Study designs eligible for this scoping review included all original research studies following JBI scoping review methodology recommendations.27 This included interventional research, such as randomized and non-randomized trials, case series, case reports, and case-control studies. Rehabilitation protocols that describe the intervention were also included. Studies published in English or French after 1995 (date of the first report on the MT technique) were deemed eligible. Retrospective studies, conference abstracts, and editorials were excluded. To be included, studies had to recruit participants who experienced PLP following upper or lower limb amputations, regardless of the level of amputation. Studies recruiting people with other pathologies such as hemiplegia were excluded. Studies had to provide information on MT as an intervention for PLP, including number of sessions, frequency and duration of treatment, exercises performed, repetitions, patient positioning, or mirror specifications. Studies had to evaluate PLP at least once, with clearly defined time points and methods of assessment, without restrictions to particular pain assessment methods. Secondary research such as reviews and guidelines were included to screen their bibliography for potential missing primary research.

We searched seven databases, including CENTRAL, Clinicaltrial.gov, Cochrane Database of Systematic Reviews, Physiotherapy Evidence Database (PEDro), PubMed, ScienceDirect, and Scopus. The detailed search strategy is provided in Table 1. Following the recommendations from the PRISMA checklist for scoping reviews,28 the initial search was conducted in March 2022 across the seven databases with an update in March 2023. Searches in CENTRAL, Cochrane Database of Systematic Reviews, ScienceDirect, and Scopus covered title, abstract, and keywords while PubMed searches were limited to title and abstract.

Two reviewers (MG and KA) independently conducted study selection and resolved disagreements through review meetings. Results from searches were stored in a shared Zotero 5.0 folder. Duplicate articles were removed during bibliographic research. Initial screening involved reviewing titles and abstracts, followed by a second screening of full-text articles to verify eligibility criteria and extract relevant data. Reference lists of included studies and systematic or narrative reviews or guidelines were also checked for potentially relevant studies missed during database research.

Assessment of risk of bias was conducted or reported exclusively for RCTs using the PEDro scale.29–31 Two reviewers (MG and KA) independently evaluated studies when the score was not available, and disagreements were resolved through consensus.

Two reviewers (MG and KA) independently conducted data extraction. A standardized Excel sheet was used to ensure consistency. Regular video conferences were scheduled between reviewers to update and discuss data extraction progress. The extraction sheet is available upon request.

Data items included authors, publication year, country, study design, participant numbers, sex distribution, amputation type and level, etiology, time since amputation, study objectives, intervention details (session number, frequency, duration, length), follow-up period, exercise content, associated therapy, control group intervention, pain evaluation scales, main results, observed effects, effect sizes, side effects, and eligibility criteria. Additionally, items from the Template for Intervention Description and Replication (TIDier)32 were sought including: who provided the treatment, where the treatment was administered, whether tailoring was involved, and any modifications made to the intervention.

Findings were summarized narratively and presented in tables. Detailed data were organized into multiple tables based on extraction categories. Values are expressed either as numbers (n) or percentages (%).

The literature search was initially conducted in March 2022, with an update performed in March 2023 across seven databases. Search synthesis is illustrated in Fig. 1.33 Of the total 1152 records identified, 44 articles were considered eligible and included in this review.

Fig. 1.

PRISMA 2009 flow diagram. PubMed: 128; Science direct: 562; Pedro: 55; Cochrane Library: 301; Clinical Trial: 20; Scopus: 829.

(0.43MB).

The included studies span from 2004 to 2023. Of the 44 included studies, 34 % of the studies (n = 15)15,34–47 were published between 2004 and 2013 and 66 % (n = 29)19–24,48–70 between 2014 and 2023. Included studies were conducted in 19 countries. Fourteen studies36,38,41,43,47–50,58,58–60,62,65,67 are from the United States of America. Four studies are from India24,53–55 and Japan39,42,44,63; 3 studies from the United Kingdom,34,35,56 Pakistan21,23,57 and Spain64,66,68; 2 studies from Iran69,70 and Turkey19,61; and 1 study from Austria,37 Cambodia/Norway,51 Ireland,46 Israel,22 Italy,15 South Korea,45 the Netherlands,20 South Africa,52 and Sweden.40

Thirty-six percent of the studies (n = 16)19–24,34–36,49–55 are randomized controlled trials (RCTs), 4 % (n = 2) are non-randomized control studies,60,62 25 % (n = 11)15,39–47,56,63–68 are case series, 25 % (n = 11) are case reports,15,43–47,64–68 and one is a case-controlled study.58 Seven percent (n = 3)48,69,70 are protocols of RCTs. This diversity of study designs reflects the heterogeneous nature of the evidence base for this treatment modality.

