1 INTRODUCTION

The most common motor symptoms of Parkinson's disease (PD) are bradykinesia and/or akinesia, stiffness, tremor at rest, postural instability, and gait disturbances (Jankovic, 2008). The gait of individuals with PD tends to be slower, characterized by narrow and short steps, flexed torso, little or no arm swing, slow and spasmodic turns (see more in Mancini et al., 2019). However, when explicitly analysing the spatiotemporal parameters of gait in individuals with PD compared with healthy individuals, studies have shown decreased speed (Cheng et al., 2014; Mondal et al., 2019; Morris et al., 1999), increased number of steps (Mondal et al., 2019), step (Cheng et al., 2014; Mondal et al., 2019), and stride length (Mondal et al., 2019) decrease; shorter duration of the swing phase and the one-leg support phase (Mondal et al., 2019); and longer duration of the double support phase (Mondal et al., 2019; Sofuwa et al., 2005). These changes in gait pattern are considered one of the symptoms that most affect the quality of life of individuals with PD and have been considered responsible for about 50% of the falls of patients (Lord et al., 2017), which may result in hospitalization and impaired motor function.

The administration of l-DOPA is one of the leading drug treatments for PD to restore dopaminergic levels. During the ON state of the medication, the self-selected gait speed of individuals with PD increases (Curtze et al., 2015; Mondal et al., 2019) regardless of how dopaminergic signalling acts (motor or motivational). However, when looking at the spatiotemporal parameters of gait individually, the medication induces different and inconsistent effects. For example, Curtze et al. (2015) showed that the ON state increased speed and stride length but did not influence cadence, step initiation, double support time, and swing time. Differently, Mondal et al. (2019) showed a decrease in the double support time and the number of steps, increasing the step and the stride length in the ON state. In that same study, the medication did not affect cadence, one-leg support time, step time, cycle time, swing time, and support base width. This variety of results makes it difficult to conclude anything about the effect of dopaminergic medication on the spatiotemporal parameters of gait. While Curtze et al. (2015) argued that l-DOPA improves gait without changing parameters related to its dynamic stability, Mondal et al. (2019) argued that parameters related to gait rhythm are resistant to l-DOPA and parameters that require caloric expenditure (i.e., stride length) are sensitive to medication (Blin et al., 1991).

Gait speed is the fundamental parameter that combines spatial and temporal gait measurement when providing information on the distance covered in a given period. In addition, gait speed influences other biomechanical variables such as joint kinematics, ground reaction force, muscle activation, moments of joint strength and power, and influences the spatiotemporal parameters of gait (Fukuchi et al., 2019; Herssens et al., 2018). The meta-analysis carried out by Fukuchi et al. (2019) showed that, in the elderly, both cadence and step length decrease when individuals walk more slowly and, as the speed increases, cadence, step size, and stride length also increase. The authors suggested that to compare the gait parameters in groups with different self-selected speeds, as occurs in the ON–OFF phenomenon, it is necessary to adjust the results according to the gait speed between the groups. For this, it is possible to use mathematical methods or models that reduce the impact of speed on the other variables analysed. Failure to adjust gait speed for comparisons between individuals with different gait speeds can influence the analysis of gait deviations and lead to misinterpretations about the impact of pathology on gait (Schreiber et al., 2018). As dopaminergic medication changes the gait speed of individuals with PD, it is possible to assume that the change in the values of the spatiotemporal gait parameters may result from the change in speed and not necessarily a direct effect of the medication on these parameters individually.

Therefore, it is necessary to understand the influence of speed on gait parameters. This study aims to analyse whether the selective ON–OFF medication effects on the spatiotemporal parameters of gait in individuals with Parkinson's can be explained by changes in gait speed in the ON and OFF states of the medication. The hypothesis is that changes in spatiotemporal gait parameters induced by the consumption of l-DOPA are associated with self-selected walking speed. Therefore, we carry out a Bayesian mediation analysis to assess which gait variables are affected by the medication state and check if those effects were mediated by speed changes induced by medication.

2 METHODS

Data on spatiotemporal gait parameters are available in Material S1.

3 RESULTS

The anthropometric and clinical characteristics of the individuals are presented in Table 1. There were significant differences between the ON and OFF conditions in the scores of the clinical scales for the total score of the UPDRS-III and items 12 of the UPDRS-III that evaluate the individuals' gait.

A significant increase in speed was observed in the ON (M = 3.56 km/h, SD = 1.02) compared with the OFF (M = 2.99 km/h, SD = 1.08) medication (p < 0.01). Figure 1 shows the dispersion of data for both medication conditions.

The Bayesian mediation analyses indicated that the medication had a general effect on five of the analysed gait variables: step length, stance phase, swing phase, total double support, and stride length (Table 2). Figure 2 shows the single mediator model. In the following analysis: c is the total effect of medication on the variable of interest (c = c’ + me); c′ or cp is the direct effect of medication; me is the speed mediation effect; pme is the proportion of the effect that is mediated by speed; a is the effect of medication on speed; b is the effect of speed on the variable of interest. The estimated total effect was positive for step length, swing phase, and stride length, indicating higher values in the ON medication; smaller values were found in the ON meditation condition for total double support and stance phase. The variables that did not present the general medication effect were not further evaluated regarding their mediation effects as the mediation loses its meaning in this case (see Material S3 for complete mediation analysis results and plots). The credible intervals indicate that speed effects mediated these effects for all of the variables that presented a significant total effect. The proportion of the mediated effects was close to 1 (fully mediated effect) for these five gait variables.

