Article Text


Developing an acupuncture protocol for treating phantom limb pain: a Delphi consensus study
  1. Esmé G Trevelyan,
  2. Warren A Turner,
  3. Nicola Robinson
  1. Faculty of Health and Social Care, London South Bank University, London, UK
  1. Correspondence to Esmé Trevelyan, Faculty of Health and Social Care, London South Bank University, 103 Borough Road, London SE1 0AA, UK; trevelye{at}


Background Little is known about how a Traditional Chinese Medicine (TCM) approach could be used to treat phantom limb pain (PLP). There is currently no standard acupuncture protocol in the literature to treat this syndrome.

Objectives To achieve consensus among a group of acupuncture practitioners on the pathology and recommended treatment of PLP and devise an acupuncture protocol for the treatment of this condition.

Methods A classical Delphi approach was used using two parallel online Delphi studies. One study focused on participants with past experience of treating PLP (TPLP, n=7) and the other on practitioners with no past experience (NTPLP, n=16). Two hypothetical case studies were provided and participants were asked for responses on how they would treat these patients. Three rounds were included. Participants were also invited to rate and comment on the finalised protocol. Round 1 data were analysed using content analysis. In subsequent rounds an a priori criterion for defining consensus was set at ≤1.75 IQR. A group median of 5–6 was considered to mean ‘agree’.

Results 19 participants completed all Delphi rounds (12 NTPLP, 7 TPLP). 108 NTPLP and 76 TPLP statements were generated and circulated in round 2; 53% of the NTPLP statements and 62% of the TPLP statements met consensus in round 2 and 45% of the NTPLP statements and 44% of the TPLP statements met consensus in round 3. Participants all agreed with the final protocol developed.

Conclusions The protocol developed does not claim to be best practice but provides a preliminary consensus from practitioners practising acupuncture for the treatment of PLP.

Statistics from


Silas Weir Mitchell (1829–1914) coined the term ‘phantom limb syndrome’, but the phenomenon was recorded medically prior to this by Ambroise Paré (1510–1590).1 Currently, most literature tends to focus on two categories of amputation-related pain: phantom limb pain (PLP) and residual pain. Phantom sensations are non-painful sensations.2 Both PLP and phantom sensations may exist simultaneously.3

The incidence of PLP in the UK has been estimated at 0.1 per 10 000 person-years,4 and recent large surveys suggest 75% of amputees suffer PLP.5 However, reports on the prevalence of PLP vary considerably, possibly due to differences in methodologies, time points of assessment, definition and cut-off values used for diagnosis and the study population.6

PLP involves multiple changes along the neuroaxis.7 Peripheral mechanisms include formation of neuroma and ectopic discharge.8 Noxious stimuli due to nerve injury sensitise central structures and cause central sensitisation, wind-up, long-term potentiation and expansion of receptive fields of the central neurons.9 Central sensitisation is characterised by reduction of inhibitory processes, increased excitability of the dorsal horn neurons and structural changes at the central nerve endings.10 Amputation alters neuronal activity in cortical and subcortical structures,11 and changes and distortion occur in cortical maps.9 Reorganisation of the somatosensory cortex occurs surrounding the area representing the deafferenated limb.12 Cortical fields deprived of input shrink and the receptive field becomes smaller. Adjacent representations from non-denervated parts of the body then take over the cortical field.9

Evidence for the successful treatment of PLP is inconsistent. In the UK, common first-line treatments include amitriptyline, gabapentin and pregabalin.4 However, the short-term and long-term effectiveness of these treatments remains unclear, with gabapentin demonstrating a trend towards short-term analgesic efficacy and amitriptyline being ineffective.13

Acupuncture has been shown to be effective in the treatment of chronic pain conditions14 but has not been well documented for treating PLP.15 The style of acupuncture and point choice varies widely in the literature.16

Three main consensus approaches are recognised in health research: the Delphi method, nominal group technique and consensus development conference.17 The Delphi technique has the advantage of not requiring face-to-face contact, so facilitating input from a wider group of participants.18 It is well suited to research in which there is incomplete knowledge about a problem or phenomenon and works well when the goal of research is to improve understanding.19 ,20 The Delphi technique was used in this study to gain agreement on acupuncture principles and to develop a protocol for the treatment of PLP which could be used in future trials and clinical practice.


As a broad understanding of views on treatment was sought, participants both with and without previous experience in treating PLP were included in the study and two parallel surveys were run (past experience of treating PLP (TPLP) and no prior experience of treating PLP (NTPLP)). A classical Delphi approach using an open first round to facilitate the generation of ideas21 was taken and delivered online. The study was approved by London South Bank University Research Ethics Committee in September 2013, piloted in October 2013, commenced in November 2013 and completed in March 2014.


