Background

Pain is the commonest symptom encountered in clinical practice often manifesting as an unavoidable consequence of medical procedures. Effective pain management is crucial to optimise medical procedures, boost patients’ satisfaction [1,2,3], reduce their anxiety, reduce hospital stay and minimise long-term analgesic dependence [4,5,6]. The use of immersive virtual reality (VR) technology has emerged as a potential tool to distract patients and to modify their perception of pain. Its adoption in clinical practice remains limited.

The search for effective, safe, and cheap analgesic treatment options is a priority accelerated in part by the emerging opiates epidemic in several countries associated with dependence risk and narrow safety profile [7, 8]. VR technology seems to offer a credible option for effective acute pain relief either as an alternative or as a combined treatment as part of a multi-modal pain relief strategy [9].

The term ‘virtual reality’ was coined by Jaron Lanier, a writer, musician, visual artist, and computer scientist, who first used it in 1986. The first application of VR in healthcare dates back to the beginning of the 1990s. It stemmed from the need to visualize complex medical data, especially when planning surgical treatment [10]. Since then, the use of VR technology in medicine proliferated into several domains including surgical training, neuropsychiatry, acute and chronic pain management, and rehabilitation [10, 11].

VR devices are designed to alter one’s perception of presence in an alternate reality and augment their immersion, and interactivity [12]. Today, several cheap and user-friendly devices offer an immersive environment largely delivered via high-resolution head-mounted displays (HMDs) with built-in sound capabilities [13]. In clinical practice, immersive VR experience aims to distract patients during medical procedures, suppressing their appreciation of immediate physical surroundings, allowing them to escape into an alternative reality away from the painful stimuli [14,15,16]. Early VR equipment had several technological barriers that limited their use in everyday practice, including high cost, relatively large size, complex operating interface, and user unfamiliarity [17]. Recent advances in audio-visual technology, driven by the wide use of smartphones, have enabled the development of affordable and user-friendly equipment [18]. Coupled with bespoke medical software, these new VR devices offer patients a versatile immersive visual and auditory experience that could be adopted across different clinical settings [11, 19].

Several meta-analyses have evaluated the efficacy of VR showing a beneficial effect with its use. Georgescu et al. [20] performed a meta-analysis for randomised trials that evaluated VR until 2018 (n = 27 RCTs, 1452 patients) showing a beneficial effect for pain reduction following medical procedure although the findings were limited by high heterogeneity and high trial risk of bias [20]. Scapin et al. [21] performed a systematic review including [22] randomised trials on the use of VR in burn patients. The findings were also supportive of the role of VR as an effective complementary drug strategy for pain relief in burn patients [21]. However, these reviews were either limited to specific clinical situations, suffered from high heterogeneity, or lacked detailed subgroup analyses to explore the reasons for heterogeneity [21].

In the year 2022, there have been 24 new randomised clinical trials (RCT) [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45] published evaluating VR technology highlighting the increased interest in this technology and offering further insight into its applicability across different medical disciplines. Still, the translation of this evidence has remained poor with respect to implementation of VR technology at scale and with variation in practice where medical specialities have taken steps towards adoption. Appreciation of the role of VR for pain relief can be aided by updated evidence synthesis [46].

In this systematic review, we conducted a comprehensive assessment of the evidence on VR efficacy as a non-invasive and non-pharmacological pain management method in patients undergoing different medical procedures. We performed an overall evidence synthesis pooling data from all relevant RCTs in addition to bespoke subgroup and meta-regression analyses to help interpret the evidence [17, 47].

Methods

We conducted this systematic review using a prospectively registered protocol (CRD 42020195919) [48] and reported in accordance with PRISMA guidelines [49].

