Effects of On- and Off-Medication Periods on Walking Performance in Parkinson’s Disease: Insights from Movement Synergies Open Access

Parkinson’s disease (PD) is an age-related neurodegenerative disorder that primarily affects older individuals but can also occur in those under 50 [1]. It is characterized by significant loss of dopaminergic neurons in the substantia nigra pars compacta, as well as in other brain regions such as the basal ganglia and cortex [1, 2]. This degeneration leads to both motor symptoms (e.g., tremors, rigidity, bradykinesia, and postural instability) and non-motor symptoms (e.g., cognitive impairment, depression, and sleep disturbances) [1, 3]. Gait abnormalities are common in PD patients and include shorter, slower steps due to hypokinesia (a reduction in movement amplitude) [4]. Arm swing movements during gait are often irregular, narrow, and slow due to bradykinesia [5], and a shuffling gait with reduced foot clearance increases the risk of tripping and falls [6]. Additionally, gait asymmetry – manifested as differences in step length and timing between the left and right limbs – further impairs walking and stability in PD patients [1]. These impairments limit locomotor adaptability and increase fall risk, highlighting the need for effective gait assessment and intervention strategies.

The primary treatment goal for PD was to alleviate motor symptoms and improve quality of life, with medications such as levodopa and dopamine agonists commonly used [7]. However, fluctuations between “on” and “off” medication periods complicate the maintenance of stable mobility [8]. During the on-medication state, dopamine restoration often improves gait performance, leading to increased stride length and improved arm swing [8]. In contrast, off-medication periods lead to a resurgence of motor symptoms, including slower walking, shortened steps, and impaired balance, which significantly reduce daily mobility and increase fall risk [8]. While many studies have examined spatiotemporal gait parameters (e.g., stride length, velocity, and cadence) across medication states in PD [9], these measures do not fully capture the neuromuscular control strategies involved in locomotion. A deeper understanding of the multi-segmental coordination and adaptive control of movement in PD is critical for optimizing rehabilitation and medication strategies.

Neuromuscular control refers to how the nervous system regulates muscle activity to coordinate movement, particularly during gait [10]. Unlike traditional gait analysis methods in PD patients [11], principal component analysis (PCA), a mathematical method for dimensionality reduction [12], can provide a different perspective on gait through a data-driven approach to identifying dominant movement patterns from complex, multidimensional kinematic data [9, 13, 14]. PCA extracts key movement components – termed principal movements (PMs) – which represent coordinated segmental interactions across different walking and running gait phases [9, 13, 15‒20]. These PMs correspond to specific gait phases, such as the swing and stance phases, offering an alternative view by assessing movement components/synergies, providing a more comprehensive representation of locomotor control [15, 16, 18, 21]. In this sense, PCA allows for an in-depth examination of gait adaptability, revealing subtle impairments that may not be detected through conventional metrics alone [22, 23]. For example, the composition and control of walking movements have been assessed by quantifying the contribution of individual PMs to total variance [15, 23] and their acceleration magnitude, reflecting the overall amplitude or energy of postural movements [13, 16]. These metrics are directly associated with system forces and myoelectric activity, demonstrating the effectiveness of PM-based variables in assessing neuromuscular control [22, 23]. Although PCA has been applied in various locomotion studies – such as investigating fall risk [18], sex differences [16], gait variability [9], clinical gait analysis [18, 21, 24, 25], and gait adaptation [13, 17, 20] – its application in PD gait analysis remains limited. Specifically, the influence of medication states on walking movement synergies in PD remains underexplored. This study aims to fill this gap by examining how on- and off-medication periods affect walking performance, offering novel insights into the neuromuscular control mechanisms that drive gait adaptations in PD.

In summary, this study aimed to examine the impact of on- and off-medication periods on walking performance, focusing on gait adaptations in PD patients through assessing the composition and control of individual PMs. The on-medication period restores dopamine levels in the brain and reduces motor symptoms, which may mitigate gait irregularities, leading to smoother, faster, and more coordinated walking, with increased stride length, arm swing, and overall mobility [26]. It was hypothesized that key PMs, particularly those reflecting the swing phase, would show increased acceleration magnitude and contribution during the on-medication state.

