Article Text

Download PDFPDF

Original research
Surgical denervation as a treatment strategy for pain in hand osteoarthritis: a systematic literature review
  1. Coen van der Meulen1,
  2. Lotte A van de Stadt2,
  3. Aniek Claassen3,
  4. Féline P B Kroon4,
  5. Marco J P F Ritt5,
  6. Frits R Rosendaal6,
  7. Sietse E S Terpstra1,
  8. Anne J H Vochteloo7 and
  9. Margreet Kloppenburg1
  1. 1Rheumatology, Leiden University Medical Center, Leiden, The Netherlands
  2. 2Department of Rheumatology, Amsterdam Rheumatology and Immunology Center, Amsterdam, The Netherlands
  3. 3Dutch Society for Rheumatology, Utrecht, The Netherlands
  4. 4Rheumatology, Zuyderland Medical Centre Heerlen, Heerlen, The Netherlands
  5. 5Department of Plastic, Reconstructive and Hand Surgery, Amsterdam University Medical Centres, Amsterdam, The Netherlands
  6. 6Clinical Epidemiology, Leiden University Medical Center, Leiden, The Netherlands
  7. 7Department of Orthopaedics, OCON Orthopedische kliniek, Hengelo, The Netherlands
  1. Correspondence to Coen van der Meulen; C.van_der_meulen{at}


Objective Surgical denervation has been proposed as a treatment for pain in hand osteoarthritis (OA). This review aimed to summarise the available evidence and to propose a research agenda.

Methods A systematic literature search was performed up to September 2022. Two investigators independently identified studies that reported on denervation for OA of the proximal interphalangeal, distal interphalangeal, metacarpophalangeal or carpometacarpal joints. Quality of studies was assessed and study characteristics, patient characteristics, details of the surgical technique and outcomes of the surgery were extracted.

Results Of 169 references, 17 articles reporting on 384 denervations in 351 patients were selected. Sixteen case series reported positive outcomes with respect to pain, function and patient satisfaction. One non-randomised clinical trial reported no difference in outcome when comparing denervation of the first carpometacarpal (CMC I) joint to trapeziectomy. Adverse events were frequent, with sensory abnormalities occurring the most, followed by the need for revision surgery. All studies had significant risk of bias.

Conclusion Surgical denervation for pain in hand OA shows some promise, but the available evidence does not allow any conclusions of efficacy and higher-quality research is needed. Techniques should be harmonised and more data regarding how denervation compares to current usual care, other denervation methods or placebo in terms of outcomes and adverse events are needed.

  • Osteoarthritis
  • Patient Reported Outcome Measures
  • Epidemiology

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information. All data relevant to the study are included in the article or uploaded as supplementary information. The used search strategies are shown in the supplementary file and can be used to reproduce the search.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

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.


  • Pain is a core symptom of hand osteoarthritis, which often cannot be fully alleviated with current treatment options. Surgical denervation has been proposed as a treatment option.


  • Surgical denervation may yield favourable outcomes, but evidence on surgical denervation is still insufficient to make a recommendation.


  • Future research should focus on (1) the innervation of the joints, (2) the best technique for surgical denervation, (3) high-quality randomized clinical trials investigating efficacy of surgical denervation versus sham or (4) comparing surgical denervation to alternative treatments and (5) the safety of surgical denervation.


Hand osteoarthritis (OA) is a common disease which causes loss of function, structural damage to the joints and, most importantly, pain.1 2 As there are currently no disease-modifying treatments, therapy is aimed at symptom control. Current guidelines of the European Alliance of Associations for Rheumatology and American College of Rheumatology (ACR) recommend non-pharmacological treatment first (eg, education, training, braces).3 4 Topical or systemic treatments with non-steroidal anti-inflammatory drugs (NSAIDs) are recommended for pain alleviation, with a preference for topical treatment due to the considerable toxicity of oral NSAIDs for elderly patients. For patients with structural abnormalities and inadequate pain control, surgery consisting of trapeziectomy with or without interposition and/or suspension arthroplasty (first carpometacarpal (CMCJ-I)), arthrodesis or arthroplasty (proximal interphalangeal joint (PIPJ) and distal interphalangeal joint (DIPJ)) is recommended as the last resort. The recommended therapies can also be combined.3 4

