PHYST-1118; No.

of Pages 10
ARTICLE IN PRESS

Physiotherapy xxx (2020) xxx–xxx

The effectiveness of graded motor imagery for reducing

phantom limb pain in amputees: a randomised controlled trial
∗
Katleho Limakatsoa, Victoria J. Maddena,b, Shamila Maniec, Romy Parkera,c,
a
Pain Management Unit, Department of Anaesthesia and Perioperative Medicine, Faculty of Health Sciences, University of
Cape Town, Cape Town, South Africa
b
Department of Psychiatry and Mental Health, Faculty of Health Sciences, University of Cape Town, Groote Schuur Hospital,
Cape Town, South Africa
c
Division of Physiotherapy, Department of Health and Rehabilitation Sciences, Faculty of Health Sciences, University of Cape
Town, Cape Town, South Africa

Abstract
Objective To investigate whether graded motor imagery (GMI) is effective for reducing phantom limb pain (PLP) in people who have
undergone limb amputations.
Design A single-blinded randomised, controlled trial.
Setting Physiotherapy out-patient departments in three secondary level hospitals in Cape Town, South Africa.
Participants Twenty-one adults ( ≥ 18 years) who had undergone unilateral upper or lower limb amputations and had self-reported PLP
persisting beyond three months.
Interventions A 6-week GMI programme was compared to routine physiotherapy. The study outcomes were evaluated at baseline, 6
weeks, 3 months and 6 months.
Outcome measures The pain severity scale of the Brief Pain Inventory (BPI) was used to assess the primary outcome — PLP. The pain
interference scale of the BPI and the EuroQol EQ-5D-5L were used to assess the secondary outcomes — pain interference with function
and health-related quality of life (HRQoL) respectively.
Results The participants in the experimental group had significantly greater improvements in pain than the control group at 6 weeks and 6
months. Further, the participants in the experimental group had significantly greater improvements than the control group in pain
interference at all follow-up points. There was no between-group difference in HRQoL.
Conclusion The results of the current study suggest that GMI is better than routine physiotherapy for reducing PLP. Based on the
significant reduction in PLP and pain interference within the participants who received GMI, and the ease of application, GMI may be a
viable treatment for treating PLP in people who have undergone limb amputations.
Clinical trial registration number (PACTR201701001979279).
© 2019 The Authors. Published by Elsevier Ltd on behalf of Chartered Society of Physiotherapy. This is an open access article under the
CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Graded motor imagery; Phantom limb pain; Amputees; Left/right discrimination; Explicit motor imagery; Mirror therapy

Introduction amputated limb [1]. Phantom limb pain occurs in up to

85% of amputees, making it the most common chronic pain
Phantom limb pain (PLP) is a debilitating condition con- dition in people who have undergone limb
char- acterised by painful sensations in the missing portion amputations [2]. The onset of PLP is often immediate,
of the although in some cases, it may be several years after
amputation, with painful sensa- tions varying between
∗ sharp, shooting, burning, throbbing, stabbing, and aching
Corresponding author at: Pain management unit, Department of
Anaes- thesia and Perioperative Medicine, D23 Groote Schuur Hospital, [3,4]. Phantom limb pain has a negative impact on
Faculty of Health Sciences, Anzio Rd., Observatory, 1925, South Africa. psychological well-being and is associated with
E-mail address: [email protected] (R. Parker).

https://doi.org/10.1016/j.physio.2019.06.009
0031-9406/© 2019 The Authors. Published by Elsevier Ltd on behalf of Chartered Society of Physiotherapy. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: Limakatso K, et al. The effectiveness of graded motor imagery for reducing phantom limb pain in
amputees: a randomised controlled trial. Physiotherapy (2020), https://doi.org/10.1016/j.physio.2019.06.009
2 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx

anxiety and depression in people who have undergone limb receive GMI. The trial provided preliminary evidence that
amputations [5]. Further, PLP interferes with sleep, GMI may reduce PLP, suggesting that a larger trial is
mobility, and work, general activities of daily living and warranted. With this back-
enjoyment of life [6].
Phantom limb pain is often ineffectively treated [7], per-
haps because of the conflicting views about the proposed
underlying mechanisms. Some evidence suggests that PLP
is primarily driven by heightened nociceptive activity from
neuromas located in the stump [8]. However, PLP has also
been reported in congenital amputees who clearly do not
suf- fer nerve trauma preceding PLP, in whom the
spontaneous discharge from a neuroma is unlikely to be a
dominant con- tributor [4]. The presence of PLP in this
group suggests that peripheral processes alone are
insufficient to account for the phenomenon of PLP.
Recent neurophysiological evidence has linked PLP to
cortical alterations in the somatosensory and motor cortices
of the brain contralateral to the amputated limb [3,9–14]. In
people with PLP, the cortical representation of the
amputated limb is ‘invaded’ by adjacent cortical areas, with
a positive correlation between the ‘invasion’ of these
cortical areas and the severity of PLP [3,11,15].
Interestingly, these cortical alterations can be reversed
using motor imagery [10,16], and there is a strong
association between the reversal of these changes and the
relief of PLP in people who have undergone limb
amputations [17,18].
The Graded Motor Imagery (GMI) programme is an
inter- vention that aims to reduce PLP using a graded
sequence of strategies including left/right judgements
(implicit motor imagery), imagined movements (explicit
motor imagery) and mirror therapy [17]. Left/right
judgements are made when one must distinguish a body
part belonging to the left side of the body from one
belonging to the right side [19]. The mental rotation
involved in performing a left/right judgement has been
shown to activate the somatosensory, premotor and
supplementary motor areas contralateral to the phantom
limb [20,21]. Imagined movements involve imagining
moving the phantom limb into various postures. Imagined
movements activate the somatosensory, premotor and
primary motor cor- tices contralateral to the phantom limb
[13,14,22]. Mirror therapy involves concealing the
amputated limb behind a mirror and simultaneously
moving the phantom portion of the amputated limb (a
movement intention) and the intact limb while observing
the reflection of the intact limb in the mirror, essentially
providing visual feedback that the move- ment intention
was successful [23]. Mirror therapy is thought to address
changes in the somatosensory and primary motor cortices
by providing visual feedback that matches the motor
intention, thus resolving a visual-motor mismatch that may
contribute to pain [12,16].
A recent systematic review of the literature by our team
[24] found only one small, randomised controlled trial of
GMI for reducing PLP after amputation [17]. Nine
amputees were included, five of whom were allocated to
ground, we conceived the current study, aiming to calculated as the mean of the seven ratings (max-
investigate whether the GMI programme is effective for
reducing PLP by testing it in a larger sample of people
who have undergone amputations.

Methods

A single-blind randomised controlled trial (RCT) was

conducted in three secondary level hospitals in Cape
Town, South Africa. The population of interest was adults
(over 18 years old) who had undergone unilateral upper or
lower limb amputations and had self-reported PLP
persisting for three months or longer. The names and
contact details of patients who had undergone amputations
between January 2013 and December 2016 were retrieved
from hospital records. Patients were then contacted
telephonically to inform them about the study and to invite
them to participate. Patients interested in the study were
screened against the inclusion/exclusion criteria using a
telephone-administered questionnaire. Patients were
excluded if they reported double amputations,
psychopathological disorders, motor problems (e.g.
tremor, dyskinesia), or severe systemic illness (e.g.
advanced cancer, lupus), visual impairment such that they
could not read unaided or with their reading glasses, or
having previously received GMI treatment. Eligible par-
ticipants were invited to the hospital for further screening,
and to complete informed consent and baseline assessment
procedures. The study sample size was calculated for 80%
power to detect a clinically meaningful between-groups
change of 3 (SD = 2.1) on a 0–10 scale for pain severity
at 6 months [17]. Statistical significance was set at 0.05.
The online sample size calculator (http://hedwig.mgh.
harvard.edu/sample size/js/js parallel quant.html) showed
that 18 participants were required (see supplementary file
1).

Outcomes

The primary outcome was pain in the phantom limb,

assessed using the pain severity scale of the Brief Pain
Inven- tory (BPI) [25]. This scale consists of four
questions that ask participants to rate the severity of their
worst, average, least and current pain on a 0 to 10 scale,
with 0 representing ‘no pain’ and 10 representing ‘pain as
bad as you can imagine’. The pain severity scale is
calculated as the mean of the four ratings (maximum: 10)
[26].
The secondary outcomes were (1) the interference of
pain with function and (2) health-related quality of life.
The inter- ference of pain with function was assessed
using the pain interference scale of the BPI. This scale
consists of seven questions that ask participants to rate the
extent to which pain interferes with their general activity,
mood, walking ability, work, relations with other people,
sleep and enjoyment of life, ona0 to 10 scale, with 0
representing ‘does not interfere’ and 10 representing
‘completely interferes’. The pain interference score is
 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx 3

Fig. 1. Consort diagram illustrating the process from recruitment to data collection (6-month follow-up).

imum: 10) [26]. Health-Related Quality of Life (HRQoL) ate a “random permutation of integers”) in a
was assessed using the Visual Analogue Scale (VAS) of the counterbalanced manner, such that group sizes were
EuroQol EQ-5D-5L, where participants rate the quality of approximately equal. The intervention period lasted 6
their health on a VAS ranging between 0 (worst imaginable weeks. The intervention group received GMI, whereas the
health state) and 100 (best imaginable health state) [27]. control group received routine physiotherapy care. Study
The individual domains of the EQ-5D-5L were completed outcomes were re-assessed on ces- sation of the
by the participants for purposes outside the current study. intervention at six weeks, as well as at three months and six
Therefore, they are not reported here. The BPI and EQ-5D- months after the intervention.
5L have been validated for use in South Africa [28,29]

Graded motor imagery intervention

Procedure
Prior to the initiation of treatment, the treating clinician
The details of the study procedure are shown in Fig. 1. briefly explained to the participant that PLP is a painful
To ensure blinding, two different researchers performed the condition associated with functional changes in the brain.
outcome assessments and treatments. The outcome assessor He explained that the cortical representation of the ampu-
was an occupational therapist trained in research methods tated limb is invaded by adjacent cortical areas, and that
and was blinded to group allocation. The treating clinician GMI is a treatment strategy which aims to reduce PLP by
was a physiotherapist trained in research methods and GMI retraining the brain in a three-step process (described
and was blinded to the outcome assessments. After the below). Each step of the GMI programme was carried out
baseline assessment of study outcomes, participants were for two weeks, during which the participant received
randomly assigned to the experimental and control groups treatment for 30 minutes on two separate days of the first
(using the random list generator at week (at least one day apart) and continued with a
www.randomization.com to gener-
structured home-exercise programme throughout the
first and second week. The par-
4 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx

ticipants received weekly telephone calls to remind them

to continue with the home programme and to record their
participation in a diary.