Literature search

We searched major electronic databases (MEDLINE, EMBASE, Cochrance CENTRAL, CINAHL, and SIGLE) for randomised trials that evaluated the efficacy of immersive VR technology equipment for pain relief from inception until December 2022. We developed a comprehensive and inclusive search strategy using MeSH search terms and combined them using the Boolean ‘AND’ and ‘OR’ (Additional File 1: Appendix S1). We applied this search strategy to individual databases after amending it to the specification of each database. We then deduplicated the results and produced a final long list of citations. We manually searched the bibliographies of relevant studies to identify any additional trials not captured by our electronic database search. We also conducted supplementary searches in Google Scholar and Trip database to identify additional studies of relevance [50]. We did not apply any search filters or language restrictions. Relevant citations in non-English were obtained and translated for assessment against our inclusion criteria.

Study selection

Five independent reviewers (JJT, DP, SH and RP, AK) completed the study screening and inclusion process in two stages. First, titles and abstracts were screened to identify potentially relevant studies following which, the full text of relevant articles were reviewed against our inclusion criteria. We included all randomised trials of any design that evaluated the efficacy of any immersive VR technology equipment for pain relief during any medical procedure, including labour and childbirth. We initially planned this review to include only adult participants and later extended this to include paediatric participants to provide a more comprehensive evidence synthesis. We excluded non-randomised studies, review articles, and animal studies. We also excluded studies that assessed distraction techniques only (e.g. a display screen with no immersive capabilities), studies in dental procedures, and those that did not assess pain using a standardised measurement tool or reported on pain scores more than an hour after the procedure. Discrepancies and disagreements between reviewers were discussed and resolved in consensus with two additional reviewers (MPR and BHA).

Data extraction

Three reviewers (JJT, DP, SH, AK) extracted data in duplicate using a piloted electronic data extraction tool. We collected data on study design (crossover vs parallel), intervention settings, population characteristics, inclusion and exclusion criteria, type of VR technology and equipment used, nature of the medical procedure or intervention, loss to follow-up, and dropouts. Our primary outcome was pain scores measured immediately after or within an hour of the procedure. We also collected data on anxiety scores where relevant. In trials including paediatric patients, we included the parents’ reported pain scores.

Assessment of risk of bias

We assessed the risk of bias in included trials in duplicate (JJT, RP, AK, DP, MPR, SH) using the Cochrane Risk of Bias assessment tool 2.0 [51]. We assessed studies in five domains: participant randomisation and sequence generation, allocation concealment, outcome assessment, completeness of outcome data, and selective outcome reporting. Due to the nature of the intervention, we did not penalise unblinded trials. Studies with a crossover design were assessed using a modified version of an established tool [52]. We assessed the risk of bias in these studies for appropriate crossover design, randomisation and order of receiving the treatment, risk of carry-over effect, data collection, allocation concealment, outcome detection, data completeness, and selective outcome reporting.

Data synthesis

We pooled data using a meta-analysis with a random effect and adjusted using restricted maximum likelihood (REML) [53]. We reported on the difference in pain scores measured using standardised mean difference (SMD) with 95% confidence intervals (CI). We assessed any detected heterogeneity using the I 2 statistics. The I 2 index is an approach to quantify heterogeneity in meta-analyses. I 2 provides an estimate of the percentage of variability in results across studies that is due to real differences and not due to chance. The I 2 index measures the extent of heterogeneity by dividing the result of Cochran’s Q test and its degrees of freedom by the Q-value itself. An I 2 of less than 25% is usually viewed as low heterogeneity, between 25 and 50% as moderate, and over 50% as high heterogeneity.

