This study uses a one way repeated measures design 28 with randomisation of the order in which the two test conditions were performed. The study follows the recommendations of CONSORT 2010 29. It has been registered with clinical trials.gov prior id°. NCT04574726
The experiment was performed with national ethical committee approval on 17/09/2020 and in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki.
b. Sample size
Cohen's d for calculating the detectable internal equivalence was performed using the TOSTER package version 0.3.4 under r statistic, "powerTOSTpaired" function.
The equivalence bounds to achieve 80 % power for a small effect: -0.35 and 0.35 correspond to a sample of 70 matched participants.
Two recruitment centres were used, Pôle St Hélier, a continued care and rehabilitation centre in Rennes, and La Musse Hospital in Evreux.
The inclusion criteria were:
healthy individuals aged between 18 and 65. We excluded people with severe vision disorders, vestibular disorders, orthopaedic disorders of the lower limbs, orthopaedic disorders requiring an upper limb brace, pregnant women and people with a score of over 25 on the MSSQ 30. The participants signed two copies of a Free and Informed Consent Form.
Each participant performed the balance tests under both conditions: in a real environment and a virtual environment. A list with the randomisation of the order in which the real/virtual tests were performed was prepared in advance by a blinded operative (random function in excel).
A real space was scanned so it could be translated into virtual reality (Figure 1). The subject was positioned in a hospital room opposite the bed. The volume of the room and its furniture, wall displays, window and shadows was reproduced. A red chair was placed to the left of the bed to provide a contrasting visual landmark. The virtual environment was loaded on to an Oculus Quest 1 (Consumer Edition, Facebook, United States). Once in position opposite the bed, each subject could follow the instructions whether in the real environment or the virtual one.
A force plate (Tyromotion Tymo plate) was positioned on the floor in front of the bed. The virtual environment was configured to be consistent with the real environment in terms of the position and distances to furnishings. The Tymo plate recorded displacements from the Centre of Pressure (COP).
Each participant went through a cycle of four successive assessments:
- static single-leg, left foot 30 sec.
- static single-leg, right foot 30 sec.
- dynamic single-leg, left foot 30 sec.
- dynamic single-leg, right foot 30 sec.
The dynamic test involved linking a series of movements. Displacement of the upper limbs and head in three dimensions were configured as follows (Figure 2):
Starting position: single-leg balance, upper limbs (UL) by the sides
- Movement 1: elevation of UL in maximum flexion for 6 seconds
- Movement 2: UL lowered towards the opposite side from the standing leg until horizontal in 3 seconds
- Movement 3: horizontal displacement from one side to the other in 6 seconds
- Movement 4: UL lowered towards the starting position in 3 seconds.
Each participant had 3 attempts for both the virtual and real environment in the randomised order.
The primary outcome measure is displacement of COP recorded during the balance tests:
- Surface of COP in cm2 measured during the tests.
- Distance covered by COP in mm
The potential side effects of immersive virtual reality were measured afterwards on the SSQ 31. “Cyber sickness” occurs where there is a conflict between vestibular and visual input demonstrating latency of the virtual environment relative to head displacement. If the SSQ scores are high, the loss of balance could be attributable to the configuration of the virtual environment.
Each participant performed the tests according to the randomisation arm they were allocated to using a previously drawn-up list. The assessors created a patient identity in the software using the enrolment number followed by the experimental arm. The assessor monitored the tests being performed and the recordings of the pilot software of the plate for capturing the centre of gravity; no data on test performance could be modified by the assessor.
h. statistical methods
The data were processed using the equivalence test: Two One-Sided Tests Procedure (TOST) 32 on paired data 33.
The statistical tests were carried out using the TOST package and the function “TOSTpaired”. To exclude confounding bias, we sought to find out whether the order of test environments had any influence on performance (t test between the two arms), and whether there was any learning over the 3 series of tests requested (Kruskall-Wallis on each series of 3 tests for 4 tests in the real environment and 4 tests in the virtual environment).