Acupuncture for the Relief of Chronic Pain: A Synthesis of Systematic Reviews

A search of electronic databases (MEDLINE, the Database of Abstracts of Reviews of Effects (DARE) and the Cochrane Library) was conducted in April 2019 and updated in July 2019 using free text search terms ‘acupuncture’, ‘chronic pain’, ‘analgesia’, ‘pain management’, ‘systematic review’ and/or ‘meta-analysis’. The search was restricted to English language databases. Systematic reviews and meta-analyses were screened for eligibility. Search results were screened by the authors, CAP and MIJ. All SRs with or without meta-analyses of studies using manual acupuncture, electro-acupuncture, dry needling or auriculotherapy (ear acupuncture) for any chronic pain condition were included. Reviews were included where acupuncture was compared with sham or placebo acupuncture, no treatment, or another intervention (pharmacological and non-pharmacological). We included Cochrane and non-Cochrane reviews and overviews of SRs. Systematic reviews containing non-RCT studies were included in order that information from RCTs could be extracted. Reviews were excluded if they did not evaluate invasive acupuncture (e.g., reviews on acupressure or laser acupuncture). Systematic reviews were excluded if they evaluated acute pain but not chronic pain (e.g., specifically focusing on postoperative pain or pain in the emergency setting). Reviews focusing on additional elements such as bee venom were also excluded. Non-English reviews were included if they contained an English abstract. However, non-English reviews were not translated. One review author (CAP) extracted information from reviews including type of pain, number of RCTs, treatments, conclusion and quality of evidence stated by the authors of each included review taken as a direct quote from the Conclusion, Abstract or Discussion sections of their manuscript. In addition, we ascribed a judgement of efficacy of each review according to whether the sample size met criteria based on the work of Moore et al. [ 16 , 17 ] and adopted by the Pain, Palliative and Supportive Care group from Cochrane Collaboration in their risk of bias assessment. They suggest that trial arms with fewer than 200 participants in RCTs or fewer than 500 participants in meta-analyses are at a high risk of bias, which seriously undermines confidence in findings. Thus, reviews were categorized as meeting our criteria for adequacy if they contained a pooled analysis of 500 events or at least one RCT with >200 participants in each arm of the trial. We categorised efficacy as: Sufficient evidence and in favour of acupuncture (+), sufficient evidence in favour of control/placebo (−), sufficient evidence but conflicting/inconclusive (=) and insufficient evidence to make a judgement (?). We also noted statements within manuscripts about RCT methodology across the following themes: The nature of placebo/sham interventions. Quality and risk of bias (including blinding). Sample size in relation to treatment effect. We used criteria developed by Dechartres [ 18 ] when commenting on adequacy of sample size as: adequately powered (≥200 patients per treatment arm), moderately powered (100–199 patients per treatment arm) and underpowered <100 patients per treatment arm). The nature of placebo/sham interventions. Quality and risk of bias (including blinding). Sample size in relation to treatment effect. We used criteria developed by Dechartres [ 18 ] when commenting on adequacy of sample size as: adequately powered (≥200 patients per treatment arm), moderately powered (100–199 patients per treatment arm) and underpowered <100 patients per treatment arm).

The World Health Organisation (WHO) recognises chronic pain as a long-term condition in its own right and as a secondary consequence of other long-term conditions [ 1 ]. It has been estimated that 28 million adults in the UK (43%) are affected by chronic pain and that the pain of 7.9 million of these adults is moderately or severely limiting [ 2 ]. The prevalence of chronic pain is higher in older age groups, with an estimated 62% of people over 75 being affected [ 2 ]. Individuals living with pain often experience a very poor quality of life, it affects their ability to work, socialise, sleep and maintain good relationships and can lead to depressive illness, decreased motivation and a reduction in physical activity [ 3 ]. As such, chronic pain represents a major challenge for health service provision and government policy. Current guidance from the International Association for the Study of Pain (IASP) recommends a biopsychosocial approach to pain utilising a multidisciplinary, multimodal, stepwise approach which combines pharmacological interventions with behavioural and non-pharmacological treatments [ 4 ]. Non-pharmacological interventions are recommended as part of a comprehensive pain management programme, including lifestyle adjustments, pain education, and physical, psychological and complementary therapies. In the UK, acupuncture has been available in some parts of the National Health Service (NHS) for decades as a non-pharmacological intervention to manage acute or chronic pain. In the NHS, acupuncture is administered by Allied Health Professionals, Nurses or Doctors. Outside the NHS, acupuncture is available from a variety of sources, including ‘traditional’ acupuncturists, sports therapists, osteopaths and chiropractors. Acupuncture is an age-old technique which became part of modern medicine in the 1970s. In modern medicine, traditional forms of acupuncture, based on the ancient Chinese concept of qi and meridians, have been superseded by acupuncture based on a neurophysiological model [ 5 , 6 ]. The unique identity of acupuncture lies in the process of inserting needles (‘acu’) in the skin (‘puncture’), although a modern definition should include the need to do this at specific points in accordance with known physiological or anatomical rationale [ 7 ]. Over the past two decades, the quantity of clinical studies on the use of acupuncture for various types of pain has significantly increased. In 2013, it was estimated that over 3000 clinical trials had been published [ 8 ] with over one hundred systematic reviews (SRs) (some with meta-analyses) attempting to synthesise available evidence. Many SRs of randomised controlled trials (RCTs) of acupuncture have been inconclusive and this has created uncertainty in clinical policy and practice. This uncertainty was highlighted in 2016 when the National Institute for Health and Care Excellence (NICE) reversed its 2009 recommendation to offer acupuncture as a first line treatment for non-specific, chronic low back pain because evidence indicated that it was no more effective than sham acupuncture [ 9 , 10 , 11 ]. Interestingly, there had been no significant change in evidence provided by RCTs between 2009 and 2016. Presently, NICE only recommends acupuncture as a prophylactic treatment for chronic tension-type headache and migraine [ 12 , 13 ]. In the face of conflicting evidence and continually changing guidance, it is unsurprising that acupuncture practitioners are finding that an intervention that, anecdotally at least, is often well received by patients in the clinic and appears to have good results, is rejected by commissioners and policy makers and regarded in some quarters as a ‘theatrical placebo’ [ 8 , 14 ]. One reason for this uncertainty may be related to the clinical research methodologies used to determine clinical efficacy. Policy makers give credence to the findings of RCTs because they are the ‘gold standard’ methodology for evaluating clinical efficacy. RCTs enable isolation of the effects (benefit and harm) associated with the active ingredient of a treatment from effects associated with the act of receiving a treatment, i.e., believing that an active ingredient of a treatment has been received. This is operationalised by using needles to puncture the skin at defined points compared with pretending to puncture the skin at defined points (i.e., a ‘placebo’ or ‘sham’ intervention). Systematic reviews and meta-analyses of multiple RCTs provide an indicator of consistency of findings between RCTs and allow for generalisability of findings [ 15 ]. Practitioners and policy makers may feel overwhelmed by the volume of SRs on acupuncture, suggesting a need to bring all this evidence together. In doing so, there is an opportunity to appraise RCT design and whether it is fit for purpose. The aim of this review is to synthesise evidence from previously published SRs of RCTs evaluating the clinical efficacy of acupuncture to alleviate chronic pain from any source. We have made judgements from a Western medical perspective. Our approach is to outline research findings through commentary rather than a comprehensive objective appraisal of SRs. We appreciate that the non-systematic approach is vulnerable to selection and evaluation biases and opinion-orientated arguments. Nevertheless, our approach enables consideration of issues surrounding the quality and adequacy of the evidence, including RCT design, and provides practitioners and policy makers with a comprehensive source of SRs published to date.

A total of 177 reviews of acupuncture for pain relief published between 1989 to September 2019 were included ( Table 1 ). There were two overviews of Cochrane reviews, ten overviews of non-Cochrane SRs and 145 non-Cochrane SRs. The earliest systematic reviews were published in 1989 by ter Riet [ 19 , 20 , 21 ]. There were 20 Cochrane SRs (including updates), with the earliest published in 2000 by Tulder et al. [ 22 ] and the most recent published in 2018 by Choi et al. [ 23 ]. Findings are presented according to the most frequent evaluations of acupuncture for different types of pain and described chronologically to provide a sense of the evolution of evidence over time. A statement of current clinical guidance from NICE is provided where available. The earliest SR that evaluated the efficacy of acupuncture across chronic pain conditions irrespective of aetiology or pathophysiology was published in 2000 and was inconclusive, although it was claimed that six or