We planned subgroup analyses to investigate potential effect modifiers (patient age group (paediatric patients defined as < 16 years old) vs adults), type of medical intervention (venepuncture-related procedures, minimally invasive medical procedures (defined as any medical procedure conducted in office setting without the need for general anaesthesia), dressing changes in burn patients, and childbirth), trial design (parallel group vs crossover trials), the trial quality as assessed using the risk of bias tool), the type of VR technology (interactive: arbitrarily defined when VR software is asking the participant to take part in specific activities compared to a passive VR experience), the VR delivery settings (inpatient vs outpatient vs emergency department) and assessed their impact on the effect estimates using a meta-regression [54]. We explored potential sources of heterogeneity using a leave-one-out analysis and a sensitivity analysis excluding potential outliers. We also investigated the risk of publication bias using Egger’s test, a funnel plot, and Galbraith plot to identify potential outliers [55]. Where publication bias was detected, we explore potential impact using the trim and fill method [56] to estimate and adjust for the number and outcomes of missing studies in the meta-analysis. We conducted a cumulative meta-analysis for selected outcomes to evaluate temporal trends and changes in effect estimate over time as new trials emerged [57]. Statistical analyses were conducted in STATA V17 (StataCorp, TX) and Open Meta-analyst software (Brown University; Providence, RI, USA).

Patient and public involvement

No input was sought from lay service consumers in the design, conduct, and reporting of this systematic review.

Results

We identified 51,140 potentially relevant citations, of which we assessed 132 studies against our inclusion criteria and included 90 articles reporting on 92 unique RCTs in our meta-analysis (7133 participants) (Fig. 1) (Additional File 1: Appendix S2. (40 studies were excluded [58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99]). No relevant citations were identified in non-English. The majority of included RCTs had a two-group parallel design (77/92, 84%), including a three-arm RCT [100], and less than one fifth had a crossover design (15/92, 16%).

Fig. 1
figure 1

Selection and inclusion process for randomised trials evaluating the effectiveness of virtual reality for pain control in medical procedures

Full size image

Forty-two of included trials enrolled adults only (42/92, 46%), three had a mixed population, and fifty included paediatric participants only (50/92, 54%). The majority of trials were conducted in high-income countries; twenty-seven trials were conducted in the USA (27/92, 29%) while nineteen were conducted in Turkey (19/92, 21%), seven trials in Australia, Canada China respectively (7/92, 8%) (Additional File 1: Table S1).

The type of VR technology and equipment used across included trials evolved over time from interactive, immersive games hosted on a personal computer to immersive environment experiences with user-controlled interactive interface and real-time feedback (Table 1). Trials conducted over the last 10 years evaluated newer VR technology with sound immersive augmentation (13/47, 26%) [101,102,103,104,105,106,107,108,109,110,111,112,113] and hand-held mobile phones or mounted goggles (27/47, 58%) [100, 102, 103, 105, 106, 110, 111, 113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131].

Table 1 Description of the VR equipment and software used in randomised trials evaluating the effectiveness of virtual reality for pain control in medical procedures
Full size table

Risk of bias

For parallel-group RCTs, the overall quality of the included studies was moderate with the majority of studies showing low or moderate risk of bias for selective reporting (73/77, 95%), outcome assessment (72/77, 94%), completeness of data (74/77, 91%) and randomisation risk of bias (70/77, 91%). Still, nine trials showed high risk for adherence to intervention groups (9/77, 12%), and none reported blinding participants or assessors (Additional File 1: Figure S1). The majority of crossover trials showed a high or unclear risk of bias, specifically for carry-over effect (13/15, 87%), completeness of data (6/15, 40%) and detection bias (12/15, 80%). The risk of bias for allocation concealment was deemed to be high in ten crossover trials (10/15, 67%) (Additional File 1: Figure S1).

Outcomes

Pain

We pooled data from 83 RCTs that reported on pain scores following any medical procedure with nine RCTs excluded from the meta-analysis due to limited outcome reporting ((Additional File 1: Appendix S3). Our meta-analysis showed a significant reduction in pain scores with the use of VR across all types of medical procedures (n = 83, SMD − 0.78, 95% CI − 1.00 to − 0.57, p = < 0.01), although heterogeneity was high (I 2 = 93%) (Fig. 2). We explored the heterogeneity using meta-regression which showed no significant effect of different covariates, including crossover trials (p = 0.53), minimally invasive procedures (p = 0.37) or among paediatric participants (p = 0.27). (Additional File 1: Table S2). We conducted a cumulative meta-analysis to illustrate the chronological change in the effect size of VR on reducing pain which showed no change in overall effect estimates with the addition of new RCTs since 2018 (Fig. 3). We also calculated the predictive intervals of the pooled effect estimated which shows that 95% of the true effect size falls between − 4.02 and 1.05 for all comparable populations.

