX are lateral to XCOM in the M/L direction.

3.3 Metrics of dynamic balance

3.3.1 COP – COM antero-posterior and medio-lateral

There were no significant differences in the COP – COM metric between the PD and CON groups at FDS, MS, or SDS in the antero-posterior (A/P) direction (Table 6). However, there was an interesting result in the medio-lateral (M/L) direction at MS, with the PD (20.24 ± 9.880 mm) metric greater than the CON (−1.30 ± 10.128 mm) metric (Table 7). The negative value for the CON group suggests that the COM was outside the base of support, i.e., the COP, that is, the CON group’s COM ventured beyond the BOS yet balance was maintained. Thus, the larger value for COP - COM at MS in PD may reflect a more stable strategy to keep the COM within the BOS. There was also a significant difference in the COP - COM M/L variable at SDS between the PD (86.52 ± 9.880 mm) and CON (35.37 ± 10.128 mm) groups (p = 0.00198). The larger value for the PD group at both MS and SDS may be a compensatory change to increase dynamic stability and balance.

	PD	CON	P-Value*	Effect size	95% CI
FDS
COP - COM	258.51 (6.785)	254.86 (6.975)	0.712	−3.655	−24.098, 16.789
COP - XCOM	−93.81 (9.060)	−109.53 (9.210)	0.239	−15.726**	−42.868, 11.416
UMax - XCOM	85.48 (9.974)	58.72 (10.186)	0.0768	−26.757**	−56.708, 3.194
MS
COP - COM	19.76 (6.785)	23.17 (6.975)	0.73	3.405	−17.038, 23.849
COP - XCOM	−265.37 (9.060)	−290.64 (9.210)	0.0661	−25.272**	−52.414, 1.870
UMax - XCOM	−202.89 (9.974)	−226.29 (10.186)	0.118	−23.397**	−53.348, 6.553
SDS
COP - COM	−193.00 (6.785)	−195.30 (6.975)	0.816	−2.293**	−22.736, 18.150
COP - XCOM	−422.51 (9.060)	−424.60 (9.210)	0.873	−2.094**	−29.236, 25.048
UMax - XCOM	−423.11 (9.974)	−421.09 (10.186)	0.889	2.020	−27.931, 31.971

Table 6.

Mean (standard deviation) expressed in millimeters for Antero-posterior dynamic balance variables for control (CON) and Parkinson’s (PD) groups during the pre-turn gait cycle.

*CON and PD A/P dynamic balance variables were determined for all participants walking at their self-selected pace before they initiated a 90-degree turn during first double support (FDS), mid-stance (MS), and second double support (SDS). P-values are for a t-test of the difference in means between CON and PD groups. The effect size was determined relative to the differences between the means for the CON and PD groups. Likewise, we are reporting the 95% confidence interval for the difference between the means.

**Negative effect sizes indicate that the CON group values were less than the PD group values.

	PD	CON	P-Value*	Effect size	95% CI
FDS
COP - COM	−25.64 (9.880)	−27.51 (10.128)	0.897	−1.865**	−31.591, 27.862
COP - XCOM	14.57 (13.527)	9.67 (14.152)	0.805	−4.903**	−46.033, 36.227
UMax - XCOM	74.96 (12.781)	62.50 (13.524)	0.511	−12.464**	−51.558, 26.629
MS
COP - COM	20.24 (9.880)	−1.30 (10.128)	0.145	−21.538**	−51.265, 8.188
COP - XCOM	50.50 (13.527)	2.91 (14.152)	0.0257	−47.586**	−88.716, −6.456
UMax - XCOM	103.83 (12.781)	55.16 (13.524)	0.0175	−48.667**	−87.761, −9.574
SDS
COP - COM	86.52 (9.880)	35.37 (10.128)	0.00198	−51.144**	−80.870, −21.417
COP - XCOM	140.98 (13.527)	29.75 (14.152)	<0.001	−111.233**	−152.363, −70.103
UMax - XCOM	204.23 (12.781)	93.49 (13.524)	<0.001	−110.746**	−149.840, −71.653

Table 7.

Mean (standard deviation) expressed in millimeters for Medio-lateral dynamic balance variables for control (CON) and Parkinson’s (PD) groups during the pre-turn gait cycle.

