Design and ethics
The presented study was conducted as a randomized cross-over trial with two arms (intervention/rest-period) in accordance with the recommendations of the “CONSORT 2010 statement: extension to randomised crossover trials” [29]. Each participant had one measurement session of about 2 h in a laboratory setting including three kinematic measurements (M1, M2, M3) and the two periods (rest/intervention) in between (Fig. 1A). The local Ethical Commission approved the study including all the described procedures (No. 10-2019). The trial was retrospectively registered in the German Clinical Trial Register (No. DRKS00029765; date of registration 27.07.2022).
Participants
The sample size calculation was done using G*power (version 3.1.9.2, Düsseldorf, Germany) for a repeated measures univariate analysis of variance. Expecting a minimal effect size (f) of 0.3, assuming an alpha error of 5% and a beta-error of 80% n = 13 participants must be measured when a repeated measures default correlation of r = 0.7 is suggested [13].
Thus, n = 13 participants were recruited; in the professional environment (physiotherapy practice) of the principle investigators (JM, LO, AN). Inclusion criteria were an age from 18 to 70 years and existing chronic non-specific low back pain (≥ 3 month), diagnosed by a physician. Participants were excluded from study participation when representing high acute low back pain (≥ 5 at a visual analogue scale (VAS) of 0–10 cm in the last seven days), diagnosed with specific back pain (e.g. fracture; herniated disc), and/or pregnancy.
The principle investigator ascertained the eligibility of participants according to the inclusion and exclusion criteria. Before starting the measurements, all participants were informed about the procedure, aim and risks of the study as well as the possibility to terminate the participation in the study at any time, without giving reasons. Then, all voluntarily signed a written informed consent form.
Randomization
Eligible participants were randomized (full randomization) in a 1:1 allocation to one of two treatment sequences-intervention-control or control-intervention. The randomisation list, generated by “randomization.com”, was kept in a locked cabinet. A research assistant not involved in the outcome assessment revealed the group allocation. All participants were blinded against the group sequence.
Intervention
The game-based real-time biofeedback training was performed using a medical device for digital (home-based) back pain therapy (Valedo Home; Hocoma, Switzerland; Fig. 1B). This system consists of two inertial measurement sensors as well as the application-based software and a tablet/smartphone. The concept of the digital back therapy is that game-based exercises for movement control and stabilization of the trunk muscles are performed by the patient. With the help of the two sensors, the patient guides and controls an avatar via body movement through different virtual worlds.
To start the intervention, the two sensors were positioned over the lower lumbar spine as well as the sternum during upright standing (Fig. 1B). Afterwards the sensors were calibrated and the individual range of motion was measured. The intervention itself started with a one-minute trunk stabilisation task during two-legged standing. This was followed by a magic mirror task, were the patient needs to imitate five given movements of the trunk/pelvis. Afterwards the movement game for lateral flexion followed. The patient had to move an avatar with focus on right- and left-sided lateral flexion in three different levels. In total, the intervention last 12 min. The resting phase comprised the same period of time.
Procedure
After receiving an anthropometric (body height (cm); body mass (kg)) assessment, all participants completed a paper–pencil-based version of the graded chronic pain questionnaire (von Korff), valid to measure the presence of chronic low back pain [30,31,32]. In addition, low back pain intensity was monitored during the entire protocol.
Participants were then prepared for kinematic analysis of trunk motion (see kinematic analysis). Afterwards, trunk motion during lateral flexion (while upright standing) was measured. Lateral flexion movement is a represent for one-sided trunk movement in daily living, e.g. one-handed lifting of a bag or box [33, 34]. In addition, it’s an isolated, controlled movement of the trunk in one plane avoiding high impact on the trunk of the patient’s analysed. Therefore, maximum angle (to the left and right side) and the individual angle reproduction were measured in a randomized order. For the assessment of maximum angle (MA), all participants performed one preparation and three test repetitions of maximum left- and right sided lateral flexion during upright standing. For angle reproduction, a target trunk angle of 20° in lateral flexion was set, as this is the reference value for lateral trunk flexion according to the neutral-zero-method. During the preparation trials, the target angle of 20° was positioned manually by a physiotherapist by use of a handheld goniometer placed on the back of each participant (baseline set at pelvis level). Each participant performed two preparation trials and ten test trials in each direction (right/left).
