Virtual reality training for management of chronic neck pain: a systematic review with meta-analysis

Abstract

Purpose

We conducted a systematic review with meta-analysis to examine the short- and long-term effects of virtual reality training (VRT) on pain, range of motion (RoM), kinesiophobia and perceived function in patients with chronic neck pain (CNP).

Methods

We searched the PubMed, Scopus, and PEDro databases for studies assessing effect of VRT against conventional therapy in CNP patients. Separate analyzes were performed for short-term (immediately after the intervention) and long-term (at follow-up) outcomes. Risk of bias was assessed with PEDro scale and the quality of evidence was assessed according to GRADE guidelines.

Results

We found 11 studies, 8 of which were eligible for meta-analysis. Most studies had good methodological quality (6–8 points on the PEDro scale). The analysis showed significant differences in favour of VRT for neck disability index (long-term), kinesiophobia (long-term), and neck flexion RoM (short-term). No significant differences were found for pain intensity in the short- and long-term assessments. Heterogeneity between effect sizes was high in most analyzes, and the quality of evidence was low.

Conclusion

Low quality evidence suggests that VRT may contribute to long-term improvements in perceived function and kinesiophobia in CNP patients compared with conventional training programs. However, VRT and conventional training programs have similar effects on overall neck RoM and pain intensity.

Keywords:

• Virtual reality
• virtual training
• neck disorders
• kinesiophobia
• neck disability index

Introduction

Up to 67% of people suffer from neck pain (NP) at some point in their lives [1]. The majority of NP is referred to as “'nonspecific”, because it results from postural and mechanical causes rather than organic pathology [2,3]. In most cases, NP resolves within 4–6 weeks, but one-third of those affected later develop chronic neck pain (CNP), which affects 10%–24% of the general population, with an even higher prevalence in older adults [4–7]. There are several factors that may contribute to NP or transition from acute to chronic pain status, including history of trauma, advanced age, genetics, female gender, beliefs, coping style, expectations, stress, anxiety and cognitive impairment [8,9]. Symptoms commonly experienced by patients with CNP (e.g. cervical spine pain, pain radiating to the back of the head/shoulders/upper limbs, headaches, stiffness, numbness or weakness in the upper limbs, balance problems) can lead to reduced work productivity, increased work absenteeism, and increased insurance costs [2,10].

Conventional CNP treatment include exercise (e.g. resistance exercise, mobility, muscle reeducation), physical modalities (e.g. low-level laser therapy), acupuncture, mind-body practice, medications for pain relief, pain education, manipulation, mobilisation or dry needling, but the improvement in function is small to moderate, with no long-term benefit across most interventions [11–13]. Another treatment option for CNP management is virtual reality training (VRT) which has become increasingly common in clinical medicine over the past decade [14]. Virtual reality (VR) allows users to view and interact with a simulated 3D virtual world (e.g. videos or games) and to immerse themselves in a multisensory experience, so that they can feel as a part of the virtual environment [15,16]. This sense of presence redirects individual’s attention towards stimuli emitted by VR system [17]. Consequently, an individual’s ability to respond to noxious stimuli and attend to nociceptive neural signals is diminished, resulting in a lower perception of pain [18]. Redirecting attention to external stimuli (rather than body movements) has also been shown to benefit motor learning and performance [19,20]. Immersion in the simulation is enhanced by integrating multisensory (e.g. visual, auditory, and tactile) experiences with various devices (e.g. a head-mounted display (HMD) or wearable haptic devices) [16,17]. Although the current evidence on VR analgesia is dominated, to a great extent, by studies on acute and procedural pain, VR has also been shown to influence pain in chronic populations, such as patients with fibromyalgia, complex regional pain syndrome, and neuropathic pain associated with spinal cord injury [21–24]. A recent meta-analysis also found significant reduction in pain intensity and kinesiophobia in patients with chronic low back pain after a VR intervention and at follow-up [25]. However, in their meta-analysis, Mallari et al. [26] concluded that in chronic pain conditions VR is effective in reducing pain intensity during and immediately after the VR exposure, but not also in the long-term.

The use of VRT in NP patients aims to improve impaired cervical kinematics (reduced RoM, accuracy or speed of movement, smoothness and stability of neck movement) and reduce fear of movement in this population [27]. In recent years, studies investigating the potential of VRT for the treatment of CNP were conducted and yielded some promising results. For instance, Rezaei et al. [28] found greater improvement in visual analogue scale (VAS) and neck disability index (NDI) scores in the VRT group compared with conventional proprioceptive training in patients with CNP after 4-week intervention. Similarly, Bahat et al. [29] reported that VR home training had benefits in improving disability, NP and kinematics in the short and medium term after a 4-week intervention compared with laser kinematic training. However, to the best of our knowledge, there is no systematic review on this topic in the current scientific literature. Therefore, the aim of this paper is to perform a systematic review with meta-analysis to explore the short- and long-term effects of VRT on pain, RoM, kinesiophobia and perceived function in patients with CNP. Consistent with previous evidence, we hypothesised that VRT would have only short-term beneficial effects in patients with CNP. The use of VR in pain management of patients with CNP could reduce the use of analgesics and their unwanted side effects, allow patients to self-treat in their home environment, reduce the socioeconomic burden, and contribute to patient motivation and adherence.

Methods

Search strategy and inclusion criteria

The review was conducted according to PRISMA guidelines. The review was not prospectively registered. The search was performed in January 2023. Both authors independently performed all search steps, and any disagreements were resolved by discussion. We searched the PubMed, Scopus and PEDro databases. For PubMed and Scopus, we used the following search string: (virtual reality OR virtual training OR virtual reality exposure therapy) AND (neck pain OR neck syndromes). In the PEDro database, we used the simple search option and separately entered three individual combinations of words (i.e. ‘neck pain virtual reality’, ‘virtual reality neck’ and ‘virtual training pain’ were used). In addition, we reviewed the reference lists of identified systematic reviews on the topic, and performed additional non-systematic search of the GoogleScholar database. The records were imported into Mendeley (version 1.19.8) to remove the duplicates, and then exported into Microsoft Excel software for further examination.