*CON and PD M/L dynamic balance variables were determined for all participants walking at their self-selected pace before they initiated a 90-degree turn during first double support (FDS), mid-stance (MS), and second double support (SDS). P-values are for a t-test of the difference in means between CON and PD groups. The effect size was determined relative to the differences between the means for the CON and PD groups. Likewise, we are reporting the 95% confidence interval for the difference between the means.

**Negative effect sizes indicate that the CON group values were less than the PD group values.

Significant differences were found for COP – COM in both the A/P and M/L directions for the main effect of condition (p < 0.001), suggesting a consistent difference in means between the PD and CON groups across phases. However, differences between groups were only seen in the M/L direction for phase:condition interaction (p = 0.001) (Table 8). Thus, in the M/L direction, the COP - COM distances were significantly different between the CON and PD groups. Furthermore, these differences were apparent at MS and SDS between the CON and PD groups (Figure 9). The most notable difference between groups appeared to be at SDS (Figure 9). The PD group had a larger median value for COP - COM M/L at SDS compared to the CON group. Since SDS occurs just prior to the turn this may be a more unstable time point in the gait cycle for the PD group and increases in MOS may assist in maintaining balance.

		Phase**		Condition*		Phase: Condition
		F	P-value	F	P-value	F	P-value
COP - COM	A/P	0.0088	0.926	10,907	<0.001	0.712	0.492
	M/L	3.4724	0.0788	267	<0.001	18.328	<0.001
COP - XCOM	A/P	1.3708	0.257	4317	<0.001	5.520	0.0046
	M/L	8.5802	0.0090	187	<0.001	52.804	<0.001
UMax - XCOM	A/P	1.3728	0.257	11,094	<0.001	10.502	<0.001
	M/L	10.673	0.0042	223	<0.001	42.560	<0.001

Table 8.

Fixed effects ANOVA of condition (PD, CON) and phase (FDS, MS, SDS) for anterior-posterior (a/P) and medio-lateral (M/L) dynamic balance variables during the pre-turn gait cycle.

*CON and Parkinsons M/L dynamic balance variables were determined for all participants walking at their self-selected pace before they initiated a 90-degree turn.

**Phase is used to categorize first double support (FDS), mid-stance (MS), and second double support (SDS).

Figure 9.

Boxplots with overlayed data points illustrating subject by color, condition (control and Parkinsons), phase (first double support (FDS), midstance (MS), and second double support (SDS)), and interaction of the two. Median values, reported in millimeters, were significantly different in the M/L direction for COP to COM.

3.3.2 COP – XCOM antero-posterior and medio-lateral

There were no significant differences between the PD and CON groups at FDS, MS, and SDS time points for COP - XCOM in the A/P direction (Table 7) or in the M/L direction at FDS (Table 8). However, there were significant differences between PD and CON for the COP - XCOM metric in the M/L direction at MS and SDS (Table 8). At MS, the mean COP - XCOM was significantly larger for the PD (50.50 ± 13.527 mm) as compared to the CON group (2.91 ± 14.152 mm) (p = 0.0257). At SDS, the mean COP - XCOM was significantly larger for the PD (140.98 ± 13.527 mm) than the CON group (29.75 ± 14.152 mm) (p < 0.001). The smaller value for the CON group suggests that XCOM was closer to the COP at both MS and SDS. For the PD group at MS and SDS, XCOM was further from the BOS, which may suggest compensatory changes to increase MOS as a strategy to maintain balance prior to a turn.

There were significant differences due to condition (p < 0.001) for COP – XCOM in the A/P and M/L directions (Table 9). Furthermore, the COP – XCOM metric showed differences in both the A/P (p = 0.0046) and M/L (p < 0.001) directions for the interaction phase:condition. Larger values for the PD group, particularly at MS and SDS (Figure 10) suggest that XCOM was further from the BOS. The most notable difference in phase and condition interaction appeared to be at SDS (Figure 10). For the PD group, increasing MOS as the gait cycle progressed from FDS to SDS may represent motor planning changes in preparation for a turn.

	PD*	CON*	P-Value	Effect size**	95% CI
FDS	14.28 (0.387)	14.59 (0.393)	0.572	0.318	−0.841, 1.477
MS	2.25 (0.387)	2.40 (0.393)	0.783	0.154	−1.005, 1.313
SDS	13.04 (0.387)	12.02 (0.393)	0.0817	−1.017**	−2.176, 0.142

Table 9.

Mean (standard deviation) expressed in degrees for COP - COM inclination angle for FDS, MS, and SDS for control (CON) and Parkinson’s (PD) groups during the pre-turn gait cycle.