Alternatively, treatment can target the nerves, disturbing the transmission of pain signals through nerves innervating the joint. An example is radiofrequency ablation, which is conditionally recommended by the ACR guidelines for knee OA.3 Surgical denervation, that is, the surgical dissection of nerves, can theoretically achieve the same goal as ablation. It was originally described in the hip and later developed for the hand and wrist,5–7 and has been proposed as an alternative surgical intervention.5 8–10 Currently, surgical denervation is increasingly being performed to treat pain in hand OA, although precise numbers are unclear.

Surgical denervation aims to dissect the nerves innervating the painful joint, thereby disabling the pain signalling without impairing joint function. Surgical denervation does not seem to be associated with a loss of function or strength, and it does not preclude further surgeries if required, which is advantageous compared with other surgical options.9 10 However, some downsides have to be taken into consideration. For the denervation to be effective, anatomical knowledge of the joints of the hand and their innervation is essential. The nerves currently known to innervate the hands are summarised in online supplemental appendix 1.11–15 The innervation of the joints in the hands is still subject to debate and more nerve branches than the ones currently treated with surgical denervation may contribute to the innervation of specific joints.11 Furthermore, the occurrence of adverse effects has been reported, including wound infection, necrosis of skin flaps, or sensory abnormalities.9 10

It is currently unclear whether the advantages of surgical denervation outweigh the disadvantages. We performed a systemic literature search, aiming to summarise available evidence on the efficacy and safety of denervation as a treatment for hand OA compared with other treatments, and to set a research agenda.


Search strategy

A systematic literature search was conducted in PubMed, Ovid and Cochrane databases from their inception up to 21 September 2022, with additional references collected from the identified publications and other systematic literature reviews. The search strategies consisted of terms for “hand”, “osteoarthritis” and “denervation” and can be found in online supplemental appendix 2.

Eligible study types were randomised clinical trials, case–control studies, cohort studies, case reports and case series. Studies on OA in other joints (including wrist) or other causes of hand pain were excluded. Studies regarding interventions other than surgical denervation were similarly excluded. Reviews, comments and editorials, as well as abstracts without a full publication, were considered ineligible for review, but have been used to gather more suitable articles from the references.

Studies of surgical denervation (intervention) in adults with hand OA (population) were included. Hand OA comprised OA or degenerative arthritis of the DIPJ, PIPJ, CMCJ-I or metacarpophalangeal joint (MCPJ). The comparator could be any other treatment for hand OA, or none, in the case of case studies. No requirements were set for the outcome measures.

Study selection, data extraction and risk of bias assessment

Two reviewers (CvdM and AC) independently screened titles and abstracts to determine eligibility for inclusion. Screening results were compared and discussed in case of disagreement. Relevant data on study characteristics, interventions (denervated joint, nerves dissected, incisions used, postoperative care), study population (sample size, diagnostic criteria, demographics and baseline characteristics) and outcomes (pain and function scores after surgery, patient satisfaction, follow-up time, strength, adverse events) was extracted (CvdM) and summarised as average and/or range. The risk of bias was assessed (CvdM and SEST). For case series the Joanna Briggs Institute (JBI) checklist for case series was used, judging studies based on inclusion, diagnosis and classification of condition, reporting of demographic, clinical and outcome information, and statistical analysis.16 The ROBINS-I was used for comparative studies, judging on confounding, selection bias, classification and adherence to intervention, missing data, outcome measurements and selective reporting.17

Some of the JBI checklist items required further specification. The following definitions have been used. For the measurement of the severity of the condition to be standard and reliable, it was required to be identical for all patients and be reproducible. The use of any validated clinical or radiological system was deemed valid. For diagnosis, a validated radiological or clinical system was similarly required. Consecutive and complete inclusion of patients was judged on whether this was explicitly stated in the paper. Clear reporting of the demographics of the patients was defined as clear presentation of at least age and sex.

The risk of bias assessments were performed independently by the two reviewers, after which the outcomes were compared. In case of disagreement, the item was discussed to reach consensus. In case no consensus was reached, it was discussed with a third reviewer (MK).

