Levodopa facilitates improvements in gait kinetics at the hip, not the ankle, in individuals with Parkinson’s disease

Abstract

Parkinson’s disease symptoms impair gait, limit mobility, and reduce independence. Levodopa improves muscle activation, strength, and coordination; thus, facilitating increased step length, but few studies have evaluated the underlying forces associated with medication-induced gait improvements. Here, we assess the effects of levodopa on gait kinetics in persons with Parkinson’s disease. Over two sessions, 13 participants with Parkinson’s disease walked on a treadmill while both optimally medicated and after a 12-hour medication withdrawal. Walking was analyzed for spatiotemporal parameters, ranges of motion, anterior-posterior ground reaction forces, joint torques, and powers using an instrumented treadmill and motion capture system. When on medication, participants increased gait speed by significantly improving step length (p = .009) and time (p = .004). Peak propulsive force (p = .001) and hip flexion torques (p = .003) increased with medication while hip extensor and ankle plantarflexor torques did not. While differences in joint power were not significantly different, the optimal medication condition showed medium to large effects, with the largest effect at the hip (d_z = 0.84). Our findings suggest the underlying forces responsible for the increases in gait speed are primarily due to increases at the hip, with limited change at the ankle. Disproportionate use of muscle force may be a limiting factor for levodopa’s use as an intervention for walking. Future interventions should consider targeting force production deficits during gait in those with Parkinson’s disease.

1. Introduction

Persons with Parkinson’s disease (PD) often experience gait dysfunction that impairs ambulation and independence. Difficulty walking can have a considerable impact on daily life; for example, the average individual with PD walks too slowly to traverse a street within the programmed crosswalk time (Ellis et al., 2016). Symptoms of postural instability, rigidity in the lower extremity, and impaired muscle activation contribute to limited range of motion (ROM), reduced muscle power production, and diminished limb advancement, significantly impairing mobility (Albani et al., 2014). At the joint level, those with moderate-to-severe PD (Hoehn & Yahr Stage (H&Y) ≥ II) display a pattern of redistributing muscle force away from the ankle and toward the hip as a potential compensatory strategy to improve stability (DeVita & Hortobagyi, 2000; Kuhman et al., 2018; Winter et al., 1990) compared to those with mild PD (H&Y < II), even though this strategy may increase energetic cost (Albani et al., 2014; Donelan et al., 2002). Collectively, these features negatively affect spatiotemporal gait measures, reducing step length and gait speed (Albani et al., 2014; Nieuwboer et al., 1998). Indeed, walking is 30% slower in those with PD compared to age-matched peers and declines markedly faster with disease progression than normal aging (Ellis et al., 2016).

Gait speed is recognized as a clinical vital sign and predicts community independence and mortality (White et al., 2013). However, speed is a single measure that can be modulated by several strategies including changes in step length and/or frequency. Since reduced step length is the primary contributor to slowed gait speed in PD (Albani et al., 2014; Morris et al., 1994), focusing on the biomechanical basis for improving step length is of particular importance. Levodopa - the gold standard treatment for PD - results in faster walking driven by increased step length, while temporal parameters, such as stride time and double support time, show varied results and may not improve (Blin et al., 1991; Bryant et al., 2011; Mirelman et al., 2019). Even when optimally medicated, PD walking remains slower than healthy controls (Ellis et al., 2016). Thus, the addition of physical rehabilitation to pharmacological interventions are common (Nijkrake et al., 2007). For example, treadmill walking has been shown to significantly improve gait, with increased step length and reduced variability of stride time and swing time (Frenkel-Toledo et al., 2005; Krystkowiak et al., 2003; Pohl et al., 2003). An improved understanding of the biomechanical mechanisms underlying changes in step length could influence physical rehabilitation techniques, including treadmill interventions, to better complement pharmacological treatment.

Step length, and consequently speed, is improved by increasing the propulsive ground reaction forces (GRF), due primarily to increased muscular force generation at the ankle and hip (Judge et al., 1996). Although it is currently unknown how levodopa alters the kinetic pattern in those with PD, differences in hip kinetics may act as a main contributor to the differences in speed between older adults with and without mobility deficits (Graf et al., 2005). When older adults with higher and lower physical performance, as determined by Short Physical Performance Battery (cut-off score: 9), walk at similar speeds, “low-performance” older adults walk with reduced ankle power, and increased hip extensor power in early stance (Graf et al., 2005). Even when the “low-performance” group increased speed, further increasing step length, ankle power remained unchanged. While levodopa improves muscle activation, coordination, and overall strength (Cioni et al., 1997; Vaillancourt et al., 2006), there are no studies to our knowledge on non-surgically treated (no deep brain surgery/stimulation (DBS)) patients that investigate the impact of medication on kinetics during walking. We hypothesized that the effect of levodopa on gait would be driven by changes at the more proximal joint, and those with PD would exhibit increased hip joint torques–resulting in improved gait performance (e.g., increased step length and selected speed) -compared to following a 12-hour withdrawal.