*CON and PD groups’ COP - COM inclination angles were determined for all participants walking at their self-selected pace before they initiated a 90-degree turn during first double support (FDS), mid-stance (MS), and second double support (SDS). P-values are for a t-test of the difference in means between CON and PD groups, the effect size was determined relative to the differences between the means for both groups. Likewise, we are reporting the 95% confidence interval for the difference between the means.

**Negative effect sizes indicate that the CON group values were less than the PD group values.

Figure 10.

Boxplots with overlayed data points illustrating subject by color, condition (control and Parkinson’s), phase (first double support (FDS), midstance (MS), and second double support (SDS)), and interaction of the two. Median values, reported in millimeters, were significantly different in the M/L direction for XCOM to COP.

3.3.3 U_MAX – XCOM antero-posterior and medio-lateral

As with the previous two metrics, there were no significant differences in U_Max – XCOM between the PD and CON groups in the A/P direction (Table 6). However, at FDS in the A/P direction, the positive values for both PD and CON indicate that the toe marker, or U_MAX, was anterior to the XCOM meaning that the XCOM was within the base of support (BOS). Yet the margin for PD was larger with an effect size of approximately 2.5 cm (i.e., 26.757 mm). This large effect size and nearly significant result (p = 0.0768) suggest a potential clinically meaningful difference.

The U_MAX - XCOM variable in the M/L direction showed significant differences between PD and CON at MS and SDS (Table 8). At MS the U_MAX - XCOM was significantly greater in the PD group (103.83 ± 12.781 mm)) compared to the CON group (55.16 ± 13.524 mm (p = 0.0175). Likewise, the U_Max – XCOM SDS for the PD (204.23 ± 12.781) group was greater than for CON (93.49 ± 13.524). The smaller values for the CON group indicate that XCOM was closer to the U_MAX at MS and SDS. This suggests that XCOM was further from the BOS in the PD group compared to CON.

For condition and phase:condition there were significant differences in the U_Max – XCOM metric in the A/P and M/L directions (p < 0.001). However, only M/L directional differences were noted in phase (Table 8).

3.3.4 COM – COP inclination angle

The COM - COP inclination angle was not significantly different between CON and PD groups at FDS, MS, and SDS (Table 9). Findings did not reveal a significant difference for phase, but did note differences for condition (p < 0.001) and phase:condition (p = 0.00129) (Table 10). The COM - COP inclination angle was larger at FDS and SDS due to the increased distance between the right and left limbs at these points during the stance phase of gait, and smaller at MS as the two limbs approached each other for both PD and CON.

	Phase**		Condition*		Phase: Condition
	F	P-value	F	P-value	F	P-value
COM - COP Inclination Angle	0.13061	0.722	2223	<0.001	6.847	0.00129

Table 10.

Fixed effects ANOVA of condition (Parkinson’s, control) and phase (FDS, MS, SDS) for COM - COP inclination angle during pre-turn gait cycle.

*Parkinson’s and Control COM - COP inclination angles were determined for all participants walking at their self-selected pace before they initiated a 90-degree turn.

**Phase is used to categorize first double support (FDS), mid-stance (MS), and second double support (SDS).

3.3.5 Relationship between the metrics of dynamic balance

COP - COM, COP - XCOM, and U_MAX - XCOM in A/P and M/L directions were the variables we used to measure MOS in this study. It is notable that the trajectories and slopes for these variables from one representative participant from the CON group were similar (Figure 11A). Overlaid curves from one representative participant from the PD group show similar trends (Figure 11B). All three variables, COP - COM, COP - XCOM, and U_MAX - XCOM, are methods to describe dynamic balance and MOS. The extrapolated COM (XCOM) accounts for the horizontal velocity of the COM during walking. The horizontal velocity component that is considered with XCOM and not COM might be one explanation for differences between COP - COM and the other MOS variables. For example, it appears that consideration of velocity appears to decrease the magnitude of the MOS (Figure 11). Thus, the larger difference between the COP - COM curve and U_MAX - XCOM or COP - XCOM curves could be due to the consideration of horizontal velocity, while the curves for U_MAX - XCOM and COP - XCOM were more similar. The margin of stability variables using XCOM demonstrated a larger change from FDS to SDS. The magnitude of change for each curve appears to be greater within the A/P direction than in the M/L direction. In the M/L direction, COP - COM, COP - XCOM, and U_MAX - XCOM were more similar with less differentiation with the use of XCOM compared to the A/P direction.

