Virtual Reality Treatment Displaying the Missing Leg Improves Phantom Limb Pain: A Small Clinical Trial

Introduction

Loss of limb is a common problem; almost 2 million people in the US have had an amputation. ¹ Up to 90% of people with limb amputation experience the persistent sensation of the missing extremity, a phenomenon known as a “phantom limb” (PL). ² Additionally, a significant proportion of individuals with a phantom limb—up to 85% in some studies—experience persistent and in some cases debilitating pain in the missing limb, a condition known as phantom limb pain (PLP). Although existing therapies provide pain relief in some cases, there is widespread agreement that current approaches fall short of bringing relief to most individuals with PLP. ²

Reflecting the many manifestations of PLP, a substantial number of theories regarding the cause of the disorder have been proposed, ³ falling into 2 major classes. The first emphasizes the role of the peripheral nervous system in the genesis of PLP and suggests, for example, that PLP may be the consequence of neuromas at the site of resection. Although this cause seems unlikely for PLP in all individuals, there is little doubt that peripheral factors contribute to PLP in some. ⁴ The second group of theories attributes PLP to changes in the central nervous system. Studies in both animals^5,6 and humans^7,8 have indicated that sensory cortex that had been responsive to input from the amputated or deafferented body part begins to respond to contiguous body parts. The “maladaptive plasticity” hypothesis suggests that cortical reorganization in primary sensory and motor cortex after amputation induces PLP. In particular, Flor and colleagues^9,10 argued that PLP is associated with contiguous cortical regions invading the cortex representing the deafferented limb. They propose that this functional takeover occurs because of unmasking previously silent pathways or because of the creation of new cortical connections. In support of this claim, Lotze et al ¹⁰ reported that severity of PLP was correlated with the degree to which motor cortex for the face (indexed by lip movements) “invaded” motor cortex formerly devoted to the missing limb. Makin and colleagues,^11-13 did not observe a relationship between PLP severity and shift of the face motor cortex,^12,13 and other researchers have found that the representation of the missing limb is largely preserved.^14,15 In the Makin studies, PLP was associated with reduced functional connectivity between brain areas^13,16 or increased activity in somatosensory areas contralateral to the amputated limb. ¹² Recent data from Raffin et al ¹⁴ suggest that the cortical remapping and functional connectivity accounts are not mutually exclusive, as it has been noted that intact cortical representation of the missing limb can coexist with the invasion of the neighboring areas. ¹⁴

Recognizing that no single account explains all of the complexities of the PLP syndrome, Ortiz-Catalan ³ proposed that PLP is caused by the “stochastic entanglement” of pain and somatosensory-motor circuitry. Following this hypothesis, amputation induces a chaotic state in the brain, which leads in a probabilistic manner to co-activation of connected brain networks. In normal circumstances, only a noxious stimulus induces the conjoint activation of pain, somatosensory, and motor networks. According to the stochastic entanglement hypothesis, the unstable state generated by the amputation leads to co-activation of the networks in the absence of noxious stimuli. Variability in PLP occurrence and characteristics across individuals is assumed to depend on the strength of the association between brain networks.

A variety of treatments targeting the sensorimotor cortices have been assessed in small studies. Mirror therapy may be the most extensively studied intervention. In mirror therapy for lower-limb PLP, participants attempt to move both legs in tandem, and they see the image of the intact limb reflected in a mirror at the location of the amputated limb, which tends to provide the sensation that the amputated limb is still present and under their control. This treatment is based on the suggestion that PLP results from the lack of correspondence between motor commands and proprioceptive and visual feedback from the amputated limb.^15,17 Motor commands continue to be generated after the amputation, but the commands do not result in the anticipated sensory feedback in the absence of the effector. The mismatch between the predicted and estimated sensory states is interpreted as an error signal that (for reasons that are not explicated) induces PLP. Vision of the “amputated” limb executing the planned movement, as seen in the mirror, is assumed to reduce PLP by providing visual feedback for the movement, thereby eliminating the mismatch between action plans and sensory feedback.^18-20 It has been shown in healthy individuals that in conditions where there is a mismatch between vision and somatosensory input, such as the rubber hand illusion, vision is predominant and action performance is generated based on visual rather than somatosensory feedback (e.g., Soechting and Flanders, 1989 ²¹ ; Ambron et al, 2020 ²² ). Therefore, manipulating vision of the limb may influence both action planning and the experience of body ownership.