The level of evidence of the individual studies was rated according to the Oxford Centre for Evidence-Based Medicine levels of evidence by CvdM.18

No review protocol was registered.


Searching PubMed, Ovid and Cochrane databases yielded 212 records. After deduplication, 169 publications remained. Screening of titles and abstracts resulted in 27 records. After full-text screening, another 13 were excluded (1 due to an unclear patient group, 12 due to type of publications). References from retrieved publications included three suitable records, leading to a total of seventeen records included (figure 1).

Figure 1

Flowchart of literature search.

Study characteristics

The included studies consisted of sixteen case series and one non-randomised comparative clinical trial. Studies are summarised in table 1. Most publications (n=11) described CMCJ-I denervation. MCPJ (n=1), DIPJ (n=1) and PIPJ (n=3) denervation were less frequent. One case series described a mixture of MCPJ, DIPJ and PIPJ denervation. A total of 384 denervations were performed in 351 patients. Twelve of these patients were described in two publications (Suresh et al19 and Tuffaha et al20). Furthermore, overlap between the patient groups in the two publications by Ehrl et al21 22 could not be excluded.

Table 1

Study characteristics

In seven of the case series, OA was diagnosed based on clinical presentation. In two a clinical diagnosis confirmed by radiography was used, in four only radiography was used. Among these four, the Eaton-Littler criteria were used in two, the Dell criteria in one and no criteria were specified in one. The method of diagnosing OA was not specified in four publications. Inclusion and/or exclusion criteria were specified in 12 publications and often included failure on conservative treatment, some description of radiological damage, response to nerve block or mobility of the joint. Previous joint surgery was used as an exclusion criterion in the majority of studies.

The non-randomised trial described a comparison of trapeziectomy to CMCJ-I denervation. Ten participants were included in the trapeziectomy arm, 35 were included in the denervation arm. Diagnosis was clinical. Inclusion criteria consisted of thumb base OA requiring surgery, with no prior surgery. For details, see table 1.

Methodological quality of included studies

Table 2 describes the methodological quality of the 16 case series. Most had clear inclusion criteria, but only in a few presence and severity of hand OA were measured in a reliable and standard way, with the assessment often not described or based on clinical judgement. Clinical assessment and radiographic staging, or a combination, were most frequently used. Radiographs were scored using either the Dell (n=1), Eaton-Littler (n=3) or unspecified (n=2) criteria. Seven reports stated inclusion had been complete, with five stating it had been consecutive. Demographic and clinical characteristics, as well as outcomes, were often presented clearly, although often few characteristics were given. None of the studies reported the demographic information of the study centre (geographic location, setting of the centre). Finally, in 11 of the case series, the statistical analysis was either absent (n=6) or inappropriate (n=5). All reviewed case series therefore had severe methodological shortcomings.

Table 2

Risk of bias assessment

The clinical trial by Salibi et al23 had serious risk of bias in multiple domains: it was an open label study, a substantial proportion of participants switched interventions from denervation to trapeziectomy and were subsequently removed from the analysis, and confounding was not taken into account in the analysis. For an overview, see online supplemental appendix 3 table A2.

Surgical techniques

The studies used a variety of techniques, summarised in table 3.

Table 3

Details surgical interventions

CMCJ-I denervation

Most studies used a single incision, either a Wagner approach (online supplemental figure A1), a transverse incision (online supplemental figure A2) or a radial S-shaped incision (online supplemental figure A3). Two groups (Donato et al24 and Loréa25) used two incisions: a palmar and a dorsal transverse incision (online supplemental figures A4 and A5). Giesen et al26 added a third incision in the fourth extensor compartment (online supplemental figure A6).

Which nerves were dissected differed between studies. Two studies did not specify the dissected nerves and one only stated that all branches leading to the joint capsule were dissected.23 27 The other nine studies all reported dissecting of branches from the radial nerve (extending from the radial sensory nerve or the superficial branch of the radial nerve), the medial nerve (from the thenar, palmar or palmar cutaneous branches) and the lateral antebrachial cutaneous nerve (from the Cruveilhier branch, or without further specification of smaller branches). Giesen et al26 additionally dissected the posterior interosseous nerve, Donato et al24 and Loréa25 a small part of the dorsal articular nerve of the first interosseous space.

