Eur Spine J DOI 10.1007/s00586-017-5323-0

ORIGINAL ARTICLE

Remote kinematic training for patients with chronic neck pain: a randomised controlled trial Hilla Sarig Bahat2 · Kate Croft1 · Courtney Carter1 · Anna Hoddinott1 · Elliot Sprecher2 · Julia Treleaven1 

Received: 4 April 2017 / Revised: 4 September 2017 / Accepted: 3 October 2017 © Springer-Verlag GmbH Germany 2017

Abstract  Purpose  To evaluate short- and intermediate-term effects of kinematic training (KT) using virtual reality (VR) or laser in patients with chronic neck pain. Methods  A randomised controlled trial with three arms (laser, VR, control) to post-intervention (N = 90), and two arms (laser or VR) continuing to 3 months follow-up. Home training intervention was provided during 4 weeks to VR and laser groups while control group waited. Outcome measures  Primary outcome measures included neck disability index (NDI), global perceived effect (GPE), and cervical motion velocity (mean and peak). Secondary outcome measures included pain intensity (VAS), health status (EQ5D), kinesiophobia (TSK), range, smoothness, and accuracy of neck motion as measured by the neck VR

* Hilla Sarig Bahat [email protected] Kate Croft [email protected] Courtney Carter [email protected] Anna Hoddinott [email protected] Elliot Sprecher [email protected] Julia Treleaven [email protected] 1

CCRE Spine SHRS, University of Queensland, St. Lucia, Brisbane, Queensland, Australia



The Department of Physical Therapy, University of Haifa, 199 Aba Khoushy Ave., Mount Carmel, 3498838 Haifa, Israel

2



system. Measures were taken at baseline, immediately posttraining, and 3 months later. Results  Ninety patients with neck pain were randomised to the trial, of which 76 completed 1 month follow-up, and 56 the 3 months follow-up. Significant improvements were demonstrated in NDI and velocity with good effect sizes in intervention groups compared to control. No within-group changes were presented in the control group, compared to global improvements in intervention groups. Velocity significantly improved at both time points in both groups. NDI, VAS, EQ5D, TSK and accuracy significantly improved at both time points in VR and in laser at 3 months evaluation in all but TSK. GPE scores showed 74–84% of participants perceived improvement and/or were satisfied. Significant advantages to the VR group compared to laser were found in velocity, pain intensity, health status and accuracy at both time points. Conclusion  The results support home kinematic training using VR or laser for improving disability, neck pain and kinematics in the short and intermediate term with an advantage to the VR group. The results provide directions for future research, use and development. Trial registration ACTRN12615000231549. Keywords  Neck pain · Virtual reality · Home training · Physiotherapy · RCT · Velocity · Kinematics

Introduction Neck along with back pain is the leading global cause of disability [1, 2]. Management directed towards physical impairments within a bio-psychosocial model is recommended [2, 3] with strong evidence for active exercises [4].

13

Vol.:(0123456789)



An important function of the cervical spine is quick and precise head movement in reaction to surrounding stimuli. In addition, kinematic impairments, such as reduced movement range, accuracy, velocity and smoothness, seem to be highly sensitive and specific to neck pain regardless of onset and symptoms [5–10]. These impairments may contribute to functional difficulties such as driving [11, 12] and may be associated with sensorimotor or psychological issues in chronic neck pain, such as dizziness or fear of neck movement [5, 11–13]. Thus management directed towards these impairments may be important. In a recent pilot trial, we demonstrated the effectiveness of 4–6 supervised sessions over 5 weeks of cervical kinematic training, using a laser beam projected onto a poster or a virtual reality (VR) device, in 32 people with chronic neck pain [14]. Patients were satisfied with therapy and significantly improved in neck pain, disability and kinematics immediately and 3 months post-intervention, with mediumto-large effect sizes. Patients in the VR group though, only accessed the device at each supervised session and continued home exercises using the laser. However, VR training has potential for remote delivery due to its inbuilt assessment and treatment capability. It can tailor treatment via dynamic feedback and can be engaging and motivating, promoting high compliance [15]. The use of VR in neck pain is in its infancy and research building on this pilot trial, to include a control group, larger sample size and evaluate home-based kinematic training in chronic neck pain with strictly VR versus laser pointer feedback is required. Thus the aims of this study were to evaluate the effect of home-based kinematic training in patients with chronic neck pain compared to a control group, and the difference in effectiveness of the two methods of delivery of this intervention in both the short and intermediate term.