In between the three (kinematic) measurements, the 12-min intervention and the 12-min-rest-time was undertaken in random order; no additional wash out time was included.
Kinematic analysis
All kinematic trunk motion outcomes were measured using a 16-camera optoelectronic 3D-motion analysis system (Optitrack, Oregan, USA; 120 Hz). For tracking multi-segment trunk kinematics, 15 markers were positioned over bony landmarks to frame two segments (thoracic segment (TS), lumbar segment (LS)) (Fig. 2A) [35]. A total of ten markers were placed on the torso; four markers were placed around the pelvis, and an additional back marker was added on the right scapula especially for improved identification of left and right side of the tracked skeleton. In addition, the whole trunk segment was analysed as the sum of the two segments (trunk segment (TR)).
Marker data were analyzed to calculate the relative angles of each segment in relation to the pelvis. Therefore, the software-based filtered data (5 Hz cut off; Motive 2.0, Optitrack, Oregan, USA) were exported to a customized software (Python, Trier University of Applied Sciences) for calculation of the main measures (maximum angle [°], movement duration [s], movement velocity [°/s]).
Primary outcomes
The primary outcome measurement was the maximum angle (MA; [°]) during left- and right sided maximum lateral flexion movement while upright standing. The secondary outcome was the individual normalized angle reproduction (AR; [°]) to a defined target movement angle in lateral flexion movement (Fig. 2B).
The maximum angle was assessed for lateral flexion as well as anterior flexion and axial rotation of each segment [36, 37]. The MA was measured as the maximum angle in all three planes while performing a maximum right-/left-sided lateral flexion described with good to excellent reliability [38]. Positive values represent trunk anterior flexion, right-sided lateral flexion and left-sided rotation. In contrast, negative values characterize trunk extension, left-sided lateral flexion and right-sided rotation.
In addition to the analysis of the MA, the angle reproduction was analysed for trunk lateral flexion described valid in low back pain patients [39]. The target angle for the reproduction task was set to 20° related to activities of daily life. To account for the manual positioning of the 20° target angle during the preparation trials by use of a handheld goniometer, the mean value out of the two preparation trials was calculated as a reference value for the test trials. The results of the ten (test) repetitions were then normalized to this mean value of the preparation trial as follows: the differences between the individual mean of the preparation trials and the ten repetition trials were calculated. In a final step, the individual mean values between preparation trial mean and the ten repetitions were calculated and are the basis for the pre/post comparison in the statistical analysis.
Secondary outcomes
Further secondary (kinematic) outcomes were movement velocity [°/s] and duration [s]. Both were calculated for the whole movement cycle from upright standing to maximum (left- or right-sided) lateral flexion and back to upright standing.
Graded pain questionnaire
The graded pain questionnaire consists of seven items, including pain intensity and disability (recently and last 3 months) [6, 30,31,32]. Six items are conducted of a numeric rating scale ranging from 0 (no pain/disability) to 10 (highest pain/disability (incapable of doing anything)). Sub-scores of pain and disability are calculated. Furthermore, participants can get classified into one of the five hierarchical pain and disability grades ranging from low pain/disability (grade 0) to high pain/disability scores (grade IV) [6, 30,31,32].
Visual analogue scale
In addition to the graded pain questionnaire, low back pain was monitored by use of a VAS (0–10 cm) in regular intervals throughout the whole measurement day (begin/after each measurement/after intervention or rest/end). A score higher seven at the VAS (acute pain) was defined as termination criterion for the whole measurement.
Data processing and statistical analysis
All non-digital data were documented in a paper and pencil-based case report form and transferred to the statistical database (JMP Statistical Software Package 14, SAS Institute®)[6]. After the plausibility check (range check and extreme/outlier value analysis for all outcomes), the data were presented descriptively (means, standard deviations, 95% confidence intervall) for all given outcomes [6].
All outcomes were checked for normal distribution with Shapiro–Wilk-Test [6]. Since the majority of the main outcomes were normally distributed, the following inferential statistical analysis was done two-stepped: Firstly, the measured data were analyzed for a potential carryover effect between the two periods (rest—intervention). Therefore, the sums of the measured outcome values after each period are calculated and an unpaired t-test for group comparison was applied [40].
Secondly, pre-to-post-intervention data were matched and analyzed by repeated measures analyses of variances to account for the potential acute interventional effects. The level of significance was set at α = 0.05 for all analyses.