The study protocol is published elsewhere and provides additional detail [43]. We conducted a single-cohort, double-blinded (participant and analysts), placebo-controlled randomized clinical trial in an online convenience national sample of community-based individuals with self-reported cLBP. Study participants were participating in a longer study of 8.5 months’ duration that involves multiple additional posttreatment assessments not reported here. The study was approved by the Western Institutional Review Board on July 2, 2020, and data collection for this report was completed in November 2020. This report is constrained to the end of treatment time point (day 56).

Community-based individuals with cLBP were recruited nationally through chronic pain organizations (eg, American Chronic Pain Association) and through Facebook online advertisements. Additionally, study advertisements were emailed to professional contacts at several medical clinics with requests to forward among medical colleagues nationally. All study advertisements directed interested individuals with cLBP to the study website for information and invitation to complete an automated online eligibility form that screened for inclusion and exclusion criteria (Textbox 1). This screening process was completed over the phone for 1 individual due to technical difficulties. Individuals determined to be eligible for the study were invited to participate in a study examining the effectiveness of an 8-week VR wellness program in helping them manage chronic lower back pain. If willing to participate, participants completed an electronic informed consent (eConsent; see Multimedia Appendix 1 for the study consent form) and provided their e-signature to complete their enrollment in the study. Figure 1 displays the participant study activities.

Enrolled participants completed a baseline survey and then received 3 pain surveys over a 2-week pretreatment assessment period. Those who completed at least one survey during this pretreatment period progressed to the treatment phase of the study, which included an 8-week VR program (therapeutic VR or Sham VR), twice-weekly surveys during the VR treatment phase, and a final survey administered at treatment completion on day 56. All study procedures occurred remotely.

Study participation was compensated in 2 ways. First, participants received US $6 per completed survey (US $126 possible; prorated) in the form of an Amazon eGift Card after the day 56 survey and upon return of their VR headset (prepaid shipping containers were provided). Second, all participants were eligible to receive a gift VR headset 6 months after their completion of treatment if they completed 16 or more study surveys, confirmed their interest in receiving a VR headset, and returned their VR headset. A total of 77 study participants met these criteria and will receive headsets (Textbox 1).

Participants were randomized 1:1 and allocated to 1 of 2 treatment groups: (1) the 56-day pain relief skills VR program (EaseVRx) or (2) a 56-day control VR condition (Sham VR). REDCap Cloud (nPhase, Inc.) was used to apply an automatic and blinded randomization program and ensure equal allocation to both groups. Participants and study statisticians were blinded to treatment group assignment. Statisticians performed blinded analysis of data sets that were randomly labeled group A and group B with statistician unblinding occurring only after posttreatment month 3 data were collected and the data set locked (posttreatment month 3 data not yet analyzed). The 2 study coordinators (LG and IM) unblinded to individual treatment group assignments were not involved in any data analyses. Study participants remain blinded to treatment group assignment until their participation in the larger study is completed (8.5 months after randomization). The larger study quantifies long-term outcomes for the current treatment study; the study protocol and details are published elsewhere [43].

Participants in both treatment groups (EaseVRx and Sham VR) received a Pico G2 4K all-in-one head-mounted VR device at no cost through postal mail. The Pico G2 4K device was used because they are commercially available, widely used, inexpensive, have minimal visual latency, and are easier for participants to use than many other devices. This hardware allows for displaying 3D images (EaseVRx) and 2D images (Sham VR). While each VR device contained software specific to the individual participant’s assigned VR treatment group, all device packaging and directions for use were common to both treatment groups. Participants in this study were provided with online access to instructional materials outlining general use and set up of the headset. Relevant to the EaseVRx group, user exhalation is measured by the microphone embedded in the Pico G2 hardware, offering biodata-enabled immersive therapeutics. Participants in both groups were instructed to complete 1 VR program session daily for 56 days. Study staff monitored participant completion of the twice-weekly surveys and device use. Study staff provided reminders to complete surveys and otherwise were available upon request for technical support. The sections below describe the elements of the study interventions.