The inclusion criteria, structured according to the PICOS tool [30], were determined as follows.

• P (population): Any adult or older adult population with chronic NP (duration > 12 weeks).

• I (Intervention): Exercise or physical therapy intervention using VR technology.

• C (Comparison): Conventional exercise or physical therapy.

• (Outcome): Outcomes describing RoM, pain (VAS), function (NDI) and kinesiophobia.

• S (Study design): For meta-analysis, we included randomised controlled trials and other clinical trials with at least two groups (VR and control exercise/therapy groups). Single-group clinical studies and feasibility studies were included only in the narrative synthesis. Articles published as conference papers, editorials, opinion papers and non-scientific journals, and any articles not published in English, were excluded.

Data extraction

The data extraction was carried out independently by both authors and disagreements were resolved through additional discussion. The extracted data included: (a) baseline, post-intervention and follow-up data (means and standard deviations) for all eligible outcome measures for VR and control groups; percent changes were considered instead of pre-post data when available (b) baseline demographics of participants (sex, age, body height, body mass, body mass index); (c) intervention characteristics (VR intervention description, duration of the intervention, weekly frequency, volume (number of exercises, sets, and repetitions), breaks, and supervision). Data were carefully collected into Microsoft Excel 2016 (Microsoft, Redmond, WA, USA). If the data were presented in a graphical form, we used GetGraphDataDigitalizer software (version 2.26.0.20) to obtain the means and standard deviations. In case of missing data, the authors of the articles were contacted by e-mail and ResearchGate. A reminder was sent after 14 days, and if no reply was received after the second inquiry, the data was considered irretrievable.

Risk of bias and quality of evidence assessment

Both authors evaluated the risk of bias of the included studies using the PEDro scale which assesses study quality based on a 10 items [31]. Potential disagreements between the reviewers were resolved by discussion. Studies scoring from 9 to 10 were considered as ‘excellent’, 6 to 8 as ‘good’, 4 to 5 as ‘fair’, and <4 as ‘poor’ quality. The PEDro scale was selected because it was specifically developed to evaluate the quality of clinical trial studies investigating physical therapy interventions [32].

The quality of evidence was assessed according to GRADE guidelines [33]. GRADE guidelines encompass 5 main domains (risk of bias, imprecision, inconsistency, indirectness, and publication bias) to categorise evidence quality. The comparisons were classified as representing either high, moderate, low or very low quality of evidence. The level of evidence was downgraded if: (a) there was a substantial heterogeneity between the studies (I² > 50% or statistically significant heterogeneity test (inconsistency); (b) more than 25% of studies were classified as fair or poor quality (risk of bias); (c) the studies assessed the effects of the intervention indirectly (e.g. population, intervention or outcome not being representative of the research question) (indirectness), (d) when a large CI was observed, or when the total number of participants was <300 (imprecision) (e) when the study results provided differed from the original protocol or study objectives (publication bias) [34].

Data analysis

The meta-analysis was carried out in Review Manager (Version 5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, London, UK). Although there is no consensus on the minimal number of studies required for a meta-analysis [35], we only performed the analyses when three or more studies could be included. Before the results were entered into the meta-analytical model, the pre-post differences and pooled standard deviations were calculated according to the following formula SD = √[(SD2pre + SD2post)–(2 × r × SDpre × SDpost). The correction value (r), which represents the pre-test–post-test correlation of outcome measures, was conservatively set at 0.75. We used continuous outcomes meta-analysis with a random-effects model and inverse variance method to obtained the pooled effects. The effect sizes were expressed as standardised mean difference (SMD), with 95% confidence intervals (CI) reported alongside to provide the precision of the overall effect estimate [36]. Statistical heterogeneity among studies was determined with the I2 statistics. According to Cochrane guidelines, the I2 statistics of 0% to 30% might not be important, 30 to 60% may represent moderate heterogeneity, 50–90% may represent substantial heterogeneity, and 75–100% indicates considerable heterogeneity. In case of substantial heterogeneity between the studies, a sensitivity analysis was performed by examining the effect of exclusion of studies that differed significantly from the pooled effect.

Results

General overview of the search results

We identified 147 records through a database search and 111 records through an additional search (GoogleScholar, reference lists). After duplicates were removed, 144 records were screened. In a next step, 24 articles were screened for eligibility. A total of 8 studies were eligible for meta-analysis. The detailed summary of the search process is shown in Figure 1. Both men and women participated in all studies. In most studies, the participants were adults (range of means for age: 26.6–53.1 years), except in the study by Beltran-Alacreu et al. [37] which examined older adults (81.8 ± 6.82 years). This study was also the only one that investigated a non-immersive type of VRT. Across studies, weekly frequency ranged from 2 to 6 and session durations ranged from 5 to 45 min. It should be noted, however, that in some programs, the program was delivered multiple times per day (e.g. 4 times per day for 5 min). Six studies also included follow-up measurements after 4 weeks to 6 months. In most studies, the control groups performed similar exercise program as the experimental group, only without VRT. In one study, the participants in the control group did not perform the exercises, but were allowed to continue with their ongoing physical therapy [38]. In all studies, the participants had to report to suffer from neck pain for at least 12 weeks. In most studies (n = 7), the intervention was supervised. In addition, 2 studies [29,38] involved home-based exercises and in one study, the participants performed the exercises under supervision, but were also encouraged to perform the exercise at home [27]. The detailed intervention descriptions are shown in Table 1, and study characteristics (patients, outcome measures, main findings) are presented in Table 2.