more sessions of acupuncture were more likely to be associated with positive outcomes [ 24 ]. The first overview of SRs was published in 2006 and concluded that acupuncture was not shown to be efficacious for a variety of pain conditions [ 25 ]. We found four other overviews of SRs of acupuncture for chronic pain irrespective of aetiology or pathophysiology. In 2010, Ernst and Lee published an overview of 30 SRs of acupuncture (319 RCTs) for ‘rheumatic conditions’ and judged there to be some evidence to support efficacy in routine care of patients with pain associated with osteoarthritis, low back pain and lateral elbow pain [ 26 ]. Hopton et al. pooled data from eight meta-analyses of acupuncture for chronic pain and concluded that acupuncture was more effective than a placebo, despite an absence of statistical significance for individual conditions, except osteoarthritis of the knee and headache [ 27 ]. Two overviews published in 2011 concluded that there was tentative evidence that acupuncture might be effective for headache, peripheral joint osteoarthritis and neck pain (overview of eight Cochrane Reviews [ 28 ], overview of 57 SRs [ 29 ]), although reviewers agreed that the quality of the primary studies was poor, with a high risk of bias. We found 20 SRs of acupuncture for chronic pain irrespective of aetiology or pathophysiology. In 2014, SRs reported that evidence supported the efficacy of wrist-ankle acupuncture and auricular acupuncture for alleviating chronic pain [ 30 , 31 ]. Since then, SRs were generally inconclusive because of methodological shortcomings and small sample sizes in primary studies [ 32 , 33 , 34 , 35 , 36 ]. In 2018, Vickers et al. concluded that evidence supported the efficacy of acupuncture for various chronic pain conditions associated with musculoskeletal disorders, headache and osteoarthritis, with beneficial effects persisting at long-term follow-up (39 RCTs, [ 37 ]). The long-term effects of acupuncture were consistent with evidence from an earlier SR by MacPherson et al. [ 38 ]. Evidence from SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for chronic pain associated with various medical conditions. There is no specific NICE guidance about the use of acupuncture for chronic pain conditions irrespective of aetiology or pathophysiology, although some guidance exists for specific pain conditions (see respective sections below). Guidance by NICE on chronic pain assessment and management is currently being developed (GID-NG10069) with publication expected in August 2020. We found one overview of Cochrane reviews of acupuncture for various pain conditions [ 28 ] (described above) that claimed there to be evidence that acupuncture was effective for tension-type headache (1 Cochrane review [ 45 ]) and migraine (1 Cochrane review [ 51 ]). The earliest SR was published in 1999 and judged there to be too few RCTs of sufficient methodological quality to determine efficacy of acupuncture for recurrent headache (22 randomised or ‘quasi’ randomised trials, Melchart [ 54 ]) or tension-type and cervicogenic headache (8 RCTs [ 47 ]). A similar pattern of ‘promising’ but not definitive evidence continued through the next decade (27 RCTs [ 53 ]; 8 RCTs [ 46 ]), including a Cochrane review of 26 RCTs of acupuncture for idiopathic headache [ 52 ]. Nevertheless, some reviewers have claimed that there is evidence that acupuncture is superior to sham for chronic headache (31 RCTs, only 2 RCTs were of high quality and adequately powered, Sun [ 48 ]), and a recent Cochrane review providing evidence of superiority of acupuncture over placebo for the prevention of tension-type headache ([ 44 ] 12 RCTs, including two adequately powered RCTs) and episodic migraine ([ 49 ], 22 RCTs, including two adequately powered RCTs). A systematic review published in 2016 is consistent with the latter finding that acupuncture was superior to sham acupuncture for migraine (10 RCTs [ 50 ]). Evidence from the SRs suggests that acupuncture prevents episodic or chronic tension-type headaches and episodic migraine, although long-term studies and studies comparing acupuncture with other treatment options are still required. The current NICE guidance (clinical guideline CG150) is that a course of up to 10 sessions of acupuncture over 5–8 weeks is recommended for tension-type headache and migraine [ 12 ]. The overview of eight Cochrane reviews of acupuncture for various pain conditions described previously [ 28 ], judged there to be evidence that acupuncture produced short-term improvements in pain based on a SR of 16 RCTs on peripheral joint osteoarthritis [ 69 ]). In 2019, an overview of non-Cochrane SRs that included a meta-analysis concluded that acupuncture was beneficial for alleviating pain associated with OAK, although RCT outcomes assessed using the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) indicated that the SR evidence was of mixed quality (12 SRs [ 55 ]). We found that the earliest SR on acupuncture for OA was published in 1997 and found studies to be contradictory with no evidence that acupuncture was more