Left/right judgements

The first two weeks of the GMI programme were intact limb positioned comfortably in front of the mirror
focused on the training of left/right judgements. Treatment (Fig. 2).
dur- ing sessions with the clinician used the Neuro
Orthopaedic Institute’s (NOI) RecogniseTM software
application. The application was set to “vanilla”, in which
photographs are presented on a plain background to
minimise distractions [30]. Fifty photographs of limbs
representing the amputated limb and the intact limb were
presented in various positions and alignments on a tablet
held by the comfortably seated participant. Each
photograph was presented for five seconds, during which
time the participant identified the presented limb as left or
right by touching a key on the tablet. The RecogniseTM
application recorded response time and accu- racy for each
trial. These left/right judgement tasks were looped for 30
minutes per treatment session with less than a minute
between loops. For the home exercise programme,
participants were provided with several magazines
containing photographs of people. They were instructed to
identify and circle the limbs that matched the side of their
own amputated limb. Participants were instructed to
perform these tasks for 10 minutes during every waking
hour from 9:00am to 9:00pm every day (12 sessions daily).

Imagined movements

The second two weeks of the GMI programme focused

on training imagined movements. From a collection of
pho- tographs on the NOI RecogniseTM software
application, participants were shown an image of a limb
that matched the side of their own amputated limb. They
were instructed to imagine moving their amputated limb
slowly and smoothly from the position they felt it to be in,
to the posture shown in the image, and then to imagine
moving it back to its origi- nal position. Participants were
explicitly advised to imagine themselves performing the
movement instead of imagining someone else performing
the movement. Imagined move- ment tasks were performed
repeatedly, using three images for each 30-minute
treatment session. For the home exercise programme,
participants were given three printed images or three
images on their smart phones and were advised to repli-
cate the instructions of the treatment session for 10 minutes
of every waking hour from 9:00am to 9:00pm every day
(12 sessions daily).

Mirror therapy

The final two weeks of the GMI programme used mirror

therapy. During mirror therapy, the amputated limb was
con- cealed behind a mirror (300
× mm 300 mm) with the
Fig. 2. A participant undergoing mirror therapy.

The patient was then shown a photograph of the unaffected

limb in an easy to attain position and instructed to simul-
taneously move the intact limb and the phantom limb into
the presented posture while observing the reflection of the
intact limb in the mirror. Mirror therapy tasks were
performed repeatedly, using three photographs for each
30-minute treat- ment session. For the home exercise
programme, participants were provided with a mirror and
advised to replicate the instructions of the treatment
session using the same pho- tographs for 10 minutes of
every waking hour from 9:00am to 9:00pm every day (12
sessions daily).

Routine physiotherapy

Participants allocated to the control group were advised

to continue rehabilitation at their respective physiotherapy
out- patient departments and continue with home
programmes of their own preference or as provided by
their clinicians, as frequently as possible. Participants
were given a diary and advised to record their activities,
specifying the nature, frequency and duration of each
activity.

Statistical analysis

Data were analysed using STATISTICA software [31].

The primary outcome was pain severity, with a clinically
meaningful reduction in pain (on a pain severity scale of
the BPI) from baseline used to classify treatment as
successful [32]. A reduction of three points or more (on a
pain interfer- ence scale of the BPI), or a pain reduction to
zero was used to indicate a clinically meaningful reduction
in pain interference [33,34]. Further analyses were
conducted on pain interfer- ence and health-related quality
of life. Because the pain severity scores at baseline were
not normally distributed, non- parametric analysis was
conducted. The efficacy of treatment at each of the
multiple data collection points was determined by
calculating the difference between median baseline and
 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx 5

Table 1 Table 3
Demographic characteristics of participants at baseline (n = 21). Between-group difference in pain severity and interference scores in the
Variable Experimental group Control group experimental and control groups.