Virtual reality (VR) has also been used to treat PLP. Some VR treatments are designed to mimic mirror therapy, so the apparent movement of the amputated limb follows the action of the intact limb.^23-27 In other studies, the visually depicted “movement” of the amputated extremity is controlled by myoelectric recording from the distal amputated limb,^28-31 surface EMG,^32-34 or motion capture^24,35,36 recorded from the residual limb.

Virtual Reality has advantages relative to mirror therapy. As interventions may take the form of diverse games rather than repetitive movements, subjects may be more likely to participate in therapy. Furthermore, VR provides participants with direct control of the motion and sense of agency of the phantom. ³⁷ Perhaps most importantly, the movements of the 2 legs need not be mirrored in VR treatments, enabling a broader variety of interactions.

We report our experience from a small clinical trial. In a previous study, ³⁸ we developed a VR system, which used immersive VR and inertial sensors mounted on participants’ intact and residual limbs to recreate and control the missing limb and intact limb independently. This system was tested in two individuals with lower-limb amputation, both of whom demonstrated pain reduction immediately after each session and across sessions. ³⁸ The study reported here uses an updated version of our system to follow up these preliminary results in a larger sample of different individuals with lower-limb amputation and test the effect of the treatment over a larger period of time.

Seven participants underwent a 9-week-long treatment protocol consisting of 2 treatment phases. In the first phase (distractor treatment), participants were exposed to a VR environment (Cool!, https://firsthand.com/health-software/) that has been demonstrated to improve chronic pain ³⁹ but that lacks the visual and auditory feedback of movement of an imputed “amputated” limb. After 5–7 sessions of Cool!, participants underwent 10–12 sessions of lower-limb treatment, in which they played a series of games using their virtual limb. The missing limb was virtually reconstructed by creating an avatar with two intact legs and visually superimposing it on the participant’s real body. Magnetic motion-tracking sensors placed above and below the knee on the participant’s intact and residual limbs tracked the real position of the limbs independently. These motions were used to control the avatar’s leg movements in the virtual world, providing participants with the sensation of control over the movements of both limbs. We predicted that both treatments would reduce PLP, but that the magnitude and duration of the benefit would be larger in the lower-limb VR than in the distractor treatment. Indeed, we expected the two treatments to differentially interact with PLP: the distractor treatment was expected to have a calming effect and move participants’ attention away from the pain; the limb treatment, in contrast, was expected to at least partially restore feedback of the missing limb, thereby reducing the mismatch between the motor command and visual feedback that may generate PLP.

Methods

Results

Single-Subject Analyses

Table 1 reports the demographics and information extracted from the Limb Deficiency and Phantom Limb Questionnaire ⁴⁹ for every participant. Amputations were performed because of vascular insufficiency, infection or trauma (see Table 1); all patients except for SUB 7 wore a rigid non-electric prosthesis. Only SUB 3 reported additional phantom-related phenomena, such as synesthesia for pain referred to the phantom limb. No subject reported telescoping (the perception that the missing limb is foreshortened).

OPEN IN VIEWER

Table 1.

Limb Deficiency and Phantom Limb Questionnaire Responses for Each Participant.

	Sub 1	Sub 2	Sub 3	Sub 4	Sub 5	Sub 6	Sub 7
Age, sex	46 F	57 M	54 M	52 F	53 F	48 M	50 M
Side of amputation	Left	Left	Right	Right	Right	Right	Left
Time from amputation to testing	11 years	3 years	2 years	2 years	8 years	4 years	6 months
Reason for amputation	Trauma	Infection	Arterial insufficiency and trauma	Arterial insufficiency	Arterial insufficiency	Trauma	Trauma
Subject’s rating of sensory function prior to amputation (rated from 0-100)	100	15	13	100	58	100	100
Subject’s rating of motor function prior to amputation (rated from 0-100)	100	17	98	77	0	51	Normal
Average pain prior amputation (0-100	78	95	99	79	100	84	87
Pain severity immediately post amputation (0-100)	100	86	78	100	0 (pain started 2 months post amputation)	100	87
PLP at its worst	88	100	74	72	65	73	100
PLP frequency	Constant	Constant	Constant	Several times daily	Several times daily	Several times daily	Several times weekly
PLP onset with respect to the amputation	Immediately after	Within 1 month	Within 1 month	Delayed for months	Delayed for months	Delayed for months	Within 1 month

The McGill VAS pre- and post-session and average for each session of the distractor and limb treatments for every subject are presented in Figures 2 and 3. Table 2 reports the average pain scores in the distractor and limb sessions for each subject. Cohen’s d for the difference between the two treatments was 5.35 (small-moderate effect) in one subject (SUB 7) favoring the limb treatment; the remaining subjects showed non-significant effect sizes for the difference between distractor and limb therapies.OPEN IN VIEWER

Figure 2.