4 DISCUSSION

This study analysed whether the change in gait speed can explain the selective effects of l-DOPA on the spatiotemporal parameters of gait in individuals with PD. Bayesian mediation analysis evaluated which gait variables were affected by the medication state and checked if those effects were mediated by speed changes induced by medication. ON–OFF medication effects were observed in 5 of the 10 evaluated gait parameters: step length, swing phase, and stride length had increased values with medication; total double support and stance phase had smaller values in the ON medication condition. All of them were mediated by speed, with proportions of mediation close to 1 (effect entirely explained by speed changes).

Our results corroborate studies that show an increase in gait speed in individuals due to the effect of l-DOPA (Blin et al., 1991; Curtze et al., 2015; Mondal et al., 2019; Rochester et al., 2011). These results show that levodopa improves gait in individuals with PD. As in the OFF medication, walking at lower speeds may be mechanically less efficient due to less recovery of elastic energy in the musculotendon complex and a greater need for stabilization (Neptune et al., 2008). However, our results differ from other studies when analysing the other spatiotemporal parameters of gait. Studies showed that l-DOPA-sensitive and l-DOPA-resistant gait parameters suggest multiple neural circuits that control gait (Curtze et al., 2015; Hausdorff et al., 2001; Mondal et al., 2019). Studies showed that temporal parameters, such as step time, single support phase, and swing phase, are l-DOPA resistant, and parameters requiring energy expenditure are l-DOPA sensitive, such as step length, total double support, and stance phase. However, there is some criticism regarding this explanation: (1) some parameters involve temporal and spatial aspects, such as the single and total double support; (2) some correlated parameters, such as the stance and swing phase. As walking speed increases, this proportion of the stance and swing phase can be altered, the duration of the stance phase decreases, and the swing phase's duration increases. This makes it inconsistent that one parameter is resistant and the other responsible to l-DOPA; (3) speed influences all spatiotemporal parameters, for example, as the speed of walking increases, the percentage of the cycle in the double support period decreases. Our results showed a mediation effect ratio close to 1, which presents an almost toon effect. This indicates that changes in gait speed, and not l-DOPA, explain changes in spatiotemporal parameters. Furthermore, Turcato et al. (2018) found that during linear walking, the spatiotemporal parameters as cadence or stride length were not different between individuals with PD and healthy subjects if spontaneously walking at the same speed of PD.

Our results provide a more robust indication of the importance of considering the effects of walking speed when comparing gait data of the ON–OFF phenomenon. Gait speed has been reported as the primary determinant of kinematic and kinetic walking changes. Spatiotemporal gait parameters, ground reaction forces, joint angles, and muscle activity have all been affected by gait speed (Bovi et al., 2011; Stoquart et al., 2008; Tirosh et al., 2013). For example, gait speed can affect the percentages to the subphases of stance, where speed increases will decrease the total double support and increase one-leg support subperiods. In a systematic review, Fukuchi et al. (2019) showed that spatiotemporal gait parameters were generally affected by walking speed in all three age groups analysed, with large effect sizes. Furthermore, their results showed that most spatiotemporal parameters analysed decreased at slower speeds and increased faster. Hence, any change in gait speed can alter the movement pattern and bias the interpretation of pathologies on gait patterns. Therefore, it is essential to consider the gait speed to interpret the results better when studying walking biomechanics.

Lee and Hidler (2008) compared treadmill walking with overground walking in healthy subjects with no known gait disorders. Overall, they found very few differences in spatiotemporal gait parameters between treadmill and overground walking (only stance phase was different). Therefore, we can hypothesize that our results can be generalized to treadmill walking. Additional studies need to be conducted to confirm this hypothesis. As a limitation of our investigation, we did not follow a formal statistical approach in calculating sample size in this study. Therefore, despite gait problems worsening as the disease progresses (Mirelman et al., 2019), our sample is heterogeneous, considering different stages of the individuals, cognitive state and the presence or not of freezing of gait. In conclusion, this study shows that l-DOPA increases gait speed, and it is the first to show that a change in gait speed better explains the changes in the spatiotemporal parameters of gait than the ON–OFF phenomenon. As an implication for rehabilitation, our results suggest that it is possible to assess the effect of l-DOPA on improving motor symptoms related to gait disorders by measuring gait speed. This considerably reduces the number of parameters measured for this purpose and simplifies the analysis.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

DBC was responsible for the study supervision and obtained funding. JAO, PRB, CENO, and DBC were responsible for the study concept and design. CENO, RCT, SMH, ELA, CB, and LSO were responsible for the acquisition of data. JAO, PRB, CENO, ACLP, and DBC were responsible for the analysis and interpretation of data. JAO, PRB, and DBC were responsible for the drafting of the manuscript. MJC and ACLP were responsible for the critical revision of the manuscript for important intellectual content. JAO and PRB were responsible for the statistical analysis. CENO, RCT, SMH, ELA, CB, LSO, and DBC were responsible for the administrative, technical, or material support.