This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. The PRISMA checklist is given in Multimedia Appendix 1. The study protocol was registered in the PROSPERO database (CRD42022382331). The Cochrane Handbook for Systematic Reviews of Interventions (version 5.1.0) was followed [27].

The inclusion criteria were as follows:

No restrictions were imposed on language or publication date.

The retrieved studies were imported into Endnote X9 software (Clarivate), which was used to eliminate duplicate studies. Two independent reviewers (authors TTZ and FW) performed the initial screening of the literature by reading the titles and abstracts of all retrieved studies, and studies that did not meet the inclusion criteria were excluded. Next, the full texts of the remaining studies were screened. Any disagreements were resolved by negotiation and discussion with a third reviewer (author XZ).

Two independent reviewers (authors FW and XL) extracted detailed information, including the name of the first author, the year and country of publication, the language of publication, study design, the number of included subjects (% female), diagnosis, and outcome indicators. Information about the characteristics of the interventions, including dose, frequency, and duration, was also collected for both the VR group and the control group. The sample size and mean (SD) of the outcome indicators in each group were collected. When the same group of participants was reported in different studies, the group with the largest sample size was selected for inclusion in this review to avoid duplicate publications [28]. For information that could not be confirmed, the authors were contacted by email. The 2 reviewers cross-checked the data at the end of the extraction, and any disagreements were resolved by negotiation.

The methodological quality of the included studies was independently assessed by 2 reviewers (authors XL and ZFH) using the Cochrane Risk of Bias tool, and the studies were classified as having a low, unclear, or high risk of bias [29]. Disagreements were resolved by consulting a third reviewer (author QD). The Egger test and funnel plots generated with Stata 14.0 software (StataCorp) were used to evaluate potential publication bias. The trim-and-fill method was used to adjust for funnel plot asymmetry due to publication bias [30]. Sensitivity analyses were performed by removing each study separately to assess the robustness of the results [29]. The overall strength of the evidence was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) criteria [31].

This systematic review and meta-analysis were performed using Review Manager software 5.4 (Informer Technologies) and Stata 14.0 software. Heterogeneity was tested using the I² statistic. A fixed effects model was selected for the outcome indicators if I²<50%, while a random effects model was used when there was significant statistical heterogeneity (I²>50%, P<.05). The effect size used to synthesize the outcome measure was the weighted mean difference (WMD). Three subgroup analyses were performed to explore the possible causes of heterogeneity among the studies: the region of CSP (CNP vs CLBP), VR types (IVR vs NIVR), and treatment duration (<4 weeks vs ≥4 weeks).

A total of 924 records were obtained from the 5 databases and the trial registration platform. A total of 394 (42.6%) duplicates were identified and removed using Endnote X9 software. After screening the titles and abstracts, 40 (7.5%) of the remaining 530 RCTs were retained, and 490 (92.5%) were excluded for the following reasons: (1) the study population included patients without CSP, (2) the intervention did not use VR therapy, (3) the type of study was a non-RCT, (4) the information was incomplete, and (5) the patients also received analgesic medication. Of the 40 studies, 15 (38%) were retained after reading the full text and 25 (62%) were excluded for the following reasons: (1) the study population included patients without CSP, (2) the intervention did not use VR therapy, (3) the type of study was a non-RCT, (4) the information was incomplete, and (5) the patients also received analgesic medication. Two additional RCTs were retrieved from the reference lists of the included studies. One RCT was retained after the full text was read, and the other was excluded due to incomplete information. A total of 16 studies were included in this review, 15 (94%) of which reported sufficient data (eg, mean [SD], sample size) on the analgesic effect of VR for CSP. Therefore, 15 studies were included in the meta-analysis. The PRISMA flowchart of selecting the included studies is shown in Figure 1.

The CSP reported in the included studies included CLBP [32-43] and CNP [44-47]. All patients had chronic pain that persisted for more than 3 months. The sample size varied from 8 to 90 participants, and the mean age ranged from 18 to 85 years. The characteristics of all the studies are summarized in Table 1.

The types of VR interventions included IVR [32,35,42,44-47] and NIVR [33,34,36-41,43], which were classified based on the degree of isolation participants experienced when interacting with the virtual environment during VR therapy. NIVR uses a wall-mounted screen or a computer monitor as the vehicle for VR content, while IVR uses a headset or head-mounted display [48]. Compared to NIVR, IVR can increase the user’s sense of presence by improving immersion through the addition of auditory or haptic feedback [49]. The duration of a single VR session ranged from 2 to 40 minutes, and the frequency of treatment ranged from 5 to 7 times a week; all the included studies ranged in duration from a single exercise session to 8 weeks. For the control groups, 5 (31%) studies performed conventional balance function training [33,34,38-40], 5 (31%) performed conventional physical therapy [36,37,41,46,47], 2 (13%) performed core training [43,44], and the remaining conducted treatments, including sham VR [32], conventional multimodal pain therapy [42], waiting lists [45], and standard care [35]. The intervention details are summarized in Table 2.

The risk of bias in the 16 (100%) studies included in the meta-analysis is presented in Figure 2. Overall, 10 (63%) studies showed a high risk of bias. In addition, 15 (94%) RCTs generated an adequately randomized sequence, and 9 (60%) of them were analyzed using a blinded method for outcome measurement. Ratings using the GRADE methodology for all outcome measurements were inconsistent and ranged from moderate to low quality (Multimedia Appendix 3). Therefore, the quality of evidence from most studies was classified as fair.