The kinematic marker data for the current study analysis were obtained from an open-access dataset [27]. The participants, a total of 22 community-dwelling individuals with idiopathic PD (5 females and 17 males), were on a stable dose of l-Dopa for at least 1 month [27]. Inclusion criteria required that participants had no neurological or physical dysfunctions other than those associated with PD and self-reported no vestibular, visual, or somatosensory dysfunctions [27]. The severity of motor symptoms of all participants was classified using the Hoehn and Yahr (H&Y) scale [28], ranging from 0 to 5, with higher scores indicating more severe symptoms and functional limitations. The study was conducted at the Biomechanics and Motor Control Laboratory at the Federal University of ABC, Brazil, and was approved by the local Ethics Committee (Approval No. 21948619.6.0000.5594). All participants provided their informed consent before participating in the study. Table 1 provides an overview of the characteristics of the PD patients included in the study.

The experimental procedures were thoroughly described in Shida et al. [27]. Briefly, the study involved two experimental sessions separated by 1 week. One session was conducted under the on-medication condition, while the other was under the off-medication condition. For the on-medication session, participants took their dopaminergic medication 1 h before the session to ensure dose stabilization. Conversely, in the off-medication session, participants abstained from their PD medication for at least 12 h prior to testing [29]. The order of the sessions was randomized across participants. Each participant was fitted with 44 reflective markers placed on anatomical landmarks, including the lower limbs, trunk, and upper limbs. Participants were instructed to walk barefoot at a self-selected, comfortable speed on an overground surface. Walking movements were captured using a motion capture system comprising 12 cameras (Raptor-4; Motion Analysis Corporation, Santa Rosa, CA, USA) operating at a sampling rate of 150 Hz. Data collection was conducted using the Cortex 6.0 motion capture system (Motion Analysis, Santa Rosa, CA, USA). While the original dataset included kinematic data from 26 individuals with PD [27], 4 participants were excluded from this study due to incomplete walking trials in either the on- or off-medication condition. As a result, data from 22 participants were included in the final analysis.

All data processing was carried out using MATLAB version R2023b (MathWorks Inc., Natick, MA, USA). To extract movement synergies, 26 markers were selected from prominent anatomical landmarks on the body. Each dataset, consisting of these 26 markers, comprises 78 spatial coordinates (x, y, z), forming 78-dimensional posture vectors [30, 31]. The kinematic datasets from 10 walking trials (5 on-medication and 5 off-medication trials) for each participant were processed for preprocessing. Centering was performed by subtracting the mean posture vector to reduce the influence of mean marker positioning on the PCA outcome [30]. Further normalization was applied using the mean Euclidean distance to account for anthropometric differences [30]. These preprocessed datasets were then concatenated into a single input matrix (5 trials × 2 conditions [on and off medication] × 22 participants) for subsequent PCA analysis, enabling a comparison between on- and off-medication periods.

PCA was conducted using the PManalyzer software [30], which employed singular-value decomposition of the covariance matrix. This process transformed the kinematic data into orthogonal eigenvectors known as “principal components” (PC_k), with the subscript k indicating the order of movement components. Each eigenvector represented a movement pattern and generated a corresponding “principal movement” (PM_k) depicted as an animated stick figure [30]. The time evolution of each PM was quantified using principle component (PC) scores, representing positions in posture space called principal positions (PP_k), i.e., in the vector space spanned by the PC eigenvectors. The first eigenvector represents the motion of the patient through the laboratory and accounts for a substantial portion of the variance. Additionally, principal accelerations (PA_k) were derived from PC scores using conventional differentiation rules, representing actual body segment accelerations [32]. Previous studies have established associations between PA_k and leg myoelectric activity, suggesting that PA-based variables provide insights into the neuromuscular control of individual PM_k [22, 23]. Specifically, the relative explained variance of PA_k (PA_k_rVAR) describes the percentage contribution of each PM to the total variance in postural accelerations. A higher PA_k_rVAR value indicates that a movement component is executed with sufficient quickness and speed to significantly impact accelerations and forces within the system [22, 31].