Figure 11.

Overlaid dynamic balance variables of a representative participant from a) control and B) Parkinson’s groups for a/P and M/L COP - COM, COP - XCOM, U_MAX - XCOM. Green line = COP - COM, black line = COP - XCOM, red line = U_MAX - XCOM. Percent (%) stance of the gait cycle accounts for FDS at 6.0%, MS at 50.0%, and SDS at 94.0%.

4. Discussion

Balance challenges, such as performing a sharp turn while walking can lead to falls and potentially debilitating injuries in persons with Parkinson’s disease. In an effort to better understand fall-related mechanics, this study investigated select biomechanical factors related to the control of dynamic stability during walking and turning. The purpose of this study was to examine the dynamic stability, as measured by the margin of stability (MOS), and other related biomechanical metrics of balance stability, during the gait cycle just preceding the 90-degree turn in persons with PD as compared to healthy age-matched controls. In general, during Pre-turn gait cycles, the PD group demonstrated significantly greater M/L MOS metrics at MS and SDS.

The inverted pendulum model has informed human balance research since its introduction to the field [19, 37]. The relationship between the center of mass and related extrapolated center of mass, and center of pressure, has been investigated under several different scenarios. The metrics derived from these relationships have been shown to be valid and reliable measures for the study of the balance of persons with neurologic disorders that affect gait [21]. Hof suggested that maintaining a constant MOS resulted in a more stable gait [24]. Buurke et al. found that interlimb coordination played a greater role in mediolateral dynamic stability, particularly in patients with balance deficits [23]. Although both intrinsic and extrinsic factors affect dynamic balance, increasing SLS time and changes in the BOS and control of the MOS are strategies used to increase stability during walking.

Since there is a complex relationship between the spatiotemporal parameters of gait and dynamic balance, it is likely that the metrics of balance during walking will change if there are changes in the ST parameters. To our knowledge, this study is the first to compare ST gait parameters of normal self-paced walking (here described as Walking) to the gait cycle preceding a turn (here described as Pre-turn). All of the ST variables examined were significantly different between Walking and Pre-turn for both the CON and PD groups. These findings may reflect that preparatory alterations in gait and balance control in anticipation of making a turn were evident both in persons with PD and healthy age-matched controls.

When comparing ST variables between CON and PD exclusively during Pre-turn, the variables were not significantly different from each other. This may indicate that in individuals with mild to moderate PD, changes in the ST parameters during turns may be more subtle or minor. Double limb support time approached significance between the PD and CON groups, as the PD group demonstrated longer double limb support times than healthy controls. A systematic review with meta-analysis of gait changes in persons with PD reported that increased double limb support time is a common result seen in studies of PD gait [38]. Corroboration of our findings and those of Zanardi et al. suggest that an increased double limb support time in persons with Parkinson’s disease may be an adaptation to improve stability by reducing single-limb support time.

When considering the effects of task: condition during Pre-turn walking (Table 6) there were significant differences between groups in double limb support and swing time, as well as cadence and velocity. These spatiotemporal differences between persons with PD and healthy controls may represent early motor changes in gait found with PD. For example, decreased swing time, and associated increases in double limb support time, may be an early symptom of an emerging hypometric gait pattern that is a hallmark sign of Parkinson’s disease. Furthermore, decreases in cycle time and cadence may be associated with an eventual decrease in gait velocity, known as bradykinesia in persons with PD [3, 39]. King et al. suggested that the neural pathways relating to bradykinesia may affect even those with very mild PD due to the highly complex nature and executive function requirements of turning while walking [16].

We anticipated that there might be differences in the A/P and M/L MOS metrics between individuals with PD and healthy controls during the gait cycle preceding a turn. To our knowledge, this is the first study to demonstrate selected differences in these metrics of dynamic balance. For the COP - COM metric in the A/P direction, there were no differences between the CON and PD groups. In contrast, there were notable differences in COP - COM in the M/L direction between the two groups. The COP-COM M/L metric at MS for the CON group had a negative value, indicating that the COM was outside the BOS, i.e., COP. This finding suggests a tendency for a less stable position. In contrast, at SDS the PD group’s COM was further from the BOS reflective of a more stable position. This finding may suggest that the healthy controls were able to maintain dynamic balance with a smaller M/L MOS compared to the PD group, or that the PD group increased their M/L MOS to adapt to changes in dynamic balance that may predispose them to a fall. The control group’s COM-COP distance was smaller, and their COM ventured beyond the COP. These findings may reflect normal and robust balance abilities, which indicates that healthy individuals can control their dynamic balance even when the COM moves beyond the BOS during this walking task demand.

