Parkinson disease is a progressive neurological condition that affects approximately 6 million people worldwide [1]. People with Parkinson disease can present with a variety of motor impairments such as tremor, slow movements, gait, and balance disorders, as well as nonmotor impairments such as reduced cognition, depression, anxiety, and sleep disorders [2]. Freezing of gait, defined as a brief, episodic, absence or marked reduction of forward progression of the feet despite the intention to walk, is a complex phenomenon that may also be present in people with Parkinson disease [3,4]. People with freezing of gait often describe their feet as being glued to the ground, which can lead to reduced mobility, falls, poor quality of life, and increased health care costs [4-8].

The pathophysiology underlying freezing of gait remains poorly understood, although it is suggested to result from dysfunction in neural networks across motor, affective, and cognitive domains [9-12]. People with freezing of gait are more likely to exhibit decreased gait automaticity and increased gait variability [13], as well as motor fluctuations and dyskinesia [14]. Furthermore, cognitive deficits and anxiety also appear to be more pronounced in people with freezing of gait than without [14-17]. Freezing of gait is also more frequent and severe in conditions of high levels of anxiety compared with low levels [18].

First-line treatments for freezing of gait consist predominantly of pharmacological interventions to maintain a good on state [4]. However, freezing of gait can persist despite these regimes. Nonpharmacological interventions are often used in conjunction with pharmacological interventions, with the most common being physiotherapy. Although several reviews have shown that these nonpharmacological interventions are effective, their results have been modest [19-23]. This might be because of the heterogeneity of the interventions, small study sample sizes, and limitations of relying on self-report to assess freezing of gait [21,24].

A recent systematic review performed subgroup analyses to determine which types of physiotherapy interventions may be the most useful in managing freezing of gait. The results showed that action observation training had a statistically significant effect on reducing freezing of gait [21]. In the 4 action observation studies, participants with freezing of gait watched videos of actors perform movement strategies designed to overcome freezing, followed by physical practice of these strategies under the supervision of a physiotherapist [25-28]. Observation of actions performed by others is understood to activate neural structures in the brain that execute the same actions, thus facilitating motor learning and performance [29]. This is then further reinforced by physical practice.

Although the results from these action observation studies were positive, the strength of this evidence was weak, and its clinical significance was unclear. This was because of the small number of moderate-quality studies included in the meta-analysis and the use of the New Freezing of Gait Questionnaire (NFOG-Q) [30] as the outcome, as this questionnaire was previously shown to be insufficiently responsive to detect small changes in freezing severity [24] and may not be a good indicator of the real symptom burden [31]. Furthermore, the implementation of action observation as an intervention for freezing of gait was limited in the following ways: (1) participants watched actors without Parkinson disease perform movement strategies, which might not be an accurate reflection of motor performance by people with Parkinson disease; (2) videos demonstrated generalized strategies that might not be relevant or appropriate for the individual; (3) videos showed the use of movement strategies to overcome freezing of gait in clinical settings, which might reduce ecological validity as patients typically present with freezing of gait at home [32]; and (4) participants watching videos on flat-screen devices could be vulnerable to distractions in their immediate environment, which might impact motor learning.

Video self-modeling, a form of observational learning that requires the observer to watch and learn from one’s own positive behavior, may be useful for people with Parkinson disease [33]. The activation of neural structures previously described is maximized when the observed actions are familiar to the observer and comprise movement that the observer is able to perform [29]. Therefore, we hypothesized that the use of video self-modeling would provide salient cues to improve motor learning and performance, as well as promote self-efficacy by strengthening beliefs in one’s ability to overcome freezing of gait [34]. People with freezing of gait may benefit further if they observe themselves using personalized strategies that address their specific motor, affective, and cognitive triggers of freezing. In addition, videos of situations at home that provoke freezing of gait (and strategies to successfully overcome freezing) may be of greater relevance. Furthermore, the use of a virtual reality head-mounted display (HMD) removes distractors in the environment, directing full attention to viewing videos.

Therefore, the aim of this study is to investigate the feasibility and acceptability of a home-based, personalized video self-modeling intervention delivered via a virtual reality HMD to reduce freezing of gait in people with Parkinson disease. The secondary aim is to investigate the potential effect of this intervention on freezing of gait, mobility, and anxiety.

