Treadmill Training with Cues and Feedback Improves Gait in People with More Advanced Parkinson’s Disease

Subjects and study procedures

PD patients were retrospectively identified among those hospitalized from January to December 2016 in our Department of Parkinson’s disease, Movement Disorders and Brain Injury Rehabilitation (“Moriggia-Pelascini” Hospital—Gravedona ed Uniti, Como, Italy) for a 4-week intensive gait rehabilitation.

Inclusion criteria were: (i) diagnosis of “probable” idiopathic Parkinson’s disease according to Gelb et al. [35]; (ii) mild to advanced stage of disease (stage 2-3-4) according to the HY scale [36]; (iii) ability to walk on a treadmill with no or minimal physical assistance; (iv) visual and hearing capacity sufficient to perceive cues and feedback; (v) no cognitive impairment (mini-mental state examination score >26) [37]; (vi) stable pharmacological treatment in the 4 weeks before admission and until the end of the training period. Exclusion criteria were: (i) diagnosis of atypical or vascular Parkinsonism; (ii) presence of cardiovascular, orthopedic, neuropathic, musculoskeletal and vestibular disorders affecting patient’s gait or balance; (iii) severe dyskinesia (iv) inability to walk on the treadmill for one minute at the belt-speed of 1.5 km/h without support.

Demographic and clinical data were derived from medical records and shown in Table 1. Data about the presence of Freezing of Gait (FOG) at baseline, evaluated by means of the Freezing of Gait Questionnaire (FOGQ) [38] were also available; the frequency of patients with FOG in our sample (patients with FOGQ score >0, indicated as FOG+ and expressed as percentage) is reported in Table 1.

On-going PD pharmacological treatments (dopamine agonists, levodopa and/or monoamine oxidases inhibitors) were converted in levodopa equivalent dose (LED), according to validated conversion tables [39]. LED was recorded for each patient at T0. No change in the total amount of dopaminergic therapy was made during the treadmill training.

Treadmill training

We used a motorized medical treadmill (Gait Trainer 3 Biodex, Biodex Medical System – 20 Ramsay Road, Shirley, New York, USA) for gait training. All patients underwent a 30-minute of treadmill training per day, divided in two 15-minute sessions, 6 days a week, for 4 weeks, with the supervision of a physiotherapist expert in movement disorders rehabilitation. Treadmill training was delivered independently from the assumption of medication. The belt speed was initially set at 1.5 km/h and progressively increased to a maximum of 3.5 km/h, depending on the physical ability of each patient. Patients were allowed to use handrails during the training, if needed. During the training visual and auditory cues and feedback were used and no weight support was provided. The visual cue was a target, delimited by two horizontal lines displayed on a screen which the patient had to reach with the stride. The software device calculated the space between the two lines for each patient, according to gender, height and age by means of an algorithm based on available normative data. The right and left footprints were shown alternatively on the screen: when they fell within the two lines, a “well done” message appeared; otherwise, patients were invited to take longer or shorter steps in order to adapt the stride length to the set target. The auditory cue and feedback consisted of musical beats with a frequency of 0.5 c/s synchronized with the visual cues. Figure 1 shows a treadmill-plus setting (A) and a detail of the screen, where visual cues and feedback are given (B).

OPEN IN VIEWER

Fig.1

A representative image of the treadmill-plus setting, showing where the screen is situated respect to the scrolling belt (A), and a detail of the screen during a training session, in which the visual cues and feedback and the recorded parameters are displayed (B).

Besides treadmill-plus training, which was the only treatment proposed for the rehabilitation of gait, we also provided daily sessions of front-to-front physiotherapy, aiming at the improvement of joints range of motion, occupational and speech therapy, lasting one hour each.

Outcome measures

The feasibility of the treadmill-plus training intervention was assessed by recording adverse events (e.g., falls, extreme fatigue, changes in PD symptoms) throughout the 4-week intervention phase. We also quantified the levels of recruitment, attendance, attrition rates and therapist compliance.

In order to standardize data collection, at T0 and T1, all clinical and instrumental evaluations were performed in the morning, one hour after taking the first dopaminergic drug dose, in medication “on” state.

Clinical data and gait parameters were collected on the same day at baseline, before the beginning of the training (admission, T0), and at the end of the 4-week rehabilitation period (discharge, T1). While treadmill training was provided until the day before the discharge, the 6MWT at T1 was performed on the last day of hospitalization, just before the discharge, in order to avoid a short-term effect of the last training on the “on-land” performance.

Clinical assessment

A neurologist specialized in movement disorders performed neurological examination of all patients and assessed the Unified Parkinson’s Disease Rating Scale (UPDRS) score [40]. A specialized physiotherapist supervised and collected results of the 6-minute walking test (6MWT) [41].