Pain scores on the McGill VAS for each participant in each session.

OPEN IN VIEWER

Table 2.

McGill VAS Pain Rating Scores From Each Patient Collected Immediately Before Starting the First Session (Pre Session 1) and Immediately After the Last Session Treatment, Average of all Within Session Rating Scores (pain Ratings Collected Before and After Each Session) for Both Distractor and Limb Treatments; and Scores From the Follow-Up Testing Sessions.

	Distractor			Limb			FOLLOW-UP
	Pre Session1	Post last session	Average of all sessions	Pre Session1	Post last session	Average of all sessions	1-week	4-weeks	8-weeks
SUB 1	8.4	3	6.8	3.6	1.2	2.8
SUB 2	8.7	6.1	7	6.9	6.4	6.04	6.3	8	10
SUB 3	3.8	1.6	4.3	2.9	1.4	2.2
SUB 4	5.9	3.5	4.5	6.3	3.7	5.5	3
SUB 5	2.4	0	.55	0	0	0	0	0	0
SUB 6	7.1	5.7	6.3	7.3	3.5	4.4
SUB 7	6.5	4.7	5.4	2.9	0.9	1	4	0	0

Virtual Reality Assessment

The games were well liked and created a sense of “presence”—that is, a feeling of being in the environment (see Figure 4). While there was not a significant difference across the activities in either presence or usability questionnaires, the ratings were generally satisfactory; no subject expressed dissatisfaction with respect to usability.OPEN IN VIEWER

We encountered no significant adverse effects. Several subjects experienced mild cybersickness, an expected and non-significant adverse effect that did not cause any subject to withdraw from the study; for SUBs 1 and 4 this was noted in the first 2 sessions. We slightly modified our VR display after these subjects and noted only mild and transient cybersickness in one subsequent subject. SUB 6 reported mild, generalized headache with the VR therapy that resolved spontaneously; his headache was subsequently attributed by his primary care physician to sinusitis.

Repeated Battery Data Analysis

From the pre-distractor time point to the post-limb treatment time point, there was a 39% reduction in BPI pain interference scores (P=.06, n=5) and a 61% reduction in the HADS depression scores (P=.06, n=5).

When looking at the distractor and limb treatments independently (see Table 3), we observed a decrease in the pain interference in daily life scale of the BPI for pre-vs post-limb treatment (28% difference, P=.03). Furthermore, we observed a trend for reduction in pre-to post-limb treatment scores for depression (60% decrease, P=.06) and anxiety (36% difference, P=.06). No other differences were observed between pre- and post-treatment measures.

OPEN IN VIEWER

Table 3.

Average Scores for Repeated Battery Assessments. Cells Highlighted in Gray Indicate the Significant (Dark Gray) or Trend (Light Gray) Differences Within Treatment Comparisons (i.e., Pre-Post).

	Pre-Distractor n=7	Post-Distractor n=7	Pre-Limb n=5	Post-Limb n=5	1 Week n=5	4 Weeks n=4	8 Weeks n=4
BPI interference	6.1 (2.5)	5.6 (3.2)	6.1 (2.8)	3.3 (2.4)	2.8 (1.9)	.7 (.9)	1.6 (2.2)
FAI	18.8 (10.4)	19.7 (7.2)	21.6 (7.6)	23 (7.8)	22.2 (8)	22.7 (9.2)	21.2 (8.8)
HADS anxiety	6.8 (2.0)	5 (3.7)	5 (3.6)	3.2 (2.7)	3.4 (2.7)	1.2 (.9)	2 (1.4)
HADS depression	6.8 (3.1)	6.4 (3.2)	6.6 (1.8)	2.6 (2.4)	3.2 (3.9)	3.5 (3.8)	1.5 (1.2)
SFHS physical	30.1 (2.8)	31.4 (8.2)	29.8 (2.9)	35 (9.7)	40.4 (5.5)	45.7 (2.7)	42.1 (8.6)
Pain catastrophizing	27.2 (11.7)	21.1 (11.3)	30.3 (9.3)	13 (4.6)	14.6 (11.4)	13.7 (12.1)	11 (18.7)

Finally, we ran permutation tests comparing the magnitude of treatment effects for pre-vs post-limb treatment as compared to pre-vs post-distractor treatment (n=5). For the BPI interference score, the decrease in scores was larger for the limb treatment (28%) than for the distractor treatment (15%; P=.03).