Figure 1. The study search and selection process.

A photograph showing the diagram of study selection and elimination process. Description: A photograph showing the diagram of study selection and elimination process. After screening PubMed, Scopus and PEDro databases as well as other sources, 144 records were screened. Of these, 26 were sought for retrieval, 24 were assessed for eligibility, and finally, 11 were included in the review. Of these 11, 9 were included in the meta-analysis.

Table 1. Intervention descriptions of included studies.

Study	General intervention description	VRT tasks	Duration of intervention (weeks)	One training duration	Weekly frequency	Supervised (yes/no)
Rezaei et al. [28]	Non-immersive VRT which included unidirectional (flexion/extension or rotation only) and two-directional neck movements using video game Cervigame^® version 1.01, where the rabbit moves at constant speed in one direction, and its movement in the opposite direction is controlled by the patient’s head movement.	The main visual component of the game was a rabbit attempting to reach carrots. The virtual carrots appeared continuously along the line of the movement pattern being trained. Patient needed to make head movements (flexion/extension/rotation or two-directional neck movements) so that rabbit could reach the carrots on the screen.	4	21 min	2 sessions/week	Yes
Bahat et al. [27]	Immersive VRT programme which contained three modules (RoM/velocity/accuracy) interspersed with kinematic training. Participants had to control virtual pilot flying the aeroplane by head motion.	During the RoM module, the participants had to align the head of the virtual pilot with yellow virtual targets by moving the head in the direction of the targets. During the velocity module, the participant first had to activate the game by positioning the pilot’s head in the centre of a red virtual ring. When participant achieved correct mid-position, a yellow target appeared in a random direction, and the participant was required to move the head in that direction within 7 s before the target disappeared. During accuracy module, a moving target was presented on a vertical or horizontal line, at a constant velocity of 10 °/s. The participant was required to maintain the pilot’s head position on the moving target as closely as possible by tracing the vertical/horizontal line.	5	EG: 30 min (15–20 min using VR + 10–15 min kinematic training) CG: 30 min	4–6 sessions in 5 weeks	Yes, but also encouraged to perform unsupervised kinematic training 3x/week for 30 min
Beltran-Alacreu et al. [37]	VR condition: Non-immersive VR training included neck flexion,extension, lateral inclincations but no rotations. ‘Active Airlines’ game was used, where participants controlled virtual aeroplane using head motions.	Flexion-extension movements moved the virtual plane down or up and lateral inclinations moved the plane to the sides. The participants had to reach targets by moving virtual plane with head movements.	12 weeks (4 weeks VR + 4 weeks washout period + 4 weeks conventional exercises).	VR: 210 sec; Conventional exercises: 30–45 min	2 sessions/week	Yes
Tejera et al. [39]	Immersive VRT which included neck movements in all directions. Two VR applications were used: ‘Fulldive VR’ (where participants had to change the images in the viewfinder by tilting their head) and ‘VR Ocean Aquarium 3D’ (where observing different marine animals was possible by making neck movements).	During the ‘Fulldive VR’ only tilting movements of the neck were necessary, while during ‘VR Ocean Aquarium 3D’ flexion, extension and rotation movements had to be performed.	4	Not specified	2 sessions/week	Yes
Nusser et al. [40]	EG1 (VR group): Immersive VRT which included neck specific sensorimotor training (e.g. finding neutral position of the head, following a virtual subject by moving head), combined with general and neck specific exercise therapy. EG2 (sensorimotor group): General and neck specific exercise therapy, combined with general sensorimotor training, that included coordination and balance exercises, small game forms (e.g. throwing and catching), partner games.	A globe was shown, moving in a virtual space on predetermined trajectories, on the monitor. The patient was asked to follow the globe by moving the head the orbital pathways of the globe. The patient had to try to track the virtual globe as closely as possible with a white circle, whose position in the virtual scene corresponded to their head position. Head Repositioning test and Head to Target Test: The subject started in an upright seated position, with their head in a neutral position, which the subject was asked to memorise. The globe then led the subject’s head to a specific position in 1 of the 8 directions and disappeared. After that, the subject was asked to first find the neutral position again without visual guidance, and then the specific end position. Dynamic exercise: In this exercise, globes travelled along specific trajectory, one after another, and the subject had to follow them in the virtual scene by moving their head precisely.	3	EG1: 20 min of VR training EG2: 30 min of sensorimotor training	EG1: 6 sessions in 3 weeks; EG2: 4 sessions in 3 weeks	Yes
Bahat et al. [29]	EG1: Immersive VR home training program which contained three modules (RoM/velocity/accuracy) directed towards increasing RoM, increasing motion velocity and increasing motion accuracy. Participants had to control virtual pilot flying the aeroplane by head motion. EG2: Kinematic home training with head-mounted laser beam aimed at a 70 by 70 poster. Tasks in laser group were similar to VR exercises (following the line with laser, moving quickly from one circle to another, etc.).	VRT was the same as in Bahat et al. [27] study (described above).	4	5 min, 4x /day	4 sessions/week	No (home training)
Bahat et al. [38]	Immersive VR home training including exercises to train fast and accurate head control using three VR modules: RoM/velocity/accuracy. During the VR the subject’s head was visualised as an animated image of a pilot flying a small aeroplane. They also continued physical therapy, if receiving any.	Not described in detail.	4	5 min	4 sessions/week	No (home training)
Cetin et al. [41]	Immersive VR consisted of ‘Ocean Rift’ and ‘Gala 360’ applications, which provide views from countries and cities or watching see animals. Patients were encouraged to move their necks during VRT.	The patients were seated and were asked to look in all directions during the VR application. They were encouraged to move their necks by expressions such as ‘follow that dolphin, the sea turtles you chose will come soon, there may be a starfish below. Now you are in front of the Eiffel Tower, you can look around’.	6	EG: 40 min (20 min motor control exercises + 20 min VR) CG: 40 min (motor control exercises)	3 sessions/week	Yes

VR: virtual reality; EG: experimental group; CG: control group; VRT: virtual reality training; RoM: range of motion.