Fig. 2
figure 2

Meta-analysis on the effectiveness of VR technology for pain control compared to routine care across different medical procedures

Full size image
Fig. 3
figure 3

Cumulative meta-analysis on the effectiveness of VR technology for pain control compared to routine care across different medical procedures

Full size image

We performed subgroup analyses across these three identified categories (trial design, type of medical procedure and participant age group). The reduction in pain scores was consistent across crossover trials (n = 13, SMD − 0.86, 95% CI − 1.23 to − 0.49, I 2 = 72%, p = < 0.01) and parallel-group trials (n = 70, SMD − 0.77, 95% CI − 1.01 to − 0.52, I 2 = 90%, p = < 0.01) (Additional File 1: Figure S2). Similarly, VR reduced pain across the different participant age groups, though the effect was higher in paediatric participants (n = 43, SMD − 0.91, 95% CI − 1.26 to − 0.56, I 2 = 87%, p = < 0.01) compared to adults (n = 40, SMD − 0.66, 95% CI − 0.94 to − 0.39, I 2 = 89%, p = < 0.01) (Additional File 1: Figure S2). The efficacy of VR in reducing pain was significant in participants undergoing venepuncture-related procedures (n = 32, SMD − 0.99, 95% CI − 1.52 to − 0.46, I 2 = 90%, p = < 0.01), minimally invasive medical procedures (n = 25, SMD − 0.51, 95% CI − 0.79 to − 0.23, I 2 = 85%, p = < 0.01.), dressing changes in burn patients (n = 19, SMD − 0.8, 95% CI − 1.16 to − 0.45, I 2 = 87%, p = < 0.01) and during childbirth (n = 7, SMD − 0.99, 95% CI − 1.59 to − 0.38, I 2 = 88%, p = < 0.01) (Additional File 1: Figure S2). The use of interactive VR technology did not yield significant difference (SMD − 0.72, 95% CI − 1.11 to − 0.34, p = 0.00) compared to using non-interactive software (SMD − 0.78, 95% CI − 0.99 to − 0.57, p = 0.00). VR was effective in reducing pain across different care settings including inpatient (SMD − 0.79, 95% CI − 1.02 to − 0.57, p = 0.00), outpatient (SMD − 0.87, 95% CI − 1.45 to − 0.28, p = 0.28), and emergency department (SMD − 0.80, 95% CI − 1.70 to 0.11, p = 0.00).

We assessed publication bias using Egger’s test, which was significant (p = 0.11). We visually inspected the variance in effect estimates for potential small study effect using a funnel plot (Additional File 1: Figure S3) and a Galbraith plot (Additional File 1: Figure S3) which identified several outliers although the overall precision in the effect estimate was high. We explored the potential impact of publication bias using the trim and fill method which did not identify any missing studies (Hedge’s g 0.00, 95%CI − 0.051 to 0.051) (Additional File 1: Figure S3).

We conducted a leave-one-out analysis, which identified five studies as potential outliers [101, 126, 131, 134, 153, 154]. We then conducted a sensitivity analysis excluding these trials, which led to a small reduction in the overall effect estimate (SMD − 0.58, 95% CI − 0.71 to − 0.45), but did not resolve the observed heterogeneity (I 2 = 82%).

Anxiety

Thirty-one trials reported on changes in anxiety between the VR group and routine care, mainly involving minor medical procedures and venepuncture procedures [101, 111, 117, 120, 127, 128, 142]. The overall effect estimate showed a significant reduction in anxiety scores with the use of VR across all populations, although heterogeneity was high (n = 31, SMD − 0.82, 95% CI − 1.09 to − 0.54, I 2 = 91%, p = < 0.01) (Additional File 1: Figure S4). The cumulative meta-analysis showed more precise effect estimates with the addition of newer trials over the last 2 years, although the confidence interval remained relatively wide (Additional File 1: Figure S4).