Discussion

A VR intervention comprising a combination of distractor and active limb treatments was successful in reducing lower-limb PLP in most participants. Group analyses showed a reduction of pain on our primary outcome measure after both treatments. Individual analyses demonstrated that the limb therapy was superior to distractor treatment in one subject, but the two treatments did not differ in other subjects. In line with previous clinical trial interventions with PLP,^30,33 some improvement was observed in the majority of individuals who took part in the present study (85% of our sample); Subject 4 demonstrated a numerical pattern of increased pain during the intervention that may have been related to active systemic lupus erythematosus during that time. The activity required in both treatments may have thus triggered musculoskeletal discomfort in this specific individual. Interestingly, despite the within-session increase in pain, this individual showed an overall decrease in her pain severity across both treatments, suggesting that VR therapy was not detrimental.

Group-level analyses revealed a significant reduction of pain after both distractor and limb treatments. Although the two treatments showed similar positive effects on PLP, only the effect of the limb treatment survived Bonferroni correction, suggesting a more robust effect of this specific treatment.

This study explored the effect of VR in a group of lower-limb PLP amputees in which treatment was provided over the course of many sessions and followed up weeks after completion of treatment. The present study is distinguished from previous work with lower-limb amputees by the use of immersive VR rather than augmented reality ³⁷ ; additionally, we report data from group analyses rather than single cases. ³² Furthermore, our VR treatment differed from some studies in that we did not employ repetitive and structured motor^32,37 training; rather, participants were presented with a variety of virtual games that engaged the use of both limbs in a naturalistic fashion. However, it is unclear whether naturalistic activity or structured motor training is more beneficial in reducing PLP; future work should explore this question. Game-based interventions such as those evaluated in the present study have the advantage of being entertaining and engaging for patients. In a randomized controlled trial study, Rothgangel et al ³⁷ assessed the effects of mirror therapy after 4 weeks of clinical training. PLP patients treated with mirror therapy showed a reduction in pain that was not statistically different from the control group, which was offered a sensorimotor treatment targeting the intact limb. In the same study, the authors also assessed the effect of classical or augmented reality-based mirror therapy after 10 weeks and 6 months of home-based training; there was not a significant effect of either treatment on PLP relative to the control group at either time point. In contrast, Lendaro et al ³² obtained results more comparable with the present study; they found a 68% progressive reduction in PLP across 12 weeks of VR treatment that remained evident at 1- and 3-months follow-up but not at 6 months. The treatment consisted of a set of motor exercises to be executed in VR in which measurements from surface electromyography controlled the motions of a visually displayed limb, which caused subjects to feel as though they were controlling their missing leg. We chose a magnetic motion-tracking system over electromyography to simplify system setup and more accurately display real limb movements in VR.

Two previous studies tested the effectiveness of multiple sessions of VR treatment in individuals with upper-limb amputation.^30,50 These studies demonstrated a reduction of PLP intensity across treatment sessions^30,50 that, in the report of Ortiz–Catalan, ³⁰ persisted for several months. In conjunction with these studies and the report of Lendaro et al, ³² our data suggest that multiple VR sessions provide a benefit that increases over sessions and, in some subjects, persists for weeks to months.

Several factors are likely to have led to an underestimation of the potential benefit of the active limb VR treatment in our study. First, all patients except one had been suffering from PLP for many years and had tried multiple alternative treatments (medication, mirror therapy) with little or no success. Only 1 participant (SUB 7) started the treatment within 6 months of amputation. This individual showed a strong reduction of pain intensity (from 6–7/10 at the beginning of the treatment to 1/10 at the end of the treatment) and frequency (from several times a week to once a month at the end of the treatment). Although they must be interpreted with caution, these data suggest that the efficacy of VR treatment should be explored early in the course of the disorder. We are unaware of studies that have addressed the effect of VR therapy in an acute or sub-acute setting.