Table 2. Characteristics of included studies.

Author (year)	Participants				Baseline values	Control group	Follow-up assessment time (if any)	Post-intervention values	Results/ Conclusions
	Sample size	Age (years) mean ± SD	Duration of neck pain (months) mean ± SD	Inclusion criteria
Rezaei et al. [28]	n = 42 EG: 21 CG:21	EG: 36.19 ± 9.80 CG: 31.23 ± 9.49	EG: 22.42 ± 15.52 CG: 22.04 ± 16.79	A history of non-traumatic NP for > 3 months; age between 20 and 55 years,	VAS (100): 47.11 ± 10.24 (EG); 38.95 ± 10.07 (CG)	Conventional proprioceptive training, consisted of: eye-follow, gaze stability, eye-head coordination and position sense and movement sense training.	5 weeks	VAS (100): 10.75 ± 8.43 (EG); 19.63 ± 9.15 (CG)	There were significant improvements in all variables in both groups immediately after and 5 weeks after the intervention. Greater improvements were observed in the VAS and NDI scores in VRT group.
					NDI (50): 13.00 ± 1.30 (EG); 12.28 ± 1.38 (CG)			NDI (50): 4.57 ± 2.39 (EG); 8.14 ± 3.13 (CG)
Bahat et al. [27]	n = 32 EG: 16 CG: 16	EG: 40.36 ± 14.18 CG: 41.13 ± 12.59	EG: 98.06 ± 96.81 CG: 87.31 ± 111.99	Age 18 years or more; prolonged NP for >3 months; NDI score greater than 10%.	NDI (%): 20.38 ± 7.6 (EG); 20.19 ± 6.5 (CG)	Kinematic training (KT) with laser pointer: active neck movements to increase RoM, quick head movements, head stability exercises, accurate neck movement training.	3 months	NDI (%): 20.38 ± 7.6 (EG); 20.19 ± 6.5 (CG)	KTVR may have some advantages in the short term for pain intensity, neck disability, RoM, and fast neck motion control and longer term for global perceived effect 3 months post intervention. Nevertheless, direct comparison between the groups suggested that the KTVR was not superior to the KT training.
					VAS (100): 35.72 ± 17.7 (EG); 35.17 ± 16.7 (CG)			VAS (100): 22.1 ± 24.1 (EG); 27.72 ± 21.9 (CG)
					TSK (0–68): 32.75 ± 6.8 (EG); 30.38 ± 5.8 (CG)			TSK (0–68): 30.13 ± 5.7 (EG); 28.64 ± 9.9 (CG)
					RoM-flexion: 38.69 ± 14.6 (EG); 43.94 ± 14.3 (CG)			RoM-flexion: 57.31 ± 11.3 (EG); 49.87 ± 17.2 (CG)
					RoM-extension: 48.72 ± 15.1 (EG); 51.09 ± 13.2 (CG)			RoM-extension: 57.45 ± 14.1 (EG); 54.19 ± 12.8 (CG)
					RoM-right rotation: 62.32 ± 13.1 (EG); 57.06 ± 16.6 (CG)			RoM-right rotation: 71.84 ± 14.0 (EG); 77.21 ± 10.6 (CG)
					RoM-left rotation: 58.58 ± 15.3 (EG); 57.94 ± 15.3 (CG)			RoM-left rotation: 77.74 ± 16.0 (EG); 72.58 ± 13.6 (CG)
Beltran-Alacreu et al. [37]	n = 14 (Cross-over study)	81.85 ± 6.82	>3 months	Over 65 years of age: NP for > 12 weeks, scored more than 5 points on NDI; have no cognitive impairment below 20 points, as assessed by the MMSE scale.	NDI: 15.77 ± 8.19	Control condition: Conventional exercises included: stairs, pedalling, pulleys, obstacle walking, cervical joint mobility exercises in all ranges without resistance.	4 weeks	VAS (10): 3.77 ± 1.92 (EG); 3.46 ± 2.22 (CG)	Both conventional exercises and the use of the serious game had the same effect in reducing neck pain. Playing the serious game first followed by conventional exercise reduced pain more than conventional exercise first followed by playing the serius game.
					VAS (10): 4.92 ± 1.88			NDI not reported.
Tejera et al. [39]	n = 44 EG: 22 CG: 22	EG: 32.72 ± 11.63 CG: 26.68 ± 9.21	Not mentioned	Non-specific CNP; age 18 to 65 years.	VAS: 4.97 ± 1.88 (EG); 4.27 ± 1.3 (CG)	Neck exercises: neck flexion/extension/lateral flexion/rotations (without resistance).	1 and 3 months	VAS: 2.67 ± 1.91 (EG); 3.11 ± 1.47 (CG)	Kinesiophobia was the only outcome that showed differences between VR and exercise at 3 months follow-up. Pain intensity, CPM, TS, RoM, neck disability, pain catastrophizing, fear-avoidance beliefs, PPT and anxiety did not show differences between both interventions. Most variables showed intra-group differences in VR.
					NDI: 13.72 ± 6.68 (EG); 14.09 ± 9.32 (CG)			NDI: 6.9 ± 6.28 (EG); 7.45 ± 5.36 (CG)
					TSK-11: 22.90 ± 7.11 (EG); 21.40 ± 6.63 (CG)			TSK-11: 18.90 ± 10.73 (EG); 18.36 ± 7.48 (CG)
Nusser et al. [40]	n = 51 EG1: 17 EG2: 16 CG: 18	EG1: 51.2 ± 8.8 EG2: 53.1 ± 5.7 CG: 49.8 ± 8.1	> 3 months	18 years or more; CNP for >3 months; no pain medication or muscle relaxants use for 24 h before the tests.	NRS (10) (during rest): 4.2 ± 2.6 (CG); 4.9 ± 2.1 (EG)	General and neck  specific exercise therapy: strengthening, mobilisation, relaxation, medical training therapy, functional gymnastics, aqua therapy, physical therapy.	/	NRS (10) (during rest): 3.2 ± 3.0 (CG); 2.2 ± 2.0 (EG)	Compared with the control group, the VR group showed significant advantages in relief of headaches, and active cervical RoM in flexion and extension. Compared with the sensorimotor group, the VR group showed significant improvements in cervical extension.