Among participants, 68.5 % were male (n = 684), 21.1 % females (n = 212), and sex was not reported for 10.3 % of the participants (n = 103). The most common etiology of the amputation, for 46.6 % (n = 472) of patients, was trauma, followed by peripheral artery disease for 22 % (n = 223), medical causes for 4.8 % (n = 49), and cancer for 3.7 % (n = 37). Less frequent 2.1 % of causes included infection (n = 21) and congenital condition (n = 1). Etiology information was missing for 20.7 % of patients (n = 209).

Among the included studies, 87 % were lower limb amputations (n = 831) and 12 % upper limb amputations (n = 113). One study did not specify the amputated limb (n = 13).59 Lower limb amputations included 58 % of amputations at the tibial level (n = 484), 30 % at the femoral level (n = 252), 2 % at the knee level (n = 16), 1 % at the hip or ankle level (n = 9), and the level was not reported in 8.4 % (n = 70). Upper limb amputations included 32 % at the humeral level (n = 36), 35 % at the radial level (n = 40), and 3.5 % at other levels (n = 4), such as wrist and shoulder disarticulation. The level of amputation was not reported for upper limb for 29 % of patients (n = 33).

Various tools were used to assess PLP, with seven different ones reported. In some studies, multiple tools were used. The Visual Analog Scale (VAS) was the most frequently used, present in 61 % of the studies15,19,21,34–37,40–45,47–51,53–56,58,59,68–70 (n = 27), followed by the Numeric Rating Scale (NRS) in 20 % of the studies20,23,24,38,39,57,61,66,68 (n = 9). Additionally, the Brief Pain Inventory (BPI) was mentioned in 9 % of the studies22,52,60,62 (n = 4) and the McGill Pain Questionnaire (MPQ) in 7 % of the studies22,34,55 (n = 3). One study reported each of the following tools: the Universal Pain Score (UPS),54 the Multidimensional Pain Inventory,59 and a 5-point Likert scale.63 Nine percent of the studies46,63,65,67 (n =4) did not report the tool used to assess PLP.

Detailed information on MT sessions, including exercises, durations, series, repetitions, and modalities, was compiled when available (Table 2).

The most common intervention duration was four weeks, used in 34.1 % of the studies (n = 15).19,21,23,49–51,54,55,58–62,66,70 Thirty-four percent of the studies (n = 15) reported the duration of the intervention being less than 4 weeks, ranging from 1 to 21 days.22,24,34,35,40–42,46,47,52,54,56,57,63,63 Twenty-five percent of the studies (n = 11) reported a duration of more than 4 weeks, ranging from 6 to 24 weeks.15,20,36,38,39,43,44,64,67–69 Intervention length was not reported in 7 % of the studies (n =3).45,53,65

Frequency of intervention was described in 84 % of the studies15,19,21–24,35,37–52,54,55,57–60,62–65,67–70 (n = 37). Reported frequencies range from 12 sessions per day52 to 1 session per week.44,67 The most common frequency was 1 MT session per day, reported in 41 % (n = 18) of studies.19,21–23,38,39,48–50,54,57–60,62,64,68,70 Two sessions per day was reported in 14 % of the studies (n = 6).15,24,37,47,51,65 Seven sessions per week was used in 7 % of the studies (n = 3),41,55,69 and 1 study reported using four sessions per week.45 Frequency was not reported in 16 % of the studies (n=7).20,34,36,53,56,61,66

A change in the frequency of sessions was used in 9 % of the studies (n = 4).19,43,45,46 For example, Darnall et al.43 used a protocol with five sessions per week for the first month, followed by three sessions per week for the next two months. MacLachlan's protocol46 was even more complex. In the first week, patients underwent two assisted sessions per day with a therapist for five days, followed by two independent sessions per day during the weekend. In the second week, patients completed two assisted sessions per day with a therapist, along with two to three independent sessions per day, and three to four daily independent MT sessions during the weekend. In the third week, a patient completed two to three daily MT sessions on their own.