Eur Spine J

All recruited participants were individuals with chronic neck pain. Eligibility Inclusion criteria Adults aged 18 years or more with neck pain for more than 3 months. Neck Disability Index (NDI) score greater than 12%, and VAS during the recent week greater than 20 mm. Lastly, VR assessment indicated a reduction of mean velocity of at least one SD from control values [16]. Exclusion criteria Existing vestibular pathology; cervical fracture/dislocation; systemic diseases, epilepsy or other neurological condition; cardiovascular, or respiratory disorders affecting physical performance; history of traumatic head injury; inability to provide informed consent; inability to complete the assessment, or pregnancy. Procedure Participants were screened for eligibility via telephone. Following informed consent, participants were then required to undertake a baseline assessment to further determine eligibility. Patients who completed the assessment and fulfilled the objective measurement criteria were then enrolled. Randomised control participants were assessed at baseline and then 4 weeks post for a second baseline before randomised into an intervention group. Randomised intervention participants were assessed at baseline, immediately post, and 3 months post-intervention. Outcome measures

Methods Design This was an assessor-blinded randomised controlled trial with concealed allocation. The first phase was three-armed with eligible participants randomised into either the control group or two intervention groups (laser or VR), and were assessed before and after the 4-week intervention (Fig. 1a). In the second phase the control group and additional recruits (to account for control dropouts), were then randomised into one of the two treatment groups (Fig. 1b). This second phase included a comparison between the two interventions at the end of the 4 weeks training and 3 months post-intervention. The randomisation scheme was generated using the software Randomization.com (http://www.randomization.com).

13

Primary outcome measures included neck disability, global perceived effect and velocity of cervical motion. These were selected as most representative, common and functional of the neck pain population. Additional secondary measures were collected and analysed to provide a comprehensive understanding of the subjective and objective effect of the provided intervention. Subjective outcome measures 1. Neck disability index (NDI) [17]: the NDI has been shown to demonstrate good validity and reliability [18– 20] with a minimal clinically important change (MCIC) of 7% [21]. 2. Global perceived effect (GPE): rated using an 11-point scale to capture perceived change in different domains

Eur Spine J

(a)

Enrollment

Assessed for eligibility (n=141)

Allocation

Randomized (n=90)

Excluded (n=25)

Did not complete the assessment (n=8) Decided not to parcipate (n=18)

VR training group (n=30)

Laser training group (n=30)

Control group (n=30)

4 weeks waiting

4 weeks training

5 drop outs: Side effects testing (n=1)

4 drop outs:

5 drop outs: Side effects testing (n=3)

Sick in hospital (n=1) Headache from exercises (n=1) Time (n=2)

Pain (n=1) Time (n=1)

Time (n=4)

Post-intervention Follow-Up VR training group (n=25)

Second baseline assessment

Laser training group (n= 26)

(b)

Control group (n=25)

Second recruitment (n=7) From controls (n=25) Total N=32 VR training group (n=48) Original allocation (n=30) Control allocation (n=14) Second recruitment (n=4)

Laser training group (n= 44) Original allocation (n=30) Control allocation (n=11) Second recruitment (n= 3)

4 weeks training 8 dropouts Side effects testing (n=3); Pain (n=1);

5 dropouts Time (n=3); Sick in hospital (n=1); Headache from exercises (n=1)

Time (n=2); Personal reasons (n=2)

Post-intervention Follow-Up VR training group (n=40)

Laser training group (n=39)

3 month Post-intervention Follow-Up VR training group (n=29) 11 dropouts- Lost contact

Laser training group (n=27) 12 dropouts- Lost contact

Fig. 1  Flow chart. a Phase 1; b phase 2

13



important to the individual [22]. Here GPE was used to measure (1) the change in neck pain since receiving treatment (from − 5 vastly worse, 0 no change, to + 5 completely recovered); and (2) patient satisfaction with the intervention (from − 5 totally dissatisfied, 0 = no satisfaction, to 5 totally satisfied) [23]. These were only obtained during phase 2. 3. Neck pain intensity during the past week, was measured by a 100 mm visual analogue scale (VAS). MCIC of 21–25 mm [21, 24]. 4. EQ-5D (EQ-5D™, http://www.euroqol.org), using a scale of 100 representing the best and 0 representing the worst possible health status [25, 26]. MCIC of 10 [27]. 5. TAMPA Scale of Kinesiophobia (TSK), a 17-item questionnaire to assess fear of movement and re-injury [28] with good validity and reliability [29, 30]. Higher summed scores (0–68) correspond to higher kinesiophobia, and scores greater than 37 indicate a high degree of kinesiophobia [31]. Smallest detectable change (SDC) for the TSK (scoring range 17–68) was 9.2 [32]. Objective outcome measures Cervical range of motion (ROM) and kinematics were collected using a customised neck VR system. Hardware included the Oculus Rift DK1 head-mounted display equipped with 3D motion tracking (https://www.oculus. com/en-us/rift/). The Oculus Rift DK1 weighs 380 g. It has a resolution of 640  ×  800 per eye, and 110° field of view. Tracking