We performed subgroup analyses across these three identified categories (trial design, type of medical procedure, and participant age group). The effect of VR technology on anxiety reduction was higher among paediatric participants (n = 15, SMD − 0.96, 95% CI − 1.37 to − 0.54, I 2 = 91%, p = < 0.01) compared to adults (n = 16, SMD − 0.68, 95% CI − 1.04 to − 0.32, I 2 = 91%, p = < 0.01) (Additional File 1: Figure S5). Reduction in anxiety was highest among trials that evaluated venepuncture-related procedures (n = 15, SMD − 0.99, 95% CI − 1.39 to − 0.58, I 2 = 90%, p = < 0.01) followed by minor medical procedures (n = 10, SMD − 0.42, 95% CI − 0.77 to − 0.07, I 2 = 84%, p = < 0.01) and childbirth (n = 4, SMD − 1.48, 95% CI − 2.19 to − 0.76, I 2 = 93%, p = < 0.01). However, the effect was not significant for dressing changes (n = 2, SMD − 0.17, 95% CI − 0.49 to 0.15, I 2 = 0%, p = 0.56) (Additional File 1: Figure S6). The reduction in anxiety was significant across parallel-group trials (n = 13, SMD − 0.85, 95% CI − 1.14 to − 0.56, I 2 = 91%, p = < 0.01) but not in crossover trials (n = 2, SMD − 0.31, 95% CI − 0.80 to 0.17, I 2 = 38%, p = 0.20) (Additional File 1: Figure S6).

Only 46 trials reported on side effects with the use of VR technology (46/92, 59%). The far majority reporting mild side effects including nausea, vomiting, and headache. No serious side effects were reported (Additional File 1: Table S1).

Discussion

Summary of main findings

Our review summarised evidence sought from different medical disciplines evaluating the efficacy of VR technology. Despite heterogeneity, the reduction in pain perception was consistent across different clinical settings, medical procedures, and patient characteristics. We identified a relatively high number of relevant trials, particularly within the last 5 years. This was associated with a gradual development in the VR equipment used moving from larger head mount display screens to lighter and cheaper smartphones interfaces [100,101,102,103, 105,106,107, 115, 116, 118,119,120,121,122, 125,126,127,128, 130, 133, 155, 159, 160, 164]. The reduction in pain scores was observed across all evaluated medical procedures, participant age groups and trial designs, which increased the generalisability of our findings.

Implications for clinical practice

The rapid progress in immersive VR technology has facilitated its evaluation within different clinical settings driven by smaller, cheaper, and more user-friendly VR equipment. VR immersion was defined as according to this point of view VR is described as ‘an advanced form of human–computer interface that allows the user to interact with and become immersed in a computer-generated environment in a naturalistic fashion’ [169].

As this technology becomes more widespread within the general population, its use within the health sector will gradually become mainstream with higher user acceptability and satisfaction [170]. Unlike other disciplines, e.g. engineering [171] and education [172], where VR use has grown organically, introducing it into healthcare requires deliberate implementation steps to ensure feasibility and patients’ safety [173]. Considering the beneficial effect observed in our meta-analysis, we argue that health policy makers should incorporate the use of VR within their pain management guidelines to enable its safe adoption [174]. This is particularly relevant for certain patient groups, such as in paediatric phlebotomy [175].

Implications for future research

Our review is focused on evaluating VR technology in acute pain relief settings, largely using non-standardised software. Such versatile and easy-to-use technology has the potential to help chronic pain patients within the community enabled by virtual reality meditation and mindfulness techniques [176]. Similarly, developing procedure or condition-specific software could also help to maximise its analgesic effect as shown by some early experimental studies [177]. Lastly, clinical implementation pathways should consider the ideal format, frequency, and timing of using VR for medical procedures as per local feasibility.