Approximately 45 % of the included studies (n = 20)19,21,23,36,40,45,47–50,55,57–60,62,63,68–70 reported MT sessions lasting 15 min. Another 25 % of the studies (n = 11)15,20,22,24,37,38,41,43,53,54,66 reported session duration between 20 and 30 min, and 4 studies39,56,64,65 used 10 min sessions. Modifications to the duration of sessions within a study were reported in 2 studies.43,52 For example, Darnall et al.43 suggested a session duration of 20 to 30 min in the first month and 30 min for the next two months. Limakatso et al.52 used a decreasing approach, starting with 30 min sessions at the hospital followed by 10 min sessions at home. Other reported session durations include five minutes,51 and 40 min.61 Duration was either not reported or unclear in 16 % of the studies (n = 7).34,35,42,44,46,67

The number of exercises varied from 1 to 10 with the most common being 3 exercises in 18 % of the studies (n = 8).47,48,50,58,63,64,67,70 Half of the studies included 3, 5, or 6 different exercises (Table 2). For detailed exercise content, see the exercise detail paragraph and Table 2. Notably, 41 % of the studies (n = 18)20,21,35,36,38,39,41,44,45,52,54,56,57,60–62,65,68 did not report the number of exercises in their protocol.

Only 18 % of the studies (n = 8) detailed the time allocated per exercise, ranging from 30 s to 5 min.37,40,42,48,51,58,59,70 For instance, Chan et al.58 outlined three exercises, each lasting five minutes for the lower limb, while Foell et al.59 described five exercises, each lasting three minutes. However, time duration per exercise was not described in 86.7 % of the studies (n = 36).15,19–24,34–36,38,39,41,43–47,49,50,52–57,60–69 For example, Wilcher et al.47 reported that the duration of the MT session should be around 15 min and should include movements such as biceps curls, opening and closing the fist, pronating and supinating the outstretched ‘arms’,47 without further detail. Because execution duration varies between participants, specificity may be lacking. Clinicians may prefer detailing repetitions instead.

Patients often inquire about the repetition of movements. Because describing the duration of each exercise can be challenging, documenting the number of repetitions could help in standardizing protocols. Unfortunately, this detail is inconsistently documented, with the number of repetitions described in only 16 % of the studies (n = 7).34,35,46,62,64,67 For instance, both studies by Houston et al.60,62 outlined 15 repetitions per movement. The study by Thomas67 reported the highest number of repetitions, with the patient performing two series of 40 repetitions for wrist and elbow movements during the first five weeks of treatment. Afterward, the patient added movements within the frontal plane (ulnar and radial deviation) with the same repetition count. The remaining four studies23,34,46,64 report using 10 repetitions per movement.

Follow-up was described in 43 % of the studies (n = 20).15,19,22,24,35,38,39,43,44,50–53,55,58–60,62,64,69 The time period varied from 2 weeks59 up to 2 years,64 with 6 studies15,19,24,35,44,52 reporting a follow-up period of six months. However, 54 % of the studies (n = 24)20,21,23,34,36,37,40–42,45–49,54,56,57,61,63,65–68,70 did not report this information (Table 2).

MT was administered either at the Hospital (41 % of the studies, n = 18),21,22,24,34,36,37,39,41,44,46,49,54–57,63,64 or home-based (18 %, n = 8).19,38,43,51,59,66,69,70 In 16 % of the studies (n = 7),20,45,50,52,53,61,65 the setting changed starting at the hospital and continuing at home. However, the setting was not specified in 25 % of the studies (n = 11).15,23,35,42,47,48,58,60,62,67,68

In terms of practice, 32 % of the studies (n = 14)20,21,34,36,37,40,42,44,49,55,57,63,64 indicated that MT was administered according to the therapist's guidance. In 23 % (n = 10),19,38,39,43,51,56,59,66,69,70 patients conducted MT independently. Notably, in 11 % (n = 5),41,46,50,52,61 the approach shifted from therapist-supervised sessions to independent patient-led sessions.41,46,50,52,66 However, 32 % of the studies (n = 14)15,23,24,35,45,47,48,53,54,58,60,62,65,68 did not provide this information.