Participants randomized and allocated to this treatment group received an immersive multimodal, skills-based, pain self-management VR program, called EaseVRx (AppliedVR), that incorporates evidence-based principles of CBT, mindfulness, and pain neuroscience education. The program content trains users on evidence-based pain and stress management strategies via immersive and enhanced biofeedback experiences. EaseVRx combines biopsychosocial education, diaphragmatic breathing training, relaxation response exercises that activate the parasympathetic nervous system, and executive functioning games to provide a mind–body approach toward living better with chronic pain. The standardized 56-day program delivers a multifaceted combination of pain relief skills training through a prescribed sequence of daily immersive experiences. Each VR experience is 2-16 minutes in length (average of 6 minutes). The VR treatment modules were designed to minimize triggers of emotional distress or cybersickness. Treatment module categories included:

In compliance with VR-CORE clinical trial guidelines, we selected an active control that utilizes nonimmersive, 2D content within a VR headset as the most rigorous VR placebo [30]. The Sham VR headset displayed 2D nature footage (eg, wildlife in the savannah) with neutral music that was selected to be neither overly relaxing, aversive, nor distracting. The experience of Sham VR is similar to viewing nature scenes on a large-screen television and is not interactive. Twenty videos were rotated over the 56 sessions, with average duration of sessions closely matching those of EaseVRx (Figure 2).

In accordance with the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) recommendations, we included multiple measures to evaluate the importance of change in outcomes across 4 recommended domains: pain intensity, health-related quality of life and functioning, and ratings of overall improvement [46-48]. Additionally, measures and individual items were included to align directly with the National Institutes of Health (NIH) Pain Consortium’s Report on Research Standards for Chronic Low Back Pain [49] or assess the domains recommended in the report. The study was performed in accordance with the CONSORT (Consolidated Standards of Reporting Trials) guidelines [50] (see Multimedia Appendix 2 for the completed checklist) and the recommended extension for reporting of psychological trials [51]. The Western Institutional Review Board approved the study (Puyallup, WA).

Data collection included electronic participant-reported measures and objective VR device use data collected from the VR devices.

Data were collected across 3 phases of the study: pretreatment (days –14 to 0), active treatment (days 1-55), and end of treatment (day 56). The 14-day pretreatment phase involved administering the pain surveys 5 times (baseline, days –10, –7, –3, and 0); these measures were averaged within participants to establish a single pretreatment score for each variable assessed. During the 8-week active treatment phase, surveys were distributed biweekly (15 total surveys during treatment) and at end of treatment on day 56. Accordingly, there were 17 time points per participant.

Table 1 outlines the timeline of variable assessment. This section details the measurement and methods used to assess each variable.

Participants were provided with study staff contact information and encouraged to contact as needed and in the event of any problems using their device or with their treatment. Similar to other studies, cybersickness was assessed at the end of treatment [15]. However, due to a problem with the electronic survey, these data were not captured, and the survey was re-administered at 1-month posttreatment.

A power analysis was performed using data from a prior RCT of a 21-day at-home VR for chronic pain compared to an audio-only version of the treatment [15]. This study revealed that an average pre–post treatment difference score in pain intensity was 1.48 for the VR group and 0.756 for the audio-only group (on an 11-point scale). Assuming an α level of .05 and 90% power, 45 participants per group would detect a treatment group × time interaction. To buffer against potential high attrition (40%), a minimum of 75 participants would be required per treatment group, with 90 participants per treatment group being ideal.

All analyses involved 2-sided hypothesis tests, with α=.05 and adjusted for any multiple comparisons within the family of tests as appropriate. Group equivalence was assessed through univariate tests of association between treatment groups (EaseVRx/Sham VR) for all baseline demographic and clinical variables with chi-square and Kruskal–Wallis applied as appropriate.

The data were analyzed in a mixed-model framework (PROC GLIMMIX in SAS) using a marginal (population-averaged) model to allow for correlated responses across the repeated measures. There were 3 explanatory factors: treatment group, time, and time × treatment group. Treatment group, EaseVRx versus Sham VR, was specified as a fixed-effects factor. Time was specified as a random-effects factor to allow for correlated response using heterogeneous compound symmetry for the covariance structure within time. The 2 effects of interest were (1) the EaseVRx versus Sham VR between-group comparison across all time points, and (2) the time × treatment group effect which tests whether the treatment group influenced the trajectory of the key variables over time.