The total distance (evaluated in meters) covered during the 6MWT was chosen as the primary outcome measure, as this has previously been shown to be a useful severity measure of patient’s self-paced gait [41, 42]. The 6MWT was conducted according to the following standardized procedure: the patient was first familiarized with the test by going once forward and backward along a straight 15-m line on the ground. Then, after a 15-minute rest, patients were instructed to walk from end to end of the line for 6 minutes and to cover as much distance as possible, independently selecting the most comfortable walking speed. No encouragement was offered during the test and no cues were used. Patients were allowed to use their habitual walking aids during the 6MWT, if needed. As secondary clinical outcome measure, total UPDRS and UPDRS III were taken into account.

Gait parameters

Gait parameters were collected using a motorized treadmill developed for gait training (Gait Trainer 3 Biodex, Biodex Medical System – 20 Ramsay Road, Shirley, New York, USA). The device is equipped with four strain gauges under the belt, sensitive to the bending of the belt caused by the weight applied by the patients’ walking. These gauges are connected to a software which enables quantification of several gait parameters. The gait evaluation was performed in the same setting conditions, both at T0 and T1, and it consisted in one-minute walking on the treadmill, without handrail support and without cues and feedback, at the scrolling belt speed of 1.5 km/h; this speed was expressively selected in order to highlight the spatiotemporal gait pattern alterations. Indeed, variables such as stride length and cadence are affected by gait velocity [43]; with lower velocities associated with shorter stride length and greater gait variability [31]. The following gait parameters were collected and included in the study as secondary outcome measures: the right and left step length (expressed in centimeters and calculated by averaging the length of all the corresponding steps), the coefficient of variance (CoV) of both right and left steps (expressing the standardized amount of variation occurring between footfalls, calculated as the ratio of the standard deviation to the mean value, in percent unit), and the step cycle (in cycle/s, average of all the step cycles of the exercise).

Statistical analysis

The Shapiro–Wilk statistic supported by visual inspection was used to test the normality of the distribution of all variables. Age, disease duration and amount of levodopa-equivalent dose did not violate the normality assumption. Accordingly, descriptive statistics are reported as mean±SD and parametric statistics (one-way analysis of variance) were used for between group comparison of these variables.

Conversely, the outcome measures violated more or less severely the normality assumption. Accordingly, descriptive statistics are reported as median (lower quartile, upper quartile) and between- and within-group comparisons were carried out by the Kruskal Wallis test and Wilcoxon signed-rank test, respectively.

Descriptive statistics for discrete variables are reported as number (percentage frequency) and comparisons were carried out by the Chi-square test or Fisher exact test, when appropriate.

A significant result for Kruskal Wallis test or parametric ANOVA, when comparing the three groups, was followed by post-hoc analysis (Dunn-Sidak and Tukey-Kramer, respectively), to compare pairs of groups.

When appropriate, false discovery rate was controlled at 5%, using the Benjamini-Hochberg method.

To assess the effectiveness of rehabilitation, the value of each outcome variable at discharge was compared with its value at admission in all patients and in the three HY groups.

To test whether rehabilitation effectiveness was dependent on disease severity, we considered the difference between values at discharge and values at admission in the outcome variables and we ran a one-factor non-parametric analysis of variance (Kruskal Wallis) on the HY factor (three groups).

The association between variables was assessed by the Spearman correlation coefficient. Since a high degree of association between HY stage, age and years of disease was expected, multivariable regression methods were used to investigate whether changes in outcome variables were associated with HY after adjustment for these covariates. All statistical tests were two-tailed and statistical significance was set at p < 0.05. All analyses were carried out using the SAS/STAT statistical package, release 9.2 (SAS Institute Inc., Cary, NC, U.S.A.).

Description of the sample

After being selected according to inclusion and exclusion criteria, the study population consisted of 206 patients, whose demographic and baseline clinical data are reported in Table 1.

Thirty-six patients were excluded because of a diagnosis of: i) vascular or atypical Parkinsonism (n = 12), ii) concurrent medical conditions for which treadmill training was contraindicated (n = 8), iii) cognitive impairment (n = 6), iv) severe dyskinesia (n = 5), v) inability to walk on treadmill for one minute at the belt-speed of 1.5 km/h without support (n = 5).

Of the 206 patients, 79 were in stage HY2, 74 in HY3 and 53 in HY4. As expected, the three groups differed for age, frequency of freezers (FOGQ >0), disease duration and amount of LED, with progressive worsening in more advanced stages. However, post-hoc analysis revealed that disease duration and LED were not significantly different between HY3 and HY4 patients. No differences in gender were observed.

Table 1 reports baseline values of primary and secondary outcome variables at baseline for the sample as a whole and stratified in 3 groups according to HY stage. As expected, a significant difference between groups at T0 was observed for all parameters (p < 0.0001 all), with worse values in higher HY stages; in particular, post-hoc comparison showed that comparing HY2 to HY3 and HY2 to HY4, all variables were significantly worse at a higher HY stage. Conversely, in the comparison between HY3 and HY4 several variables did not differ significantly (step cycle, left step length, left and right step CoV).

