Interacting With Virtual Objects via Embodied Avatar Hands Increases the Illusion of Presence, Increases Distraction, and Modulates VR Analgesia Effectiveness: A Randomized Crossover Trial.

The current study introduces a new experimental environment and apparatus for exploring cognitive factors in pain. Interacting with virtual objects via embodied avatar hands increased the illusion of “being there” in the virtual world, increased VR analgesia for acute pain, and reduced accuracy on an attention demanding task. Twenty-four healthy volunteer college students participated in this within-subject randomized crossover design study. During Phase 1, each participant received brief thermal pain stimuli during interactive embodied avatar VR vs. passive VR (no avatar and no interactivity), VR treatment order randomized. After each pain stimulus, participants provided subjective 0-10 ratings of pain. Compared to the passive VR condition, during the interactive avatar VR, participants reported signicant reductions in 1). worst pain, 2). pain unpleasantness, 3). time thinking about pain and 4). they had signicantly more fun during the pain stimulus (p = .000 for each). During Phase 2, participants performed a divided attention task in each VR condition. Participants made signicantly more errors on the divided attention task during the interactive avatar VR condition, compared to passive VR, implicating an attention mechanism for how virtual reality reduces pain and helping understand how VR inuences pain perception.


Introduction
The opioid crisis makes development of non-opioid pain control techniques a national priority. [41] Pain levels during medical procedures are frequently excessive. [31,57] Opioid analgesics help reduce pain, and become more effective at higher doses, [2,45,50,55,74] but opioid side effects limit dose levels. [10] Furthermore, unhelpful psychological in uences can exacerbate/increase pain intensity during wound care, [3,5,11,13,17,18,58,61,62] and excessive acute pain can lead to chronic pain. [69] Fortunately, adjunctive psychological treatments can help reduce pain [59], with few or no additional side effects. For example, immersive virtual reality, rst reported by our team in the 1990s [24], is emerging as an unusually powerful adjunctive non-pharmacologic analgesic. [6,9,15,21,26,34,36,39,40,42,47,48,49,62,70] fMRI brain scans show that in addition to reducing participants subjective experience of pain, VR also reduces painrelated brain activity. [29] A follow up fMRI study showed that VR reduces pain as much as a moderate dose of hydromorphone. [30] However, the mechanism of how VR reduces pain and why some VR systems are more effective than others, is not well understood and is the focus of the current study.
Considering VR Analgesia to be a divided attention task may help understand how VR reduces pain and why some VR systems reduce pain more effectively than other VR systems.
During VR analgesia, instead of devoting all of their attention to pain, during VR, some of the patient's attention is used to process the information coming into their brain from the VR system, and simultaneously, some of their attention is used to process information coming into their brain from the pain receptors (e.g., in their skin). In other words, during VR, some of the brains attentional resources are allocated to VR, and some are allocated to pain. The more attention grabbing the VR system, the less remaining attention the patient's brains have available to process incoming nociceptive signals, and the more effective the VR analgesia treatment (see Fig. 1).
In the current context, embodiment (e.g., "in a body") involves giving participants a sense of ownership over a computer generated representation of the participant, aka, an avatar [43,60], for example "these cyberhands I see in VR, are my hands." I predicted that interacting with virtual objects via embodied avatar hands would increase the subject's illusion of "being there" in the virtual world, would make VR more attention grabbing, would increase VR analgesia, and would reduce accuracy on an attention demanding task, implicating an attentional mechanism for how VR reduces pain.

Materials And Methods
Participants Twenty-four college students from the University of Washington Department of Psychology (18-24 years old, mean = 18.93 years old, SD = 1.36) were included in the main study, and an additional different 12 students participated in a side project to test the assumptions of our pain paradigm (age range [18][19][20][21][22][23][24].