Age in years [median (IQR)] 63 (53 to 65) 62 (59 to 67) Odds Ratio (95% CI)
Gender [n (%)] Pain severity
Male 8 (73) 8 (80) 6 weeks 10.50 (1.36 to 81.12)
Female 3 (27) 2 (20) 3 months 4.50 (0.63 to 32.30)
Medical information 6 months 15 (1.21 to 185)
Reason for amputation [n (%)] Pain interference
Diabetes 9 (82) 7 (70) 6 weeks 78 (3.31 to 1849)
Infection 1 (9) 2 (20) 3 months 10 (1.43 to 81.11)
Trauma 1 (9) 1 (10) 6 months 63 (3.34 to 1194.81)
Time since amputation [median (IQR)]
Months 17 (13 to 28) 20 (12 to 36)
Site of amputation [n (%)] a median age of 62 [IQR (interquartile range): 59 to 65]
Lower limb 11 (100) 9 (90)
Upper limb 0 (0) 1 (10) years. The most common reason for amputation was
Level of amputation [n (%)] diabetic complications (n = 16), followed by trauma (n = 3)
Below knee 6 (55) 4 (40) and infec- tion (n = 2). Only one participant had undergone
Above knee 5 (45) 5 (50) an upper limb amputation (above the elbow); and an equal
Above elbow 0 (0) 1 (10) number of participants had undergone above-knee and
Outcomes [median (IQR)]
Pain severity 5 (3.75 to 5.75) 3.25 (2.50 to 5.75)
below-knee amputations (n = 10 for each). The results for
Pain interference 5.43 (2.43 to 8.30) 1.73 (1.14 to 4.14) PLP severity and interference are summarised in Tables 2
HRQoL 70 (60 to 90) 75 (60 to 95) and 3.
Abbreviations: IQR, Interquartile range; HRQoL, Health-Related Quality
of Life; n, number of participants. Phantom limb pain (primary outcome)

post-intervention pain severity and interference scores. The The participants in the experimental and control groups
Odds Ratio (OR) was calculated to test for a significant had improved pain severity at 6 weeks (P = 0.007; P =
median difference between groups at each time point. For 0.002) and 3 months (P < 0.001; P = 0.001). However, only
all analyses, the alpha was set≤at P 0.05. The Fragility the par- ticipants in the experimental group had further
Index (FI) was calculated to determine the robustness of improvements in pain severity at 6 months (P < 0.001; P =
the results at different data collection points [35]. The FI 0.58) (Fig. 3). The participants in the experimental group
indicates the number of participants with a bad outcome had significantly greater improvements in pain than the
that would convert the results from being significant to not control group at 6 weeks (P = 0.02) and 6 months (P =
significant (P > 0.05). To explore clinically meaningful 0.03). No between-group difference was seen at 3-months
improvements in pain, the Number Needed to Treat (NNT) (P = 0.14). At all three follow-up time points, the fragility
AR
[ 1 ] was calculated, with a 3-point
R reduction in pain index for pain severity was 1. The NNTs were: 2 [95% CI:
considered to be clinically mean- ingful. The missing data 1.10 to 6.50] at 6 weeks,
of participants lost at follow-up were handled by carrying 3 [95% CI: 1.91 to 7.13] at 3 months and 2 [95% CI: 1.11 to
forward the last observed measure [36]. 7.10] at 6 months.

Pain interference (secondary outcome)

Results
The participants in the experimental group had improved
The demographic characteristics of participants are pre- pain interference with function at 6 weeks (P < 0.001), 3
sented in Table 1. The study included 21 participants with months (P < 0.001) and 6 months (P < 0.001) (Fig. 4). The
participants in the control group had no improvement in
pain

Table 2
Pain severity and interference scores in the experimental and control groups at each time point.
Baseline 6 weeks 3 months 6 months Median difference (95%
median (IQR) median (IQR) median (IQR) median (IQR) CI) baseline-6 months
Pain severity
Experimental 5 (3.75 to 5.75) 0.75 (0 to 2.75) 0 (0 to 0.50) 0 (0 to 1) 5 (3.06 to 5.80)
Control 3.25 (2.50 to 5.75) 1.50 (0.75 to 4) 1.88 (0 to 4.50) 5.63 (0.37 to 6.63) −1.37 (−7.43 to 9.75)
Pain interference
 Experimental 5.43 (2.43 to 8.30) 0 (0 to 2) 0 (0 to 1.86) 0 (0 to 0) 5.43 (3.10 to 6.56)
Control 1.73 (1.14 to 4.14) 1.65 (0.29 to 6.30) 2.64 (0 to 4.60) 4.17 (0 to 6.75) −2.44 (−3.34 to 3.28)
Abbreviations: IQR, interquartile range; CI, confidence interval.
The median difference was calculated by subtracting 6 months pain scores from baseline scores. Therefore, a positive difference indicates a decrease in pain.
6 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx

Fig. 3. Change in pain severity scores over time for the experimental (n = and 6 months (U = 46; P = 0.16).
11) and control (n = 10) groups. Each black dot represents a single
participant’s score.

Fig. 4. Change in pain interference scores over time for the experimen-
tal (n = 11) and control (n = 10) groups. Each black dot represents a single
participant’s score.

interference at all data collection points. The participants in

the experimental group had significantly greater improve-
ments than the control group in pain interference at 6
weeks (P = 0.007), 3 months (P = 0.02) and 6 months (P =
0.006). At all three follow-up time points, the fragility
index for pain interference was 4. The NNTs were: 2 [95%
CI: 1 to 2.13] at
6 weeks, 2 [95% CI: 1.12 to 6.50] at 3 months and 2 [95%
CI: 0.91 to 2.14] at 6 months.