A single-group pre-post mixed methods pilot trial was conducted from April 2019 to April 2020. Ethics approval was obtained from the University of Sydney Human Research Ethics Committee (project number 2018/893), and written informed consent was obtained from all participants. The trial was registered with the Australian New Zealand Clinical Trials Registry (ANZCTR12619000139178).

A total of 10 participants were recruited from existing databases of people with Parkinson disease at the University of Sydney and from Parkinson’s New South Wales support groups. The following inclusion criteria were used: (1) diagnosis of idiopathic Parkinson disease, (2) presence of freezing of gait (defined as having a score of ≥1 on question 2 and score of ≥2 on question 4 of the NFOG-Q), (3) stable dopaminergic medication regime for at least 4 weeks before commencing the study, (4) ability to walk independently with or without a walking aid, and (5) living in the greater Sydney metropolitan area. Participants were excluded if they had (1) any medical conditions that would interfere with the study safety and conduct, such as unstable cardiovascular disease and neurological conditions other than Parkinson disease; (2) cognitive impairment defined as having a score of <24 on the Mini Mental State Examination [35]; (3) newly commenced deep brain stimulation or changes in stimulation parameters within 6 months before participating in the study; and (4) significant head tremor or motion sickness limiting the ability to use a virtual reality HMD.

The intervention protocol (including printed information and instructions to the participants) is described in detail in Multimedia Appendix 1 [23,25,36-38] and illustrated in Figure 1. In brief, a physiotherapist (LG) delivered 6-8 home visits over 6 weeks, with each visit lasting approximately 60 minutes. The interventions were delivered, and practice was completed when participants were in their on phase, that is, when their medications were working optimally.

Before the first home visit, the physiotherapist reviewed the participants’ medical history, including disease and freezing of gait severity (NFOG-Q), mobility status, and the circumstances of any falls. The physiotherapist also evaluated the baseline results of participants’ Characterizing Freezing of Gait Questionnaire (CFOG-Q) to determine their personal motor, cognitive, emotional, and environmental triggers of freezing of gait, as well as any strategies previously used to overcome freezing [36].

During the first home visit, participants identified a situation at home, where freezing of gait was troubling. Examples of situations included turning around in tight spaces and walking through doorways while performing an additional motor task. The physiotherapist assessed the participants’ performance in this situation and worked with participants to develop a suitable personalized strategy to help them overcome their freezing. Examples of strategies included stepping in time to the beats of an external rhythmic auditory cuing device, such as a metronome typically set at lower than usual cadence; self-initiated movement strategies, such as counting, shifting weight from side to side, and/or simplifying complex tasks; and progressing by increasing the complexity of tasks where appropriate.

Once a situation was identified and a strategy developed and practiced by the participants, a 3D camera (Mirage, Lenovo) was used to produce 180° videos showing the participants using their personalized strategy to successfully overcome freezing of gait in the situation identified. The videos were edited to show the participants’ successful performance 3 times and lasted 2-5 minutes. The videos were created to be compatible with a virtual reality HMD (Oculus Go, Facebook Technologies). Participants were instructed on the use of the HMD and asked to watch their personal video using the HMD while seated twice a day, 5 days per week for 6 weeks. After one of the 2 daily video viewings, participants performed the physical practice of their strategy, as shown in their video for 10 minutes. Participants were also provided with a printed guide on how to use and navigate the virtual reality system.

The physiotherapist monitored participants’ progress during home visits. Once participants were able to consistently use their strategy to overcome freezing of gait in the first situation, a second situation was identified, and a second strategy was developed. A second video was created for each participant. This second video involved the participant performing the same task in a different environment, the same task with increased complexity, or a different task altogether. Strategies used to overcome freezing of gait were adjusted as appropriate depending on the situation, the person-specific freezing of gait triggers, and preferences. Participants were then asked to watch their second video and perform its associated physical practice 4 times a week, and the first video and its associated physical practice once a week. If participants continued to make further progress, a third situation was introduced, and this process continued for the duration of the intervention period. The introduction of additional videos was determined by the physiotherapist and was based on clinical judgment and participants’ self-assessment of their progress.