Participants in the side study received passive VR for both test phase thermal pain stimuli (instead of one passive VR and one interactive cyberhands VR condition used in the main study). All data was collected between late January and early March 2020. All participants included in each analysis were in their originally assigned treatment order groups. In addition to a baseline "no VR" thermal pain stimulus and pain rating, each of the 24 participants in the main study rated their pain during "passive VR with no cyber hands and no interactivity" during one thermal pain stimulus, and after a brief inter stimulus interval wash out period, rated their pain again during "yes interactive cyberhands avatar VR" during a second thermal stimulus. To help balance the number of participants in each treatment order, VR treatment order was block randomized using random number sequences from www.random.org. The random numbers were written on 2" paper squares, and each paper square was put into its own sealed sequentially numbered opaque envelope and was opened by the researcher as the subject was reading the consent form. To reduce bias, the researcher only learned the treatment order of an individual participant during the consent process. Participants knew they would receive two VR treatments but they only learned the details of each VR treatment (passive vs. interactive cyberhands) during the brief instructions immediately preceding each VR treatment. Using an AB/BA within subject design, with treatment order randomized, some participants received "interactive cyberhands VR" rst and "passive VR with no cyberhands" second, and others received "passive VR with no cyberhands" rst and "interactive cyberhands VR" second, (see participant enrollment owchart in Figure 2). "The particular strength of the simple AB/BA crossover design is that both interventions are evaluated using the same participant, which allows comparison at the individual rather than the group level" (Dwan et al., 2019, page 2 [12]). The within-subject design reduces nuisance variance. The VR analgesia + odd numbers protocol described herein (see appendix for full research Protocol) was developed to measure for the rst time how effectively a given virtual reality system reduces acute pain, and the relation between the amount attention allocated to VR, and the VR analgesia effectiveness of that VR system. The current experiment involves the identi cation of a painful but tolerable temperature (baseline pain, no VR). After the baseline pain temperature is identi ed, Phase 1 begins. Phase 1 measures pain during VR. Participants experience a High Tech VR treatment during one brief pain stimulus, and they experience a Low Tech VR condition during another brief stimulus (VR treatment order randomized). Phase 2 (attention during VR) measures performance on a brief cognitive test designed to quantify how much attention is paid to the test during No VR, during High Tech VR and during Low Tech VR, by measuring the participants accuracy on the attention demanding "odd number" task during No VR, during Low Tech VR, and during High Tech VR. (See Figure 3). Phase 1: Quantitative Sensory Testing.
Following the paradigm used by our team in several previous studies, [29,30] a commercially available Medoc Quantitative Sensory Testing thermal (heat) pain stimulator (Ramp and Hold program) was used to deliver brief 10 second heat stimuli to their right foot, at "painful but tolerable" temperatures individually pre-selected by each participant. During Phase 1, each student was allowed to choose the temperature they wanted to use during the test phase. The mean stimulus temperature selected by participants was 45.2•C (SD = 1.40, range = 44 -48.0•C).
Participants received the following instructions. "During Phase 1, you will receive several 10 second pain stimuli. Starting at a relatively low temperature, we will slowly work our way up, one 10 second stimulus at a time, to a temperature you nd painful but tolerable. You get to select which temperature to use and once you select the temperature you want to use during the test phase, you will receive only two additional 10 second stimuli at one of the temperatures you pre-approved. Phase 1 typically lasts around 10-15 minutes. We are not trying to see the highest pain you can tolerate, we are just wanting you to nd a temperature that you nd 'painful but tolerable for 10 seconds', that you are willing to experience for two additional 10 second stimuli, later in the study." The instructions to participants for GRS ratings were as follows: "Please indicate how you felt during the most recent 10-second pain stimulus by making a mark anywhere on the line. Your response does not have to be a whole number." The above two sentences of instructions were presented to the subject on a conventional laptop computer screen, using a computer animated text-to-speech character software program https://ttsdemo.com/. The researcher previously typed the above two sentences of instructions into the text to speech website, and when the researcher clicked start, the animated character on the laptop screen said the brief instructions to the participant. The animated text to speech character was used to make the animated "research assistant" blind to treatment condition for at least part of the study, to help reduce bias.