2. Methods

The current study represents a secondary analysis from an original dataset and detailed methodology can be found in (Roemmich et al., 2014). Thirteen participants with idiopathic PD, whom were being treated with stable doses of orally-administered levodopa, were recruited. University Institutional Review Board approved, written informed consent was obtained. Participants came for testing at the same time of day on two days, so that on day 1, half of the participants were “On medication” (ON) and the other half were “Off medication” (OFF) with the opposite on day 2. During ON, participants came in their self-reported, optimally medicated state and during OFF were at least 12-hours withdrawn from any anti-parkinsonian medication (time since last dose: 15 ± 2 h). Participants disease severity was assessed using the Unified Parkinson’s Disease Rate Scale motor section (UPDRS-III) and H&Y via video-recordings scored by a movement disorders neurologist blinded to medication condition. The videographer scored rigidity in-person to complete the exam.

Thirty-five retroreflective markers were placed on participants according to the Plug-in-Gait full body marker system. A 7-camera motion capture system (120 Hz; Vicon, Oxford, UK) with instrumented treadmill (1200 Hz; Bertec Corporation, Columbus, OH) generated kinematic, including spatiotemporal parameters (STP), and kinetic data. Participants walked on the treadmill at their comfortable, preferred speed for 5 min while holding the handrails (Dingwell and Marin, 2006). Handrails do not significantly alter the kinematics in those with PD (Ambrus et al., 2019; Bello et al., 2010), but it is unknown how handrail use alters kinetics in PD or how levodopa affects use of handrails in PD. Treadmill speed was chosen independent of medication status or overground walking speed. Participants were naïve to their self-selected speed. Overground comfortable pace across a 12-m walkway was recorded before the protocol.

The last minute of walking was analyzed and gait events (foot-strike and foot-off) were identified. The following STP were calculated: step length, time, width, and single support (percentage of the gait cycle). Kinematic and kinetic data were computed using the Plug-in-Gait model and filtered using 4th-order, low-pass Butterworth filters with cutoff frequencies of 10 and 20 Hertz, respectively (Roemmich et al., 2014). Variables of interest included: sagittal ROM of the ankle and hip, as well as peak GRF braking and propulsive force, peak plantarflexion torque and power, and peak hip extensor and flexor torques and powers. The knee was not included in this analysis since it does not meaningfully contribute to forward progression (Judge et al., 1996). All kinetic variables were calculated as the peak value during stance (foot-strike to foot-off). To ensure consistency of peak power generation of the hip, hip power generation was limited to early stance (H1), hip extensor power generation (Eng and Winter, 1995; Winter et al., 1990). ROM was calculated as the difference between the maximum and minimum angles between foot-strike to the subsequent ipsilateral foot-strike.

Normality was verified (Kolmogorov–Smirnov test) and paired t-tests were performed between medication status for UPDRS-III, H&Y, overground speed, and average vertical GRF on the more affected limb. Considering PD is a movement disorder characterized by asymmetric motor symptom onset and progression, this investigation focused on the more affected limb, designated by UPDRS-III. Four repeated multivariate analysis of variance tested differences between medication state on each set of variables: 1) STPs, 2) ROMs, 3) force components (GRF and torques), and 4) force-time components (powers). Multivariate significance was set at p < .05. For univariate follow-up, Dunn-Bonferroni adjustments were used (STP: p = .01, ROM: p = .025, force components: p = .01, powers: p = .017). Cohen’s d_z estimated the within-subject repeated-measures effects using 0.2, 0.5, and 0.8 as a small, medium, and large effects, respectively.

3. Results

Thirteen participants were analyzed (height: 1.74 ± 0.07 m, mass: 74.1 ± 8.4 kg). H&Y stage was not different between collections (Table 1; t(12,1) = 1.897, p = .082). UPDRS-III scores improved ON levodopa (t(12,1) = 2.891, p = .014) and preferred overground walking speed was significantly faster ON (t(12,1) = 3.015, p = .011).

Table 1.

Demographic and disease severity ratings for each participant.