A second factor that may have led to an underestimation of the effect of limb VR therapy is that all subjects received an alternative therapy prior to limb VR therapy. Indeed, we observed a positive effect for both distractor and limb treatments. PLP was reduced by the distractor treatment, and the limb treatment produced additional pain reduction. As the limb treatment always occurred after the distractor treatment, it is difficult to partition out the effect of the limb treatment from the overall VR protocol and the placebo effect. However, both individual and group data suggest that the beneficial effect of the limb treatment was somewhat larger than that of the distractor treatment. Additional studies with larger samples will be required to further assess this possibility.

Although both the distractor and limb treatments produced positive results, we speculate that they differ with respect to mechanism. It is likely that the distractor therapy reduced pain at least in part by diverting attention from the pain^39,51; as there was a reduction between sessions, however, this account does not adequately explain the benefit and suggests that the relaxing VR environment may have a persistent down-regulating effect on pain perception.

The pain reduction with the limb treatment is likely to rely upon an additional mechanism, namely the provision of visual feedback that is able, in part, to substitute for the missing proprioceptive and sensory afferent information. Indeed, our system aimed at reducing the mismatch between the intact motor commands and visual feedback from the amputated limb. The evidence of a significant reduction of PLP after the limb treatment supports the hypothesis that PLP is related, in part, to the incongruence between predicted and expected sensory feedback and that the restoration of the impaired representation of the limb can reduce PLP. The results of Lendaro et al ³² are consistent with this speculation.

In addition, the battery data revealed a numerically beneficial effect of the limb treatment in the reduction of depression (HADS) and in pain interference with daily life (BPI). A decrease in PLP interference with daily activities had previously been observed in a clinical study with upper-limb PLP ³⁰ ; we provide evidence of this effect in subjects with lower-limb amputation. Although the small sample size mandates a cautious interpretation, the effects are encouraging. These data suggest that being able to see and have control of the amputated limb may improve attitudes towards pain, help in the daily management of pain and improve emotional sequelae of pain.

There were no significant or unanticipated adverse effects of the VR treatment. Two subjects experienced moderate cybersickness when playing the Quest for Fire game. This game might have increased motion sickness because participants rotated 360° in the VR environment using only the motion of their legs; the VR background moved accordingly, while their actual body remain stationary. Indeed, previous work had suggested that the lack of correspondence between absence of self-motion, provided with vestibular information, and visual motion generate cybersickness, proposing the control and reduction of scene movement as a possible way to reduce this symptom. ⁵² Although descriptive, we note that the limb treatment was engaging and entertaining. Several participants were not keen to stop the limb treatment, as they wanted to continue to play; several asked if they could purchase our limb treatment.

We note that the games used in the limb treatment sessions received adequate scores with respect to “presence” and system usability; in fact, scores were slightly higher for our games such as Quest for Fire and Dog Food as compared to the commercially available Cool! Efforts to enhance these games are underway.

Our study suffers from a number of limitations, including the small sample size and the fact that only subjects with below-the-knee amputations were included. Feedback from some potential participants suggested that the number of visits to the lab was an impediment to participation. We are currently exploring home-based treatments in individuals with PLP with above- and below-knee amputations.

Another limitation of the present study is that the distractor treatment, which we hypothesized at the outset would have smaller effects than the limb treatment, was always presented first. This choice was predicated on the anticipated small sample size, which was insufficient to assess order effects as would be required had we used a crossover design with balancing of order across subjects. Additionally, the design we used quasi-randomly assigns subjects to baselines (or initial treatments) of staggered lengths such that changes after the switch point between treatment A and B can be attributed to the effect of treatment rather than the passage of time or non-specific practice effects. ⁵³ This design is optimal to use with randomization tests and permutation analyses, which assess whether the obtained data could have occurred in Phase A or B by chance. Moreover, we did not include a washout period. We are therefore unable to assess whether the apparent benefits of the limb treatment are instead attributable to additive effects of both treatments. Future studies should investigate the long-term effects of both treatments independently.

Finally, as both the distractor and active VR treatments reduced PLP, one might speculate that the pain reduction was attributable to a placebo effect. We elected to compare active VR therapy to a treatment demonstrated to reduce pain (Cool!) to ensure a stringent test of VR therapy. The fact that active VR treatment reduced pain to a greater degree than distractor therapy by some measures argues against—but does not disprove—an account based on a general placebo effect. Future studies, including those currently planned, will address the role of placebo effects in VR therapy for PLP.