					NRS (10) (during motion): 5.3 ± 2.2 (CG); 5.3 ± 2.6 (EG)			NRS (10) (during motion): 3.5 ± 2.2 (CG); 3.7 ± 2.0 (EG)
					RoM-flexion: 45.8 ± 12.9 (CG); 40.9 ± 14.6 (EG)			RoM-flexion: 42.9 ± 12.6 (CG); 48.5 ± 13.3 (EG)
					RoM-extension: 43.1 ± 13.3 (CG); 35.4 ± 12.8 (EG)			RoM-extension: 39.8 ± 14.7 (CG); 44.6 ± 12.9 (EG)
					RoM-left rotation: 59.3 ± 10.1 (CG); 57.4 ± 12.7 (EG)			RoM-left rotation: 59.9 ± 9.3 (CG); 65.2 ± 16.9 (EG)
					RoM-right rotation: 61.8 ± 15.3 (CG); 55.6 ± 10.4 (EG)			RoM-right rotation: 64.0 ± 15.2 (CG); 59.5 ± 14.0 (EG)
					NDI (0–50): 18.2 ± 6.7 (CG), 18.7 ± 5.2 (EG)			NDI (0–50): 13.7 ± 7.0 (CG), 13.7 ± 7.9 (EG2); 11.4 ± 7.5 (EG)
Bahat et al. [29]	n = 90 EG1: 30 EG2: 30 CG: 30	EG1: 48.0 ± 5.4 EG2: 47.6 ± 6.7 CG: 47.9 ± 6.9	>3 months	Adults aged 18 years or more with NP for more than 3 months; NDI score greater than 12%, and VAS during the recent week greater than 20 mm.	NDI (%): 32.88 ± 12.5 (EG);24.72 ± 10.7 (CG)	Waiting.	3 months	NDI (%): 23.75 ± 15.7 (EG); 26.88 ± 14.0 (CG)	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 advantages for using the VR compared to the laser were not enough to recommend its use in its current state.
					VAS (100): 47.79 ± 20.9 (EG); 45.78 ± 21.5 (CG)			VAS (100): 31.1 ± 23.6 (EG); 45.78 ± 21.5 (CG)
					TSK (68): 35.22 ± 7.4 (EG); 32.64 ± 7.2 (CG)			TSK (68): 33.26 ± 7.8 (EG); 33.96 ± 6.2 (CG)
					RoM-flexion: 60.54 ± 10.1 (EG); 62.98 ± 7.3 (CG)			RoM-flexion: 61.38 ± 7.1 (EG); 64.52 ± 6.0 (CG)
					RoM-extension: 63.17 ± 11.9 (EG) 65.42 ± 11.2 (CG)			RoM-extension: 64.67 ± 11.0 (EG); 67.11 ± 7.9 (CG)
					RoM-right rotation: 73.77 ± 16.2 (EG); 78.87 ± 12.2 (CG)			RoM-right rotation: 75.28 ± 13.3 (EG); 75.64 ± 12.2 (CG)
					RoM-left rotation: 76.86 ± 14.5 (EG); 78.93 ± 16.0 (CG)			RoM-left rotation: 77.43 ± 11.9 (EG); 77.14 ± 11.4 (CG)
Bahat et al. [38]	n = 45 EG: 22 CG: 23	EG: 30 ± 5.8 CG: 28 ± 5.1	Median (Q1, Q3) EG: 36 (5.0, 84.0) CG: 12 (6.0, 48.0)	Fighter and helicopter pilots from the Israeli Air Force with acute, subacute, or chronic neck pain; with or without referral to the upper limbs; average NP intensity (during the past week) of at least 20% [on a VAS/100 mm].	VAS (100): 36.4 ± 22.8 (EG); 49.5 ± 21.1 (CG)	Physical therapy if  receiving any.	6 months	VAS (100): 25.7 ± 24.0 (EG); 26.9 ± 22.3 (CG)	Findings showed a positive effect on kinematic performance in the training group as compared to the controls, and a positive relationship between exercising time and improvement in neck associated disability. However, results showed no change in self-reported measures such as pain intensity and disability.
					Median (Q1, Q3): NDI (%): 15 (12–22) (EG); 16 (10–20) (CG)			Median (Q1, Q3): NDI (%): 10 (6–26) (EG); 16 (8–20) (CG)
					Median (Q1, Q3) RoM-flexion: 65.2 (65.1, 65.6) (EG); 65.2 (63.6, 65.9) (CG)
					RoM-extension: 73.5 (70.9, 76.3) (EG); 73.3 (72.6, 75.2) (CG)
					RoM-right rotation: 81.5 (80.2, 83.1) (EG); 80.9 (80.1, 81.8) (CG)
					RoM-left rotation: 80.1 (78.1, 81.7) (EG); 80.7 (80.2, 81.3) (CG)
Cetin et al. [41]	n = 34 EG: 17 CG: 17	EG: 40.0 ± 11.88 CG: 41.94 ± 10.76	>6 months	Age between 18 and 65 years with a minimum of 6 months of NP; a baseline NDI score of at least 20% (10 points); and the neck region as the primary pain area.	RoM-flexion: 49.05 ± 7.89 (EG); 44.64 ± 9.92 (CG)	Motor control (MC)  exercises: strenghtening of the deep cervical flexors/extensors and axioscapular muscles, stretching exercises, postural correction exercises.	/	(Reported as % change) RoM-flexion: 10.64 ± 7.44 (EG); 8.47 ± 7.21 (CG)	The results indicated that VR in addition to MC exercises was more effective than MC exercises alone in improving PPTs (C1/2 and C5/6), JPSE, and functional limitation. However, neither intervention was superior in improving RoM, pain, muscle performance, symptoms, psychosocial status, or quality of life.
					RoM-extension: 62.47 ± 11.88 (EG); 61.23 ± 9.44 (CG)			RoM-extension: 10.05 ± 9.25 (EG); 7.7 ± 7.05 (CG)
					RoM-right rotation: 56.76 ± 9.83 (EG); 51.47 ± 8.43 (CG)			RoM-right rotation: 6.05 ± 13.6 (EG); 8.52 ± 6.06 (CG)
					RoM-left rotation: 55.0 ± 9.18 (EG); 49.11 ± 8.33 (CG)			RoM-left rotation: 7.35 ± 9.0 (EG); 9.41 ± 9.98 (CG)
					VAS (10): 5.77 ± 1.39 (EG); 5.98 ± 1.93 (CG)			VAS (10): −3.69 ± 1.85 (EG); −2.44 ± 2.14 (CG)