ID	M/F	Age (years)	LEDD (mg/day)	Disease duration (months)	H&Y		UPDRS-III		Overground Speed (m/s)		Treadmill Speed (m/s)
					OFF	ON	OFF	ON	OFF	ON	OFF	ON
1	M	79	800	24	3.0	3.0	47	43	0.94	0.89	0.65	0.80
2	M	69	825*	72	2.5	2.5	39	42	0.94	1.14	0.81	0.81
3	M	76	1250	n/a	3.0	3.0	58	46	1.31	1.35	0.89	1.26
4	M	65	488*	48	2.5	2.5	31	28	0.98	1.03	0.85	0.97
5	M	65	1100	60	2.5	2.5	40	38	0.99	1.29	0.93	1.03
6	M	65	413*	48	2.5	2.0	40	39	1.07	1.22	0.78	0.80
7	F	54	380*	28	2.5	2.5	29	27	1.07	1.12	0.81	1.01
8	M	66	750	60	2.5	2.0	45	39	0.82	1.3	0.81	1.08
9	M	77	500*	36	3.0	3.0	46	41	1.36	1.36	1.04	1.04
10	F	49	1775*	60	2.0	2.0	24	23	1.16	1.17	1.11	1.35
11	M	65	n/a	36	3.0	3.0	44	40	0.96	1.08	0.95	0.95
12	M	70	n/a	108	3.0	2.5	32	34	1.00	1.02	0.74	0.82
13	M	72	642*	120	2.5	2.5	41	37	1.02	1.35	0.68	0.96
Mean		67.1	975	58.3	2.65	2.5	39.7	36.7	1.05	1.18	0.85	0.99
SD		8.5	239.8	29.8	0.32	0.38	9.0	6.8	0.16	0.15	0.13	0.17

ID – Participant Identification, LEDD – Levodopa Equivalent Daily Dose, H&Y – Hoehn & Yahr stage, UPDRS-III - Unified Parkinson’s Disease Rating Scale Motor Score (III), Overground Speed – preferred pace of walking overground, n/a – not available

^*dopamine agonist in addition to levodopa

STP:

Multivariate results revealed a significant effect of medication (Table 2; λ = 0.810, F(5,8) = 5.109, p = .021). The ON condition increased preferred speed (F(1,12) = 16.995, p = .001), increased step length (F(1,12) = 9.574, p = .009), and reduced step time (F(1,12) = 12.336, p = .004). Step width and single support percent were not different between conditions (p ≥ 0.783).

Table 2.

Gait Parameters for ON and OFF at preferred Treadmill speed. Mean (Standard Deviation).

	OFF	ON	p-value	Cohen’s dz
Spatiotemporal *
Selected Treadmill Speed (m/s) †	0.85 (0.13)	0.99 (0.17)	.001	1.15
Step Length (m) †	0.51 (0.08)	0.55 (0.08)	.009	0.85
Step Time (s) †	0.66 (0.07)	0.61 (0.05)	.007	1.07
Step Width (m)	0.22 (0.03)	0.22 (0.03)	.872	0
Single Support (% cycle)	37.7 (2.0)	37.5 (1.9)	.783	0.08
Gait Cycle Range of Motion *
Ankle (Degrees) †	24.0 (5.5)	27.9 (8.8)	.015	0.80
Hip (Degrees) †	41.3 (3.8)	44.3 (5.5)	.003	1.02
Peak Force Based Kinetics *
Propulsive Force (%BW) †	0.13 (0.02)	0.15 (0.03)	.001	1.31
Braking Force (%BW)	−0.07 (0.03)	−0.09 (0.06)	.041	0.57
Ankle Plantarflexor Moment (Nm/kg)	1.29 (0.14)	1.26 (0.18)	.549	−0.22
Hip Flexor Moment (Nm/kg) †	−0.45 (0.24)	−0.74 (0.36)	.003	1.03
Hip Extensor Moment (Nm/kg)	0.85 (0.19)	1.14 (0.46)	.059	0.60
Peak Force-time Based Kinetics
Ankle Power Generation (Watts/kg)	1.48 (0.45)	1.87 (0.59)	.017	0.79
Hip Power Generation (Watts/kg)	0.71 (0.12)	1.31 (1.01)	.053	0.59
Hip Power Absorption (Watts/kg)	−0.40 (0.22)	−0.73 (0.43)	.009	0.84

^*Significant at Multivariate Level

^†Significant at Univariate Level with Dunn-Bonferroni Correction Note: Hip Power Generation is limited to early stance.

ROM:

Medication significantly affected ROM at the ankle and hip (λ = 0.401, F(2,11) = 8.203, p = .007). ROM at the ankle (F(1,12)= 8.13, p = .015) and hip (F(1,12) = 14.07, p = .003) increased with dopaminergic therapy.