The Egger test indicated significant publication bias in the results for pain intensity (P=.03; Figure 5). The sensitivity analysis for pain intensity revealed that removing each study separately did not significantly affect the pooled results, thus indicating that the results are robust (Figure 6). The trim-and-fill method was performed, and it was estimated that there were 4 missing studies. The pooled estimates (95% CIs) calculated for the fixed effects model and the random effects model were –2.30 (–2.42 to –2.18) and –2.06 (–2.50 to –1.61), respectively (Figure 7). No significant changes in the results were observed before or after pruning or filling, indicating that our results are robust and plausible.

The primary purpose of this meta-analysis was to compare the relative efficacy of VR therapy and other therapies (eg, conventional therapy, sham stimulation, and standard care) for treating CSP. The results indicated that VR therapy can effectively relieve CSP. The results of subgroup analyses showed that VR is a beneficial pain management strategy for patients with CNP and CLBP. For different types of VR, subgroup analyses showed that compared to the control group, IVR and NIVR both significantly improved CSP. No statistically significant differences were found between patients who underwent VR treatments for a duration of <4 weeks and a duration of ≥4 weeks. VR was associated with a significant improvement in inflammatory marker levels but not in the fear of movement, spinal ROM, or disability level. VR was found to be well tolerated among these patients.

The primary result suggested that VR reduces self-reported pain intensity, which might be explained by several implicating mechanisms [52,53]. A previous study reported that abdominal muscle strength is significantly lower in people with LBP [54], and a lack of strength in the core trunk muscles can lead to a decrease in intra-abdominal pressure, affecting spinal stability [55]. VR, as a novel human-computer interaction approach, can stimulate and mobilize the sensory system during training and results in changes in neuroplasticity and enhanced performance of relevant muscle groups, promoting a new motor learning process and leading to increased spinal stability [37,56], which would benefit pain relief. Furthermore, previous studies have reported that an intervertebral disc undergoes aging or pathological changes in the adjacent region in patients with CSP, exposing cells within the nucleus pulposus to macrophages, resulting in an inflammatory response that might trigger pain [7,8]. VR therapy may enhance the activity of disc fibroblasts and increase the thickness of the multifidus muscle [39,57], which is beneficial for relieving pain intensity. Furthermore, pain is an unpleasant subjective sensation associated with actual or potential tissue damage and is correlated with the degree of patient attention given to the pain area [58-61]. The various virtual game environments and real-time feedback methods are the most eye-catching features in the VR training process; these methods can be used to attract the patient’s visual and auditory attention to achieve motor performance, while relatively less attention has been given to the effects of VR on pain [62,63].

Although the high heterogeneity of the primary outcome and the results of the subsequent subgroup analyses suggest that the region of CSP, VR type, and treatment duration may play a role in the heterogeneity, the results of the sensitivity analysis indicate that these differences are more likely to be caused by 6 studies [33-35,37,38,40], which included participants of different ages.

VR therapy significantly improved the levels of inflammatory markers, including CRP, TNF-α, and IL-6. Numerous studies have previously reported an association between CSP and changes in inflammatory cytokines, such as IL-1 and TNF-α, which are thought to be closely related to the pathogenesis of disc herniation and degeneration [64,65]. Similarly, Nambi et al [66] reported that 4 weeks of VR training could significantly decrease pain intensity, increase functional impairment, and improve CRP, TNF-α, IL-2, IL-4, and IL-6 levels. However, the limited number and low quality of the included studies need to be noted, and further RCTs with large samples and rigorous study designs are needed to elucidate these results.

Patients with CSP may engage in fear/avoidance behaviors to avoid pain and protect themselves by limiting spinal motion, which ultimately affects spinal mobility and the speed of movement [67,68], with the degree of pain catastrophizing being proportional to the degree of disability [69,70]. However, we found no statistically significant differences in fear avoidance beliefs after the VR intervention but at the 3-month follow-up [47]. A systematic review and meta-analysis reported that VR therapy enhances spinal ROM and physical functioning in patients with CNP [26]. We failed to observe significant differences in the spinal ROM or disability level after VR intervention compared to those in the control group, which may be attributed to the relatively short duration (0-8 weeks) of the VR intervention (the reported mean duration was 4.81 weeks).

Several limitations need to be addressed in this meta-analysis. First, the pooled analysis of the studies may be imprecise due to the large heterogeneity and the low quality of evidence from most of the included studies, and the results should be interpreted with caution. Second, the optimal duration of treatment for CSP could not be determined. Third, the effectiveness of VR therapy in patients with CSP and its analgesic effects in long-term follow-up must be further explored in high-quality studies. Fourth, indicators related to quality of life, such as depression and anxiety, should be emphasized and investigated in depth in future studies of patients with CSP.

VR therapy is an innovative and effective analgesic method that has beneficial effects on inflammatory markers in patients with CSP compared to other therapies (sham stimulation, usual care, conventional treatment). However, this approach may not have significant effects on the fear of movement, spinal ROM, or disability level. Notably, the quality of the evidence from the RCTs included in this study ranged from moderate to low. Therefore, we recommend that readers interpret the results of this study with caution. Future trials with large sample sizes, rigorous designs, and long-term follow-up periods are needed to explore the clinical significance of these differences and key issues in patients with CSP and to elucidate the underlying mechanisms of VR.