The present study is an assessor-blinded monocentre parallel-group randomized controlled trial (ClinicalTrials.gov identifier: NCT04467554) that obtained ethical approval (Ethische Toestingscommissie Jessa Ziekenhuis, Belgian registration number; B2432020000014). To improve transparency we report this study according to the CONSORT guidelines for pilot or feasibility trial (25)
Recruitment, randomization and blinding
Participants were eligible when 1) suffering from first stroke, 2) more than six months ago, 3) 18 years or older, 4) had an impairment in trunk function (≤19 on trunk impairment scale (26)), 5) were able to maintain a seated position independently for more than 10 seconds, 6) were able to travel to the study location, 7) had no significant comorbidities other than stroke affected trunk function 8) having sufficient cognitive and language capacity to perform and understand study protocol participation, and 9) provided written informed consent. Participants were excluded if they did not meet the inclusion criteria.
Participants were recruited between July and November 2020. Leaflets and posters with study information and contact details were distributed in rehabilitation centres and at practices of physiotherapy located near the study location. Written approval was given by the potential participant to be contacted by the investigator (signed informed consent for contact). One investigator contacted potential candidates for further explanation of the study. After confirming eligibility, a signed informed consent was obtained.
The study took place in a dedicated room in an outpatient centre in Belgium. For this study, we aimed to recruit 30 participants in the chronic phase after stroke. Because of the pilot design of this study a sample size calculation was not performed. However, according to previously conducted trials with similar design, and recommendations by Whitehead et al (27), a sample of 15 participants in each arm of the trial was considered in order to be able to answer the research question.
The principal investigator (GV) randomly allocated participants, after consent, to two different groups: the experimental group and the control group. The principal investigator (GV) used the coin flip randomization method (28) without having any contact with the therapist or participants. Allocation was concealed. The information concerning the group allocation was provided to the therapist (EV). Therapist (EV) and participants were aware of the allocated groupings. The assessor and data analysist (LT) was blinded throughout all the assessments (three measurement points) and analysis.
Both groups received usual care comprising of physiotherapy and/or occupational therapy with strength exercises, conditional training, and task-oriented therapy. The usual care intensity was, on average, 3 sessions of 30 minutes to 2 hours therapy per week.
Participants in the control group received usual care only. No therapy time was spent on sitting balance therapy.
In the experimental group, participants received usual care plus additional technology-supported sitting balance therapy. The experimental therapy consisted of 12 one-hour individual sessions within four weeks at a ratio of three to four times per week. Each session consisted of 42 minutes of active sitting balance and trunk training and 8 minutes of cooling down in seated position. Time of intervention was monitored using a stopwatch and excluded all rest periods and set-up time. Sitting balance therapy was conducted in a seated position and consisted of predefined, standardised exercises including reaching training, lateral trunk lengthening and shortening, weight-shift training, pelvic tilt and training while sitting on an unstable surface. One therapist (EV) trained all participants. The therapist scored safety during and at the end of each training session. Participants rated tiredness of leg and trunk muscles after each session. When training was scored safe and tiredness was moderate, that is, scoring less than five out of ten points on a fatigue visual analogue scale, training difficulty was increased to the next level, according to a standardised scheme. Additional file 1 supplies a detailed description of the exercises and cooling down of the first session of each week.
T-chair seating and gaming
Therapy in the experimental group occurred on a novel rehabilitation technology prototype, called T-Chair (Figure 1). T-Chair is an instrumented chair that provided visual feedback. This prototype is specifically developed to train sitting balance. The seating provided a stable and unstable surface and allows for movements of the seating surface in the anterior-posterior and lateral direction(29). In the seating, 64 sensors (FlexiForce A401 force sensors, Tekscan, United States) permanently measures patient's responses to external stimuli, such as movements of centre of pressure. The T-Chair provided visual feedback of range of motion of forward, backwards and lateral movements during therapy. The T-chair included specifically designed gaming to stimulate and activate participants. The goal of the game (boat game, Figure 2) was to keep balance and improve range of motion during weight shifts according to targets visualised on the screen. To maintain safety, T-Chair is equipped with a safety belt and two emergency stop buttons. The therapist continuously supervised participants during this pilot trial.