Four groups added additional procedures during the surgery: Dellon27 injected the nerve endings with lidocaine and removed volar osteophytes, Ehrl et al21 22 performed synovectomy and excised osteophytes, followed by saline irrigation of the joint. Arenas-Prat28 performed periosteal resection of the first metacarpal and Suresh et al19 report anaesthesia with lidocaine of articular branches.

PIPJ denervation

Several incision techniques were used. A Brunner incision was used in all four studies investigating PIPJ denervation (online supplemental figure A7). Among the four, only Servasier et al29 also used other approaches: a double lateral (online supplemental figure A8) or straight incision on the dorsal aspect of the joint (online supplemental figure A9).

Again, the specific nerves that were dissected varied. Servasier et al29 stated they dissected all joint afferents, Madsen et al30 dissected the articular branches of the radial, ulnar and dorsal digital nerves, Jiménez et al31 dissected the palmar articular nerve, and the articular branches of the common and dorsal digital nerves, and Braga-Silva and Calcagnotto32 dissected the branches from the palmar digital nerves.

DIPJ denervation

A skin flap extending from the eponichium to the joint was used in both studies on DIPJ denervation (online supplemental figure A10). Both described dissecting the articular branches of the digital nerves, with Arenas-Prat33 specifying the volar and dorsal digital nerves.

MCPJ denervation

MCPJ denervation was done using two incisions in both studies describing it: a volar Brunner (online supplemental figure A11) or Chevron incision (online supplemental figure A12) and a dorsal linear incision (online supplemental figure A13). Both groups dissected articular branches of the digital nerves (dorsal and volar for Arenas-Prat,34 radial and ulnar by Madsen et al30), Arenas-Prat further dissected the branches from the deep branch of the ulnar nerve. Madsen et al dissected the branches from the digital branch of the radial sensory nerve and the dorsal sensory branch of the ulnar nerve.

Postoperative care

When described, most recovery plans were comparable: gentle return to activities as pain allows, with the main difference between studies being the time before return to activities (0–3 weeks).

Patient characteristics and outcomes

Baseline characteristics

The studies encompassed a patient group with average age from 55 to 65 years (range 30–87 years) (table 4). The study by Jiménez et al31 stood out, having a population with a lower median age (52 years). The percentage of female participants was around 60%–75% in most studies, with a very high percentage (98%) in the study by Servasier et al,29 and a very low percentage (36%) in the study by Jiménez et al.31

Table 4

Patient characteristics and outcomes

Most studies presented baseline pain scores, often on a 0–10 Numeric Rating Scale (NRS). Average baseline pain scores varied from 7.5 to 8.7 on NRS. Giesen et al26 split the baseline scores into pain at rest, light activity and demanding activity. They reported median pain scores of 5, 7.5 and 10, respectively. Madsen et al30 used a 5-point scale, reporting a median of 5/5 (table 4).

Of note, the study by Servasier et al29 included arthropathies of various types: degenerative (43 joints) and inflammatory (11 joints). Inflammatory was further divided into rheumatoid arthritis (seven joints), ankylosing spondylarthritis (one joint), psoriatic arthritis (two joints) and undetermined inflammatory rheumatism (one joint).


Fifteen out of seventeen studies reported the follow-up duration, which ranged from 4 to 152 months. In most studies, a physical follow-up was used, with the exception of the study by Ehrl et al,22 in which 23 of the 60 patients only had a follow-up over the telephone, and Giesen et al,26 where the final follow-up consisted of a questionnaire only, with a physical follow-up halfway through the follow-up period.


Fifteen of the seventeen groups reported postoperative pain outcomes, using a Visual Analogue Scale (VAS) or NRS, or using their own scales or descriptions. Almost all studies reported good results, with 56%–92% patients experiencing some measure of pain relief.19 20 24 25 34 35 The mean change in NRS score ranged from 3 to 8.1 on a 10-point scale.21 22 24 26 27 29–32 Salibi et al23 reported no differences in pain when comparing denervation to trapeziectomy from baseline to months 6, 12 and 60.

