Persons with Parkinson’s disease show impaired interlimb coordination during backward walking

Abstract

Introduction:

Although there is growing literature supporting the implementation of backward walking as a potential rehabilitation tool, moving backwards may precipitate falls for persons with Parkinson’s disease. We sought to better understand interlimb coordination during backward walking in comparison to forward walking in persons with Parkinson’s disease and healthy controls.

Methods:

We assessed coordination using point estimate of relative phase at each participant’s preferred walking speed.

Results:

Persons with Parkinson’s disease demonstrated impaired interlimb coordination between the more affected arm and each leg compared to controls, which worsened during backward walking.

Conclusion:

For those with Parkinson’s disease, inability to output smooth coordinated movement of the more affected shoulder may impair coordination during forward and, especially, backward walking. Our findings provide new information about backward walking that can allow clinicians to make safer, more effective therapeutic recommendations for persons with Parkinson’s disease.

Introduction

Coordination between arm and leg swing is a fundamental feature of human gait. When walking at a moderate pace, the arms swing out of phase with one another and in phase with the respective contralateral leg. This interlimb coordination (ILC) contributes to the maintenance of stable gait, as disruptions in ILC have been shown to increase the risk of falling [1]. Changes in coordination are well established in populations affected by gait impairments and who exhibit high fall risks, including persons with Parkinson’s disease (PwPD) [2–4]. PwPD exhibit more variable ILC compared to healthy age-matched peers, and show increased variability in the timing of arm swing [3] during forward walking (FW). These features of impaired coordination have been associated with clinical ratings of gait and posture impairment [5]. Understanding how PwPD coordinate their limbs across various walking tasks is critical to determining how coordination contributes to mobility and the incidence of falls in PwPD.

Safely ambulating in everyday life requires the ability to successfully walk in multiple directions, including backward, making BW performance valuable. Although risk of falling is related to postural instability regardless of direction of movement, the inability to walk backward or take compensatory backward steps in response to a perturbation is associated with an increased risk of falling [6]. While both PwPD and healthy older adults walk backward slower with shorter strides compared to young adults [6,7], PwPD are worse at performing a BW task than older controls [6]. Thus, BW is increasingly used as an evaluation tool in older adults and Parkinson’s disease populations [8,9], including as a measure of diagnosing fall risk in older adults [9]. Interventions using BW have also been explored in PwPD [10,11] with evidence showing BW training improves mobility in PwPD [11]. However, coordination of the limbs during BW in PwPD has not been explored. This information is critical for enhancing our understanding of the training benefits of BW when trying to potentially reduce fall risk.

The purpose of this study was to explore how ILC is affected during BW as compared to FW in PwPD and age-matched controls (CON). Gait speed impacts ILC [3,12], thus we aimed to target a cohort of CON with walking speeds similar to PwPD. As PwPD show impaired ILC during FW [5], we hypothesized PwPD will show more impaired ILC during BW than during FW as compared to CON. Since more challenging walking tasks heighten differences in gait performance between PwPD and older adults [13], we also hypothesized PwPD will exhibit greater impairments in spatiotemporal gait parameters compared to CON during BW than during FW.

Methods

Eighteen PwPD and 18 age- (±2 years) and gender-matched older adult controls (Table 1) participated. PwPD were recruited from the Program for Movement Disorders and Neurorestoration at the Fixel Institute for Neurological Diseases at UF Health. Older adult participants were recruited from the north Florida region. Inclusion criteria for PwPD included: 1) a clinical diagnosis of idiopathic Parkinson’s disease made by a movement disorders neurologist, 2) modified Hoehn & Yahr (HY) stage of 2.5 or better in the “on medication” state, 3) a stable regimen of antiparkinsonian and psychotropic medications for 30 days prior to participation. PwPD were excluded if they had 1) evidence of secondary or atypical parkinsonism, 2) significant cognitive impairment (MMSE score < 24) or major psychiatric disorder, 3) presence of significant motor fluctuations (unpredictable “on” and “off” fluctuations), and 4) taking cognition-altering drugs. Exclusion criteria for older adult participants included: 1) having a diagnosed neurological condition, unstable or progressive cardiac or pulmonary disease, insulin-dependent diabetes, or any other condition affecting ambulation, 2) significant cognitive impairment (MMSE score ≤ 24), 3) a self-selected gait speed > 1.15m/s [14], and 4) inability to walk comfortably without assistance. As gait speed influences ILC [3,12], we recruited CON with mobility characteristics similar to those with early-stage Parkinson’s disease, specifically targeting CON participants who walked slower than 1.15m/s [14]. This criteria ensured any differences observed in ILC are not due to physical performance, but underlying pathology. All participants read and signed an informed consent form approved by the University of Florida’s Institutional Review Board prior to testing.