Data were 95.05% complete; missing values were not imputed for estimation of effects, but the predicted means were used in the graphical description. The primary outcomes were the time course of DVPRS (pain scale) from baseline (defined as the average of 3 pain ratings obtained during the 2 weeks before enrollment/randomization), at 8 weekly time points (twice per week) across the 8-week treatment period, and immediately posttreatment (Day 56). A linear mixed model was used with the treatment group (EaseVRx versus Sham VR) as an independent groups factor (ie, a between-subjects factor) and time of measurement as a dependent groups factor (ie, a within-subjects factor). DVPRS-II measures were analyzed using the same approach. Effect sizes for the EaseVRx versus Sham VR for the between-groups comparison were calculated using the standardized mean difference version of Cohen d [72]. The effect sizes for the within-subjects comparison (baseline to immediately posttreatment completion at day 56) were computed by treatment group using d_rm, an adaptation of Cohen d to suit the repeated measures design [73].

Figure 3 shows the CONSORT diagram for the study (see Multimedia Appendix 2 for the CONSORT checklist). In total, 1577 individuals were assessed for eligibility and 1389 were excluded with the primary reason of having met the threshold for depressive symptoms. A total of 188 individuals were enrolled, randomized, and allocated to the treatment group. After randomization, 9 individuals discontinued participation, 5 were unable to receive a VR device, 1 returned their unopened device due to a recent medical diagnosis, and 4 voluntarily withdrew for unknown reasons. A total of 179 individuals received a VR device with their assigned treatment (EaseVRx [n=89] or Sham VR [n=90]). Because intention-to-treat analyses were performed, the analytic data set includes 11 individuals who did not provide complete data. Nearly 94% of participants in group A (84/89) and 93% of participants in group B (84/90) completed the day 56 assessment (end of treatment).

Table 2 displays the baseline demographic characteristics by treatment group. The sample included 179 participants from 40 states. RUCA codes were applied to categorize participants by zip code. In total, 76.5% (n=137) resided in highly urban or metropolitan areas, 13.4% (n=24) participants resided in metropolitan or micropolitan areas, 9.5% (n=17) were from small town or rural areas, and 0.6% (n=1) had no rural–urban identifier information. The sample was predominantly female (76.5%, 137/179), Caucasian (90.5%, 162/179), with at least some college education (91.1%, 163/179), and a mean age of 51.5 years (SD 13.1; range 18-81). No significant between-group differences were observed for any demographic variable, thus demonstrating that randomization was effective.

Table 3 presents the baseline pain and clinical characteristics for the sample by treatment group. The sample duration of back pain was 5 or more years, and the mean pain intensity score was 5/10 (SD 1.2; range 1-9). No significant differences were observed between treatment groups for all variables assessed.

Device use data revealed nonsignificant between-group differences for treatment engagement: EaseVRx participants completed a total of 43.30 (SD 15.91) experiences (average 5.4 per week) and Sham VR participants completed 48.06 (SD 24.78) experiences (average 6.0 per week).

Of the 147 participants who completed the 1-month posttreatment survey, 7/72 (9.7%) from the EaseVRx group and 5/75 (6.7%) from the Sham VR group reported experiencing nausea and motion sickness during the treatment phase of the study (P=.50). Participants were encouraged to contact study staff with any problems experienced during treatment; however, no participants contacted study staff to report adverse events of any type, including nausea and motion sickness.

Common analyses and data visualization were applied for all primary outcomes. The x-axis represents time (days), with days –14 to 0 averaged and labeled “day 7” to represent the pretreatment phase, days 1-55 were the active treatment phase, and day 56 was the end of treatment and the primary endpoint. The color bands represent the 95% CI values for the mean after correcting for multiple comparisons (Tukey–Kramer). Overlapping bands indicate nonsignificant treatment group differences (P-value) of simple main effects within each time point. The corresponding model effects for each primary outcome are displayed in Figures 4-8.