Health-related quality of life (secondary outcome)

The participants in the experimental group had improved

HRQoL at 6 weeks (P = 0.003), 3 months (P = 0.002) and
6 months (P = 0.02) (Fig. 5). The participants in the control
group had no improvement in HRQoL at any of the data
collection points. No between-group differences were
shown at 6 weeks (U = 35; P = 0.16), 3 months (U = 46; P
= 0.16)
Fig. 5. Change in EQ-VAS scores over time for the experimental (n = 11)
and control (n = 10) groups. Each black dot represents a single
participant’s score.

Discussion

The aim of this study was to investigate whether the

Graded Motor Imagery programme (GMI) is effective for
reducing PLP in people who have undergone amputations.
Here, a 6-week GMI programme was compared to rou-
tine physiotherapy care and found to reduce PLP and pain
interference with function. GMI was superior to routine
phys- iotherapy for producing clinically meaningful pain
reductions for up to six months after the intervention.
However, the improvement in these outcomes was not
reflected by an improvement in quality of life.
The participants in the GMI group had pain reductions
greater than three points on a 10-point VAS. A cut-off of
two points is widely accepted as an indication of a
success- ful treatment [37,38]. In this study however, a
conservative estimate of three points was used as an
indication of a success- ful treatment. Considering the
conservative approach taken during data analysis, our
results indicate that only the par- ticipants who received
GMI had clinically meaningful pain reductions at all three
follow-up time points (median differ- ences for GMI
group: 4.25 to 5; control group: 1.37 to 1.75). The NNT to
achieve a clinically meaningful pain reduction at these
time points was 1.90 (95% CI: 1.12 to 7.16). To eval- uate
the significance of this NNT, it is worth considering in
relation to the NNTs for other interventions recommended
for the treatment of PLP (Table 4). The previously
reported NNTs for GMI [2 (95% CI: 1 to 5)] and the
NNT obtained
in the present study [1.90 (95% CI: 1.11 to 7.10)] compare
favourably with the NNTs for lidocaine [3.80 (95% CI:
1.93 to 16.61)], amitriptyline [5.20 (95% CI: 3.62 to
9.11)] and
pregabalin [8 (95% CI: 5.94 to 32)], all of which are
recom- mended for the treatment of PLP [39]. The low
NNTs for GMI suggest that treatment of patients with PLP
could be optimised with the application of this non-
pharmacological intervention.
The NNT results are supported by the odds ratios. The
odds ratio at 6-month follow-up indicates that participants
who underwent the GMI programme had a 15 times
greater
 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx 7
that the ordered application had a superior effect for reducing
Table 4 pain [40]. One explanation for this is that imagined
Numbers Needed to Treat for different interventions used for the manage-
ment of PLP.
Treatments for PLP NNT (95% CI)
Graded Motor Imagery (current study)
Immediate 2 (1.10 to 6.50)
3-months follow-up 3 (1.90 to 7.10)
6-months follow-up 2 (1.11 to 7.13)
Graded Motor Imagery [17]
Immediate 3 (2 to 6)
6-months follow-up 2 (1 to 5)
Lidocaine infusion [46]
30 minutes 4 (1.94 to 16.60)
Memantine [47]
3-weeks 6 (2.13 to 10.62)
Amitriptyline [48]
6-weeks 6 (3.60 to 9.13)
Pregabalin [49]
4-weeks 8 (5.91 to 32)

