Clinical features of patients who might benefit more from walking robotic training

1 Introduction

Stroke is the leading cause of disability in the industrialized world causing long-time motor and cognitive disabilities. To restore the possibility of walking is one of the main goal of the multidisciplinary neurorehabilitation for improving quality of life of patients and their participation to the community life (Mauritz, 2009). Although motor rehabilitation is effective, there is the need to boost rehabilitation outcome: Donnan and co-workers refer to neurorehabilitation as the sleeping giant of stroke medicine (Donnan, 2013). In this line, a recent meta-analysis has highlighted that a training that is task specific, task oriented and with a sufficient intensity could be the best option to treat subacute as well as chronic patients after stroke (Veerbeek et al., 2013). Robots and electromechanical devices could help physiotherapists to provide an intensive walking training for non- ambulant patients in a safety manner, usually through an integrated body weight support system (Iosa et al., 2013).

During the past 20 years, robots rapidly increase their diffusion in the rehabilitation gyms despite the use remained between optimism of some clinicians and scepticism of some other ones (Morone et al., 2017). A recent updated Cochrane advised that robotic users on the importance of patient’s characteristic, highlighting that subacute patients should be a better target than chronic patients (Mehrholz et al., 2017). Other studies reported that the more severe is the motor deficit, the more helpful could be the robotic therapy (Morone et al., 2011, Morone et al., 2012). However, despite many studies compared conventional therapy versus robotic therapy (Mayr et al., 2007; Chua et al., 2016), and many others investigated the prognostic factors of conventional rehabilitation outcome (Weimar et al., 2002; Masiero et al., 2007; Paolucci et al., 2008; Lee et al., 2015), poor attention has been deserved in investigating if patients treated with robotic therapy have some specific prognostic factors, or these factors are the same of conventional therapy. Conversely, to detect patients who could benefit more from a specific robotic treatment is important both for ethical and economic purposes. To do that, it is fundamental to know the clinical features depicting a patient as a good candidate for robotic therapy.

The aim of the present trial is to identify the clinical characteristics of patients that could benefit from robotic walking training with respect to conventional walking training.

2 Methods

To identify the prognostic factors of patients undergoing robotic therapy and comparing with those of patients conventionally treated, we continued the enrolment of a Randomized Controlled Trial the results of which were previously published (Morone et al., 2011). The published data were related to 48 patients (24 per group). We planned to collect new data for arriving at about the double of the number of patients, arriving at 100 patients who completed the rehabilitation treatment (50 per group). All subjects with stroke admitted to the department of intensive neurorehabilitation hospital of Santa Lucia foundation IRCCS from January 2008 to March 2015 were screened. Inclusion criteria were: hemiplegia/hemiparesis in the subacute phase (see Table 1) with significant gait deficits (FAC <3) (Wade, 1996) caused by a first-ever stroke, lesion that were confirmed by computed tomography or magnetic resonance imaging, and age between 18 and 80 years. Exclusion criteria were: the sequelae of prior cerebrovascular accidents or other chronic disabling pathologies, orthopedic injuries that could impair locomotion, spasticity that limited lower extremity range of motion to less than 80%, sacral skin lesions, Mini-Mental State Examination (MMSE) score <24 (Wade, 1996) and severe hemispatial neglect, as evaluated by a neuropsychologist.

Subjects enrolled were randomly allocated (following a randomization list electronically generated) into the robotically assisted group (RG) or the control group (CG). Both groups received our standard rehabilitation treatment (nearly 3 h/d for 5 d/wk), starting immediately after admission into our neurorehabilitation hospital (for being admitted the patients should be medically stable). In particular, 2 daily physiotherapy sessions were performed by all groups. After 1 week postadmission, RG participants started to perform 20 robotic sessions (5 times per week for 4 weeks) instead of a second session of standard physiotherapy. This robotic session lasted 40 minutes, 20 of which were active therapy (the remaining 20 minutes were allocated for the patient’s preparation, parameter setting, and rest breaks if needed). The duration of robotic treatment was 4 weeks, in accordance with previous studies. The device used for robotic therapy was a Gait Trainer (GT, Reha-Stim, Berlin, Germany) (Hesse et al., 2000). The GT is formed by a system for BWS and a controller of endpoint feet trajectories. Each of the patient’s feet is fixed by straps to a plate moved by the GT engine, which simulates physiological swing and stance gait phases, by means of a crank and rocker system that includes a planetary gear system. The result is a gait-like movement, in which the values of the spatiotemporal parameters are selected by the physiotherapist and imposed on the patients by the device, whereas ioint kinetics and kinematics are not imposed but influenced by the spatio-temporal parameters selection (Iosa et al., 2011). During all robotic sessions, 1 physiotherapist assisted patient training and verbally encouraged the patient to perform the task with correct posture, when necessary. The step length was adapted following the anthropometric characteristics of the patients. The walking speed was selected to be around 1 to 1.5 km/h at the first GT session and was gradually increased as soon as possible in accordance with comfortable gait for each patient. BWS was 0% to 50% of body weight. Hands were positioned on the handrail solely for balance. The Body weight support (BWS) was selected and adjusted following a progressive decreament during rehabilitation, avoiding patient discomfort, and according to the patient’s ability to extend his hip and sufficiently bear weight on the paretic leg (Morone et al., 2011; Morone et al., 2012; Iosa et al., 2011). During the GT session, a rest period was possible, if required by patients.