Fifteen out of seventeen studies reported other outcome measures. Six reported on patient satisfaction, which ranged from 70% to 92%.24 28 32–35 Three studies reported on range of motion of the joint after surgery; mean increase ranged from 3.5° to 27°.29 31 32 Three other studies reported on grip strength, with average increase around 3.9 kg20 26 or 12 foot/lb24 for grip strength and 2.1 kg for pinch grip strength.20 26 Finally, various questionnaires and other physical examinations were reported; most showing beneficial results.19 21 22 26 30 31 For details, see table 4.

Adverse events

Rates of adverse events varied between 0% and 75% (table 4).20–22 24–34 Most studies (n=9) reported rates of 20% or lower. The most frequently occurring complications were (temporary) sensory disturbances, such as pain, paraesthesia or numbness.20–22 24–26 28–34 Other reported complications included wound infection, skin necrosis and complex regional pain syndrome type I.21 22 24 29 34 Three studies (by Salibi et al, Suresh et al and by Servasier et al19 23 29) reported numbers of patients undergoing an additional type of surgery due to dissatisfaction. In the study by Salibi et al 9 out of 35 (26%) underwent a trapeziectomy, Servasier et al reported 7 out of 54 (13%) joints underwent either arthrodesis (n=2) of arthroplasty (n=5), Suresh et al reported 3 out of 9, which underwent arthroplasty.


Pain in hand OA remains difficult to treat, and new therapies are required. Surgical denervation has been proposed as an option. In this review, we gave an overview of the available literature describing the efficacy and safety of surgical denervation for OA in the PIPJ, DIPJ, MCPJ and CMCJ-I. No meta-analysis was performed due to the heterogeneity of the surgical techniques and the reported outcome measures.

The overall quality of the evidence was low. Most studies were case series (with the inherent shortcoming that there was no blinding or randomisation) and all studies had methodological shortcomings. Most common were lack of clarity regarding the inclusion of patients (consecutive or not, stringency of diagnosis, use of diagnostic criteria) and statistical analysis of the results. Despite the fact that most included studies were case series, these shortcomings still stand out and therefore the results of these studies may be biased. The varying use of diagnostic criteria hampers generalisability. Furthermore, properly performed randomised clinical trials were unavailable, precluding adequate comparison to other (non)pharmacological treatments or surgical methods, as well as comparison to usual care or sham. In particular, it would be valuable to have trials comparing surgical denervation to other interventions targeting nerves, such as for example, radiofrequency ablation. Ablation (using radiofrequency or cryoneurolysis) is conditionally recommended by the ACR guidelines for knee OA,3 based on two randomized clinical trials (RCTs) comparing it to sham36 37 and two RCTs comparing it to intra-articular injection with corticosteroids38 or platelet rich plasma and hyaluronic acid,39 as well as one comparing it to oral analgesics.40

The trial by Salibi et al23 started out as randomised, but diverted from this design due to slow inclusion, which arose from a strong patient preference for denervation. As such, their results are likely to be biased due to among others regression to the mean and placebo effects, as well as publication bias favouring positive outcomes. In total, the current state of the literature does not allow for definitive conclusions. However, the reported results of this intervention are generally positive. Most studies showed pain reduction, high patient satisfaction and retained or improved function. The pain reduction after denervation exceeded the minimal clinical important difference for NRS pain in most studies.41 This legitimises further evaluation in randomised clinical trials.