We observed a significant treatment effect (P=.001); on average, the EaseVRx group had lower pain intensity compared to the Sham VR group (Cohen d=0.49). Separately, we observed a time effect; average pain intensity significantly decreased over time for both treatment groups (time effect, P<.001). Most importantly, the decrease was greater for EaseVRx versus Sham VR (treatment × time effect, P<.001). Pain intensity reduced by an average of 42.8% for the EaseVRx group and 25.1% for the Sham VR group. The d_rm for EaseVRx was 1.31, with combined results showing large effect size and moderate clinical importance. The VR Sham group d_rm was 0.75, with combined results showing a large effect size and minimal clinical importance. As much as 65% (55/84) of EaseVRx and 40% (34/84) of Sham VR participants achieved 30% or more reduction in pain intensity. For EaseVRx, 46% (39/84) achieved 50% or more pain reduction, while for Sham VR 26% (22/84) reached that threshold.

We observed a significant treatment effect (P=.004) on pain interference with activity; on average, the EaseVRx group had lower activity interference compared to the Sham VR group (Cohen d=0.44). We also observed a time effect; pain interference with activity decreased over time for both treatment groups (time effect, P<.001). Most importantly, the decrease was greater for EaseVRx versus Sham VR (treatment × time effect, P=.013). Pain interference with activity reduced by an average of 51.6% for the EaseVRx group and 32.4% for the Sham VR group. The d_rm for the EaseVRx group was 1.27, with combined results showing large effect size and moderate clinical importance. As much as 71% (60/84) of EaseVRx and 57% (48/84) of Sham VR participants achieved 30% or more reduction in pain-interference with activity, and 56% of (47/84) of the EaseVRx participants achieved 50% or more reduction. The VR Sham group d_rm was 0.97, with combined results showing a large effect size and moderate clinical importance.

We observed a significant treatment effect (P=.005) on pain interference with mood; on average, the EaseVRx group had lower mood interference compared to the Sham VR group (Cohen d=0.42). We also observed a time effect; pain-interference with mood decreased over time for both treatment groups (time effect, P<.001) and the decrease was greater for EaseVRx versus Sham VR (treatment × time effect, P=.010). Pain interference with mood reduced by an average of 55.7% for EaseVRx and 40.04% for the Sham VR. The d_rm for the EaseVRx was 1.18, with combined results evidencing a large effect size and substantial clinical importance. As much as 74% (62/84) of EaseVRx participants and 60% (50/84) of Sham VR participants achieved 30% or more reduction in pain-related interference with mood, and 61% (51/84) of the EaseVRx participants achieved 50% or more reduction. The d_rm for the VR Sham group was 0.79, with combined results showing a moderate effect size and moderate clinical importance.

We observed a significant treatment effect (P=.004) on pain-interference with sleep; on average, the EaseVRx group had lower sleep interference compared to the Sham VR group (Cohen d=0.44). We also observed a time effect; pain-interference with sleep decreased over time for both treatment groups (time effect, P<.001). However, there was no difference between treatment groups over time (P=.755). Pain interference with sleep reduced by an average of 54% for EaseVRx and 39.2% for the Sham VR. The d_rm for the EaseVRx was 0.95, with combined results showing large effect size and substantial clinical importance for symptom reduction. As much as 70% (59/84) of EaseVRx and 60% (50/84) of participants achieved 30% or more reduction in Pain-related interference with sleep, and 60% (50/84) of the EaseVRx participants achieved 50% or more reduction. The VR Sham group d_rm was 0.87, with combined results showing a large effect size and moderate clinical importance.

We observed a significant treatment effect (P=.009) on pain interference with stress; on average, the EaseVRx group had lower stress interference compared to the Sham VR group (Cohen d=0.40). We also observed a time effect; pain interference with stress decreased over time for both treatment groups (time effect, P<.001). Most importantly, the decrease was greater for EaseVRx versus Sham VR (treatment × time effect, P=.004). Pain interference with stress reduced by an average of 59.9% for the EaseVRx group and 38.3% for the Sham VR group. The d_rm for the EaseVRx was 1.17, with combined results evidencing a large effect size and substantial clinical importance. As much as 76% (64/84) of EaseVRx participants and 56% (47/84) of the Sham VR participants achieved 30% or more reduction in pain-related interference with stress, and 63% (53/84) achieved 50% or more reduction. The VR Sham group d_rm was 0.77, with combined results showing a large effect size and moderate clinical importance.