Seven different comparators are reported in 48 % of the studies (n = 21).19–24,34–36,40,42,48–55,58,69 The most common were (i) covered/obscured mirror34,36,40,49,50,55 and (ii) mental visualization,19,36,49,53,69 followed by the standard physical therapy treatment.23,35,52,70 Other comparators like tactile discrimination of the residual limb,51 sensorimotor exercise of the intact limb,20 transcutaneous electrical nerve stimulation,54 and transcranial direct current stimulation (TDCS)22 were reported once (Table 2).

Table 3 summarizes session content, including motor and sensory exercises. Our scoping review found that exercises were not reported in 31 % of the studies (n = 14).20,21,35,36,38,40,41,44,45,53,54,61,65,68 All included studies described 15 lower limb motor exercises, encompassing mobility exercises for the hip, knee, foot, and toes. Foot flexion and extension were the most common exercises reported in 27 % of the studies (n = 12),19,22–24,34,37,43,46,48,51,58,64 foot rotation and inversion/eversion were reported in 16 % (n = 7), and knee flexion and extension were reported in six studies23,24,34,38,46,55 (i.e., referred to as straighten and bend in some publications). Additionally, 14 % of the studies (n = 6) instructed patients to move their limb without specific instructions.15,21,39,53,60,62

For the upper limb, 14 movements were detailed, involving thumb to elbow mobility. No shoulder movements were described. The most common exercises were pronation/supination (11 % of the studies, n = 5)47,49,55,59,66 followed by elbow flexion/extension (9 %, n = 4),47,49,55,67 and opening/closing of the hand.66 Some exercises focused on functional tasks, such as “Touch the fingertips with the thumb; tracing figures with fingers” described by Foell et al.,59 or moving the hand toward and away from the mirror as used by Imaizumi et al.63

Sensory exercises were less prevalent, featured in 9 % of the studies (n = 4),15,19,43,50 all focused on the lower limb. Six different sensory exercises were identified, with four originating from Clerici et al.’s15 case report: looking at the limb, touching it, caressing it, and scratching it. Additionally, Darnall et al.62 described an exercise involving touching the big toe, and Anaforoğlu et al.19 detailed muscle relaxation following foot and ankle muscle contraction.

Methodological assessment of the 16 included RCTs19–24,34–36,49–55 using the PEDro scale is reported in Table 4. The mean PEDro score was 5.5/10, with a standard deviation of 1.75. Only three studies34,42,50 received a score of six which corresponds to a high-quality study, while 9 studies19,22,24,34,35,49,52,54,55 obtained a score of 4 or 5 corresponding to a moderate quality study, and 4 studies23,36,51,53 were of poor quality with a score below 4.

While MT shows promise in non-invasive therapy for cortical maladaptation,59,73 alternative approaches should be considered to tailor treatment to individual patient needs. The true effect size of MT may be overestimated, and limitations could arise due to its restricted range of motion. Other non-invasive methods like sensory discrimination,72 TENS,74 hypnosis,75 vibration therapy,76 and acupuncture77 show promise, but evidence supporting their efficacy is limited. Ongoing research explores novel approaches like virtual reality.78–83 Integrating these techniques rather than viewing them as competing modalities could benefit patients. MT could serve as an initial step followed by VR-MT, leveraging virtual reality's advantages in diverse exercises and enhancing engagement. VR could be particularly useful for patients facing challenges using a real mirror due to unresolved feelings about their amputation. It enables the modeling of a simplified arm, whose movements and appearance are enough to create an illusion and alleviate PLP. However, research in this field remains limited to case studies and case series,78–84 necessitating further investigation with higher levels of evidence.

Invasive approaches for alleviating PLP are also used. Techniques like deep brain stimulation,85 targeted muscle reinnervation (TMR),86 targeted sensory reinnervation (TSR),87,88 Regenerative Peripheral Nerve Interface (RPNI),89–91 and Agonist-antagonist Myoneural Interface (AMI)89,92 show promising results. Combining invasive and non-invasive methods may offer favorable outcomes, allowing selection from various therapeutic approaches to accommodate individual patient preferences.