Differential gait adaptation patterns in Parkinson’s disease – a split belt treadmill pilot study

Parkinson's disease (PD) is a neurodegenerative disorder. As the disease progresses it disrupts the patient's motor skills and causes a notable gait asymmetry, most likely stemming from asymmetrical neural dopaminergic degeneration of the nigrostriatal pathways [1]. Asymmetrical PD gait is expressed, for example, by both step length [2] and swing time [3] asymmetries, which were also implicated with the prevalence of the freezing of gait symptom and falls among persons with PD [4, 5]. Thus, in recent years, asymmetry has been a target for intervention in PD and other neurological disorders patients, utilising neural plasticity by means of conventional physiotherapy [6] or by the use of split belt treadmills (SBTM) [4, 7‐9].

A SBTM is a useful tool to study neural adaptation mechanisms addressing bilateral function of gait, i.e., gait asymmetry (GA) and left–right stepping coordination [10, 11]. By inducing uneven speeds to the two belts, participants are forced to adapt their gait to the changing conditions by altering stance/swing times relations and thus, step length [12‐14]. These changes are different for each leg, thus result in a consequent GA modification [10, 11, 14]. Often, in studies, participants are exposed to the split belt walking condition for a duration of 5–15 min, known as the 'adaptation period', which allows them to gradually adapt to the new walking pattern. Following the adaptation period, the belt speeds are reset back to equal speeds and effects of post-adaptation become apparent (i.e., induced change in GA) [4, 14‐23]. In the case of intensive SBTM training programs, after-effects may be evident for weeks or months [24, 25]. SBTM walking can be administrated in different ways, i.e., increasing the right belt speed, decreasing it, or manipulating the left belt’s speed. In cases where disease’s motor symptomology is asymmetric, complexity increases when changes are made with reference to the more\less affected side. A recent systematic review [9] aimed to summarise the existing evidence on SBTM paradigms and effects on gait in PD compared to healthy controls. It concluded with a strong recommendation for a standardisation of current SBTM protocols, due to a large variation across studies in methodological aspects such as SB intervention parameters. The aim of this pilot study is to systematically compare the adaptation patterns associated with split-belt walking, with respect to the type of manipulation (increase\decrease speed) and side of the manipulated belt in reference to PD-related asymmetry. We hypothesised that stronger adaptation effects would be observed by augmenting the initial asymmetry by either reducing the belt's speed of the less affected side or increasing the speed of the more-affected side.

The present study, for the first time to our knowledge, evaluates gait adaptation patterns in all four SBTM conditions. We found that stronger adaptation effects are obtained by decreasing a belt’s speed, as opposed to increasing the speed of the opposite belt. This distinction is rarely addressed in SBTM studies.

Nenhoe-Mahabier et al. found no significant differences between BSI and WSI and treated the two SB conditions interchangeably [30]. Fasano et al. distinguished between two non-matching SB conditions, BSD and WSD, and found BSD the most effective condition for GA adaptation, in agreement with our results [4]. Further to their study, we observed that the matching condition (WSI) had weaker effects on GA.

We found that the BSD/WSI matching conditions bears the highest potential impact on the post adaptation period (compare panels A & D in Fig. 2), by first inducing and increase in baseline asymmetry during the early adaptation period, i.e., by placing the leg with the longest step on the relatively slower belt (BSD), thereby initially producing a longer step length in response to the asymmetric belt speed [31]. Even though our participants were relatively symmetric at baseline, when asymmetry was induced via SBTM intervention, a change was observed in the opposite direction (Figs. 1 and 2).

Differences within pairs of matching conditions were further observed (e.g., BSD compared to WSI), both as a better SB initial effect (Fig. 2A) and during adaptation stage (i.e., comparing LA with EA) (Fig. 2B). As individuals are walking in their comfortable pace, reducing a belt’s speed is likely to keep them in their “comfort zone”, thus leaving room for appropriate gait modification. Increasing a belt’s speed, however, introduces a less manageable situation, in which one is asked to react to a perturbation that may exceed his/her physical limitations. Another explanation stems from left–right speed ratios. Increasing one belt’s speed by 50% results in a 2:3 speed ratio, while decreasing a belt’s speed by 50% results in a 1:2 ratio, a more aggressive intervention which may result in a more robust change to the gait pattern.

Interestingly, we found that SBTM walking produced similar trends of adaptation patterns in arm swinging asymmetry (Fig. 1B), presumably reflecting spinal inter-segmental neuronal synchronisation between upper and lower limb central pattern generators (CPG) [29]. Specifically, the BSD/WSI matching condition resulted in a more symmetrical arm swinging in the first 30 s after returning to TB setting (i.e., EPA), compared to BL. Input to the CPG is delivered partly by supra-spinal structures, such as the cerebellum and basal ganglia. It is likely that the split-belt intervention induces a neural change at the level of these structures [18, 21]. Moreover, the sub thalamic nucleus and dopaminergic motor-related systems exert less influence on the executive circuitry responsible for arm movements, but have a preferential action on those for the gait cycle of the lower limbs [23]. Such notion may serve as an explanation for the relatively higher level of dysfunction of the upper limbs in terms of asymmetry, relatively to the lower limbs in our cohort. On the other hand, gait adaptation is a form of supervised, error-based motor learning, attributed mainly for the cerebellum [24], which may explain the similar upper/lower limb trends.

None of the outcomes was significantly different in the LPA period, compared to BL. Moreover, the trends for both step length and arm swing asymmetries were consistently directed back towards BL values. This absence of post-adaptation long-term effects might be attributed to the relatively short SB duration and would warrant future studies aiming to establish a more beneficial and long-lasting therapeutic effect.

Two major limitations must be acknowledged. The first is the small sample size. This was an exploratory study and results should therefore be interpreted with caution. Given that persons with PD show high amounts of variability in their gait features, a larger scale investigation is warranted in order to make the results generalisable. Secondly, our sample was comprised of individuals with relatively symmetrical step-lengths. While this is not an optimal baseline to assess SB-induced asymmetry change, the similar trends observed in the upper limb suggest that similar effects would have been observed for initially-asymmetrical individuals. This, however, should be corroborated in future studies and potentially by including a control group(s). In the current study design, each participant provided their own baseline, resulting in paired comparisons of SBTM manipulation effects presumably providing a sufficient comparator. Still, future studies, that will include a control group, might be required to confirm our finding on the superiority of BSD among other asymmetrical patient groups. For example, involving a group of persons with lower limb amputation, whose observed asymmetry is of orthopedic origin rather than neurodegeneration. Our results, although exploratory, provide insights regarding strategies for the implementation of SBTM interventions when attempting to address gait asymmetry in persons with neurological deficits, for either research or physiotherapy [4, 7, 11, 13, 15, 16, 24]. Particularly, we posit that the speed of the belt of the less affected side should be decreased in order to obtain a stronger adaptation, and consequently a stronger post-treatment effect. The latter, however, could not be confirmed due to short adaptation and post adaptation periods used. Our results should also be corroborated on initially asymmetrical individuals, which unfortunately was not the case with our cohort.

Our results demonstrate that in order to achieve stronger SB adaptation effects on GA, the speed of the best side belt should be decreased, rather than increasing the speed of the worst side. Future clinical studies would benefit from considering such methodological planning of split-belt sessions for maximising intervention outcomes. In addition, studies with larger sample sizes are needed in order to further elucidate the residual influence of SB interventions on arm swinging. Finally, post-adaption effects, while not observed in the current investigation, must be further explored by means of longer-duration interventions, in order to maximise the therapeutic outcomes of SBTM-based treatments.