After each brief thermal stimulus, participants indicated how painful they found the stimulus using Graphic Rating Scales (GRS), validated by the measures' strong associations with other measures of pain intensity, as well as through the measure's ability to detect treatment effects, [22,37,38,68,72]. GRS ratings were used to measure "worst pain", "pain unpleasantness", and "time spent thinking about pain" that correspond to three separable components of the pain experience; sensory pain, affective pain, and cognitive pain, respectively.
The question regarding "to what extent did you feel like you 'went into' the virtual world," was adapted from Slater et al. [65] . Similar presence measures have been shown to be reliable [19] and able to detect treatment effects. [25,29,63] Equipment.
The VR laptop computer MSI GeForce GTX 1080 8 GB, Intel Core i7 7th (2.80 GHz), 16 GB RAM, Windows 10 operating system was connected to a VRGineering.com XTAL VR helmet, 5120 x 1440 (2560 x 1440 per eye), with extra wide 170-180 degrees eld of view, using Two Quad HD high-density OLED displays.
A VR demo Architecture World (vrgineering.com) that came with the helmet, included optical hand tracking of participants' real hands to simultaneously control their cyberhands in VR. The XTAL VR helmet came with LEAP Motion camera based optical hand tracking https://www.leapmotion.com/. The VR system was designed to give participants the illusion of "being there" in a 3D computer generated virtual house named Architecture World (VRgineering.com).
In both VR treatment conditions, each subject could see a living room in virtual reality, with a green living room chair, and two books sitting on a coffee table next to the green chair. Architecture World had no audio or sound effects in the current study.
Phase 1: Measuring pain during passive VR vs. during interactive cyberhands avatar VR.
For the passive VR condition, participants received the following instructions. "For this next part of the study, please put on the VR helmet, you will see a living room, and will look at a green chair, and a coffee table with books on it. While you are in VR doing this, you will receive another 10 second pain stimulus at one of the temperatures you have approved. In other words, we won't go any higher than that last stimulus you approved". After receiving a brief thermal stimulus while in virtual reality, participants took off the VR helmet and answered the pain ratings, using the pen and paper GRS pain and presence ratings. The pain ratings and subsequent instructions phase lasted approximately 5 minutes, and served as the "wash out" period between VR treatments, to help minimize carryover effects.
Participants received the following instructions for the Phase 1 interactive embodied cyberhand avatar VR analgesia condition. "For this next part of the study, please put on the VR helmet, you will see a living room, and will look at a green chair, and a coffee table with books on it. Please reach out with your hand and grab one of the virtual books, drop it on the oor, grab the second virtual book, drop it on the oor and then wiggle your ngers the rest of the time. While you are in VR doing this, you will receive another 10 second pain stimulus at one of the temperatures you have approved. In other words, we won't go any higher than that last stimulus you approved." While in virtual reality, during the interactive embodied cyberhands avatar condition, participants could interact with objects in virtual reality using their hand movements in the real world (see Figure 4). For example, they could wiggle the ngers of their cyberhand in VR by wiggling their real ngers whose detailed motion was detected by a miniature head mounted video camera attached to the XTAL VR helmet (the video camera facing outward towards the real world). With the camera based LEAP Motion system, patients did not have to hold any controllers or hardware in their hands, and did not have to wear cybergloves. The patient wearing virtual reality goggles could see their virtual hands in the virtual living room, and could use their cyberhands to interact with virtual objects in the computer generated world (pick up virtual books and drop the books onto the virtual living room oor, see Figure 4). All participants were able to perform all tasks. Afterwards, they took off the VR helmet and answered the pain ratings, using the pen and paper GRS pain and presence ratings.