For the COP - XCOM metric, there were no differences between groups in the A/P direction; however, there were significant differences between groups in the M/L direction at both MS and SDS. At these points in the stance phase, the PD group’s XCOM was further from the BOS suggesting a more stable position. The PD group may have made different adaptations to preserve their balance than healthy controls in preparation for a turn. In contrast to our findings, the study by Mellone et al. found that participants with mild to moderate PD had a narrower BOS than age-matched healthy adults, resulting in a smaller mean distance between XCOM and BOS. This would indicate a level of instability [8]. These ideas may suggest that individuals with PD could increase their BOS as a compensatory strategy to maintain balance and therefore, increase their MOS. Practitioners might use these results to inform clinical practice by demonstrating the importance of a wider BOS during turns and ambulatory tasks.

In the medio-lateral direction, the U_MAX - XCOM was significantly larger in PD compared to CON at MS and SDS. The PD group’s XCOM was further from the BOS, once again suggesting a more stable position. It has been proposed that in persons with stroke-related balance impairments, individuals are more likely to achieve balance following a perturbation by changing their COM position rather than their COM velocity (XCOM) compared to their BOS [40]. Since XCOM is velocity-dependent, it is possible that an interaction between COM and velocity could reduce the U_MAX - XCOM metric. Buurke et al. suggested that individuals with impaired dynamic balance can improve M/L MOS by improving inter-limb coordination [23]. PD-related gait changes that adversely affect inter-limb coordination are another possible explanation for differences found in MOS. Further research on this metric of dynamic balance in persons with PD-related gait impairments is warranted.

There were no differences found in the COM-COP inclination angle between the PD and CON groups; however, this metric did differ between the three time points of the stance phase within the PD group. The COM - COP inclination angle increased at FDS and SDS as the lower extremities were further away from each other. Whereas the COM - COP inclination angle decreased at MS as the limb approached single leg stance and COM approached the COP. In previous studies examining individuals with balance impairments related to vestibular deficits and hemiparesis, researchers found that those with balance impairments had a significantly reduced COM - COP inclination angle compared to controls [41]. This contrast to the results of our study may be due to the mild disease state in the PD group in our study.

There were notable differences in dynamic measures of balance during the gait cycle preceding a 90-degree turn. The largest differences in MOS between PD and CON appeared to be at SDS in the M/L direction, with the PD group demonstrating larger MOS distances than the control group. This finding was unexpected as a previous study with the same cohort that examined the MOS variables during the 90-degree turn found the most notable difference between groups during MS [20]. The midstance phase of gait has been identified as the most unstable point of the gait cycle, due to the single limb stance stability demands during this phase of gait [10]. The COM or XCOM approaching the COP at MS might be expected as the participant is in single-leg balance. Individuals in the PD group could have increased M/L MOS in a preemptive attempt to avoid loss of balance. A study by Nilsson et al. also identified that persons with PD experience turning hesitations due to fear of falling [42]. The Nilsson et al. finding, in addition to the slower gait velocity seen in the PD group in our study, may support our speculation. The larger differences found at SDS could be attributed to dynamic balance and gait motor planning demands during this gait phase immediately before the turn. These differences at SDS may be one factor contributing to increased fall risk during turns in persons with PD. Further research is warranted to investigate our premise about individuals with PD with a fall history or more advanced disease stage.

We note several limitations in this study. One limitation is that participants in the PD group were not classified into disease stages based on the Hoehn and Yahr scale or by score on the Movement Society-Unified Parkinson’s Disease Rating scale, due to the researcher’s limitations in training and formal certification in the administration and scoring of these standardized measures. However, a standardized assessment of balance and gait function was assessed with inclusion criteria that reflect a sample of individuals with mild to moderate PD. Participants in this study were community ambulators and did not have a positive fall history. Therefore, our results cannot be generalized across disease stages or to individuals with high fall risk. The small sample size may have contributed to a reduced statistical power, which framed several conservative conclusions, even among the statistically different findings that we presented. Since data were collected during ON medication times, our findings may not accurately represent dynamic balance during OFF medication times or during periods where medications are wearing off in persons with PD. Finally, Terry et al. [43] and Kazanski et al. [44] have suggested that Hof et al’s. [19] original work on the margin of stability for static postures may need to be modified for use in dynamic gait, therefore additional work is needed in this area.