The CG patients performed 2 daily physiotherapy sessions. One of them was dedicated to upper limb training and the other one to walking training. Considering the patient’s ability, this walking therapy was focused on trunk stabilization, weight transfer to the paretic leg, and walking between parallel bars or on the ground. As well as for robotic treatement, also conventional therapy has been adapted along the rehabilitation pathway in accordance with the occurred improvements, progressively increasing the weight transfer on paretic leg, the number of performed steps in each session, and the passing from walking between parallel bars to walking with mechanical walker, up to independent walking.

The standard physiotherapy, shared by both groups in one session per day, was focused on the facilitation of movements on the paretic side and upper-limb exercises, and improving balance, standing, sitting, and transferring. Hence, both groups performed the same amount of therapy and also the same amount of therapy dedicated to walking recovery.

The protocol was approved by the local independent ethics committee, and all participants provided written informed consent. The primary outcome measure of the study was walking ability (as measured by FAC), assessed after the 4 weeks of treatment and at the end of the entire rehabilitation program. The remaining outcomes included assessments of mobility function and ability level, evaluated by Trunk Control Test (TCT) and Barthel index (BI) (Wade, 1996). A physician, blinded to the treatment, took measurements immediately after random allocation (which occurred in the first week after admission), and at hospital discharge.

3 Results

A total of 110 subject affected by subacute stroke were enrolled and randomized. Ten patients were drooped for medical complications not related to rehabilitation therapy. Table 1 reports the demographical and clinical data for the entire sample and for the two groups of patients. At admission in the rehabilitation hospital, the parameters were not significantly different between the two groups. At the end of rehabilitation, RTG had significantly higher values in terms of FAC-score (p = 0.001), TCT-score (p = 0.012). The differences in terms of BI-score only approached the statistically significant threshold (p = 0.061).

Table 2 reports the results of Forward Binary Logistic Regression for the entire sample of 100 patients, using treatment (robotic vs. control therapy) as an independent variable. The parameters entered into the model were a higher BI-score at admission, a higher TCT-score at admission, a short time from stroke at admission in rehabilitation hospital, and the type of treatment, revealing higher chances of recovery for RG. The variance explained by these four factors was of 78%. The variables not entered into the model did not have a significant effect on the FAC-score at discharge (age: p = 0.533, affected side: p = 0.517, type of stroke: p = 0.199, sex: p = 0.722).

When the analysis was separately performed between the group RG and CG, different prognostic factors were highlighted, as shown in Table 3. For Control group, the BI-score at admission and time from stroke at admission entered into the model, explaining the 80.9% of variance. TCT-score at admission remained out of the model (p = 0.285). Conversely, for RG, the only prognostic factor playing a significant role was the TCT-score at admission, as shown in Table 3. Despite only one factor entered into the model, it explained the 67.3% of variance. BI-score at admission remained out of the model with a border line level of p (p = 0.072), as well as time from acute event and beginning of rehabilitation (p = 0.096). Figure 1 shows the FAC-score at discharge for the CG and RG divided by high (>15) or low BI-score at admission, and divided by high (>40) or low TCT-score at admission. The FAC-score at discharge was mainly dependent by BI-score at admission in CG, whereas in RG the TCT-score at admission was more important for obtaining different FAC-score at discharge, confirming the results of forward binary regression analysis.