chance of having a clinically meaningful reduction in pain

than those who received routine physiotherapy (P = 0.03).
These results are reinforced by those of an RCT and a
retro- spective case-series on the effects of GMI on PLP in
people who have undergone amputations [17,18]. The RCT
of nine participants showed that the GMI group (n = 5) had
a clin- ically meaningful reduction (GMI group 3.2; control
group 0.6, o n a 0 to 10 VAS) in pain at 6 weeks [17]. In that
study, the NNT for a 50% decrease in pain at 6 weeks was
3 [95% CI: 2 to 6]. Further improvements were seen at 6
months, with those receiving GMI continuing to report
clinically meaning- ful reductions in pain (GMI group 3.8;
control group 0.7, on a 0 to 10 VAS), with an NNT of 2
[95% CI: 1 to 5]. Between- group comparison at these
follow-up time points showed that GMI was superior to
routine physiotherapy for reducing PLP (t(7) = 2.60, P =
0.04). Similarly, a recent retrospective case series of GMI
in four amputees revealed clinically meaning- ful pain
reductions (3.30; 0 to 10 VAS) with three of the four
patients reporting no pain at all at the end of the GMI pro-
gramme [18]. Interestingly, these NNTs for the use of GMI
for PLP resemble those of studies in Complex Regional
Pain Syndrome Type 1 (CRPS 1) – the condition for which
GMI was first developed [17,40] – which range from 2 to 3
[17,40]. Although the consistent positive findings from the
literature may reflect a publication bias, the results from
these studies suggest that GMI may be a viable treatment
for PLP.
The effectiveness of the GMI programme in reducing
PLP might be because it targets the cortical mechanisms
proposed to underlie PLP. The rationale for the three
components of the GMI programme is that they
progressively correct cortical changes associated with PLP
by activating the cortical rep- resentation of the amputated
limb in a graded manner [17]. This graded progression is
credited for the success of the GMI programme. A previous
study comparing the ordered applica- tion of the
components of GMI to an unordered application of the
same components in people with CRPS 1, demon- strated
 the low FI could be due to the relatively small study sample
movements or mirror therapy alone seem to be capable rather than the unreliability of the results [35]. Whichever
of aggravating chronic pain [40,41]. To avoid the the reason, these results should be interpreted with caution.
triggering of a protective pain response from the brain, a Further, the OR at 6 weeks and 3 months should be
less demanding exercise (left/right judgements) may be
needed first. Perhaps, in a sense, left/right judgement is
to imagined movements what standing is to walking,
and imagined movements is to mirror therapy what
walking is to running.
The current results also indicate that GMI may be
effec- tive for reducing the interference of pain with
function at 6 weeks and this effect is sustained at 6
months. The NNTs immediately after intervention and at
6 months indicate that one of two patients who have
undergone GMI will have sig- nificant reductions in
pain interference with function. This, too, is supported
by the results of the only pre-existing RCT of GMI for
PLP [17]. That RCT used a patient-specific func- tional
scale where patients selected five activities that they
regularly performed before amputation but now found
dif- ficult because of PLP, and recorded the degree of
difficulty on a numerical rating scale (NRS: 0 to 10). At
the end of that trial, the GMI group had significantly
greater improve- ments in function than the control
group, with an NNT of 2 [95% CI: 1 to 5]. Together,
these results support that GMI may reduce the
interference of PLP with function in people who have
undergone amputations. Surprisingly, the reduction in
pain interference in the current study was not reflected
by improvements in health-related quality of life —
perhaps due to the many varied contributors to health-
related quality of life that are unrelated to pain. For
example, a systematic review which investigated factors
influencing the quality of life of lower limb amputees
found that depression, the inability to walk and low self-
reliance were associated with a low quality of life [42].
In addition, it has been reported that these factors are
primarily a consequence of the loss of a limb, and that
PLP is only a contributor [6,43]. This suggests that
reducing PLP may not necessarily result in significant
improvements in the quality of life due to limb loss,
because, the patient may still have depression, activity
limitations and participa- tion restrictions. Perhaps
rehabilitation programmes that go beyond merely
treating the PLP and which aim to restore patients to
their previous level of function may improve the quality
of life in patients who have undergone limb amputa-
tions [44].

Limitations

Our results are less robust than would be hoped. The

FI of 1 for pain severity indicates that only one
participant with a bad outcome would convert the results
from being significant to not significant (P > 0.05) [45].
Perhaps the loss of statistical significance at three
months could be due to the worsening of one participant
in the GMI group (Fig. 6). Alternatively, although the
current sample size was based on a priori cal- culation,
8 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx

Fig. 6. Pain severity score at four time points, coloured by participant.

interpreted with caution because this study had been

powered to detect a clinically meaningful between-group Key messages
change at 6 months. What the article adds to the current literature:
Further limitations include the lack of monitoring of
treat- ment adherence, the small sample, the restricted - The results of this study suggest that the graded
setting (Cape Town, South Africa), a single treating motor imagery programme is superior to routine
clinician, a lack of post-trial blinding assessment and lack physiother- apy for reducing phantom limb pain in
of a sham treatment. A larger multicentre trial using a valid people who have undergone limb amputations.
sham treatment and a more representative sample of What new knowledge is added by this study?
amputees, conducted in multiple sites by different
clinicians may provide a defini- tive conclusion regarding - Graded motor imagery is effective for reducing
the effectiveness of GMI within a broader context. phan- tom limb pain in people who have undergone
limb amputations regardless of the cause of the
amputation.

Conclusion
Ethical approval: This study was granted ethical approval
The results of the current study showed that GMI is by the Human Research Ethics Committee of the
better than routine physiotherapy for reducing PLP. Based University of Cape Town (HREC: 244/16).
on the clinically meaningful reduction in PLP and pain Funding: This study was funded by PainSA. KL received
interference within the participants who received GMI, scholarships from the Oppenheimer Memorial Trust and
and the ease of application, GMI may be a viable treat- National Research Foundation of South Africa. VJM was
ment for treating PLP in people who have undergone limb supported by an Innovation Postdoctoral Fellowship from
amputations. the National Research Foundation of South Africa.
 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx 9

[14] Lotze M, Montoya P, Erb M, Hülsmann E, Flor H, Klose U, et al.