5. Conclusion

This study’s findings provide some insight into the biomechanical control of dynamic stability during ambulation and turns in persons with Parkinson’s disease. The PD group demonstrated significant differences for all spatiotemporal parameters between Pre-turn and Walking conditions. Decreased step and stride length, as well as cadence and walking velocity for the PD group during Pre-turn, compared to Walking, suggest an adaptation in anticipation of a turn.

Our findings regarding the MOS variables demonstrate important differences that reflect dynamic stability changes during Pre-turn in individuals with mild-moderate PD. In the M/L direction, the PD group’s COM is further from the BOS (or COP) as compared to healthy controls. The PD group demonstrated significantly increased MOS in the M/L direction at MS and SDS for all three variables and at FDS for U_MAX - XCOM. In the A/P direction for U_MAX - XCOM the PD group demonstrated significantly increased MOS at FDS. These changes in the MOS metrics may reflect adaptation within the PD group to increase dynamic stability in anticipation of a turn.

Persons with mild to moderate Parkinson’s disease demonstrated changes in spatiotemporal parameters and their margin of stability, particularly M/L stability, during ambulation prior to a 90-degree turn. This information may provide insight into possible factors that contribute to fall risk in this population during walking and turning tasks in this population. Rehabilitation clinicians working with persons with PD may want to carefully assess changes in gait and balance control both in preparation for and during the turn while walking.

Further research in this area should include a larger sample size and persons with moderate to severe PD symptoms who have a greater fall risk or freezing of gait during walking mobility tasks.

Note: This project was part of a larger single-session data collection and some methodological details were not provided in this paper but can be found in Alderink et al. [20].

Acknowledgments

We would like to thank the participants of this study, as well as their families, for volunteering their time to contribute to this research.

Conflict of interest

The authors declare no conflict of interest.

Appendix A: Statistical analysis

Marker Labels	Definition	Position on Patient
LFHD	Left front head	Left temple
RFHD	Right front head	Right temple
LBHD	Left back head	Left back of the head (defines the transverse plane of the head, together with the frontal markers)
RBHD	Right back head	Right back of the head (defines the transverse plane of the head, together with the frontal markers)
C7	7th cervical vertebra	On the spinous process of the 7th cervical vertebra
T10	10th thoracic vertebra	On the spinous process of the 10th thoracic vertebra
CLAV	Clavicle	On the jugular notch where the clavicles meet the sternum
STRN	Sternum	On the xiphoid process of the sternum
RBAK	Right back	Anywhere over the right scapula (no equivalent marker on the left side)
LSHO	Left shoulder	On the acromio-clavicular joint
LUPA*	Left upper arm	On the upper lateral 1/3 surface of the left arm (place asymmetrically with RUPA)
LELB	Left elbow	On the lateral epicondyle
LFRM	Left forearm	On the upper lateral 1/3 surface of the left arm (place asymmetrically with RUPA)
LWRA	Left wrist marker A	At the thumb side of a bar attached to a wristband on the posterior of the left wrist, as close to the wrist joint center as possible
LWRB	Left wrist marker B	At the little finger side of a bar attached to a wristband on the posterior of the left wrist, as close to the wrist joint center as possible
LFIN	Left finger	Just proximal to the middle knuckle on the left hand
RSHO	Right shoulder	On the acromio-clavicular joint
RUPA*	Right upper arm	On the upper lateral 1/3 surface of the right arm (place asymmetrically with LUPA)
RELB	Right elbow	On the lateral epicondyle approximating the elbow joint axis
RFRM*	Right forearm	On the lower lateral 1/3 surface of the right forearm (place asymmetrically with LFRM)
RWRA	Right wrist marker A	At the thumb side of a bar attached symmetrically to a wristband on the posterior of the right wrist, as close to the wrist joint center as possible
RWRB	Right wrist marker B	At the little finger side of a bar attached symmetrically to a wristband on the posterior of the right wrist, as close to the wrist joint center as possible
RFIN	Right finger	Just proximal to the middle knuckle on the right hand
SACR*	Sacral	On the skin mid-way between the posterior superior iliac spines (PSI) and positioned to lie in the plane formed by the ASIS and PSI points
LASI	Left ASIS	Left anterior superior iliac spine
RASI	Right ASIS	Right anterior superior iliac spine
LPSI	Left PSIS	Left posterior superior iliac spine (immediately below the sacroiliac joints, at the point where the spine joins the pelvis). This marker is used with the RPSI marker as an alternative to the single SACR marker.
RPSI	Right PSIS	Right posterior superior iliac spine (immediately below the sacroiliac joints, at the point where the spine joins the pelvis). This marker is used with the LPSI marker as an alternative to the single SACR marker.
LTHI	Left thigh	Over the lower lateral 1/3 surface of the left thigh in line with the hip and knee joint centers
LKNE	Left knee	Laterally, on the flexion-extension axis of the left knee
LTIB	Left tibia	Over the lower 1/3 surface of the left shank
LANK	Left ankle	On the lateral malleolus along an imaginary line that passes through the transmalleolar axis
LHEE	Left heel	On the calcaneus at the same height above the plantar surface of the foot as the toe marker
LTOE	Left toe	Over the second metatarsal head, on the mid-foot side of the equinus break between the forefoot and mid-foot
RTHI	Right thigh	Over the upper lateral 1/3 surface of the right thigh
RKNE	Right knee	Laterally, on the flexion-extension axis of the right knee
RTIB	Right tibia	Over the upper 1/3 surface of the right shank
RANK	Right ankle	On the lateral malleolus along an imaginary line that passes through the transmalleolar axis
RHEE	Right heel	On the calcaneus at the same height above the plantar surface of the foot as the toe marker
RTOE	Right toe	Over the second metatarsal head, on the mid-foot side of the equinus break between the forefoot and mid-foot