The primary outcome measures assessed the feasibility and acceptability of the intervention. Measures of feasibility included recruitment rate, retention rate, and adherence to the intervention (by recording the number of daily video viewings and physical practice) using self-report logbooks and adverse events associated with the intervention. Measures of acceptability included a modified Players Experience of Need Satisfaction (PENS) Questionnaire [39] and a semistructured interview at postintervention. In the modified PENS Questionnaire (Multimedia Appendix 2 [39]), participants were asked to reflect on their experience of using the virtual reality system, which comprised the HMD, the handheld control, and the experience of navigating and viewing the videos. Participants rated their interest or enjoyment, sense of presence or immersion, competence, and intuitiveness of the controls and rated 2 additional items that were added to the PENS. These were the presence of motion sickness and whether they would use the virtual reality system for the management of their freezing of gait in the future if it was available. The questionnaire included 18 questions across the categories listed above, with each question scored on a 7-point Likert scale, where 1=strongly disagree and 7=strongly agree (higher score is better). In the interviews, participants were asked to describe their experiences of the intervention, including both video viewing and physical practice (Multimedia Appendix 3).

Secondary outcome measures were collected to assess any potential effects of the intervention on freezing of gait, mobility, and anxiety. These measures were collected at baseline and postintervention. Participants also completed the Goal Attainment Scale, with goals in relation to managing freezing of gait at home set at baseline and goal attainment evaluated at postintervention [40].

To obtain freezing of gait measures, participants were videotaped by performing 2 freezing of gait provoking tests: the Ziegler test [41] and the turn-in-place test [42]. In the Ziegler test, participants began in a seated position 3.4 m from a closed door. Participants were asked to stand up, walk forward 1 m to a square outlined with tape on the ground (40×40 cm), perform two 360° turns (clockwise and counter-clockwise) within the square, and walk forward a further 2 m to open the door and walk through the doorway, before returning to sit in the chair. Participants were asked to perform the test as fluently as possible under 3 conditions in the following order: (1) no additional task, (2) with an additional motor task (ie, carrying a tray with a cup of water), and (3) with additional motor and cognitive tasks (ie, carrying a tray with a cup of water and counting backwards by 7 from 100). In the turn-in-place test, participants were asked to turn 360° on the spot, alternating right and left at a self-selected pace for 1 minute.

Two assessors (KAEM and JS) determined the percentage of time frozen for each freezing of gait provoking test, plus the time taken to complete the Ziegler test, via offline video analyses [43,44]. They were blinded to the baseline and postintervention testing conditions. A detailed protocol is described in Multimedia Appendix 4 [4,43]. Interrater reliability between the assessors was excellent. The intraclass correlation coefficients (two-way mixed effects, absolute agreement) were as follows: Ziegler test percent time frozen=0.962 (from analyses of 29 videos), Ziegler test duration=1.000 (from analyses of 29 videos), and turn-in-place test percent time frozen=0.984 (from analyses of 20 videos). The videos used to determine intraclass correlation coefficients were from baseline on and off performances of 5 participants where postintervention measures were not available.

Participants also completed the NFOG-Q, which reports the severity and impact of freezing of gait (range 0-28), and the CFOG-Q, where section 2 reports the frequency of freezing of gait triggers (range 0-48). Lower scores indicate less severe freezing of gait in both measures.

The mobility measures included comfortable walking speed (measured over 10 m) [45] and the Timed Up and Go Test [46] in single- and dual-task conditions (ie, counting backward from 100 by 3), with lower scores indicating better mobility. Anxiety was measured using the Parkinson Anxiety Scale (PAS; range 0-48), with lower scores indicating lower levels of anxiety [47].

Participants were assessed at baseline within a week before the start of the intervention and postintervention within a week of completing the intervention. The following demographic information was also collected at baseline: age, gender, severity of Parkinson disease using the Movement Disorder Society—Unified Parkinson’s Disease Rating Scale Section III [48] and the Hoehn and Yahr stage [49], cognitive function using the Trail Making Tests A and B [50], and current medication regimen. Assessments were conducted at a university laboratory when participants were in their on phase. The freezing of gait provoking tests (both Ziegler and turn-in-place tests) were also repeated in participants’ homes when they were in their off phase after 12 hours withdrawal of their levodopa medication overnight.