Table 1.

Subject demographic information; values are presented as mean ± standard deviation

	PwPD	CON
Men	15	15
Women	3	3
Age	69 ± 6	70 ± 5
Height (cm)	171. 9± 4.9	171.2 ± 9.3
Combined Limb FW Gait Speed (m/s)	0.94 ± 0.13 m/s	1.08 ± 0.22 m/s
Hoehn & Yahr	2.0 ± 0.5	-
UPDRS	35 ± 10	-
Disease Duration (years)	4 ± 4	-

Three-dimensional motion capture was used to measure gait patterns. Reflective markers were placed on bony landmarks according to the Vicon Nexus Plug-in-Gait full body model (Nexus, version 2.3, Oxford, United Kingdom). A ten-camera motion capture system (Vicon Nexus, Oxford, United Kingdom) recorded marker positions at 120 Hz. Participants completed a battery of walking tasks as part of a larger study, where the tasks were in a fixed order. Only FW and BW were analyzed for the present study, FW was always preformed first, followed by BW. Participants walked five times across an eight-meter walkway at a “comfortable, preferred speed” for both FW and BW.

Sagittal shoulder and hip angles were calculated in Vicon Nexus as the relative angles between the upper arm/trunk segments and the thigh/pelvis. ILC was quantified using the point estimate of relative phase (PERP) between body segments at the maximum angle of each segment of the ipsilateral and contralateral hip and shoulder angles. PERP and range of motion (ROM) were calculated with custom MATLAB software in accordance to Roemmich et al [5]. PERP is an angular measurement of the phase relationship between limbs; therefore, PERP values are typically close to 180° for ipsilateral arms and legs since these limbs move out of phase, and close to 0° for contralateral arms and legs moving in phase. We included an analysis of ROM to ensure any differences in PERP did not result from differences in functional flexibility, the ROM required for performing daily activities such as walking of the joints. Contralateral interlimb coordination was defined as the coordination of contralateral joints, such as more affected hip/less affected shoulder (MA/LA) and less affected hip/more affected shoulder (LA/MA). Ipsilateral interlimb coordination was defined as the coordination of ipsilateral joints, such as the more affected hip/more affected shoulder (MA/MA) and less affected hip/less affected shoulder (LA/LA). More affected and less affected sides were identified using the motor section of the Unified Parkinson’s Disease Rating Scale (UPDRS-III). Specifically, the more affected side was determined by subtracting the average score on questions 20, 21, 23, 24, 25, and 26 for the right side of the body (upper and lower limbs) from the average score on questions 20, 21, 23, 24, 25, and 26 for the left side of the bod (upper and lower limbs) in section III of the UPDRS.

For CON, we identified dominant limb by asking participants what leg they would use to kick a ball and matched the non-dominant limb with the more affected side of those PwPD. The following spatiotemporal gait parameters were also analyzed: gait speed, step length, step width, cadence, and percentage of the gait cycle spent in double limb support [15].

Measures for a single limb were analyzed across all spatiotemporal gait variables to avoid violating assumptions that may arise from averaging the two limbs. Differences in gait speed, step length, step width, cadence, and double support time for the more affected side in the PwPD group and non-dominant side in the CON group were analyzed in a 2 (Group: PwPD vs CON) × 2 (Condition: FW vs BW) repeated measures MANOVA. For each combination of limbs (MA/MA, LA/LA, MA/LA, LA/MA), PERP and ROM were evaluated in two separate 2 × 2 (Group × Condition) repeated measures MANOVAs. We used the most stringent level of Median Absolute Deviation method to detect outlying values, as 95% of data should lie within ~2 deviations of the median if normal [16–18]. In both walking conditions, we identified and replaced outlying PERP values for all combinations of limbs and hip and shoulder ROM values on both sides with two deviations from the median [19]. For our PwPD group, 8 participants had at least one PERP or ROM value replaced, and for our CON group, 5 participants had at least one PERP or ROM value replaced (Table 2). Significance was set a priori at p<0.05, and postlarge [20].