Fig. 2  The ROM module. The three VR training modules were all based on the same concept. The virtual red airplane is controlled by head motion. The participant is required to align the head of the pilot with the yellow targets by moving the head in the direction of the targets. The order of targets’ appearance in the various directions is randomised so that the participant cannot foresee where the next

13

Eur Spine J

sensors include a gyroscope, accelerometer, and magnetometer at 60 Hz refresh rate, and 1000 Hz update rate, with a CMOS Sensor for positional tracking (http://riftinfo.com/ oculus-rift-specs-dk1-vs-dk2-comparison). The software was developed using the Unity-pro software, version 3.5 (Unity Technologies, San Francisco). The developed software included the virtual reality and motion tracking data analysis real-time. Three modules were developed, including range of motion (ROM), velocity and accuracy modules. These modules enable elicitation of cervical motion by the patient’s response to the provided visual stimuli. A full kinematic report for each patient was generated after completion of the modules. During the VR session, the virtual pilot flying the red airplane was controlled by the patient’s head motion and interacted with targets appearing from four directions (to elicit flexion, extension, right rotation, left rotation). The VR software did not elicit side flexion movement. The VR modules are illustrated and described in Figs. 2, 3 and 4. For all kinematic measures, motion initiation was determined as the point in time when 5% of peak velocity was obtained [10]. Data were low-pass filtered (frequency 6 Hz, order 4) and sampling rate was 60 Hz. The cervical movement kinematic variables were calculated for each trial in each of the four directions assessed (F, E, RR, LR). The following are the definitions of the cervical motion kinematics outcome measures. 1. Velocity (deg/s) was the primary objective measure consisting of mean and peak velocity of cervical motion (Vmean, Vpeak). It was calculated as the mean/peak angu-

target will appear. The VR software was designed to challenge the participant’s head range of movement (from mid-position) by gradually increasing the ROM required following a successful hit of target. After three consecutive failures to reach a target in a specific direction this recording was completed

Eur Spine J

Fig. 3  The velocity module. During the velocity module 16 yellow ball targets were randomly displayed in four different directions (flexion, extension, right and left rotation). a Capture from the beginning of each trial, when the participant has to activate the game by positioning the pilot’s head in the centre of a red ring. The ring changes colour from red to green once the correct mid-position is achieved for 3  s. Once the ring turns green, a yellow target appears, and the

participant is required to move the head in that direction within 5  s before the target disappears. Target’s life time is visualised using a green circle around the target that diminishes gradually and functions as a timer (b). This feature aims to motivate the participant to move quickly towards the target before it disappears. During this dynamic part of the velocity module, velocity, number of velocity peaks (NVP), and time to peak velocity percentage (TTP%) were recorded

3. Number of velocity peaks (NVP) was counted from motion initiation to target hit, and represented motion smoothness, MDC 0.74 units [33]. 4. Accuracy error (°) was measured during the smooth head pursuit task in the accuracy module, and consisted of the difference between the target’s and the player’s position in degrees. The accuracy error was measured in the plane of motion, i.e. the X axis for rotation, and Y axis for flexion–extension. These measures of accuracy were shown to be significantly sensitive and specific [5]. MDC for these measures has not been investigated. 5. Cervical ROM was the maximal active ROM achieved in each direction. This methodology has good repeatability and sensitivity [10, 34], MDC 6.5° [35]. Fig. 4  The VR accuracy module. The VR accuracy module included a head pursuit task, to train cervical motion control and accuracy. The participant had to keep the head of the virtual pilot on the target as it moved in its route up, down, right and left

lar velocity of three maximal results achieved from each direction. It has previously demonstrated good repeatability, MDC for mean velocity 14.31 deg/s [33]. 2. Time to peak velocity percentage (TTP%) was the time from motion initiation to peak velocity moment, as a percentage of total movement time, representing the ratio between the acceleration to deceleration phase in the velocity profile, MDC 7.31% [33].

ROM was collected with the ROM module, velocity, NVP and TTP with the velocity module, and accuracy error with the accuracy module of the VR device. Interventions The baseline assessment provided initial kinematic training values in the VR group. Each participant in either intervention group was provided with a training plan directed towards: (a) increasing ROM; (b) increasing motion velocity; (c) increasing motion accuracy in smooth head pursuit, or a combination. They were taught how to train by a qualified physiotherapist during a 20-min session at baseline. A head-laser beam and poster (Fig. 5a), or VR hardware and software (Fig. 5b) was provided for home use.