The presence of motion sickness experienced while using the virtual reality HMD and any falls, injuries, and fatigue during the intervention and assessments were monitored. Participants were asked to record any adverse events in a logbook and report the events to the researchers. At each home visit, the researcher questioned the participants about the occurrence of any adverse events.

Descriptive statistical analyses of feasibility and secondary outcome measures were conducted using SPSS Statistics for Windows, version 26.0 (IBM Corporation). Interview data were audio-recorded and transcribed verbatim by independent transcribers who were external to the study. NVivo 12 software (version 2, QSR International) was used to code the interview data, and inductive thematic analysis was used to interpret the results [51]. One researcher coded all of the interviews (LG) and 2 other researchers coded parts of the interviews (CGC and NA) such that all data were coded independently by at least 2 researchers. The codes were then compared and grouped to form the main themes through an iterative process. Any differences were discussed in depth until a consensus was reached among the 3 researchers.

Individual and aggregate participant background information is presented in Table 1. Overall, participants (9 males and 1 female) had a mean age of 70.6 years (SD 7.7 years), were diagnosed with Parkinson disease for an average of 13.3 years (SD 5.2 years), and had moderate to severe disease severity with a mean Movement Disorder Society—Unified Parkinson’s Disease Rating Scale Section III score of 37.3 (SD 13.3) [52]. All participants had moderate to severe freezing of gait (NFOG-Q range 10-24/28), and 5 participants had significant anxiety (PAS>14/48). A total of 6 participants were considered recurrent fallers (defined as having more than 2 falls in the past 12 months), with 2 participants experiencing particularly high rates of falls. A total of 2 participants had not fallen in the past 12 months (Table 1).

To the best of our knowledge, this pilot study is the first of its kind to examine the feasibility and acceptability of video self-modeling (delivered via a virtual reality HMD) plus physical practice to help people with Parkinson disease manage their freezing of gait. The results of this study showed that the intervention is a feasible and acceptable option for addressing freezing of gait in people with Parkinson disease.

This intervention was safe to deliver with no participants experiencing falls during the intervention or assessments, despite freezing of gait, meaning that participants were at a high risk of falls [37]. The physiotherapist who delivered the intervention supervised the physical practice during the home visits until the participant was deemed safe to practice either independently or with a trained carer. Future larger randomized studies of video self-modeling plus physical practice to address freezing of gait should consider investigating falls as an outcome given the potential for this intervention to ameliorate fall risk.

As virtual reality motion sickness (due to sensory conflicts between visual, vestibular, and proprioceptive inputs) was a possible adverse event, susceptibility to motion sickness was an exclusion criterion. Although all participants were screened for motion sickness, 1 participant gradually developed intolerance of the HMD after 1 week. This was because of an exacerbation of his head and neck dyskinesia when wearing the HMD, which contributed to motion sickness. The camera and virtual reality system used in this study were commercially available devices selected for their accessibility to clinicians and relatively low cost. Although virtual reality motion sickness may be minimized by improving the technical aspects of the viewing experience, such as 6 degrees of freedom tracking, using dynamic depth of field, and providing multimodal sensory information [54], these improvements would require further development and involve higher costs. Future research to determine the incidence and intensity of virtual reality motion sickness, as well as the impact of varying technical aspects and costs of virtual reality systems, is needed in this population. Nevertheless, the use of a flat-screen device was found to be acceptable to the participant who required it.

The retention rate and adherence to intervention were high in this study. Participants completed most of the prescribed video self-modeling and physical practice sessions, with adherence rates comparable with those of other home-based exercise interventions [55]. Interestingly, the adherence rate of video viewing was less than that of physical practice. It was likely that repetitive viewings of the videos were more tedious compared with physical practice and that participants placed more value on physical practice over video viewing, even though both aspects of this intervention were designed to be complementary. This interpretation is supported by interview data, where several participants suggested reducing the amount of video viewing but did not suggest reducing the amount of physical practice. The amount and distribution of video viewing and physical practice should be further investigated in future studies to determine the best combination to improve performance while remaining engaging.

When comparing the recruitment rate of this study with others specifically targeting freezing of gait [28,56-59], our recruitment rate was lower. This was likely because of our recruitment from a general database of people with Parkinson disease, where a large proportion of individuals did not meet the inclusion criteria for this pilot study. It is also possible that potential participants with freezing of gait were deterred by the time needed to commit to the study and unfamiliarity with the technology.