OPEN IN VIEWER

Fig.1

the FAC-score at discharge for the Control Group (gray) and Robotic Group (dark) divided by high (>15) or low BI-score at admission (left), and divided by high (>40) or low TCT-score at admission (right). FAC, Functional Ambulation Category; BI, Barthel Index; TCT, Trunk Control Test.

The aim of the present trial was to identify the clinical characteristics of patients who could benefit from robotic walking training. We observed that a good outcome, in terms of functional ambulation at discharge (FAC-score), depended by BI-score at admission and time from stroke in patients who performed conventional therapy, whereas this dependency was not significant in patients who performed robotic therapy. So, the reduced autonomy was a negative prognostic factor for conventional therapy, as already reported by previous studies (Paolucci et al., 2008), whereas it was not for robotic therapy. The outcome of robotic intervention was affected only by the capacity of patients of controlling their trunk.

More in general, the results of our study also confirmed that subjects performing robotic intervention in add-on to standard therapy had more chances to regain autonomous ambulation. Considering that patients enrolled in our study were in subacute phase of stroke, these findings resulted in line with those highlighted by a recent update Cochrane including 36 trials involving 1472 participants (Mehrholz et al., 2017). As noted in that Cochrane review, the effect of the different devices is still unclear. Until now there are two most important categories of walking robot or electromechanical devices reflecting two different biomechanical and rehabilitative approaches for walking training (Masiero et al., 2014): end effector approaches and exoskeletons. In a recent review we have suggested how the different machines should be used in a different frame or conditions of patients affected by stroke (Morone et al., 2017). This idea originated from our first trial in which we found a significant difference in the efficacy of robotic training with respect to conventional therapy only in most severely affected patients (Morone et al., 2011; Morone et al., 2012; Morone et al., 2014). The present results confirmed these findings, and added the information for identifying the best candidates for an effective robotic gait training: most severely affected patients who maintained a certain degree of trunk control. Trunk competencies are crucial for walking ability (Iosa et al., 2014; Sawacha et al., 2013) and also for reduce risk of falls in overground walking (Kao et al., 2014; Morone et al., 2014). The trunk, in fact, has a role in attenuating movement-related forces that threaten to challenge the body’s postural control system (Cole et al., 2017). In add, these results may reflect the specific kind of electromechanical device used in our study (Gait Training), that do not provide strict constrictions, except the harness of BWS system that is the conditio sine qua non for training patients not able to walk (Iosa et al., 2011). The other side of the coin is that due to this degree of freedom, subjects that do not have a sufficient trunk control may have a reduced benefit from robotic gait training. It is important to note how the outcome of electromechanical assisted training, differently from conventional gait training, is less influenced by the general initial level of disability. This is an important result that stress the usefulness of these devices not for all subjects, but in a selected kind of patients in which conventional therapy could be less effective (Morone et al., 2017). Our observation is similar to the result obtained by Duret et al. in the study regarding robotic arm training, in which the efficacy of arm robot training was correlated with severity of the disease. (Duret et al., 2015).

Our results should be considered at light of some limits of our study. The first one is the using of functional ambulatory category score as dependent variable. Despite it was previously used also in other studies about prognostic factors in stroke rehabilitation (Viosca et al., 2005; Masiero et al., 2007), it may suffer by substantial ceiling effect (Lord et al., 2004). Further research should use a more robust parameter, such as for example the timed results of ten meter walking test or six minute walking test. Another limit of our study is related to the use of binary regression analysis: despite it is the most common analysis performed to identify prognostic factors in stroke (for example: Weimar et al., 2002; Paolucci et al., 2008; Lee et al., 2015), it implies the need of dychotomizing both dependent and independent variables, with the risk of reducing sensibility of the analyses and also to introduce some biases related to the selection of the dychotomization thresholds.

Our results showed that robotic assisted walking training in stroke is a promising option especially for most severely affected patients that have few chances to perform an intensive overground walk-oriented training.

5 Conclusions

Robotic assisted walking training in subacute stroke is an effective intervention that improves walking recovery, even in patients most severely affected, even more in those who maintained a preserved residual trunk control. In particular, our results demonstrated that target population for walking robotic training is more severe affected subject. Differently from conventional training severity of disease and time from stroke to beginning of rehabilitation does not affect robotic training efficacy.

Funding

No funding to be reported.

Conflict of interest

None of the authors has a conflict of interest to declare.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from both the healthy individuals and the patients’ legal guardians participating in the study.