We conducted the first placebo-controlled RCT of home-based therapeutic VR in a national sample of individuals with cLBP. We hypothesized that an 8-week pain relief skills VR program (EaseVRx) would be superior to Sham VR at posttreatment (day 56) for our primary outcomes: average pain intensity and pain-related interference with activity, mood, sleep, and stress. While both study groups had significant reductions in pain and all domains of pain-related interference, EaseVRx evidenced superior treatment effects for all primary outcomes except sleep interference; the between-groups Cohen d effect sizes ranged from 0.40 to 0.49. For EaseVRx, large pre/posttreatment Cohen d effect sizes ranged from 1.17 to 1.3 and demonstrated moderate to substantial clinical importance for reduced pain intensity and pain-related interference with activity, mood, and stress at end of treatment. A greater proportion of participants in the EaseVRx group exceeded thresholds for clinical importance of effects. For moderate clinical importance in pain reduction (≥30% reduction in pain), 65% (55 of 84) in EaseVRx versus 40% (34 of 84) in Sham VR met this threshold. For substantial clinical importance in pain reduction (≥50% reduction in pain), 46% (39 of 84) in EaseVRx versus 26% (22 of 84) in Sham VR met this threshold. These effects for therapeutic VR exceeded effects reported for a 3-week skills-based VR program for chronic pain [15] and effects reported for CBT studies involving in-person 8-week treatment with a trained therapist [4,6,7].

Both treatment groups evidenced moderate to substantial reductions for pain-related interference with sleep; however, no between group differences were found. By contrast, EaseVRx was superior to Sham VR for reducing sleep disturbance (secondary outcome), suggesting that therapeutic VR is particularly effective for reducing general versus pain-related sleep disturbance.

The secondary outcomes yielded interesting findings. First, physical function significantly improved for both treatment groups, with superior improvements found for EaseVRx. We note that the therapeutic content included no kinematic elements, nor did it include direction for activity, movement, or goal setting for either. Accordingly, improvement in physical function may be a product of substantially reduced pain-related interference in activity. Next, our hypothesis that therapeutic VR would be superior to Sham VR for improving pain coping (eg, reducing pain catastrophizing and improving pain self-efficacy and chronic pain acceptance) was unmet. The lack of effect for pain catastrophizing was particularly striking because this is a malleable construct that is highly responsive to behavioral treatments broadly [4,6,7] (Darnall et al, unpublished). However, our results align with the prior 3-week VR program study that similarly found no effect on pain catastrophizing and pain self-efficacy. Findings suggest that increased treatment time and additional focal content were insufficient to affect these factors significantly. These findings also highlight differences in treatment efficacy between therapeutic VR and CBT. Multiple studies have shown that CBT imparts its largest effect on pain catastrophizing [4]. By contrast, therapeutic VR evidenced its largest effects for reducing pain intensity and pain interference across several key domains.

We found no differences for either treatment group for change in prescription opioid use, and note that prescribing changes are unlikely to occur within the 2-month timeframe of the study. Similarly, we found no changes in “as needed” opioid use. We found substantially reduced use of OTC analgesic medication at posttreatment for the EaseVRx group only. While additional research is needed to replicate this finding, this finding offers a promising suggestion that therapeutic VR may reduce need for analgesics. Future VR research should examine medication use in greater detail and with higher-frequency data capture.

The study’s methodologic rigor was strengthened by a placebo treatment (Sham VR), which evidenced equivalent participant engagement as therapeutic VR. The extant literature on digital behavioral health research has reported participant treatment engagement rates ranging from 20% to 60% [13,74-76]. Strikingly, the current trial evidenced a 90% engagement rate in both groups, thus suggesting that efforts to enhance the face validity of the Sham VR were effective. These results also highlight the public interest in home-based VR as a chronic pain treatment modality. Therapeutic VR was rated significantly higher than Sham VR for satisfaction, likelihood to recommend to others, and likelihood to continue using the device after the 8-week treatment phase if it was made available. Combined with high participant engagement data and device usability ratings, these data extend prior work [15] supporting the utility, user satisfaction, and efficacy of home-based VR for chronic pain.

