The effects of augmented visual feedback during balance training in Parkinson’s disease: study design of a randomized clinical trial

This study is a pilot RCT comparing two treatment groups of patients with PD (Figure 1). Patients are allocated to either a five-week training program with balance exercises containing augmented VF (experimental group), or a five-week balance training program that follows existing guidelines (control group). Assessments will take place before intervention (T0), at six weeks (T1), and at twelve weeks (T2). Intervention assignment will be concealed by drawing opaque sealed envelopes by an independent investigator not involved in the study. A blinded researcher will perform the assessments. Patients will be asked to refrain from mentioning the nature of the intervention they receive to the researcher. All assessments will be performed in the ON-phase of levodopa medication. To minimize the effects of (changes in) PD medications on the patients’ physical performance, the time of day at which patients are assessed will be kept constant throughout the study.

The protocol has been approved by the Medical Ethics Committee of the VU University medical center (VUmc) Amsterdam and is registered with Current Controlled Trials under ISRCTN47046299.

A convenience sample of 36 patients will be recruited from the Neurology Department and from patient databases of the Department of Rehabilitation Medicine, VUmc. Patients who are likely to meet the criteria for inclusion will be invited to participate in the RCT.

Inclusion criteria will be (i) a diagnosis of idiopathic PD, mild to moderate stage (i.e. Hoehn & Yahr stages II and III), (ii) able to participate in either of the training programs, and (iii) written and verbal informed consent. Exclusion criteria are the presence of neurological, orthopedic, or cardiopulmonary problems that can impair participation, insufficient cognitive function (Mini Mental State Examination, MMSE < 24), an unstable medication regime, and any condition that renders the patient unable to understand or adhere to the protocol such as cognitive, visual, and/or language problems.

The intervention will contain ten treatment sessions of 60 minutes each over a period of five weeks. In order to efficiently organize the training, different workstations related to standing balance will be organized in a circuit allowing six patients to train simultaneously. To capitalize on the benefits of action observation [24] patients will work in pairs at each workstation. Patients will take turns performing the exercise while the other person observes and/or rests. The training paradigm will be applied to both the experimental and the control group.

In the experimental group VF is explicitly integrated in each workstation (see below for details). In the control group workstations will consist of balance exercises that follow the current guidelines for physical therapy in PD [9]. The exercises in both groups will focus on controlling body posture in forward and sideways direction, exploring limits of stability, weight-shifting, sit-to-stand exercises, and dual-task exercises. Which specific workstation will be used for which treatment session will be decided before the start of the training program. Two expert therapists (CdG and IB) will define training goals, monitor the training intensity during the sessions, and keep time. Throughout the training program, the therapists will monitor individual progress and progressively adjust personal training goals. If desired, exercise complexity and/or workload will be increased throughout the sessions. All training sessions will take place in an outpatient setting of VUmc. All participants will be asked to keep a training and fall log during the duration of the training program. Interventions will be rated by the participants in terms of the perceived exertion over the entire training session.

The content of each treatment session will be controlled for type and duration as shown in Table 1. Warming-up and cooling-down exercises of about 5 minutes will be carried out as a group at the beginning and conclusion of each session, respectively.

In the experimental group VF will be integrated explicitly in each exercise. Three workstations (Motek Medical, Amsterdam, The Netherlands) will be set up in a gym. Figure 2 illustrates the key features of these workstations. They consist of a conventional PC with a 42" flat-panel LCD monitor and movement registration hardware. The latter will be used to map body movements to movements of an object (‘avatar’) on the monitor, by which patients can interact with balance games that are run on the workstations. Six different games will be used, each creating different challenges for the patient within a different virtual environment. Four games focus on controlling body position in space, one on stepping movement, and one on performing a sit-to-stand transfer movement. More accurate and/or faster control of the avatar results in accrual of points.

Movement registration will be accomplished using inertial sensors (Xsens, Enschede, The Netherlands) and a force plate (ForceLink, Culemborg, The Netherlands). Inertial sensors are attached to the patient’s chest using neoprene bands with a snap-on system, providing data about acceleration and rotation of the upper body. Monitoring the sit-to-stand transfer also requires an inertial sensor at the upper leg. The sensors can be easily transferred between patients throughout the training session. The force platform is used to obtain data about center-of-pressure displacements (COP).

The user interface on the workstations is deliberately kept simple and can be operated by a conventional mouse. Patients will receive detailed instructions about the operation of the software including the adjustment of game difficulty. They are encouraged to operate the workstations as autonomously as possible.

Participants in the control group will receive equally-dosed balance exercises according to current guidelines for physical therapy in PD [9].

All assessment sessions will take place in a motion laboratory at VU University Amsterdam, Faculty of Human Movement Sciences. During each session the clinimetric assessments will be performed first, followed by combined posturographic and EEG recordings. An assistant will stand by at all times to prevent falls in the case of a loss of balance.

The clinimetric test battery has been tested extensively in a previous study [27]. The tests will be carried out in accordance with current guidelines [9, 40].