13



Eur Spine J

Fig.  5  a Kinematic training with a head-mounted laser beam aimed at a 70 by 70  cm poster. Tasks in the laser group were similar to the VR exercises, such as following the line with the laser, moving quickly from one circle to another, etc. The laser beam provided vis-

ual feedback relating to head motion, but unlike the VR, laser training velocity was not controlled. b Kinematic home training using the virtual reality head-mounted display, and customised software with the virtual airplane controlled by head motion

Both training groups were instructed to train in short trainings up to 5 min continuously to avoid side effects. Dosage was 5 min, 4 times a day, i.e. 20 min a day, 4 times a week, for 4  weeks. Each group was provided with an illustrated handout of the exercises and ways to progress. The physiotherapist contacted each participant weekly to: progress training difficulty, help to solve problems and reemphasise the exercise regime. Records of exercise compliance were collected in two ways—laser participants completed an exercise diary and VR sessions were retrieved from the computers when returned.

scores. Bonferroni adjustment was performed on a withinparameter basis, i.e. p values were multiplied by 4, to account for separate tests of the two groups at two different time points. Change scores were constructed for each of the two posttreatment times with reference to the pre-treatment scores and compared using independent groups t tests, again Bonferroni corrected, i.e., p values multiplied by 2 for the two time points. Cohen’s d was calculated to demonstrate size effect of observed between-group changes [36] and interpreted as suggested by Cohen with d ≥ 0.2 indicating small size effect, 0.5 medium, and d ≥ 1 large size effect [36]. Cohen’s d was calculated for the post- minus pre-intervention differences, using the following formula (M—mean, SD—standard deviation): Cohen’s d = (M1 − M2)/SQRT((SD12 + SD22)/2). Number needed to treat (NNT) was calculated to represent the number of patients over the given treatment time period that one would need to treat to achieve one additional study endpoint [37]. NNT calculation did not include imputed values, and threshold for successful outcome was

Statistical analysis Data were analysed on an intention-to-treat basis. Missing data were imputed separately for each parameter and group, using a correlation-based method utilising all non-missing data. Basic treatment effects were examined using separate paired t tests for each parameter for each group on each of the post-treatment time points versus the pre-treatment

13

Eur Spine J

defined as 10% for NDI [21], and 14.3 deg/s for mean velocity [33]. Significance levels were set at 0.05. JMP software (SAS Institute, Cary, NC) was used for data analysis. Data were collected at the University of Queensland. Ethics approval was obtained from the human medical research ethics committee at the University of Qld. The trial was registered in the Australian New Zealand clinical trials registry, Trial Id: ACTRN12615000231549.

point reduction in NDI. Overall success rate for velocity was 40–60%. NTT was better in sagittal movements. Table 4 represents baseline, post-treatment and 3 months post-treatment measures for both intervention groups (phase 2) (Table 5). Results demonstrated a few differences between the two intervention groups with all advantages to the VR group including pain intensity, health status rating (EQ5D) and selected velocity and accuracy measures at both time points.

Results

Compliance and perceived effect

Subject flow for both phases is depicted in Fig. 1a, b. One hundred and forty-one participants were initially screened for enrolment. Ninety participants were randomised to VR, laser or control group (phase one). Phase 2 included 48 subjects in the VR and 44 in the laser intervention. There were 14 (15%) dropouts during the intervention period, and 23 (29%) dropouts to the 3 months follow-up (Fig. 1a, b). Approximately 16% of data were missing in phase I (postintervention), and 26% in phase II (3 months follow-up). Examinations of data distributions did not indicate any substantial deviation from normality which would affect parametric statistics employed. Table  1 presents the characteristics of study populations. No group differences were found at baseline in age (p > 0.05) or gender (Chi-square likelihood ratio = 1.288), or any of the clinical characteristics for phases 1 (Tables 1, 2).

The number of home exercise sessions was significantly higher for the laser group (mean 18.29 ± 8.63 or 4.5 times per week) compared to the VR group (14.36  ±  5.78 or 3.5 times per week). Overall, participants in both groups reported improvement in their neck pain, and satisfaction from treatment (Table 4). There were no group GPE differences. There were a few cases of side effects from the VR use. Out of 14 dropouts at post-intervention assessment, 5 were due to VR-associated sickness and headache. To determine the percentage of subjects who demonstrated some improvement and satisfaction with treatment, we classified the 11-point GPE into 3 categories following Evans et al. [22] in those who had completed the study: − 5 to 0 = not improved; 1–2 = improved; 3–5 much improved. The results showed that the majority of patients perceived some improvement and were satisfied with treatment at both time points (Table 6).

Pre–post changes The within-group analysis showed no changes in the control group for any variables, compared to improvement in 13–14 variables (p