Professional acupuncture associations (British Acupuncture Council, Association of Traditional Chinese Medicine and Acupuncture Association of Chartered Physiotherapists) and universities teaching acupuncture were provided with information about the study and asked to forward information to members/past students. Any participant who contacted the researcher about the study was provided with information. An email sent back by the participant was taken as consent to participate. Convenience and snowball sampling was used to capture as many participants as possible with past experience of treating PLP. Inclusion criteria included: completion of recognised training in acupuncture, registered with a professional body, at least 3 years clinical experience, and the ability to communicate in English. With Delphi studies it is recommended that, for a homogeneous sample, a small sample size is appropriate, such as 10–1520 or 8–12 participants,21 and the study aimed to recruit approximately this number in each group.


The study was quasi-anonymous (the principal researcher (ET) knew which response came from which participant but participants did not). An a priori criterion of three rounds of Delphi was set. A fourth anonymous round was included involving rating and commenting on the finalised protocol. Participants were asked to respond to each round within 7–10 days and each round closed after 2 weeks. Subsequent rounds were sent out 2 weeks after closure of the previous round. All questionnaires were developed using the Bristol Online Survey (

Round 1 questionnaire consisted of two hypothetical case studies and 12 open-ended questions per case. Rounds 2 and 3 included statements generated from analysis of round 1. A ‘no comment’ option was provided if participants did not feel qualified to rank a statement. Statements were ranked on a 6-point Likert scale (from strongly disagree to strongly agree). Round 3 also provided additional information: response from the last round, median and IQR. To minimise participant fatigue, questions for which ≥50% of the ranked statements had met consensus in round 2 were not included in round 3 unless agreement had not been met.

Analysis of data

Qualitative data were analysed using NVIVO 10 and quantitative data using SPSS V.21. In round 1 qualitative content analysis was used, using thematic criteria.22 Raw data were condensed, and grouped and condensed statements which only appeared once were excluded from round 2. Two researchers (ET and NR) discussed and reviewed the grouped statements to ensure meaning was not lost or biased through researcher interpretation. A consistency check of coding was applied by testing intra-rater reliability 10–14 days after initial coding (ET). Inter-rater reliability was checked through a second coder (NR) who coded a portion of the data. In subsequent rounds an a priori criterion for defining consensus and agreement was set. A group median of 5–6 was considered to mean ‘agree’ and an IQR ≤1.75 was considered indicative of consensus.23 Stability of results was assessed using a Wilcoxon matched pairs signed rank test. The results were deemed stable when there was no significant difference between items.23 A p value ≤0.05 was used in this study. To maximise validity, SD was recorded to demonstrate convergence of results. Round 3 NTPLP and TPLP results were compared to allow for comparison and triangulation between groups.


Figure 1 shows participant flow throughout the Delphi process. A total of 37 practitioners initially expressed interest in participating; 23 completed round 1 (the TPLP group only managed to recruit seven participants), 19 completed both rounds 2 and 3 (12 acupuncturists, 6 physiotherapists practising acupuncture, 1 acupuncturist and physiotherapist) and 17 rated the final protocol. The reason for attrition by the 14 practitioners who did not complete round 1 was sought. A total of 12 responded with 7 reporting lack of time. Table 1 shows the demographics of the participants. All participants had completed acupuncture training in the UK; most practised a combination of Traditional Chinese Medicine (TCM) and 5 Element acupuncture. In the TPLP group, six practised TCM±other acupuncture styles and, in the NTPLP group, 12 practised TCM±other acupuncture styles. Those who did not practise TCM practised Western medical acupuncture (n=5). In the TPLP group, six were acupuncturists±physiotherapists and one was a physiotherapist. In the NTPLP group, eight were acupuncturists and eight were physiotherapists.

Table 1

Participant demographic data

Figure 1

Flow chart of participants through the study. TPLP, practitioners with past experience of treating phantom limb pain; NTPLP, practitioners with no past experience of treating phantom limb pain.

Figure 2 provides data on the number of statements included in each round and the number of statements meeting consensus throughout the rounds. In round 1, 619 NTPLP statements and 292 TPLP statements were generated which were condensed down to 204 and 135 statements, respectively. Of the condensed statements, 96 were omitted in the NTPLP group as they did not meet the threshold for inclusion in round 2 (ie, only appeared once) and, in the TPLP group, 59 were omitted for the same reason. Of the statements included in round 2, 53% of NTPLP statements and 62% of TPLP statements met consensus and, in round 3, 45% of NTPLP statements and 44% of TPLP statements met consensus. The ‘no comment’ option was selected <21% of the time in the NTPLP group and <10% in the TPLP group.