To address noise amplification resulting from differentiation, a Fourier analysis was performed on the raw PC scores [30]. This analysis revealed that the highest power was concentrated in the 2–5 Hz frequency range, with noticeable power extending into the 5–10 Hz range. Consequently, a 3rd-order zero-phase 10-Hz low-pass Butterworth filter was applied to the PCA-based time series before differentiation. Additionally, leave-one-out cross-validation was conducted to assess the robustness of individual PM_k and the PCA-based dependent variables, thereby ensuring the validity of the input data matrix. The study selected the first five PCs that demonstrated robustness after cross-validation for hypothesis testing.

Two PCA-based variables were computed for the first five PMs. The first variable, participant-specific relative explained variance of principal positions (PP_k(t)), or PP_k_rVAR, was calculated to quantify the percentage contribution of each PM_k to the overall variance in postural positions [22, 30]. This variable provides insights into how much (in percentage) each PM_k is shaping the movement structures of overground walking [30]. Differences in PP_k_rVAR, such as those between on- and off-medication periods, could indicate variations in the coordination structure of overall postural movements [30].

The second variable, participant-specific root mean square of principal accelerations (PA_k(t)), or PA_k_RMS, was calculated to represent the overall magnitude of acceleration over a given time interval [13, 16, 17]. To account for the influence of walking speed on movement intensity, PA_k_RMS was normalized by each participant’s individual walking speed [13, 16]. This normalization ensures that any observed changes in PA_k_RMS are not confounded by variations in walking speed, allowing for a more accurate comparison of neuromuscular control across participants and trials [13, 16]. Acceleration measures the rate of velocity change, with higher values indicating quicker movements or stronger forces acting on the body [13]. The root-mean-square (RMS) calculation quantifies the intensity or energy of these acceleration signals [33], providing insights into the magnitude of neuromuscular control exerted over individual PMs. A higher RMS of acceleration typically reflects more forceful or physically demanding movements [34, 35]. Since PA(t) is linked to myoelectric activity [23], PA_k_RMS serves as a valuable variable for quantifying neuromuscular control, capturing both the intensity and variability of movement – key components of motor control [13, 16, 17]. As discussed in the review paper [36], neuromuscular control theories propose that individuals demonstrate structured muscle activity patterns within a theoretical solution space. These patterns exhibit variability, which reflects the motor system’s adaptability to external demands. In this sense, PA_k_RMS provides valuable insights into this aspect of neuromuscular control, illustrating how the system adjusts muscle forces to maintain stability and efficiently perform motor tasks.

The IBM SPSS Statistics software version 26.0 (SPSS Inc., Chicago, IL, USA) was employed for conducting all statistical analyses, with the alpha level set at α = 0.05. Shapiro-Wilk tests were utilized to assess the normal distribution. The Wilcoxon signed-ranks test was conducted to examine the effects of on- and off-medication periods in the two PCA-based variables (PP_k_rVAR, and PA_k_RMS) as indicators of walking performance. Cohen’s d was calculated to determine the effect size.

Figure 1 summarizes the descriptive characteristics and eigenvalues of the first five principal movements (PM_1–5) derived from pooled on- and off-medication trials. The first principal movement (PM₁) accounted for the largest proportion of variance in PP, primarily representing whole-body movements in the direction of walking (PP₁_rVAR: 98.6 ± 0.6). PM₂ was associated with the swing phase movement (PP₂_rVAR: 1.7 ± 0.6; PA₂_rVAR: 30.9 ± 7.6), contributing to the largest proportion of variance in PA. PM₃ and PM₄ represented mid-stance phase movements, with PM₃ involving anti-phase lower-limb movements coupled with mediolateral sway of the upper body (PP₃_rVAR: 0.1 ± 0.1), while PM₄ featured trunk rotation (PP₄_rVAR: 0.06 ± 0.04). Finally, PM₅ corresponded to the double-leg support phase, characterized by synchronized ankle and knee flexion-extension movements (PP₅_rVAR: 0.1 ± 0.1; PA₅_rVAR: 26.1 ± 5.8). Although PM₅ accounted for the smallest proportion of variance in PP, it contributed significantly to PA variance. The animated stick figures of these PM_1–5 are available in online supplementary Video S1 (for all online suppl. material, see https://doi.org/10.1159/000546733).