Conversely, adverse events were frequent, with only one study reporting no adverse events.27 Sensory abnormalities frequently occurred, as well as postoperative infections and the need for other surgical interventions. However, the frequency and severity of adverse events after surgical denervation can currently not be assessed with certainty, as the described studies lack sufficient quality. Nine cases from the studies developed complex regional pain syndrome. This severe adverse event may be worse than hand OA, and as such should be taken seriously. Of these patients with complex regional pain syndrome, two were described to be settled with hand therapy and analgesics (by the same authors, which may describe the same patient in two papers).21 22 Of the remaining seven, three diagnoses were doubtful and showed swift regression of symptoms, one resolved within 6 months and three resolved within 12 months.29 Another potential adverse event of denervation could be negative effects on the joint structure, given the concerns raised previously that removal of or interfering with pain signalling in the joint may exacerbate cartilage damage, both in clinical and basic science.42 43 Development of Charcot joint has also been described as a potential adverse outcome of surgical denervation, but was not seen in the studies covered in this review.9 10 The knowledge gap concerning adverse events needs to be addressed before denervation surgery can be recommended as a standard part of treatment for hand OA, as no sufficient risk-benefit analysis can be done without adequate information on adverse events. Specifically, comparisons of the adverse events after surgical denervation compared with other surgical interventions and other therapies targeting nerves are needed.

Another aspect of surgical denervation to consider is the possibility of a second surgery should the denervation fail, previously described as a strong potential benefit.9 10 44 45 Although it may technically be possible, it is unclear whether the outcomes of such an intervention are comparable to the outcomes of the same intervention without a preceding denervation. The reviewed studies offered insufficient evidence to answer this question.

Finally, techniques employed for denervation still vary greatly between surgeons. The studies included here differed in the incisions used, in the nerves targeted for denervation and in additional interventions performed. This makes direct comparison of the results difficult, and a consensus on at least the nerves to dissect and potential additional interventions to perform should be reached. For example, Giesen et al26 decided to make an extra incision to dissect the posterior interosseous nerve in addition to the nerves innervating the CMCJ-I. More uniformity in the surgical techniques may aid interpretation and evaluation of the effects of surgical denervation. The selection of the surgical techniques should be based on the innervation of the joint, and as such this innervation needs to be known.

There is an increasing understanding of the complexity of hand OA pain, which is thought to be nociceptive, but also nociplastic or neuropathic in nature, with central and peripheral sensitisation influencing it.46 Studies in this review did not assess the type of hand pain. So, it is currently unknown for which type of hand pain surgical denervation might be beneficial. This lack of results stratified by pain phenotype needs to be addressed in future studies.

In conclusion, we currently cannot be sure the benefits of surgical denervation outweigh the harms to treat patients with hand OA, given the small number of cases and overall low quality of the evidence. Thus, we do not recommend denervation surgery for pain relief in hand OA. However, the available results indicate the outcomes may be favourable, although a considerable number of complications were reported. To further evaluate the use of surgical denervation in hand OA, we propose future studies should investigate (1) the innervation of the joints, (2) the best surgical technique to dissect all relevant nerves, (3) perform high-quality randomised clinical trials to investigate the efficacy of surgical denervation in comparison to sham in different patient groups, (4) to investigate other (non-surgical) therapies targeting the nerves, and finally (5) the safety of surgical denervation.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information. All data relevant to the study are included in the article or uploaded as supplementary information. The used search strategies are shown in the supplementary file and can be used to reproduce the search.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.


We would like to thank the working group for the Dutch recommendations for the management of hand osteoarthritis.


Supplementary materials

  • 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.


  • Twitter @OAKloppenburg

  • Contributors MK, AC, MJPFR and AJHV designed the study. CvdM, AC, SEST and MK collected the data. CM, LAvdS and MK analysed the data. CvdM, LAvdS, AC, FPBK, MJPFR, FRR, SEST, AJHV and MK interpreted the data and wrote the report. All authors approved the final version of the manuscript. CvdM is the guarantor of this manuscript.

  • Funding For the current study, MK reports funding from the SKMS, paid to the institution.

  • Competing interests MK reports the following, all outside the current study: Grants from IMI-APPROACH and the Dutch Arthritis Society, paid to the institution. Royalties or licences from Wolters Kluwer and Springer Verlag, paid to the institution. Fees for consulting/advisory boards by Abbvie, Kiniksa, Galapagos, CHDR, Novartis, UCB, all paid to the institution. Payment or honoraria for lectures or presentations from Galapagos and Jansen, paid to the institution. Roles on the OARSI board (member), EULAR council (member advocacy committee EULAR) and presidency of the Dutch Society for Rheumatology. For the current study, MK reports funding from SKMS, paid to the institution.

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

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.