Figure 2

Number of statements generated and included in each round and number of statements meeting consensus in each round. CS1, case study 1; CS2, case study 2; NTPLP, practitioners with no past experience of treating phantom limb pain; TPLP, practitioners with past experience of treating phantom limb pain.

Table 2 provides data on the questions which were included in round 3:13 questions in the NTPLP group and nine in the TPLP group. After completion of round 3, statements had met consensus and agreement for 7/12 questions (both for case studies 1 and 2) in the NTPLP group and, in the TPLP group, statements had met consensus for 8/12 questions (case study 1) and 10/12 questions (case study 2). The Wilcoxon matched pairs signed rank test showed good stability of results with no significant difference between rounds in 93.02% of NTPLP statements and 97.22% of TPLP statements.

Table 2

Round 2 data on number of statements under each question which did and did not meet consensus and questions included in round 3

Details of the hypothetical case studies and statements which met consensus and agreement are provided in the online supplement.

Comparison of the NTPLP and TPLP statements showed that there were differences. The NTPLP group did not meet consensus on whether to treat using body acupuncture or body and auricular acupuncture and whether to treat the contralateral or both lower limbs. The TPLP group, however, achieved consensus on using auricular and body acupuncture on the contralateral limb. Additionally, the NTPLP group specified the acupuncture points: GV20, SP10, Four gates (a term for the bilateral point combination LR3 and LI4) and points on the lower back, taking a segmental approach to dermatomal pain (see final protocol, figure 3).

Figure 3

Acupuncture protocol for the treatment of phantom limb pain.

The study achieved consensus in nine of the 12 questions across case studies on the pathology and recommended treatment of PLP. By combining the four case studies and identifying similarities between cases, a protocol was developed (figure 3). Statements which appeared across case studies were included in the final protocol (see online supplement for the final list of statements which met consensus and agreement). The final protocol was rated using a 6-point Likert scale (from 1=strongly disagree to 6=strongly agree); 76.5% of participants agreed with the protocol, 11.8% strongly agreed and the remaining 11.8% somewhat agreed (median 5, IQR 5–5).

Six participants provided feedback on the protocol. Feedback included that it was considered interesting and surprising that consensus was not met in all areas and that the protocol allowed room for manoeuvre. One participant stated that he/she would not use GV20 and another felt needle retention time was excessive. One participant felt using the point pair LI4 and LR3 bilaterally (Four gates) was problematic, and it was proposed that Yintang could be used to replace LR3 on the amputated limb. One participant felt the protocol should be referred to as guidelines. One participant took the protocol to a team meeting and feedback was that a protocol was not needed as each patient would present and should be treated as an individual.


The aim of the study was to achieve consensus among a group of acupuncture practitioners on the pathology and recommended treatment of PLP and to devise an acupuncture protocol for the treatment of this condition. Consensus was established in 9 of the 12 areas under investigation and a protocol developed. It is acknowledged that, due to the background of some participants in the study, the protocol does include some statements which are specific to the constructs of TCM. However, practitioners from a Western paradigm would still be able to use the protocol.

Across case studies, participants achieved consensus on the pathology diagnosed. A common item was ‘Qi and blood stagnation’ (stagnation of energy and the circulation of energy in the body). Similarly, for both case studies there was consensus on the treatment principle to ‘move qi and blood’ and ‘resolve pain’. Historically, in TCM, physical trauma (such as amputation) is referred to as ‘causing the local stagnation of qi or blood’, which causes pain24 and this could be associated with peripheral pathology (such as the occurrence of neuroma and ectopic discharge) involved in PLP.

In this study the TPLP group reached consensus using a combination of body and auricular acupuncture. From a review of the literature, it is not clear whether auricular acupuncture, body acupuncture or a combination of these approaches is most efficient for treating PLP.16 A systematic review suggests that auriculotherapy may be effective for the treatment of pain,25 providing support for having auricular acupuncture within this protocol. Shenmen was the most commonly used point for the treatment of pain in a previous systematic review, which also supports its inclusion in this protocol.25

In keeping with this study, the acupuncture literature does tend to recommend treating the contralateral limb.15 ,16 Treating painful conditions through the contralateral side is common practice in TCM.26 Laboratory studies have compared the efficacy of ipsilateral and contralateral acupuncture and found that the degree of effectiveness is similar.27 ,28 Clinical studies have found that both ipsilateral local needling and contralateral distal needling can reduce carpal tunnel syndrome.29 Spinal interneurons may have an important role in reducing pain through contralateral needling.30