As shown in Figure 2, the results comparing the “on” and “off” conditions for the PP_k_rVAR and PA_k_RMS measures across different movement components (PM_1–5) revealed significant findings in specific PMs. For PP_k_rVAR (Fig. 2a), significant differences were found in PM₁, which represents whole-body movement in the direction of walking (PP₁_rVAR, p < 0.001, Cohen’s d = 0.883). The off-medication period showed a greater contribution to this PM compared to the on-medication period. In contrast, the on-medication period exhibited a greater contribution to PM₂, which represents the swing phase movement (PP₂_rVAR, p = 0.001, Cohen’s d = 0.870), and PM₅, which represents the double-leg movement (PP₅_rVAR, p = 0.016, Cohen’s d = 0.779) compared to the off-medication period. However, no significant differences were found for PP₃_rVAR (mid-stance with mediolateral sway; p = 0.527, Cohen’s d = 0.172) and PP₄_rVAR (mid-stance with trunk rotation; p = 0.058, Cohen’s d = 0.267), suggesting a lesser impact of the on-medication period on these phases.

For the PA_k_RMS measures (Fig. 2b), significant differences were observed in PM₄, which represents single-leg support coupled with trunk rotation (PA₄_RMS, p = 0.026, Cohen’s d = 0.564). In contrast, no significant differences were found for PA₁_RMS (p = 0.095, Cohen’s d = 0.332), PA₂_RMS (p = 0.685, Cohen’s d = 0.115), PA₃_RMS (p = 0.082, Cohen’s d = 0.345), and PA₅_RMS (p = 0.685, Cohen’s d = 0.115).

The current study investigated the effects of on- and off-medication periods on overground walking performance in PD patients, focusing on the composition and control of PMs. PMs, which are movement components or synergies extracted using PCA, were analyzed to understand how medication impacts walking performance in PD patients. Two PM-based variables were computed for each PM: PP_rVAR (composition) and PA_RMS (control). In partial agreement with the hypothesis, the current findings showed that medication increases the contribution and acceleration magnitudes of specific movement components. Specifically, an increased contribution (greater relative explained variance [rVAR]) was observed in PM₂, representing the swing-phase component, while increased movement acceleration magnitude (greater RMS) was observed in PM₄, characterizing the mid-stance phase coupled with trunk rotation.

Regarding the movement structure or composition of walking movements, individuals with PD exhibited a greater contribution (higher rVAR) to the swing-phase movement component (PM₂) during the on-medication period. This suggests several positive implications. An increased rVAR indicates that medication enhances leg movement dynamics during the swing phase, which may help alleviate symptoms such as freezing of gait – crucial for achieving forward propulsion during locomotion [37]. Improved medication efficacy can lead to smoother and more coordinated movements, facilitating stride initiation and execution, ultimately manifesting as increased walking speed [38]. Since individuals with PD often struggle with initiating and executing the swing phase, leading to shorter and less effective strides [39], medication may promote a more natural gait pattern and reduce gait abnormalities like shuffling. Additionally, an increased rVAR during the swing phase may suggest improved gait efficiency and propulsion, enabling patients to walk more effectively with reduced fatigue. Medication may also enhance energy efficiency by lowering energy expenditure, allowing patients to walk longer distances with less effort [40]. This is particularly important as reduced muscular strength is associated with slower walking velocity and an increased risk of falls in individuals with PD [40]. However, while a higher rVAR generally reflects improved motor function and gait dynamics, it is essential to ensure that these enhancements do not come at the cost of increased instability or fall risk. In summary, an increased rVAR during the swing phase suggests that medication effectively enhances motor function, contributing to improved gait efficiency. However, further assessment is necessary to determine the balance between these benefits and potential risks associated with instability.