effective than sham (9 RCTs [ 71 ]). SRs on acupuncture for peripheral joint OA were inconclusive in 2006 (18 RCTs [ 70 ]), superior to waiting list controls in 2010 ([ 69 ], Cochrane review of 16 RCTs) and associated with reductions in pain intensity, improvement in functional mobility and quality of life in 2014 (12 RCTs [ 68 ]). In 2018, a Cochrane review by Manheimer et al. found little evidence that acupuncture significantly reduced pain associated with OA of the hip [ 67 ]. To date, the majority of SRs have evaluated the clinical efficacy of acupuncture for OA of the knee (OAK). We found that the earliest SR on acupuncture for OAK was published in 2001 and was inconclusive (7 RCTs, with 4 of low methodological quality [ 66 ]). Chronologically, SRs in 2002 (4 RCTs, [ 65 ]) and 2007 (11 RCTs [ 64 ]) were inconclusive, whereas a SR by Bjordal et al. in 2007 found statistically significant and clinically relevant short-term pain relief from two to four weeks of intensive electroacupuncture (7 RCTs [ 63 ]). In 2008 and 2012, two further SRs (10 RCTs [ 62 ]; 14 RCTs, [ 61 ] respectively) reported superiority over sham acupuncture. More recently, SRs in 2016 (10 RCTs [ 59 ]), 2017 (17 RCTs [ 72 ], 11 RCTs [ 58 ]), 2018 (16 RCTs [ 57 ]) and 2019 (8 RCTs [ 56 ]) judged there to be evidence that acupuncture provides relief of pain associated with OAK when administered alone or in combination with other treatments. These positive findings are supported by a network meta-analysis published in 2013 that evaluated 22 treatments, including acupuncture (11 RCTs [ 60 ]), and judged there to be evidence of short-term efficacy. This finding was confirmed in another network meta-analysis published in 2018, which found that needle or electro-acupuncture decreased pain compared with other treatments (16 RCTs [ 57 ]). Nevertheless, reviewers consistently mitigate these positive findings by describing RCTs as having low methodological quality, thus reducing confidence in judgements. The most recent evidence from a Cochrane review of 16 RCTs suggests that acupuncture is not superior to sham acupuncture for OA of the hip [ 67 ], although in contrast, evidence from non-Cochrane reviews suggests that there is moderate-quality evidence that acupuncture may be effective in the symptomatic relief of pain from OA of the knee. Why there should be a difference in evidence between the knee and the hip is not known. Interestingly, guidance from NICE (CG177) states: “Do not offer acupuncture for the management of osteoarthritis” Section 1.4.6. [ 198 ]. The overview of eight Cochrane reviews of acupuncture for various pain conditions described previously [ 28 ] included one Cochrane review on low back pain [ 88 ] and judged there to be evidence that acupuncture might be an effective adjunctive intervention for low back pain. However, the quality of the primary studies was low. An overview of 16 SRs on acupuncture for low back pain published in 2015 judged that acupuncture either in isolation or as an adjunct to conventional treatment had short-term benefits but again, the quality of the included reviews was variable [ 78 ]. We found that the earliest SR on acupuncture chronic low back pain was published in 1989 and evaluated the clinical efficacy of acupuncture for neck and/or back pain (22 studies including 16 RCTs [ 19 ]) but the findings were inconclusive. The earliest SR that evaluated the clinical efficacy of acupuncture specifically for chronic low back pain was published in 1998 by Ernst et al. (12 RCTs [ 94 ]), and found evidence that acupuncture was superior to various control interventions, but insufficient evidence to judge whether it was superior to placebo. The included studies were mostly of high-quality, but sample sizes were inadequate. In 2000, Smith et al. published a SR that found no evidence that acupuncture was effective for either chronic neck or low back pain (13 RCTs [ 99 ]). Throughout the following decade, SRs reported insufficient high-quality evidence to make any judgement on efficacy of acupuncture in treating low back pain (4 RCTs [ 93 ], 11 RCTs [ 22 ], 5 RCTs [ 91 ], 6 RCTs [ 95 ], 33 RCTs [ 89 ], 35 RCTs [ 88 ] and 10 RCTs [ 90 ]). In 2010, Trigkilidas published a SR that evaluated the clinical efficacy of acupuncture for low back pain (4 RCTs [ 85 ]) that judged there to be evidence of superiority of acupuncture compared with usual care in treating chronic low back pain, especially when patients have positive expectations about the intervention. Between 2011 and 2017, none of the published SRs provided compelling evidence to support the efficacy of acupuncture for chronic low back pain (2 RCTs [ 84 ], 7 RCTs [ 83 ], 13 RCTs [ 82 ], 32 RCTs [ 81 ], 8 RCTs [ 79 ] and 7 RCTs [ 77 ]), or lumbar spinal stenosis (12 studies, 6 RCTs [ 80 ]). In 2018, Hu et al. published a SR that evaluated the clinical efficacy of acupuncture for low back pain (16 RCTs [ 75 ]) and found evidence that dry needling was more effective for low back pain than conventional acupuncture or sham immediately post treatment, but at follow-up, was