Phase 2: Measuring the amount of attention used by passive VR with no avatar vs. interactive avatar cyberhands VR.
After Phase 1, the thermal pain stimulator was removed from the participants foot, and participants were told that Phase 1 was now over, and they received the following instructions. "You have just completed Phase 1 and you can now remove the thermal pain stimulator from your foot. Please take off the thermal stimulator and put your shoe back on now. There are no more pain stimuli, but we will now begin Phase 2 of the study. In Phase 2, no pain stimuli will be administered. Instead you will go into virtual reality again and while you are in the virtual world, you will monitor a string of numbers from 1 to 10, and will say 'now' any time you hear three odd numbers in a row, (the odd numbers are 1, 3, 5, 7 or 9). For example, if you hear "1, 9, 3", you would say "now". The researcher will keep track of your answers and will be measuring your accuracy on the odd number task. You will receive a total of three divided attention tasks which each last 2 minutes per task. The rst two minute session is with No VR. Are you ready to begin the odd number task? Just say "now" any time you hear 3 odd numbers in a row.
The researcher then played the pre-recorded auditory string of numbers (see below) via a digital audio le on a laptop with high quality Dolby Sound laptop speakers. The traditional odd number divided attention task [8,25,35,46] was adapted for use in the current VR study, using our own new number set (shown below) customized for the present study (with 10 odd number triads during a 2 minute session). 1 2 3 5 7 5 2 4 1 9 3 6 5 8 1 1 4 3 5 1 2 3 2 1 7 9 2 5 2 2 1 9 5 6 2   4 6 7 1 8 2 6 4 7 5 3 1 6 7 2 1 9 3 5 1 8 2 1 5 3 8 1 3 2 9 7 6 5 3 1  3 5 1 4 The string of numbers (i.e., the odd number task) lasted 2 minutes. The same identical two minute audio le was played a total of three times. The rst time, it was played with No VR (baseline). The second time the odd number task was played during VR (e.g., passive VR) and the third time, it was played in VR again (e.g., interactive embodied avatar cyberhands VR). The VR treatment order used for the Phase 2 divided attention task was the opposite order as the VR treatment order used in Phase 1 (thermal pain stimuli), and VR treatment order was thus also randomized. Phase 2. To summarize, the Odd Number Task is a divided attention task that involves monitoring auditory numbers during No VR (for 2 minutes) vs. passive VR (for 2 minutes) vs. during interactive cyberhands avatar VR (for two minutes). During the "dual task", participants must perform the two tasks at the same time, being in VR was one task, and monitoring the odd number task was a competing task. In each condition, participants listened to an auditory string of numbers from 1 to 10, and said "now" every time they heard three odd numbers in a row. They were told that the researcher would be monitoring their accuracy on the odd number task. During the odd number task, the researcher had a printout of the number sequence, and the researcher made a mark to indicate every time the participant said "now" in each of the following conditions.
No VR + odd number task.
Participants did not wear a VR helmet during the No VR condition. They monitored the odd number task with No VR.
Passive VR Condition + odd number task.
During the passive VR condition, participants went into a virtual living room, and were instructed to look at the green chair, coffee table and the books on the coffee table in VR. They could not see any cyberhands or avatar in virtual reality in this treatment condition (passive, no cyberhands). While they were in passive VR during the Phase 2 divided attention task, as they were in VR looking at the living room chair scene, they also listened to an auditory string of numbers, and said "now" any time they heard three odd numbers in a row.
Interactive embodied cyberhands avatar condition + odd number task.
While they were in VR during the Phase 2 divided attention task, they looked at the living room chair scene, they used their cyberhands to pick up books in VR, and wiggled their cyber ngers at the same time they listened to an auditory string of numbers, and said "now" any time they heard three odd numbers in a row.