Table 2.

Breakdown by group of the combined number of PERP values replaced with MAD for all sets of limb combinations, and the combined number of hip and shoulder ROM values replaced with MAD for both sides.

	PERP Values Replaced with MAD		ROM Values Replaced with MAD
	BW	FW	BW	FW
PwPD	9	9	2	4
CON	2	2	4	1

Results

Spatiotemporal Analysis

We found no significant Group × Condition interaction (λ=0.810, F(5,30)=1.41, p=0.250, η_p²= 0.19) for gait speed, step length, step width, cadence, or double support time. When comparing PwPD to CON (λ=.474, F(5,30)=6.65, p<0.001, η_p²= 0.526), there were no significant differences in gait speed (p=0.020, η_p²= 0.149), step length (p=0.393, η_p²= 0.022), step width (p=0.004, η_p²= 0.218), cadence (p=0.044, η_p²= 0.114), or double support time (p=0.561, η_p²= 0.01) after applying Bonferroni corrections. However, when comparing conditions (λ=.063, F(5,30)=689.909, p<0.001, η_p²= 0.937), gait speed was significantly slower (p<0.001, η_p²= 0.674), step length was significantly shorter (p<0.001, η_p²= 0.895), step width was significantly wider (p<0.001, η_p²= 0.607), and double support time was significantly higher (p<0.001, η_p²= 0.407) during BW compared to FW. Cadence was not significantly different between FW and BW (p=0.550, η_p²= 0.011) (Table 3, Supplementary Figure 2).

Table 3.

Mean (standard error) for the gait cycle parameters for the more affected/non-dominant side.

	PwPD		CON
	BW Mean (SE)	FW Mean (SE)	BW Mean (SE)	FW Mean (SE)
Gait Speed (m/s)	.637(.048)	.920(.037)	.793(.065)	1.08(.052)
Step L (m)	.366(.022)	.566(.016)	.408(.028)	.574(.019)
Step W (m)	.105(.007)	.071(.006)	.135(.008)	.095(.007)
Cadence (steps/min)	100.79 (4.35)	96.77(4.11)	107.57(3.42)	108.05(2.67)
DSP (%)	30.46(1.99)	22.67(1.25)	29.13(1.88)	26.28(.95)

Step L: step length; Step W: step width; DSP: double support percent Gait speed: linear velocity of the heel marker during FW and toe marker during BW; Step Length: horizontal distance in the sagittal plane between heel markers during FW and toe markers during BW; Step Width: horizontal distance in the frontal plane between heel markers during FW and toe markers during BW; Cadence: number of steps taken per minute; Double Support %: percent of gait cycle spent in double support

Range of Motion Analysis

Multivariate tests revealed no significant Group × Condition interaction (λ=0.894, F(4,31)=0.922, p=0.463, η_p²=0.106) for hip and shoulder ROM. Although there were no differences in hip or shoulder ROM between PwPD and CON for any joint (λ=0.788, F(4,31)=2.09, p=0.106, η_p²=0.212), we did find that when comparing BW to FW (λ=0.174, F(4,31)=36.80, p<0.001, η_p²=0.826), ROM was significantly smaller (p<0.001) during BW for the more affected hip (η_p²=0.793), less affected hip (η_p²=0.754), more affected shoulder (η_p²=0.439), and the less affected shoulder (η_p²=0.568) (Table 4, Supplementary Figure 2).

Table 4.

Mean (standard error) for hip and shoulder range of motion (ROM) measured in degrees and mean (standard error) for the point estimate of relative phase (PERP) measured in degrees.