The context of the COVID-19 pandemic may have influenced the participant engagement rate. This trial occurred entirely at a time when people were adhering to strict social distancing measures and were environmentally isolated. Indeed, for many people receipt of medical care is worryingly low due to limited availability or unavailability of outpatient treatment options. These circumstances likely supported interest in effective home-based pain care. The COVID-19 context and the home-based study design support the ecological validity of the study findings. Notably, the study was conducted remotely and did not benefit from any in-person contacts or enhanced placebo effects that occur when research involves high-touch protocols or is conducted in medical treatment settings (ie, halo effects).

Strengths of this study include methods that attended to the IMMPACT recommendations and the NIH Research Standards for Back Pain. The study was conducted in a national sample drawn from 40 states, was well-distributed geographically, and included participants from urban and rural settings. Additional aspects of methodological rigor included participant and analyst blinding, intention-to-treat analyses, randomization, and a rigorous placebo control group that adhered to recommended specifications for an optimal VR sham [77].

Several limitations bear consideration when evaluating the study results. With the exception of device use metrics, all data were self-reported. The study was untethered from medical care and thus, there was no ability to confirm pain diagnoses or analgesic prescription information. The study sample was predominantly female, white, college educated, and internet savvy; thus, findings may not generalize to individuals with disparate demographic characteristics. These findings are consistent with previous evidence showing that highly educated females are more likely to use self-care mobile health technologies, particularly those with mindfulness-based content [78], and this dovetails with a general female predilection to seek treatment for pain and other health concerns. As eHealth literacy and awareness increase in clinicians and the general population, it is likely that health technologies (including therapeutic VR) will benefit other demographics [79]. Additionally, the study was conducted in individuals with cLBP and findings may not generalize to other chronic pain conditions.

Data on cybersickness were collected at 1 month posttreatment and this lag introduces potential for recall bias, despite others documenting that participants readily recall cybersickness due to its specificity and salience [15]. Overall attrition was low (n=11) and it is possible that the 2 EaseVRx participants who left the study after receiving their headset did so due to cybersickness. No participants contacted study staff to report adverse events of any type. Data on sex differences for cybersickness are mixed, with some reporting a female preponderance, while a recent meta-analysis suggests no sex effect [80]. Within the context of our limitations, we highlight low reports of cybersickness in a predominantly female sample. Future research may better capture potential VR adverse effects by assessing these factors in the first week of treatment. Interpretation of data on prescription opioid use was limited by low-frequency sampling methods that are subject to recall bias and poor data accuracy. Opioid prescriptions often allow “as needed” flexibilities in medication use and future research designs may benefit from high-frequency sampling methods which improve data accuracy. Further, in quantifying analgesic medication use (prescription opioids and OTC analgesics) we did not assess or control for life events or circumstances that may have influenced medication use (eg, acute injury or surgery). Finally, our threshold for depression screening and inclusion was applied to provide greater specificity [44] yet is noted to be lower than what is reported for many individuals with chronic pain. While our purpose in applying a low threshold for depressive symptoms was to minimize poor engagement and attrition resulting from anhedonia or avolition (and therefore poor data quality to determine treatment efficacy), recent research in cLBP suggests such concerns may be unfounded for mild to moderate depressive symptoms (Darnall et al, unpublished).

As a fully self-administered and on-demand treatment, VR is a promising and effective option that can transcend many traditional barriers to nonpharmacologic pain treatment; however, currently access is limited. With future reimbursement and commercial availability, therapeutic VR could become affordable and widely accessible to consumers.

An 8-week self-administered home-based pain relief skills VR program appears effective for reducing pain intensity and pain-related interference in activity, mood, and stress posttreatment. Treatment effects ranged from moderately to substantially clinically important. Therapeutic VR had high rates for engagement and user satisfaction. Additional studies are needed to determine effects in demographically diverse populations and in other pain conditions. Data suggest therapeutic VR is not operating through traditional pain coping mechanisms. As such, additional research is needed to characterize mechanisms of treatment effects and durability of effects. Home-based VR appears to provide effective and on-demand nonpharmacologic treatment for cLBP.