In terms of movement control, principal accelerations (PAs) are directly associated with muscle activation signals, reflecting neuromuscular control and demonstrating how the nervous system coordinates muscle activity to manage postural control and movement [23]. When the body undergoes complex, whole-body movements, the accelerations of individual body segments (captured by PAs) offer valuable insights into how these movements are regulated by the neuromuscular system [22]. Focusing on the current findings, medication administration was shown to boost movement acceleration magnitudes during the single-leg support phase (PM₄) of the gait cycle, particularly during mid-stance. This increase in movement dynamics may reflect greater adjustments of the body to maintain equilibrium [33] or alternatively, increased instability [38]. Gait and balance are not singular functions; they involve distinct and overlapping neural circuits with varying sensitivity to l-Dopa [38]. While increased movement amplitudes and enhanced postural muscle responsiveness could improve gait efficiency, they also place greater neuromuscular demands on individuals with PD, particularly those with impaired postural reflexes or balance deficits [41, 42]. Excessive or uncoordinated dynamic movements might generate destabilizing forces, raising the risk of falls, especially during challenging walking tasks or in unfamiliar environments [38, 41, 42]. Moreover, while levodopa-induced dynamic improvements enhance mobility, such as increasing walking and turning speed, they may also present challenges to stability, further influencing fall risk [38]. These findings underscore the need to consider whether increased movement dynamics, as indicated by higher RMS values, might elevate fall risks in individuals with PD. This warrants further investigation to better understand the balance between improved mobility and increased instability.

From a clinical point of view, the current empirical findings emphasize the potential benefits of medication in improving movement composition by increasing the contribution of the swing phase. When PD patients are on medication, specific exercise strategies can further amplify gait improvements [43]. During the swing phase, exercises such as dynamic leg swings with resistance bands and controlled high knees with added weights can help improve strength, range of motion, and hip flexor coordination [44]. Although medication seems to be adverse for the stance phase due to potentially increasing instability, exercises like weighted single-leg stances and heel-to-toe walks with resistance bands may enhance balance, postural stability, and lower limb strength [44]. Additionally, a constraint-focused agility exercise program, incorporating movement principles from Tai Chi, boxing, and Pilates, can provide a progressive sensorimotor challenge by improving neuromuscular coordination, postural control, and resistance to balance disturbances [41]. Understanding walking movement synergies can help in optimizing rehabilitation strategies for PD patients.

In conclusion, this study highlights the impact of medication on overground walking in PD, showing increased movement contribution during the swing phase (PM₂) and greater acceleration magnitudes in mid-stance (PM₄). These findings suggest improved gait efficiency but also raise concerns about potential instability. While medication enhances mobility, its effects on balance warrant further investigation. Future research should explore tailored interventions to optimize both gait performance and stability in PD patients.

The current study has several limitations. While the dataset provided the total daily levodopa equivalent dose [27], it lacked detailed information on medication types and dosages, which could influence gait dynamics. The absence of a control group [27] further limits comparisons with individuals without PD, making it challenging to isolate the specific effects of PD and its treatment. Another limitation is the lack of head movement data as head and neck movements contribute to postural control and balance. Future studies should incorporate data from all body segments for a more comprehensive understanding of neuromuscular dynamics [22]. Additionally, the study did not assess tasks such as backward walking, which is often impaired in PD. Including such tasks could provide further insights into the effects of medication on gait in real-world scenarios [45]. Finally, the study’s short duration and controlled environment may not fully capture the long-term effects of medication. Future research should incorporate longitudinal assessments to evaluate how medication use influences gait over time.

The current open access datasets used in this study were conducted in accordance with the Declaration of Helsinki, and the study protocol was approved by the Ethics Committee of Federal University of ABC, Brazil (Approval No. 21948619.6.0000.5594) [27]. Written informed consent has been obtained from the patients [27].

The authors have no conflicts of interest to declare.

This study was funded by University of Phayao and Thailand Science Research and Innovation Fund (Fundamental Fund 2025, Grant No. FF68-UoE5030/2567).

Conceptualization, software, and writing – review and editing: A.P. and P.F.; methodology, validation, formal analysis, data curation, writing – original draft preparation, visualization, project administration, and funding acquisition: A.P.; and supervision: P.F.

The original data used in the current study are openly available in Shida et al. [27]. Further inquiries can be directed to the corresponding author.