An illustration of the combined posturographic and EEG recordings is provided in Figure 3. The patient will be asked to stand on a 600 mm × 400 mm force plate embedded in the floor (Kistler 9281B; Kistler, Ostfildern, Germany). A monitor providing feedback of the COP position is positioned at eye-height at a distance of ~75 cm from the subject. After familiarization with the setup the patient will perform twelve trials (4 conditions × 3 repetitions), with the possibility to rest after each trial. Each trial will consist of a sequence of 20 s quiet stance, 100 s rhythmic tracking, and 20 s quiet stance. During the rhythmic tracking segment a target will oscillate horizontally on the monitor at a frequency of 0.5 Hz. The patient is asked to match the target as accurately as possible by making whole-body swaying motions from side to side. VF of the COP will be provided by means of a black bar on the monitor. The experimental conditions during the swaying task will include conditions in which VF is withheld, is presented real-time, or is presented with a finite, constant delay of 250 or 500 ms (see Figure 3). From previous studies it is known that VF which is not provided to the subject directly (i.e. real-time) but with a finite delay, may destabilize performance on (tracking) tasks [41‐43]. In that case, the better a subject is able to decouple from the VF (i.e., the less the subject relies/depends on visual information), the better he/she is able to perform the task. It should be noted that both the task and the VF during assessments will differ from what the patients experience during the training sessions.

During the postural tasks, 64-channel EEG recordings will be collected from Ag/AgCl electrodes that are placed on the scalp by means of a nylon head cap (TMSi, Enschede, The Netherlands). All signals will be high-pass filtered at 0.1 Hz and low-pass filtered at 100 Hz before being sampled at a rate of 2048 Hz (REFA amplifier, TMSi, Enschede, The Netherlands).

Anatomical MRI scans of the brain will be used to substantiate off-line source localization of the EEG signals. If a patient’s scan is not available, a new scan will be taken at VUmc, Department of Radiology.

We calculated the sample size required to obtain sufficient power for finding a difference of 9 cm on the FRT. Pooled estimate of the common variance, calculated from data by Ashburn [44], was entered in a formula for two independent groups with paired observations [45]. A total sample size of twenty-four patients is necessary to find a difference between groups with at least 80% power. To account for smaller effect sizes and an estimated attrition rate of 10%, we plan to include thirty-six patients in total, eighteen in each group.

Group differences in baseline patient characteristics will be tested using χ²-tests for categorical data and t-tests for continuous variables. Continuous outcome measures that are measured at all three time points will be tested for normality using the Shapiro-Wilk test. If assumptions for parametric testing are met, a mixed design ANOVA with between-subjects factor group (experimental vs. control) and within-subjects factor time (T0, T1, and T2) will be used to analyze the differential effect of the two interventions and the effect of time. If normality is rejected, or if outcomes are measured on the ordinal level, change scores between T1 – T0 and between T2 – T1 will be subjected to nonparametric Mann–Whitney U-tests. P < 0.05 will be set (two-tailed) as threshold for significance. Relative phase will be assessed through circular statistics using, in particular, Rayleigh tests to analyze group differences of relative phase uniformity between sources [46]. Correlation coefficients will be compared to investigate the association between outcomes.

Statistical analyses regarding EEG outcomes will be conducted using the methods outlined by Houweling and co-workers [47], with adaptations to EEG data regarding the accompanying lead field. Sources will be determined via pseudo z-scores of the beamformers that will be tested for within-group consistency through permutation tests [48]. The (relative) phases of the source projected EEG activity in the beta frequency band will be assessed through circular statistics (see above).

Posturographic analyses of balance using a setup of force platform with VF enable the quantification of postural performance and allow for a detailed investigation of its dynamical structure. In the past, researchers have characterized postural sway as chaotic [50, 51], purely stochastic [52], or fractional stochastic [36, 53‐55]. Analyses that quantify the non-linear and stochastic temporal evolution of postural sway have been used to assess the effects of health status and rehabilitation in a number of clinical studies [56‐58]. In PD it has been found that postural control mechanisms are characterized by an increase in random fluctuations (short- and long-term diffusion coefficients) in the mediolateral direction [37]. These measures are associated with a history of falls and poor performance on clinical measures of balance. Due to its quantitative and unbiased nature, stabilogram-analysis will enable us to test specific hypotheses involving outcome measures that may be much more sensitive to changes in the control exerted during the postural tasks. Combining the VF setup with a delayed feedback paradigm [41, 43] can characterize the extent to which subjects are coupled to the VF. As such we anticipate that these outcomes will form an important adjunct to the clinical outcomes of the proposed RCT.

The positive findings associated with movement strategies that rely explicitly on external stimuli seem to reflect the extent to which the central nervous system capitalizes on the potential to reroute or restructure functional networks. For instance, some of the movement-related neural activity might be rerouted to different brain regions or networks such as the premotor cortex and the cerebellum [12, 13, 59]. Other researchers have pointed out that internally and externally guided movements may require distinct neural processing [10, 11, 60‐62]. For instance, movements that are externally guided using visual information seem to preferentially activate neurons in the premotor cortex [61], an area which receives visual information directly from the visual cortex and projects directly to the spinal cord (and hence participates in both the visual and the motor network).

Alternatively, it has been suggested that giving explicit feedback or displaying environmental cues may involve attention-related mechanisms [49, 63]. This is supported by the finding that some of the benefits of cueing-based therapy have been shown to carry-over to situations without cues [15, 16, 63, 64]. This observation is difficult to explain if it were the availability of cues alone that is instrumental to functional improvement.

The present proof-of-concept trial will certainly help to highlight benefits of a balance-training program based on augmented visual feedback. How improvement is achieved in terms of altered motor control will be determined through the co-registered posturographic recordings. We believe that the complex nature of PD calls for a multimodal assessment approach to unravel the underlying mechanisms that influences the effects of VF on the quality of balance control.