Conflict of interest: None declared. Activation of cortical and cerebellar motor areas during executed
and imagined hand movements: an fMRI study. J Cogn Neurosci
1999;11(5):491–501.
[15] Raffin E, Richard N, Giraux P, Reilly KT. Primary motor cortex changes after
Acknowledgements amputation correlate with phantom limb pain and the ability to move
the phantom limb. NeuroImage 2016;130:134–44.
[16] Diers M, Christmann C, Koeppe C, Ruf M, Flor H. Mirrored, imag-
We thank Halalisani Cebolenkosi Mbatha for his contri- ined and executed movements differentially activate sensorimotor
bution as the blinded assessor throughout the study. We are cortex in amputees with and without phantom limb pain. PAIN®
grateful for his commitment to this work. 2010;149(2):296–304.
The authors thank the staff of Somerset, Khayelitsha [17] Moseley GL. Graded motor imagery for pathologic pain A randomized
controlled trial. Neurology 2006;67(12):2129–34.
and Victoria hospitals for their assistance with recruitment
[18] Hinkel M. Graded motor imagery for the treatment of phantom limb
of participants. pain. Arch Phys Med Rehabil 2017;98(10):e72.
[19] Wong CK, Wong CK. Limb laterality recognition score: a reliable clini- cal
measure related to phantom limb pain. Pain Med 2017;19(4):753–6.
[20] Parsons LM. Imagined spatial transformation of one’s body. J Exp
Appendix A. Supplementary data Psychol Gen 1987;116(2):172.
[21] Parsons LM, Fox PT. The neural basis of implicit movements used in
Supplementary data associated with this article can be recognising hand shape. Cogn Neuropsychol 1998;15:583–616.
found, in the online version, at https://doi.org/10.1016/ [22] Gerardin E, Sirigu A, Lehéricy S, Poline J-B, Gaymard B, Marsault C, et
al. Partially overlapping neural networks for real and imagined hand
j.physio.2019.06.009. movements. Cereb Cortex 2000;10(11):1093–104.
[23] Ramachandran VS, Rogers-Ramachandran D. Synaesthesia in phantom limbs
induced with mirrors. Proc R Soc London. Series B: Biol Sci
1996;263(1369):377–86.
References [24] Limakatso K, Parker R, Madden V. The effects of graded motor
imagery and its components on phantom limb pain and disability in
[1] Siddle L. The challenge and management of phantom limb pain after upper and lower limb amputees: a systematic review and meta-
amputation. Br J Nurs 2004;13(11):664–7. analysis. 2018 [submitted for publication].
[2] Sherman RA, Sherman CJ. Prevalence and characteristics of chronic [25] Ilfeld BM, Moeller-Bertram T, Hanling SR, Tokarz K, Mariano ER,
phantom limb pain among American veterans: results of a trial survey. Loland VJ, et al. Treating intractable phantom limb pain with ambu-
Am J Phys Med Rehabil 1983;62(5):227–38. latory continuous peripheral nerve blocks: a pilot study. Pain Med
[3] Flor H, Nikolajsen L, Jensen TS. Phantom limb pain: a case of mal- 2013;14(6):935–42.
adaptive CNS plasticity? Nat Rev Neurosci 2006;7(11):873. [26] Cleeland C, Ryan K. Pain assessment: global use of the Brief Pain
[4] Ehde DM, Czerniecki JM, Smith DG, Campbell KM, Edwards WT, Inventory. Singapore: Annals, Academy of Medicine; 1994.
Jensen MP, et al. Chronic phantom sensations, phantom pain, residual [27] Wang P, Luo N, Tai E, Thumboo J. The EQ-5D-5L is more discrimina-
limb pain, and other regional pain after lower limb amputation. Arch tive than the EQ-5D-3L in patients with diabetes in Singapore. Value
Phys Med Rehabil 2000;81(8):1039–44. Health Reg Issues 2016;9:57–62.
[5] Desmond DM, MacLachlan M. Prevalence and characteristics of phan- tom [28] Parker R, Jelsma J, Stein DJ. Pain in amaXhosa women liv-
limb pain and residual limb pain in the long term after upper limb ing with HIV/AIDS: Translation and validation of the Brief Pain
amputation. Int J Rehabil Res 2010;33(3):279–82. Inventory–Xhosa. J Pain Symptom Manage 2016;51(1), 126-32. e2.
[6] van der Schans CP, Geertzen JH, Schoppen T, Dijkstra PU. Phantom [29] Jelsma J, Mkoka S, Amosun L, Nieuwveldt J. The reliability
pain and health-related quality of life in lower limb amputees. J Pain and validity of the Xhosa version of the EQ-5D. Disabil Rehabil
Symptom Manage 2002;24(4):429–36. 2004;26(2):103–8.
[7] Angarita M, Alejandra M, Carrillo Villa S, Ribero G, Fernando O, [30] Wajon A. RecogniseTM Hands app for graded motor imagery training
García RG, et al. Pathophysiology and treatment of phantom limb pain. in chronic pain. J Physiother 2014;60(2):117.
Revista Colombiana de Anestesiología 2014;42(1):40–6. [31] StatSoft I, Tulsa, USA STATISTICA (data analysis software system),
[8] Vaso A, Adahan H-M, Gjika A, Zahaj S, Zhurda T, Vyshka G, version 6; 2001 http://www.statsoft.com.
et al. Peripheral nervous system origin of phantom limb pain. PAIN® [32] Lee JS, Hobden E, Stiell IG, Wells GA. Clinically important change in the
2014;155(7):1384–91. visual analog scale after adequate pain control. Acad Emerg Med
[9] Karl A, Birbaumer N, Lutzenberger W, Cohen LG, Flor H. Reorgani- 2003;10(10):1128–30.
zation of motor and somatosensory cortex in upper extremity amputees with [33] Dworkin RH, Turk DC, Wyrwich KW, Beaton D, Cleeland CS, Farrar
phantom limb pain. J Neurosci 2001;21(10):3609–18. JT, et al. Interpreting the clinical importance of treatment outcomes
[10] MacIver K, Lloyd D, Kelly S, Roberts N, Nurmikko T. Phantom in chronic pain clinical trials: IMMPACT recommendations. J Pain
limb pain, cortical reorganization and the therapeutic effect of mental 2008;9(2):105–21.
imagery. Brain 2008;131(8):2181–91. [34] Wong K, Zeng L, Zhang L, Bedard G, Wong E, Tsao M, et al. Minimal
[11] Flor H, Elbert T, Knecht S, Wienbruch C, Pantev C, Birbaumers N, et al. clinically important differences in the brief pain inventory in patients
Phantom-limb pain as a perceptual correlate of cortical reorganization with bone metastases. Support Care Cancer 2013;21(7):1893–9.
following arm amputation. Nature 1995;375(6531):482–4. [35] Walsh M, Srinathan SK, McAuley DF, Mrkobrada M, Levine O, Ribic
[12] Foell J, Bekrater-Bodmann R, Diers M, Flor H. Mirror therapy for C, et al. The statistical significance of randomized controlled trial
phantom limb pain: brain changes and the role of body representation. results is frequently fragile: a case for a Fragility Index. J Clin Epidemiol
Eur J Pain 2014;18(5):729–39. 2014;67(6):622–8.
[13] Lotze M, Flor H, Grodd W, Larbig W, Birbaumer N. Phantom move- [36] Kenward MG, Molenberghs G. Last observation carried forward: a
ments and pain An fMRI study in upper limb amputees. Brain crystal ball? J Biopharm Stat 2009;19(5):872–88.
2001;124(11):2268–77.
10 K. Limakatso et al. / Physiotherapy xxx (2020) xxx–xxx