Power Analysis: A power analysis to determine the number of participants needed to test our primary hypothesis was computed apriori, using the statistical program GPower 3.10. The following assumptions were used in the power analyses, all determined from pilot data and Wender et al., 2009, [71] an effect size (d) of .78, power of .95, and an alpha of .05. Under these conditions, we would require 24 participants in the main study to be able to detect a signi cant treatment effect, and to show that interactive cyberhands treatment was more effective than passive VR for reducing acute worst pain during the brief thermal pain stimuli.

Results
Pre-analysis for carry-over effects.
When using a within-subjects design, carry over effects sometimes occur when the effect of the rst treatment condition persists, and in uences the responses of the second treatment, so that the observed difference between the treatment conditions depends on the treatment order. 12 The following pre-analysis was conducted to see if there were carry over effects (ie., to see if the difference between passive VR and interactive avatar VR was stronger depending on which treatment order (passive 1 st or passive 2 nd ). A two-way Mixed ANOVA was used to test if there were undesired carryover treatment order effects. Passive VR vs. interactive cyberhands was the repeated measure, and treatment order was the between groups factor (people who received passive VR 1 st were considered one group, and those who received passive VR second were considered a second group). No signi cant interaction was found between treatment order and worst pain ratings (i.e., no signi cant treatment order effects for Worst pain), F(1,21) = 1.01, p = .33 NS, partial η2 = .05. No signi cant interaction was found between treatment order and pain unpleasantness, F(1,21) < 1, NS, p = .72, partial η2 = .01. No signi cant interaction was found between treatment order and participants' ratings of Time spent thinking about pain during the thermal stimulus, F(1,21) = 2.80, p = .11 ns, partial η2 = 0.12. And nally, no signi cant interaction was found between treatment order and participants ratings of Fun during the thermal stimulus, F(1,21) = 2.43, p = .13 ns, partial η2 = 0.10.
Since there were no carry over effects, all of the following analyses were collapsed across treatment order (see Figure 5 and Figure 6).

Worst pain (the primary dependent variable).
A one way repeated measure ANOVA showed a signi cant main effect of No VR vs. passive VR vs. interactive cyberhands VR for the primary dependent measure, Worst Pain, F(2,46) = 28.73, p = .000, MS = 29.24, partial η2 = 0.55. Post-hoc paired comparisons (paired t-tests) are shown below, 1). Comparing No VR vs. passive VR, worst pain ratings were NOT signi cantly lower during passive VR, but showed the predicted pattern. 2). Comparing No VR vs interactive cyberhands VR, worst pain was signi cantly lower during interactive cyberhands VR. And 3). Most importantly, comparing passive VR vs. interactive cyberhands VR, worst pain was signi cantly lower during interactive cyberhands VR.

Primary dependent variable
Worst pain, mean ratings on a zero to 10 graphic rating scale (SD in parentheses) Pain Unpleasantness (a secondary pain measure).
Post hoc paired comparisons (paired t-tests) for the variable "pain unpleasantness" are shown below. 1). Comparing No VR vs. passive VR, pain unpleasantness was signi cantly lower during passive VR. 2) Comparing No VR vs. interactive cyberhands VR, pain unpleasantness was signi cantly lower during interactive cyberhands VR. And 3). Most importantly, as predicted, comparing passive VR vs. interactive cyberhands VR, pain unpleasantness was signi cantly lower during interactive cyberhands VR.
Unpleasantness, mean ratings on a zero to 10 graphic rating scale (SD in parentheses). Post hoc paired comparisons (paired t-tests) for the variable "Fun" are shown below. 1). comparing No VR vs. passive VR, participants reported having signi cantly more fun during passive VR. 2). Compared to No VR, interactive cyberhands VR was signi cantly more fun. And 3). Most importantly, compared to passive VR, interactive cyberhands VR was signi cantly more fun.