Force based kinetics:

Medication significantly changed the kinetics of individuals walking on the treadmill (λ = 0.174, F (5,8) = 7.581, p = .007). Peak hip flexion torque in stance (F(1,12) = 14.08, p = .003) and peak propulsive force (F(1,12) = 21.009, p = .001) significantly increased in magnitude. Peak hip extension torque (p = .059), peak plantarflexion torque (p = .549), and peak braking force (p = .041) were not significantly different after corrections.

Force-time based kinetics:

Although the multivariate level was found to be not significant for medication changes in power (λ = 0.551, F(3,10) = 2.719, p = .101), univariate p-values can be found in Table 2.

4. Discussion

Unsurprisingly, optimal medication status increased preferred speed on the treadmill with corresponding changes to reduced step time and increased step length. The underlying kinetic changes included increased propulsive force and increased hip flexion torque in stance with no significant change in hip extensor or plantarflexor torque. Yet, the medium effect size between the ON and OFF conditions for hip extensor torque further suggests that the hip extensors in early stance potentially contributes more effectively than the ankle toward maintaining position on the treadmill and progressing forward. No statistically significant differences were found in the variables of joint power when analyzed together, but all variables showed medium to large effects eluding to potentially meaningful effects. These findings are consistent with the distal-to-proximal redistribution found in healthy older adults (DeVita & Hortobagyi, 2000; Graf et al., 2005; Kuhman et al., 2018). Collectively, these results identify further gait-related measures that are resistant to anti-parkinsonian medication that could be specifically targeted with supplementary interventions.

The underscaling of amplitude of movements in PD is thought to be more affected than the regularity or timing (Albani et al., 2014), therefore improvements in gait speed are primarily exhibited by increased step length (Albani et al., 2014). Yet similar to Krystkowiak et al., 2003, optimal medication state improved both temporal and spatial parameters of walking with significant increases in joint ROM with large effects. Additionally, the coordination of phasic action, shown with single support percent, was nearly identical between medication states and was within the normative values for healthy control subjects, further supporting that underlying control for gait may remain intact in PD (Hollman et al., 2011; Morris et al., 1994). The use of the treadmill may further aid in controlling coordination as a constant speed is required, as shown with the lack of differences in single support and improvement of both step length and time. As all participants used the handrails, we expected no differences in step width and the present observations confirmed this. While the kinematic variables support the use of levodopa to improve gait performance through increasing amplitude of steps, multiple variables including selected speed, step length and time are still below healthy control norms suggesting additional improvement may be possible (Hollman et al., 2011).

The biomechanical changes related to the increase in speed were driven primarily by a large effect of hip flexor torque increases accompanied by a slightly decreased ankle plantarflexion torque. Joint power was statistically unchanged, similar to the results found by DeVita and Hortobagyi, 2000, perhaps due to the large variability within group. While our conservative statistical approach limits the interpretation of the univariate effects, the differences showed medium to large effects, including peak ankle power, warranting further exploration of joint power. While we analyzed peak hip extensor power generation, which stabilizes the trunk and aids propulsion, the hip flexors also provide considerable power absorption to decelerate the limb during extension and power generation to forward motion during pre-swing (Eng & Winter, 1995; Winter et al., 1990). Interestingly, optimal medication state increases activation of distal leg muscles better than proximal muscles (Cioni et al., 1997), this pattern is not reflected in the kinetic changes observed, as ankle joint moment slightly decreased. While Ferrarin et al. tested overground gait associated with both medication and DBS, their levodopa-specific results match ours, with increases in hip torque and power with no change in ankle torque or power when speed increased (Ferrarin et al., 2005). Of note, all subjects in the mentioned study (Ferrarin et al., 2005) had DBS and thus, even in the OFF medication and OFF stimulation condition, participants had surgical lesions that may improve symptoms of PD (Siegel & Metman, 2000; Walter & Vitek, 2004). No participants in the current study had DBS. While anti-parkinsonian medications may improve neuromuscular activity of the distal musculature (Cioni et al., 1997), optimally-medicated individuals with PD still display reduced triceps surae activation and increased proximal muscle activation compared to healthy older adults (Cioni et al., 1997; Dietz et al., 1995; Islam et al., 2020). Although we did not analyze joint redistribution or muscle activation, medication may result in disproportionate reliance on the hip flexors and extensors to improve propulsion and speed, regardless of the gains provided with dopaminergic therapy (Cioni et al., 1997). As this proximal redistribution may be an adaptive compensation (Kuhman et al., 2018), hip and ankle function during walking should be further investigated in this population to better understand the potential positive or negative impacts associated with this gait pattern. Improved gait continues to be an outcome of importance for those with PD and future studies should specifically target impaired force production, in conjunction with joint kinetic redistribution, to deliver novel therapeutic interventions that could improve mobility.