	PwPD		CON			PwPD		CON
	BW Mean (SE)	FW Mean (SE)	BW Mean (SE)	FW Mean (SE)		BW Mean (SE)	FW Mean (SE)	BW Mean (SE)	FW Mean (SE)
MA/Non-Dominant Hip ROM (°)	25.85(.860)	37.87(1.35)	30.15(1.32)	39.58(1.17)	MA/MA PERP (°)	100.29(8.75)	138.80(5.44)	152.49(4.89)	153.48(4.44)
MA/Non-Dominant Shoulder ROM (°)	13.43(2.68)	21.13(3.36)	19.48(1.42)	30.06(2.38)	LA/LA PERP (°)	118.02(10.05)	141.10(5.36)	137.30(8.22)	155.51(4.26)
LA/Dominant Hip ROM (°)	27.41(1.20)	36.97(1.89)	31.61(1.48)	40.54(1.46)	MA/LA PERP (°)	45.13(6.15)	23.40(4.10)	26.07(5.56)	21.43(3.25)
LA/NonDominant Shoulder ROM (°)	13.16(1.75)	22.44(2.62)	17.58(1.63)	25.20(1.97)	LA/MA PERP (°)	57.83(6.34)	34.53(4.64)	21.06(3.21)	22.76(3.27)

ILC Analysis

Multivariate tests revealed a significant Group × Condition interaction (λ=0.585, F(4,31)=5.508, p=0.002, η_p²=0.415) on PERP. During BW, PERP values were significantly smaller (p<0.001, η_p²=0.248) in the more affected hip/more affected shoulder, and significantly larger (p<0.001, η_p²=0.202) in the less affected hip/more affected shoulder combination for PwPD disease compared to controls, with no significant differences in PERP between PwPD and CON during FW (p≥0.044) after applying Bonferroni corrections (Table 4, Supplementary Figure 1). In the more affected hip/less affected shoulder (η_p²=0.095) and less affected hip/less affected shoulder combinations (η_p²=0.004), PERP was not significantly different between groups during both BW (p≥0.028) and FW (p≥0.043). Compared to CON (λ=0.499, F(4,31)=7.77, p<0.001, η_p²=0.501), PwPD exhibited significantly worse PERP for the more affected hip/more affected shoulder (p<0.001, η_p²=0.431) and less affected hip/more affected shoulder (p<0.001, η_p²=0.427) compared to CON after corrections were made. There were no significant differences in PERP between groups in the less affected hip/less affected shoulder (p=0.036, η_p²=0.123) or the more affected hip/less affected shoulder combination (p=0.053, η_p²=0.106) after corrections were made. When comparing walking direction (λ=0.691, F(4,31)=3.46, p=0.019, η_p²=0.309), PERP was reduced during BW compared to FW for the ipsilateral limb combinations (all p≤0.005, MA/MA η_p²=0.267, LA/LA η_p²=0.206) and increased during BW compared to FW for the MA/LA limb combination (p=0.006, η_p²=0.199).PERP was not significantly different between BW and FW in the less affected hip/more affected shoulder combination (p=0.016, η^p2=0.159).

Overall, for PwPD impairments in ILC were exacerbated during backward walking compared to forward, especially when compared to CON. Furthermore, any combination of limbs which included the most affected shoulder exhibited the greatest impairment in coordination.

Discussion

In this study, we assessed measures of ILC and spatiotemporal gait parameters during backward and forward walking in PwPD and older adult controls with similar gait speed. Our findings confirmed that spatiotemporal gait parameters and forward ILC were not statistically different between groups. However, ILC during BW was significantly impaired in those with PD. Specifically, coordination between the more affected shoulder and the more affected hip, and more affected shoulder and less affected hip were impaired when walking backward. This disruption in coordination may contribute to the mobility and activity of daily living issues often reported during backwards walking in PwPD.

Our age- and gender-matched Parkinson’s disease and older adult groups demonstrated large differences in ILC for the more affected hip/more affected shoulder, more affected hip/less affected shoulder, and less affected hip/more affected shoulder during BW, but not forward walking compared to CON. This suggests ILC impairment during BW may be due to an underlying mechanism of Parkinson’s disease, and not the result of mobility impairments seen with normal aging. As we only found differences in ILC between groups during BW, it is possible when comparing persons with mild-to-moderate Parkinson’s disease to a healthy control group with similar mobility characteristics, measures of ILC during FW or spatiotemporal measures of gait during FW, may not be sensitive enough to detect impairment. It is also possible when looking at PERP, we did not find differences in FW ILC between groups because the cohort of older adults in our study were more mobility impaired than older adult cohorts in previous studies of ILC [5].