Table A1.

Plug-in gait markers.

Note: Marker labels marked with an asterisk (*) are optional; however, using them improved marker tracking during dynamic trials. The model was modified in the present study with the addition of markers placed on the heads of the left and right fifth metatarsals. Additionally, for the subject calibration trial, markers were placed (later removed for walking trials) on the apex of the right and left medial femoral condyles and right and left medial malleoli.

Appendix B: Statistical analysis

This is an exploratory study utilizing a relatively small convenience sample of subjects. As is common for biomechanical studies, numerous gait cycles are observed for each subject with a selection of ‘good’ gait cycles chosen for analysis. Measurements are then made on each gait cycle. Thus the observational unit is a gait cycle. Gait cycles are considered independent but must be nested within subjects to control for subject differences.

The primary purpose of an exploratory study is to identify potential relationships, not prove clinical practice. As such there is a minimal cost of false positives since follow-up studies are anticipated. False negatives however could discourage follow-up work. We chose to use a standard alpha of 0.05 but note ‘interesting’ results with higher p-values, especially if the estimated effect size is clinically significant.

Prior studies have treated measurements at multiple points in a gait cycle as separate dependent variables. We chose instead to model them as repeated measures in order to get a better overall picture. This approach allowed the identification of surprising differences in the gait cycles based on condition.

The margin of stability (MOS) model chosen is a three-level mixed model Figure B1. Level one measurements (primary variables) are taken at three points in the gait cycle called phases (FDS, MS, and SDS), i.e. they are repeated measures within a gait cycle. Gait cycles are level two and random within a subject. Subjects are level three and random with condition (control = CON/Parkinson’s = PD) as an attribute. Condition, phase, and their interaction are the fixed effects of interest. Random effects (subject and gait cycle) are used to remove these effects from the model to get appropriate modeling of variance and allow for better detection of effects due to the presence of Parkinson’s.

Figure B1.

Margin of stability multilevel mixed model. FDS = first double support; MS = midstance; SDS = second double support.

The spatiotemporal model does not have multiple measurements for each gait cycle. Instead, a fixed effect of task (Pre-turn/Walking) is included in the model. The structure is two level with gait cycle at level one within subject level two, both are random. Condition (CON/PD), task (Pre-turn/Walking), and their interaction are the fixed effects of interest.

Several interactions were significant so post hoc t-tests and confidence intervals within levels of condition for MOS variables and Task and Condition for ST variables were performed. Rather than a standardized effect size like Cohen’s d, we chose to use the estimated difference since it has direct clinical meaning in this context.

This is a considerably more sophisticated statistical model than often used in these studies but is justified to make the best use of large quantities of data collected on limited subjects. Future statistical work could apply Bayesian statistics and/or consider more extensive multivariate modeling beyond the simple repeated measures used here.