Age was significantly associated with the 6MWT and with treadmill-related gait parameters at baseline (spearman r: 6MWT = –0.51, step cycle = 0.30, left step length = –0.30, right step length = –0.29, left step CoV = 0.19, right step CoV = 0.18, p < 0.01 all). A slightly weaker association between disease duration and outcome measures was observed (spearman r: 6MWT = –0.31, step cycle = 0.16, left step length = –0.22, right step length = –0.13, left step CoV = 0.25, right step CoV = 0.26, p < 0.05 all).

No right-left side difference was recorded in step length and CoV in the whole sample (p = 0.21 and p = 0.26, respectively).

Treadmill training improves both self-paced and slow-speed gait in advanced PD

All subjects performed baseline treadmill evaluation and no dropouts occurred during the 4-week training, with an attrition rate of 0% and a 100% attendance recorded of all the scheduled training sessions; therefore, all 206 patients completed the study. No training-related adverse events were reported, such as severe fatigue or exacerbation of PD symptoms. Both FOG+ and FOG- patients could undergo treadmill training. No difficulties emerged regarding physiotherapist compliance to exercise supervision and data collection, which was complete for all patients. The study did not require any supplementary cost, as all the equipment was already available in our rehabilitation department.

Table 2 reports the difference of outcome variables (values at T1- values at T0) in the whole population and in the three groups.

After rehabilitation all outcome variables significantly improved, considering the whole population of the study (p < 0.0001 after adjustment for false discovery rate). The same result was obtained stratifying the patients according to the HY stage, with a significant improvement with respect to baseline in all disease stages (all p < 0.0001, except left CoV showing p = 0.003, p = 0.0004 and p = 0.003 for HY2, HY3 and HY4 respectively and right CoV showing p = 0.003, p = 0.029 and p = 0.003 for HY2, HY3 and HY4 respectively). Figure 2 shows the difference (values at T1- values at T0) of the gait parameters in the three groups, highlighting that a significant improvement was recorded after treadmill treatment both in gait endurance, evaluated with the 6MWT (Fig. 2A) and in gait regularity (expressed with the improvement of step length, cadence and step CoV, Fig. 2B-F), both in early and in more advanced disease stages.

OPEN IN VIEWER

Fig.2

Baseline (T0) and discharge (T1) gait parameters demonstrate the effectiveness of treadmill training on gait disturbances in every disease stage. ^*p < 0.05 compared to the respective baseline value.

Finally, comparing the improvement after treatment among HY stages in outcome variables, Kruskal Wallis ANOVA revealed a significant difference only for total UPDRS (p = 0.01). Post hoc analysis revealed that the only significant difference in effectiveness was between HY2 and HY4 (p < 0.01), with better improvement in HY4 patients. No significant differences in treatment efficacy were observed in all remaining variables (p values ranging from 0.09 for changes in left step length to 0.66 for changes in the 6MWT).

Investigating the relationship between the improvement in outcome variables and HY stage, age, sex and disease duration, none of these variables was independently associated with the amount of improvement, with the exception of total UPDRS, where disease duration was the only independent predictor of improvement (p = 0.003, longer disease duration associated with better improvement), and HY stage became borderline significant.

Study limitations

This is a retrospective study analyzing a prospectively collected database. Even though the lack of a prospective randomized controlled trial design might constitute a potential limitation, data collection procedures were carefully designed and all relevant information was recorded on a structured protocol.

The lack of a direct comparison between traditional treadmill and treadmill-plus training does not allow quantifying the specific contribution of cueing techniques to the observed gait improvement; further studies will be planned to compare the two different strategies.

No follow-up data are available at the moment, thus limiting our knowledge about the duration of gait improvement. Another limitation arises from the analysis of gait parameters performed using the gait trainer belt sensors. Compared to over-ground walking, in fact, treadmill walking induces slightly shorter stride lengths and slightly higher cadence [57], particularly in PD patients [58]. However, all subjects were tested in the same experimental setting; the observed differences in spatiotemporal gait parameters among groups should not be affected by this bias.

Finally, another limitation concerns the speed of the belt scrolling that we set at 1.5 km/h: as already demonstrated [59], spatiotemporal gait variability depends on self-generated walking speeds. Thus, the selected speed of the belt scrolling could have influenced data about the coefficient of variation in this study.

Conclusions

In conclusion, our study supports the feasibility and the usefulness of a treadmill-plus training in patients at more advanced PD stages. The improvements in spatiotemporal gait parameters found using a slow-speed treadmill were recorded together with better performances in the 6MWT at the end of the training. Further studies will be needed to clarify the mechanism underlying the effectiveness of treadmill-plus training on gait disorders, to define its long-lasting effects and its role within a tailored rehabilitative protocol for advanced PD patients.