Similar to previous studies investigating differences in BW compared to FW [6,7], both groups in the current study demonstrated slower gait speed, shorter step length, wider step width, and higher double support percent during BW compared to FW. However, there were no significant differences between groups in spatiotemporal gait parameters after correcting for multiple comparisons. Although not statistically significant, the effect sizes of our results ranged from small (η_p² = 0.01) to large (η_p² = 0.937), highlighting the overall worse gait performance during BW for our cohort of PwPD. Studies within neurologically impaired populations have also found similar reductions in gait speed and step length, as well as increases in double support time during BW compared to FW [21]. This suggests BW may be better at discerning mobility impairments than FW in pathologic and aging populations [21]. Interestingly, we found that PwPD in our study exhibited smaller step widths during both BW and FW walking, and smaller double support percent during FW compared to controls. It should be noted that these differences were not statistically significant. Although the participants were of similar age and mobility characteristics, we did not collect the rehabilitation history of either group. It is possible that our findings are due to participants in our PD group undergoing physical therapy treatments to improve characteristics of their gait, such as step width.

Due to the speed-dependent nature of ILC, we specifically targeted older adults with gait speeds comparable to those with early PD. Of note, other than the reduced gait speed targeted, these CON were otherwise healthy. This cohort of PwPD had an average HY score of 2.0 and a forward gait speed of 0.94m/s (this value falls within population normative values for those with PwPD [22]). Previous work suggested that forward gait in early and/or optimally medicated Parkinson’s disease may not be markedly distinguishable from older adult controls [23–25]. Our study replicates these findings with this specific group of healthy CON, as we did not find any significant differences in spatiotemporal measures of gait between groups during FW.

In contrast to the results of Roemmich et al. [5], the current study failed to detect significant differences in ROM of the shoulder or hip between groups regardless of walking direction. As gait speed influences ROM [26], the lack of difference in ROM between groups may be attributed to the much slower gait speeds exhibited by the two groups in our cohort compared to those in previous studies of ILC [5]. Even with the lack of differences in ROM seen between older adult controls and PwPD, it is important to note coordination was still significantly impaired during BW in our Parkinson’s disease participants compared to our older adult participants. The lack of statistical difference in ROM suggests any differences in PERP between groups are unlikely to be due to differences in joint flexibility.

Reduced limb movements, such as arm swing while walking, in PwPD may be influenced by rigidity [27], potentially limiting functional ROM, the ROM required for performing daily activities such as walking. Arm swing is essential for successful gait performance because reduced arm swing increases the metabolic cost of walking and decreases gait stability [28]. In our study, the greatest impairments in ILC during BW were in the more affected hip/more affected shoulder and less affected hip/more affected shoulder combinations. Functional ROM of the more affected shoulder may be impacting the ability of PwPD to swing their arm while walking backward. To address this assumption, we investigated associations between these combinations of PERP and the associated ROM of the shoulder. The analyses identified only weak correlations between the more affected hip/more affected shoulder PERP and more affected shoulder ROM (p=0.517, r=0.164), and the less affected hip/more affected shoulder PERP and more affected shoulder ROM (p=0.155, r=−0.350) (Supplementary Figure 3). The lack of association may suggest functional ROM of the more affected side in PwPD is not driving impairments in ILC. More range of motion at the shoulder or hip does not necessarily improve coordination as measured by PERP. Rather, this impaired coordination of the arms and legs may be due to poor neural control of smooth coordinated movements of the limbs because of dysfunction in the basal ganglia.

The present study was limited to persons with mild to moderate PD. A PD cohort with more severe gait deficits may show greater ILC impairment. Future studies should include individuals at different stages of disease progression to investigate how coordination during backward walking is impacted by more severe PD. Including fall history could help better understand the relationship between backward walking performance and incidence of falls.

Our results demonstrate this cohort of PwPD have impaired coordination during BW, while FW ILC was similar to older adults. Impaired coordination in PwPD while walking backward may be due to dysfunction in controlling the movement of the more affected shoulder. Such dysfunction could contribute to the increased incidence of falls while trying to complete common activities of daily living that include backward walking. With BW on the rise as a popular and proposedly effective gait rehabilitation exercise [8,10,11,29], it is important to understand how the coordination PwPD differs from an older adult during this task. Clinicians can target arm swing when asking PwPD to perform a backward walking exercise, allowing PwPD to benefit from training a multidirectional movement such as backward walking.

• Point estimate of relative phase was used to assess coordination in those with PD.

• Those with PD had significantly impaired limb coordination while walking backward.

• More affected shoulder was involved in impaired coordination in backward walking.