Previous systematic reviews [20, 178,179,180] called for larger trials to address the perceived heterogeneity. Our trim and fill analysis suggests that larger trials are unlikely to nullify the depicted cumulative beneficial effect across the trials included in our analysis, thus offering low added value.

The majority of the included trials in our review focused on acute pain control following medical intervention. VR could be a game-changer to convert several inpatient procedures to outpatient settings, thus driving down cost, hospital stay, and in-hospital complications [120].

The reduction in pain management cost alone could offer a substantial advantage to reduce the length of hospital stay and associated costs, which was estimated at around $5.4 per patient (95% CI − 11 to 156) with VR use compared to routine care [181]. In this case, VR will prove dominant without the need for a formal cost-effectiveness study.

Most of the included trials used varied pain scales with no clear justifications, which may have led to higher heterogeneity at evidence synthesis. Adopting available standardised and validated outcome measurement tools would enable precise evidence synthesis and help to eliminate across trial heterogeneity. Leveraging the advances in VR user interfaces could enable interactive and contemporary built-in outcomes assessment, thus eliminating assessment bias in future studies.

Strengths and limitations

The main strength of our review stems from our comprehensive approach to evaluating the efficacy of VR technology across different medical disciplines in contrast to previous reviews that focused on particular patient demographics or medical conditions [21]. We undertook a prospective registration, employed an exhaustive search strategy, and evaluated the sources of bias. We followed an established methodology to explore potential sources of heterogeneity and evaluated the risk of publication bias.

Our findings suffered some limitations, most notably the heterogeneity of effects among included trials. We explored this heterogeneity in a meta-regression which suggested a higher effect in minor procedures and in trials involving children. However, the observed beneficial effect pertaining across all evaluated subgroups with relatively narrow confidence intervals supports the overall benefit of VR technology for pain control. We explored this heterogeneity using a cumulative meta-analysis which confirmed that future trials are unlikely to change the certainty in the beneficial effect of VR in reducing pain following medical procedures. The prediction intervals also suggest that most population would see a benefit from using VR although a small portion might not observe this benefit (Additional File 1: Figure S7).

A potential source of heterogeneity could stem from the assumed variation in the reported common comparator (routine care). Several analgesic agents, doses, and frequencies could have been used in the control group across included studies which we were unable to adjust for in our analysis.

Several factors could drive this heterogeneity, including variations in the common comparator, background, type of software (e.g. interactive vs static), hardware fidelity, procedure and exposure duration, patient morbidity and pain tolerance, and measurement assessment tools. Exploring these effect modifiers is only possible using individual patient data. However, such analysis might fail to add significant value especially when evaluating a subjective outcome such as pain, even within the context of an individual patient data meta-analysis [182].

Most of the included studies had a small sample size, with some evident outliers identified on the funnel plot. To address the risk of publication bias, we conducted a cumulative and one-out trial analysis, excluding obvious outliers, which helped us to refine the effect estimates. While some of the included crossover RCTs suffered from risk of bias [124, 125, 131], our subgroup analysis supported the overall beneficial effect of VR across both crossover and parallel-group RCTs. Majority of the included crossover trials only reported on the effect estimates after the final crossover step which limited our ability to adjust for the potential risk of bias when pooling data from such trials. We explored the limitation of evidence sought from crossover trials using a subgroup analysis which demonstrated a wider confidence intervals compared to evidence from parallel-group trials. However, evidence of reduction in pain scores remained significant (Additional File 1: Figure S2).

Lastly, we were unable to report on the planned secondary outcomes in our protocol due to limitations in reporting across the included trials. 

Conclusions

Immersive VR technology offers effective pain control across various medical procedures, albeit statistical heterogeneity, albeit statistical heterogeneity. Further research is needed to inform the safe adoption of this technology across different medical disciplines.