Figure 2 Screenshots of boat game exercise: boat game (left) the participant has to navigate the boat through weight distribution to the left to catch the arrows and then move the boat at the port. Boat game (right), the participant is on the left side of the canal, against the bank, and by weight distribution, the participant can move the boat to the right where a new target is located.
T-chair mechanical properties
The rotational axis of the T-Chair seat is placed approximately 30 cm above the seat level height. The chair has following characteristics regarding range of motion (ROM): forward and backwards movements of 10° (71 mm) each and sideways bending of approximately 10° (73 mm). The seating can be positioned at a horizontal plane or at a stable inclined plane of maximally 10°. The chair has a height of 50 cm, a width of 55 cm and a depth of 90 cm. T-Chair development is based on structured input from participants and clinical experts. A previous study evaluated usability of this training prototype and feedback by therapists and participants after stroke led to further improvements (30).
T-chair features and electronical properties
The training prototype contains emergency stop buttons. During an emergency situation the button can be pushed by the participant as well as the therapist. All actions are immediately interrupted, making it possible to move the seat manually to all directions, choose to return to the starting position of the training prototype or to remove the participant from the seat without further movement of the T-Chair. The main voltage remains active on the prototype during this emergency action. All parts of T-Chair continue to be powered to prevent a new potentially unsafe situation.
A PC application is used together with the chair. The training protocols of the application can be applied to the chair via an RFID badge (USB Desktop reader evohfv2, idtronic, Germany). In the training protocol the therapist can choose an exercise, adapt the duration of the exercise, the direction and the number of repetitions and synchronise it with the badge. Before starting the therapy, the chair homes to the starting position followed by placing the badge on the badge reader (NiniX Technologies, Belgium) located on the T-Chair. The T-Chair has five controllers (Figure 3) each with their own software, 1) the master controller is the main controller of the T-Chair and communicates with the touch display-, motion-, game- and sensor controller, 2) the motion controller calculates the center of gravity, the acceleration and controls the drives of the motors, 3) the game controller provides all needed range of motion and sway measurements, trainings and exercises (= protocols) and also provides the ability to play games on the T-Chair, 4) the sensor controller controls all sensors that are embedded in the seat of the T-Chair. It sends its data to the motion controller for processing and 5) the touch-display controller provides all the needed features for the therapist to interface with the T-Chair.
Descriptive baseline characteristics and testing time points
Baseline data such as age, type and location of stroke, comorbidities, dominant hand, educational level and gender were collected. Participants were screened for neglect (star cancellation (31)), cognition (Montreal cognitive assessment (32)) and level of depression (patient health questionnaire (33)). Testing was performed three times for all participants. Two time points pre-intervention, with an interval of 2 weeks, established baseline (baseline and pre time point). This evaluated stability in our outcomes in the chronic stage. The third time point was post-intervention, i.e., four weeks after pre-intervention. All outcomes were assessed using clinical measurement tools or questionnaires.
The primary aim of this study was to examine feasibility of the intervention. We evaluated feasibility in terms of recruitment and retention, participation, adherence, acceptability and enjoyment, safety and adverse events, and device development or modification suggestions after each therapy session in the experimental group.
The number of contacted and eligible participants characterized recruitment and retention. We defined recruitment rate as the number of participants in the trial divided by the number of contacted participants. Retention is the number of recruited participants that completed all 12-therapy sessions divided by the total numbers of trial participants that entered the trial.