Fun during the thermal stimulus, mean ratings on a zero to 10 graphics rating scale (SD in parentheses). To test some important assumptions of our thermal pain paradigm, pilot data collected from 12 new participants from the same subject pool, who were not involved in the main study, received No VR during baseline, passive VR during thermal Test 1 vs. passive VR again during thermal Test 2. As predicted, participants' pain ratings from the thermal pain stimulations were stable over repeated test pain stimulations for people who received one baseline pain and two test pain stimuli with passive VR, using the same thermal pain paradigm as the main study.

Discussion
In the current study, as predicted, interacting with virtual objects via embodied avatar hands (i.e., avatar VR) signi cantly increased the participant's illusion of "being there" in the virtual world, increased VR analgesia, and increased fun during the pain stimulus. The current study also measured for the rst time, how much attention was diverted by avatar VR. Consistent with the notion that interactive embodied immersive VR is unusually attention grabbing, during Phase 2, participants made signi cantly more errors on the divided attention task during the interactive avatar VR condition, compared to passive VR. These results implicate an attentional mechanism for how VR reduces pain, and help understand how VR in uences pain perception.
Although immersive virtual reality can often help reduce the acute pain of patients during painful medical procedures, some medical procedures are so painful that extra strength VR analgesia may be needed. [14,22,31] The current study measured whether a stronger (more immersive) dose of virtual reality could increase analgesia. Our results indicate that the immersiveness of a VR system can be increased substantially, (e.g., via avatars) with little or no increase in VR side effects, unlike opioids, which show a dose-response increase in side effects (e.g., increased nausea and constipation) with higher doses, and opioid side effects linger for hours after the medical procedure. Immersive interactive cyberhand avatar embodiment with extra wide eld of view VR helmet (180 degrees eld of view, high resolution) appears to have considerable potential as a short acting extra strength VR analgesia system. A number of studies have now shown that increasing the immersiveness of the VR system increases VR analgesia for acute pain. [1,24,32,33,71] However, according to Trost et al., 2020,[70] there is currently a gap in the scienti c literature on the use of immersive VR avatars to treat acute pain. To date, nearly all studies involving virtual embodiment have targeted treatments designed for chronic pain patients, (e.g., phantom limb pain) [22,23,51,52,53,54,67,70]. The current study is one of the rst VR studies designed to treat acute pain that includes an avatar, [20,24,27,28] and is the rst acute pain VR analgesia study to manipulate embodiment, and to measure whether interactive embodiment with cyberhands avatars increases analgesia (and reduced accuracy on the odd number task, an objective measure of attention), compared to passive VR with no avatar, a promising direction for future research.
The current study has a number of limitations that should be taken into consideration when interpreting the results. The within-subjects design is statistically powerful and allows each participant to compare the different treatments, but has the drawback that participants remain aware of the different treatment conditions, and this awareness has the potential to in uence the results. [4] The current study should thus be replicated using a between groups design, ideally with participants blinded to VR treatment group conditions, [64] and using a larger sample size. Another limitation is that the current study is an analog laboratory pain study, with only brief thermal pain stimuli and brief VR treatment durations. Whether the current results generalize to clinical settings (e.g., severely burned children during 20 minute burn wound debridement sessions) is an important research topic for future clinical studies with patients during painful medical procedures, e.g., [22,24].
Despite these limitations, the results of this study could also have other important clinical implications. According to Keefe et al., [41] the epidemic of opioid related overdose deaths [7,16,73] has greatly increased the urgency to develop effective non-drug pain control techniques that can help reduce the medical community's current heavy reliance on opioid analgesics for pain control. Although to date VR analgesia has typically been used adjunctively in addition to traditional pain medications, a stronger version of VR has the potential to reduce the use of opioids [14,44,56] and/or to help compensate for increasing tendencies to under-medicate patients, as opioids become much more strictly controlled, due to recent increases in federal regulation of opioid prescription. Major computer companies' multibillion dollar investments into virtual reality technology are making VR goggles widely available and affordable for medical applications. Additional research and development of VR analgesia is recommended.