VR: virtual reality; EG: experimental group; CG: control group; n: numerus; VRT: virtual reality training; VAS: visual analogue scale; NDI: neck disability index, RoM: range of motion; KT: kinematic training; TSK: TAMPA scale of kinesiophobia; KTVR: kinematic plus virtual reality training; MMSE: mini-mental state examination; CPM: conditioned pain modulation; PPT: pain pressure threshold; NRS: numerical rating scale; JPSE: joint position sense error; IRP: Interdisciplinary Rehabilitation Programme; MC: motor control, NP: neck pain; CNP: chronic neck pain; TS: temporal summation.

Assessment of study quality

The assessment of study quality according to Pedro Scale is shown in Table 3. Most studies were of good quality, with 6 reaching a score between 6 and 8. One additional study was of fair quality (4 points) and one was of poor quality (3 points).

Table 3. Assessment of study quality with Assessment of study quality with Pedro Scale.

	TOTAL SCORE	Eligibility criteria	Random allocation	Concealed allocation	Baseline comparability	Blind subjects	Blind therapists	Blind assessors	Adequate follow-up	Intention-to-treat analysis	Between-group comparisons	Point estimates and variability
Razaei et al. [28]	6	1	1	0	1	0	0	1	1	0	1	1
Bahat et al. [27]	7	1	1	1	1	0	0	1	1	0	1	1
Beltran-Alacreu et al. [37]	3	1	1	0	0	0	0	1	0	0	0	1
Tejera et al. [39]	6	1	1	0	1	0	0	0	1	1	1	1
Nusser et al. [40]	6	1	1	0	1	0	0	0	1	1	1	1
Bahat et al. [29]	7	0	1	1	1	0	0	1	0	1	1	1
Bahat et al. [38]	8	1	1	1	1	0	0	1	1	1	1	1
Cetin et al. [41]	4	1	1	0	1	0	0	0	0	0	1	1

Effects of VR on neck pain (VAS)

Eight studies included in the meta-analysis (161 and 163 participants in VR and control groups, respectively) assessed neck pain with VAS scales (Figure 2). Overall, there was low-quality evidence for a moderate, but non-significant (CI crossing the 0) effect of VRT (SMD = −0.50; CI = −1.10 to 0.09) The study effects were highly heterogenous (I2 = 85%). Removing studies of poor and fair quality did not affect the SMD nor the precision of the effect estimate (SMD = −0.60; CI = −1.27 to 0.16). Five studies (102 and 99 participants in VR and control groups, respectively) also included a follow up measurement. The pooled effect on VAS was not statistically significant (SMD = −0.39, CI = −0.73 to 0.15), with high heterogeneity between the studies (I² = 89%). In summary, there is low quality evidence that VRT is not superior to conventional therapy in short-term and at follow-up assessment.

Figure 2. The meta-analysis for neck pain (VAS).

The figure shows a forest plot, combining studies that assessed the effects of virtual reality training on neck pain. Description: The figure shows a forest plot that contains 8 studies that assessed the effects of virtual reality training on neck pain. The overall effects favour virtual reality training over control interventions, but not statistically significantly. The studies are very heterogenous, some showing large effect and some pointing in the opposite direction.

Effects of VR on neck disability index (NDI)

Six studies included in the meta-analysis (125 and 127 participants in the VR and control groups, respectively) reported NDI scores (Figure 3). Overall, the studies tended to show a moderate but not significant effect VR (SMD = −0.49; CI = −1.05 to 0.06). Again, the study effects were highly heterogenous (I² = 78%). Five out of six studies (98 and 99 participants in the VR and control groups, respectively) also reported the NDI in a follow up period. Here, the overall effect was large and statistically significant (SMD = −0.95; CI = −1.70 to −0.20), but the heterogeneity between the studies was high (I² = 83%). In summary, low-quality evidence suggests that VRT training is not superior to conventional therapy in terms of NDI in the short term but may provide additional benefits at follow-up.

Figure 3. The meta-analysis for neck disability index (NDI).