The Pittsburgh rehabilitation participation scale (34) assessed participation. The therapist judged participation on a six-point Likert scale, ranging from poor to excellent participation. Adherence was evaluated using the Clinician Rating of Compliance Scale (35)(36). This is a seven-point ordinal scale that assesses the level of adherence of the participant. A score lower than five is defined as non-adherent. A score of six stands for moderate participation with some knowledge and interest, no prompting required and a score of seven represents active participation and the participant shows responsibility for the therapy regimen. Participants scored level of enjoyment during the therapy by means of the physical activity enjoyment scale (37). This scale contains 18 items, each with a seven-point Likert scale, with a range of 18-126. In this scale the maximal score of 126 represents total enjoyment. Also, all interferences of the therapist to pursue the safety of the participants were noted after each therapy session. The therapist asked for fatigue using Visual Analog Fatigue Scale, ranging from zero to ten (38). A score of zero stands for none fatigue and a score of ten for worst possible fatigue. Level of fatigue was asked for leg, trunk and general. Borg Rating of Perceived Exertion (39) evaluates fatigue and exertion. The scale ranges from zero to 20. A score of six represents no exertion and a score of 20 maximal exertion. Feedback from the participants and therapist to improve the prototype and protocol were noted after intervention by a questionnaire containing 16 questions with three open questions and 13 categorical questions on a five- and seven-point Likert scale (Additional file 2). This questionnaire was only administered after the last therapy session and asks, among other things, whether therapy with the prototype has an additional benefit for rehabilitation and whether it was easy to use.
At the last assessment session all participants completed the Self-Reported Patient Global Impression of Change (40), which evaluated participants’ believe in improvement and rate the change as for instance ‘very much improved’ to ‘no change’ to ‘very much worse’.
At all-time points an experienced, blind assessor (LT) conducted all the assessments.
We investigated trunk function using the Trunk Impairment Scale (TIS) (26) and sitting stability using the Modified Functional Reaching Test (41). For this task a participant sat on a stable surface next to a measuring tape on a wall, leaning on against the wall was not allowed. The participant was instructed to reach as far as he/she can with their non-affected hand, without losing stability, towards four directions: forwards, to the affected side, to the less affected side and backwards. The distance measured in each direction was recorded in centimeters (cm).
Gait was assessed on different levels; gait capacity, gait speed and endurance. The Functional Ambulation Categories (42) (FAC) examined walking capacity. This scale is a 6-point ordinal scale which scores level of independent walking. Ten Meter Walk Test (43) measured comfortable and maximum gait speed. The two-minute walk test scored gait endurance. The Fugl-Meyer Lower Extremity assessment (44) evaluated selective movements of the lower extremities. The Berg Balance Scale (45) scored functional balance. The Functional Independence Measure (46) and the Modified Barthel Index (47) measured the level of independence in activities of daily living.
We measured trunk and leg strength in Newton with a hand-held dynamometer (MicroFet 2, Hoggan Health Industries Inc., USA) of trunk extensors, flexors and lateral flexors, hip extensors, flexors, abductors and adductors, knee extensors and flexors and ankle plantar and dorsal flexors. This protocol was based on previous trials (48–50) and adapted to reduce compensations.
Tone of different muscle groups was evaluated using the Modified Ashworth Scale (51); elbow flexors and extensors, hip flexors and adductors, knee flexors and extensors, and ankle plantar flexors. We composed a total score for the affected and the non-affected side.
For all clinical scales, a higher score represents a better outcome, except for the Modified Ashworth Scale.
All participants received a calender to note the number of falls and their accompanied circumstances, to monitor their usual care and included sporting activites.
The feasibility and safety results are presented descriptively as distribution of response frequencies to the questionnaires and scales.
Regarding changes in clinical outcomes, we evaluated normality of pre-intervention evaluations using the Shapiro-Wilk test and visual inspection of Q-Q plots because of the sample size. Change scores and their variability (pre-intervention minus baseline and post minus pre-intervention) in both groups were calculated, both mean and standard deviation or median and interquartile range, depending on whether data was normally distributed or not. Differences between groups were then analysed using parametric one-way analysis of variance or non-parametric Mann-Whitney test with a two-sided p-value<0.05. If data of change scores were normally distributed, we applied parametric testing, in the other case non-parametric analysis. Analyses were conducted with IBM SPSS Statistics for Windows, version 27 (IBM Corp., Armonk, N.Y., USA). Analyses was by intention-to-treat and included all randomized participants in the groups to which they were assigned. Dropouts were included if there was a post intervention assessment independent of the number of treatments the participants received. This is an exploratory pilot study, hence we did not conduct a multiple testing correction for incorporating multiple variables.