Specific body acupuncture points recommended in this study included LR3 and the Four gates (LR3 and LI4). These points are commonly used in case studies16 and, in clinical practice, the Four gates are commonly used due to their calming effect and ability to treat pain.24 The use of local acupuncture points may have met consensus because needling close to the injured tissue has the potential to elicit strong analgesic effects and encourage peripheral neuropeptide release (producing local vasodilation and modulating the local immune response).31 Distal points which share innervation with the injured tissue or close to the spinal level that shares innervation with the injured part may have met consensus because they influence the segment via the dorsal rami.31

Consensus was met on advice to obtain de qi and to retain needles for 20–30 min. This was in keeping with acupuncture trials on treating PLP15 and was not unexpected as many acupuncture textbooks recommend needle retention for this period of time.32 In clinical practice, when treating elderly or constitutionally weak patients and depending on the style of acupuncture, needle retention time may vary. These possibilities may not have been addressed while developing this protocol.

No consensus was met on needle manipulation, possibly because needle manipulation can vary even between different acupuncture points within the same treatment depending on the intent of the practitioner and the specific function of each acupuncture point. No consensus was met on whether to use electroacupuncture or some other form of adjunctive treatment, and there was minimal convergence between rounds 2 and 3 with these statements. This may have been due to participants’ familiarity with different treatment techniques and patient response to treatment. In the literature, adjunctive treatments are rarely used in conjunction with acupuncture to treat PLP.16 In this study, consensus on treatment frequency and the total number of treatments was similar to UK case studies published on acupuncture treatment for PLP.16

The study had limitations. Consensus is a contentious component of Delphi studies. IQR is generally accepted as an objective and rigorous way of determining consensus and an IQR of ≤2 on a 10 unit scale or ≤1 on a 4–5 unit scale is considered a suitable indicator of consensus.23 An IQR of ≤1.75 as used in this study may have been a slightly greater indicator of consensus but was chosen to avoid participant fatigue.

Also to avoid participant fatigue, not all statements were recirculated in round 3, meaning that the stability of the results could not be calculated for all items.

The protocol involved the opinion of both practitioners with (n=7) and without (n=16) past experience of treating PLP. Moreover, it is recognised that, although all practitioners were experienced, even those with past experience of treating PLP were not necessarily experts in this field. However, acupuncture is not commonly used for the treatment of PLP. In past surveys only 1% of amputees reported ever having used acupuncture.33

Professionals with both Western medical and traditional acupuncture backgrounds were included and therefore the protocol included a mixture of concepts. However, as the aim of this study was to capture a broad approach to treatment including expertise from different disciplines, this was not seen as a limitation to the study.

Potentially useful information was lost by introducing a threshold for inclusion of statements in round 2. However, due to the large number of statements generated in round 1, this process of reduction was deemed necessary.

This study elicited valid opinion from professional acupuncture practitioners. However, the Delphi technique does not produce right, wrong or definitive answers21 and, although consensus was obtained and a protocol developed, it should not be interpreted as the best treatment approach.

This study did achieve consensus and develop an acupuncture protocol for the management of phantom limb syndrome in lower limb amputees and practitioners’ views on the pathology and treatment of PLP were established. The protocol developed in this study is novel and provides guidelines which have not been previously available for acupuncture treatment of PLP. The proposed protocol is not claiming to be the best or the most correct approach to treatment, but does give some guidelines for practitioners on which to base their treatment. The protocol allows practitioners to take a pragmatic approach to treatment which is in keeping with TCM while still providing a structure to treatment.

Further study is required to gain consensus on areas such as needle manipulation, the use of electroacupuncture and other adjunctive treatments and the use of channel pairing when mirroring points on the contralateral limb. This protocol will form the basis of a future randomised controlled trial being carried out by the authors.

Summary points

  • Various acupuncture approaches are used to treat phantom limb pain and symptoms.

  • We used Delphi consensus methods with (finally) 17 practitioners to devise a treatment protocol.

  • The protocol is pragmatic but also consistent with a traditional approach.


The authors thank the clinical staff at the Amputee Rehabilitation Unit, Lambeth Community Care Centre, London and Gateway Acupuncture Clinic, Lambeth Community Care Centre, London for their input and advice when developing the survey, the participants who took part in this study and the associations who disseminated the information to their practitioners.


View Abstract
  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors All authors were involved in the design of the study and analysis and interpretation of data. All authors reviewed and commented on the manuscript and gave final approval of the version to be published.

  • Competing interests None.

  • Ethics approval Ethics approval was obtained from London South Bank University UREC 1327.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.