The figure shows a forest plot, combining studies that assessed the effects of virtual reality training on neck disability index. Description: The figure shows a forest plot that contains 6 studies that assessed the effects of virtual reality training on neck disability index. The overall effects favour virtual reality training over control interventions, but not statistically significantly. The studies are very heterogenous, with one study showing especially large effect.

Effects of VR on Tampa scale for kinesiophobia

Three studies included in the meta-analysis (68 participants in the VR and control groups each) reported the Tampa Scale for Kinesiophobia (TSK) scores. The overall effect was small and statistically non-significant (SMD = −0.19; CI = −0.52 to 0.15), while the studies were very heterogenous (I² = 0%). All three studies also reported the scores at follow-up. Here, the overall effect was moderate and statistically significant (SMD = −0.69; CI = −1.34 to −0.03) with moderate heterogeneity between the studies (I² = 67%). In summary, moderate-quality evidence suggests that VRT provides no additional benefit in the short term compared with conventional therapy, whereas low-quality evidence suggests that VRT may provide additional benefits in the follow-up period.

Effects of VR on neck RoM

Four studies included in the meta-analysis (80 and 79 participants in the VR and control groups, respectively) reported on neck RoM. For neck flexion, the studies tended to show a positive overall effect of VR (SMD = 0.53; CI = 0.01–1.06), while moderate heterogeneity between the studies was present (I² = 67%). For the neck extension, the effect was slightly smaller and non-significant (SMD = 0.41; CI = −0.09 to 0.91), but still tended towards a positive effect of VR, while moderate heterogeneity between the studies was present (I² = 58%). For left and right rotation RoM (SMD = 0.06 and −0.29, there was no statistically significant effect CI = −0.25 to 0.38 for left rotation, CI = −0.70 to 0.11 for left rotation). Overall, low-quality evidence suggests a possible additional benefit of VRT over conventional therapy for neck flexion RoM, and no additional benefit for other RoM measures.

Neck RoM was also assessed in a randomised clinical trial, investigating VR exercises in comparison to neck exercises without VR [39]. This study was not included in the meta-analysis because they reported combined flexion and extension RoM, while the other studies reported RoM for each direction separately. While they reported a significant time effect for rotation RoM (p < 0.01), the interaction between group and time was not significant (p = 0.630), indicating no additional effect of VR over the control group. There were no statistically significant time effects (p = 0.46–0.71) nor for interactions (p = 0.18–0.99, p = 0.18, respectively) for flexion-extension and lateral flexion movements. In addition, the post-hoc tests showed no within-group differences in either group between baseline and follow up measures for rotation motion [39].

Discussion

The objective of this systematic review and meta-analysis was to examine the short-term (immediately after intervention) and long-term (at follow-up) effects of VRT on pain, RoM, kinesiophobia and perceived function in patients with CNP. In the meta-analysis, we compared the short and long-term effects of VRT with conventional exercise programs, performed by control groups. Interventions included a variety of exercises, including proprioceptive training, fast and accurate head control training, cervical joint mobility exercises, strengthening or stretching of the deep neck muscles, stair climbing, aqua therapy, pedalling, pulleys and obstacle walking. Some control groups also performed exercises with lasers. The VRT groups generally performed exercises similar to those in the control groups, but in a VR environment. Eleven studies were included in this review and eight were eligible for the meta-analysis. Low-quality evidence was found to suggest possible differences in favour of the VR interventions for NDI in the long term, for TSK in the long term, and for neck flexion RoM in the short term. However, no significant differences were found in pain intensity in the short and long-term, in NDI and TSK in the short-term, in neck extension RoM in the short-term and in both rotations in the short-term. VRT may achieve significantly better results in improving perceived function and kinesiophobia in the long term compared with conventional exercise programs. However, VRT and conventional exercise programs have got similar effects on overall neck RoM and pain intensity. Most evidence found in this review was of low quality, which calls for further research. Given the high statistical heterogeneity in the meta-analysis and many differences between the studies in terms of intervention and population, future studies should specifically assess the effects of different types and applications of VRT, explore the effects of VRT with respect to patient characteristics (e.g. pain duration, pain intensity, kinesiophobia), and include longer intervention and follow-up periods.

The tasks performed in the VR interventions varied widely, but were mostly intended to improve RoM, proprioception, motion accuracy and motion speed, and extinguish the movement-pain association (see Table 1 for details). Regardless of the task performed by subjects, by shifting attention to external stimuli, VRT aims to reduce pain and allow patients to experience movements they would otherwise be afraid of. With one exception [28], studies have used an immersive VR system in which subjects wore head-mounted displays to engage with the VR world. The diversity of VR interventions indicates that a critical evaluation of the broader literature is needed not only in terms of treatment effects, but also in terms of the technology, user interface, and purported mechanisms by which VRT provides additional benefits over conventional movement therapy.

Effects of VR on NP

Neck pain was assessed in eight studies included in the meta-analysis. Five of these studies also included a follow-up measurement. The meta-analysis showed that the pooled effect favoured VRT over conventional exercise both in the short term and at follow-up, but the effect was small to moderate and not statistically significant. Therefore, we can conclude that VRT has similar effects on neck pain as conventional exercise (i.e. the difference between VRT and convention exercise is not significant). In one study [28], the effect of VRT (relative to control group) was very large both after the intervention and at the 5-week follow-up. In this study a VRT program (a game called Cervigame^®) reduced NP about twice as much as a conventional proprioceptive exercise program, performed in the control group. This larger effect compared to other studies could be due to an effective VR program, or perhaps to a sample that included relatively young participants (VRT group: 36.19 ± 9.80 years; control group: 31.23 ± 9.49 years) with only mild functional disability according to NDI. We assume that younger patients may adapt more easily to modern forms of treatment, such as VRT, which may be the reason for the better outcomes. In their study, Liu et al. [42] investigated the differences in VR experience and acceptance between older adults and college students and suggested that VR headsets may be too complicated for older people to use on a daily basis. In addition, older people were more likely to report dizziness and anxiety during VR use [42]. Older adults may need more time to adapt to VR [43], thus, it could be hypothesised that VR would have more positive effects for them on the long-term, but the evidence for this hypothesis is currently lacking.