This intervention was widely accepted. Most of the participants considered the intervention appropriate for their needs and would choose to do this again in the future. Although video self-modeling may be a useful tool for learning movement strategies, its use in people with Parkinson disease should be carefully considered. When people with Parkinson disease view themselves, they can bring up powerful emotions that are both positive and negative. Video self-modeling may improve self-efficacy by providing an opportunity for individuals to see their potential to overcome their freezing of gait. On the other hand, other forms of observational learning may need to be considered for people with Parkinson disease who are uncomfortable viewing themselves. Future studies are required to determine whether video self-modeling plus physical practice is superior to more generic observational learning plus physical practice for addressing freezing of gait.

Difficulties operating and navigating the virtual reality system to view videos may hinder the learning of the movement strategies and reduce adherence to the intervention. People with Parkinson disease may have tremors and dyskinesia, which makes it difficult to use a small controller, especially when vision is eliminated. They may also have difficulties donning and doffing an HMD that requires multiple adjustments to fit securely. In addition, people with significant cognitive impairments are likely to have issues using and engaging with a complex virtual reality system. Simple and intuitive platforms that provide people with Parkinson disease with large icons for selection and sufficient time to respond facilitates navigation of the applications.

Overall, the interpretation of secondary outcomes was limited by the small sample size and the variability of the results. Although there were minimal changes in the severity of freezing of gait, this intervention may be effective in reducing anxiety in participants with high levels of anxiety. Of the 5 participants with significant levels of anxiety at baseline (PAS>14/48), 4 participants (P6, P7, P8, and P10) had significant reductions in their anxiety scores. One additional participant (P2), who did not meet the cut-off for significant anxiety, also demonstrated a significant reduction in his anxiety score (PAS from 9/48 to 0/48). This finding is supported by interview data, in which participants reported increased confidence in performing tasks where they previously experienced freezing of gait. They also reported feeling less anxious when they experienced freezing of gait and felt better equipped with strategies to overcome freezing when it was triggered.

Given that anxiety is a risk factor and a significant predictor of developing freezing of gait [18], reducing anxiety may alleviate freezing of gait by reducing the load on attentional or cognitive resources needed to control gait in people with Parkinson disease. Our results suggest that individuals with high levels of anxiety regarding their freezing of gait may be more suited and more likely to benefit from this intervention. It also suggests that personalized rehabilitation for gait impairments is required, where individual characteristics help determine intervention modality and delivery to achieve the best outcomes [60,61].

Interestingly, 2 participants in this study (P5 and P6) specifically identified tasks and set goals to overcome their freezing in their off phase, as they experienced minimal freezing of gait during their on phase when their medication was working optimally. Even though the intervention was completed when the participants were on, the effects of the intervention appeared to carry over to the off phase. In their interviews, both participants reported that they were able to successfully implement their personalized strategies when they experienced freezing and reported achieving their goals on the Goal Attainment Scale.

Given its episodic nature and the myriad of factors that can trigger or alleviate freezing of gait, the assessment of freezing is exceptionally challenging. Common outcome measures used to assess the severity of freezing of gait, such as the Freezing of Gait Questionnaire [62] and NFOG-Q, have limitations because of their subjective nature and poor responsiveness and should be used in conjunction with other freezing of gait measures [24,31]. Several objective clinical measures of freezing of gait have been developed to complement self-report measures but freezing of gait is still not reliably captured, especially in the home environment [32]. The freezing of gait measures used in this study provided a good indication of the participants’ severity of freezing but might not be sufficiently sensitive to reflect the frequency and severity of freezing in daily life. Given that this intervention was delivered in participants’ homes with a highly personalized approach, the development of protocols and technology to measure freezing throughout the day in the home environment using wearable sensors is urgently required.

Video self-modeling using an immersive virtual reality HMD plus physical practice of personalized movement strategies is a feasible and acceptable method of addressing freezing of gait in people with Parkinson disease. Future larger randomized controlled trials could explore the use of wearable sensors to measure freezing of gait at home for home-based intervention, the use of different platforms to deliver video self-modeling, and the impact of this intervention on freezing of gait, anxiety, and falls.