[37] Farrar JT, Young Jr JP, LaMoreaux L, Werth JL, Poole RM. Clinical Wong CK, Ehrlich JE, Ersing JC, Maroldi NJ, Stevenson CE, Varca MJ.
importance of changes in chronic pain intensity measured on an 11- Exercise programs to improve gait performance in people with lower limb
point numerical pain rating scale. Pain 2001;94(2):149–58. amputation: a systematic review. Prosthet Orthot Int 2016;40(1):8–17.
[38] Bombardier C, Hayden J, Beaton DE. Minimal clinically impor- [45] Carter RE, McKie PM, Storlie CB. The fragility index: a P-value in
tant difference. Low back pain: outcome measures. J Rheumatol sheep’s clothing? Eur Heart J 2017;38(5):346–8.
2001;28(2):431–8. [46] Wu CL, Tella P, Staats PS, Vaslav R, Kazim DA, Wesselmann U, et al.
[39] Haslam C, Nurmikko T. Pharmacological treatment of neuropathic pain in older Analgesic effects of intravenous lidocaine and morphine on postam-
persons. Clin Interv Aging 2008;3(1):111. putation painA randomized double-blind, active placebo-controlled,
[40] Moseley G. Graded motor imagery is effective for long-standing crossover trial. J Am Soc Anesth 2002;96(4):841–8.
complex regional pain syndrome: a randomised controlled trial. Pain [47] Maier C, Dertwinkel R, Mansourian N, Hosbach I, Schwenkreis P,
2004;108(1–2):192–8. Senne I, et al. Efficacy of the NMDA-receptor antagonist memantine
[41] Moseley GL, Zalucki N, Birklein F, Marinus J, van Hilten JJ, et al. in patients with chronic phantom limb pain–results of a ran-
Thinking about movement hurts: the effect of motor imagery on pain domized double-blinded, placebo-controlled trial. Pain 2003;103(3):
and swelling in people with chronic arm pain. Arthritis Care Res (Hobo- ken) 277–83.
2008;59(5):623–31. [48] Robinson LR, Czerniecki JM, Ehde DM, Edwards WT, Judish DA,
[42] Davie-Smith F, Coulter E, Kennon B, Wyke S, Paul L. Factors influ- Goldberg ML, et al. Trial of amitriptyline for relief of pain in amputees:
encing quality of life following lower limb amputation for peripheral results of a randomized controlled study 1. Arch Phys Med Rehab
arterial occlusive disease: a systematic review of the literature. Prosthet 2004;85(1):1–6.
Orthot Int 2017;41(6):537–47. [49] Finnerup NB, Otto M, Jensen TS, Sindrup SH. An evidence-based
[43] Padovani MT, Martins MR, Venancio A, Forni JE. Anxiety, depression algorithm for the treatment of neuropathic pain. Medscape Gen Med
and quality of life in individuals with phantom limb pain. Acta Ortop 2007;9(2):36.
Bras 2015;23(2):107–10.

[44]

Available online at www.sciencedirect.com

ScienceDirect