A study included in our meta-analysis investigated the effects of a home-based VR program in fighter and helicopter pilots with NP and showed that VR training had negative effects on VAS [38]. The authors of the study suggested this may be due to the busy schedules of the pilots, which led to poor compliance in self-exercising [38]. The VR home exercise program was also investigated in another study, which concluded that the benefits of using VR compared with traditional exercise were insufficient to recommend its use [29]. Therefore, supervised VR programs are recommended, because they show a more significant decrease in NP compared to home-based VR programs.

Heterogeneous conclusions regarding the effects of VRT on pain intensity in CNP patients are also found in other studies not included in our meta-analysis. While Zauderer et al. [44] reported a decrease in neck pain in their single-group study, Glavare et al. [45] found a significant increase in pain during the VR sessions and a nonsignificant increase in pain after intervention. Furthermore, Harvie et al. [46] found no differences in neck pain between baseline and post VR exercise assessment in their short-term study. Glavare et al. [45] attributed the increase in pain to a lack of inhibition and enhanced excitability in central nervous system processing of pain, which is present in several chronic pain conditions. An imbalance between inhibition and excitation in the central nervous system may promote pain during treatment in this patient group [45,47]. Note, however, that this effect was observed only during, not after, exercise [45].

Effects of VR on NDI

NDI was assessed in six studies included in the meta-analysis. Five of the six studies also reported the NDI in a follow-up period. The meta-analysis showed that the pooled effect favoured VRT compared with conventional exercise for both in the short-term and at follow-up, but the effect was statistically significant only at follow-up. Since we are mainly interested in long-term changes in patients with chronic pain, we can conclude that VRT can have a significant impact on perceived function in patients with CNP in comparison to conventional exercise. For the patients with chronic musculoskeletal conditions, the pain is an ongoing rather than a transient experience. Therefore, interventions that focus more on restoring function rather than escaping the experience of pain per se are more appropriate for them [48]. We believe that the improvement in perceived function after VRT is a more important outcome than improvement in pain from the perspective of CNP patients.

Effects of VR on TSK

TSK scores were reported in three studies included in the meta-analysis. All three studies also reported TSK scores at follow-up. The meta-analysis showed that the pooled effect statistically significantly favoured VRT only at follow-up. TSK was also assessed in two single-group studies that were not included in meta-analysis. Zauderer et al. [44] reported a decrease in TSK score after intervention and at follow-up, but did not indicate whether the difference was statistically significant. Glavare et al. [45] came to similar conclusions, reporting a significant decrease in TSK score after the VRT intervention. In their meta-analysis, Wang et al. [49] concluded that VR technology has the potential to reduce the level of kinesiophobia, especially when combined with exercise intervention. Furthermore, Asiri et al. [50] found that kinesiophobia was significantly correlated with pain intensity, proprioception and functional performance in CNP patients. In chronic pain patients, reducing kinesiophobia is crucial because confronting the fear during exercise or active exercises can accelerate recovery [49].

Effects of VR on neck RoM

Four studies assessed neck RoM after the VRT intervention. Meta-analysis showed that the pooled effect statistically significantly favoured VRT only for neck flexion, where the overall effect was just above the significance threshold. Neck RoM was also examined in two single-group studies that were not included in the meta-analysis. Tejera et al. [39] found no additional effect of VR over the control group performing neck exercises without resistance. Zauderer et al. [44] reported possible increases in neck flexion and left axial rotation, but no statistical analysis was performed in their study. From our results, we can conclude that VRT may have a positive effect on perceived function, but this does not mean that objective measures of RoM also improve significantly. Consistent with our conclusions, Rudolfsson et al. [51] observed weak associations between RoM measures and greater self-reported head movement related problems, activity limitations, symptoms and lower physical activity in people with neck pain.

Strengths and limitations

According to the Pedro scale, most of the studies included in the meta-analysis (six out of eight) had good methodological quality, and only one had fair and one poor quality. However, it should be noted that studies had small sample sizes (total sample size across studies ranged from 14 to 90). Most meta-analyses favoured VRT, with no statistically significant pooled effect, which may be the consequence of underpowered studies. In addition, study effects were highly heterogeneous. For the most part, the interventions were similar across the studies, with the main differences in the exercises performed in control groups, and methods used for VRT. In addition, studies differed in regard to the level of supervision (e.g. two studies included in meta-analysis performed home-based programs). It is possible that different conclusions would have been drawn if the studies had been more homogeneous and included a larger sample size.

Conclusions

The objective of this systematic review and meta-analysis was to investigate the short-term and long-term effects of VRT on pain, RoM, kinesiophobia and perceived function in patients with CNP. Low quality evidence for effects in favour of VR interventions on NDI (long-term), TSK (long-term), and neck flexion RoM (short-term) was found. However, no significant differences were found for pain intensity for short- and long-term, for NDI and TSK for short-term, for neck extension RoM for short-term, and for both rotations for short-term. VRT can achieve significantly better results in improving perceived function and kinesiophobia in CNP patients in the long term compared with conventional training programs. However, VRT and conventional training programs have similar effects on overall neck RoM and pain intensity. The quality of evidence in this review must be considered low, mainly because of the large heterogeneity between studies and the small number of studies.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

We would like to acknowledge the support of the Rectors Fund of the University of Primorska (internal post-doctoral project RAVOTEZ). The funder only provided part of the salary for one of the authors (Z.K.), but was not involved in any stage of manuscript preparation.