equal to acupuncture. In 2018, Tang et al. published a SR that evaluated the clinical efficacy of acupuncture for the relief of pain associated with lumbar disc herniation (30 RCTs [ 76 ]), which found insufficient robust evidence to draw firm conclusions because of methodological shortcomings in primary RCTs. However, there was tentative evidence that dry needling was more beneficial than lumbar traction, drug therapy or Chinese herbal medicine. In 2019, Xiang et al. published a SR on acupuncture for non-specific low back pain (14 RCTs [ 74 ]). There was moderate evidence of benefit but confidence in the results was diminished due to heterogeneity and small sample sizes in the included studies. Evidence suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for low back pain. In 2009, NICE published guidance for the management of non-specific low back pain that recommended a course of acupuncture as part of first line treatment [ 10 ]. This guidance produced much debate. Subsequently, NICE have updated guidance for the management of low back pain and sciatica in people over 16 (NG59) and currently recommend in Section 1.2.8 “Do not offer acupuncture for managing low back pain with or without sciatica”, even though the evidence had not significantly changed [ 9 ]. We found that the earliest SR on acupuncture (dry needling) to alleviate pain associated with myofascial trigger points (MTPs) was published in 2001 (23 RCTs [ 116 ]) and found no evidence to demonstrate the efficacy of any needling technique beyond placebo. In 2009, Tough et al. published a SR that evaluated the efficacy of dry needling acupuncture (7 RCTs [ 115 ]) which produced insufficient evidence to determine efficacy. A further systematic review by Tough et al. published in 2011 (3 RCTs [ 113 ]) had similar conclusions. In 2013, a SR found that acupuncture was superior to sham or placebo in reducing pain associated with upper quadrant myofascial pain immediately post-treatment and at four weeks, although the quality of the primary studies was low (12 RCTs [ 112 ]). In 2014, Ong and Claydon published a SR that evaluated the clinical efficacy of dry needling to alleviate pain associated with MTPs in the neck and shoulders (5 RCTs [ 111 ]) and found that there was no significant difference between dry needling and lidocaine. In 2017, Espejo-Antúnez et al. published a SR that evaluated the clinical efficacy of dry needling to alleviate pain associated with myofascial trigger points (15 RCTs [ 106 ]) and found a possible short-term benefit following dry needling. In 2017, SRs have found tentative evidence that acupuncture alone or combined with other therapies improved outcomes associated with myofascial pain syndrome (10 RCTs [ 108 ]; 33 RCTs [ 107 ]), although substantial heterogeneity and a high risk of bias, including inadequate sample sizes in the primary RCTs, undermined confidence in the findings. Evidence from SRs suggests that dry needling acupuncture might be effective in alleviating pain associated with myofascial trigger points, at least in the short-term, although there are insufficient high-quality RCTs to judge the efficacy with any degree of certainty. There is no guidance from NICE on the management of myofascial pain syndrome. We found one overview of SRs of acupuncture for palliative and supportive cancer care that included 7 SRs [ 125 ], but only one systematic review on cancer-related pain [ 126 ]. We found that the earliest SR on acupuncture for pain associated with cancer and/or its treatment (7 studies with 3 RCTs [ 126 ]) concluded that there was evidence of efficacy for chemotherapy-induced nausea and vomiting, but insufficient evidence to judge efficacy for cancer-related pain. In 2010, a SR of 7 RCTs provided tentative evidence that that acupuncture alleviated cancer-related pain [ 124 ], and in 2011, the first Cochrane review on acupuncture for cancer pain judged there to be insufficient evidence to determine the efficacy (3 RCTs [ 123 ]), and this was confirmed in an update in 2015 (5 RCTs [ 119 ]). Subsequently, non-Cochrane SRs in 2012 (15 RCTs [ 122 ]), 2013 (11 RCT [ 121 ]), 2014 (33 studies, 6 RCTs [ 120 ]) and 2016 (20 RCTs [ 120 ]) provide promising but inconclusive evidence of efficacy. In 2017, Chiu et al. published a Cochrane review that evaluated the clinical efficacy of acupuncture for cancer-related pain, which included treatment-related or surgery-related pain, and judged there to be evidence that acupuncture alleviated pain associated with malignancy (29 RCTs [ 117 ]), but there was a high risk of bias due to inadequate sample sizes. Evidence from the SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for cancer-related pain and more high-quality, appropriately designed and adequately powered studies are needed. The most recent guidance from NICE (CSG4) recognises that patients who are receiving palliative care often seek complementary therapies, but it does not specifically recommend acupuncture. It recognises that “Many studies have a considerable number of methodological limitations, making it difficult to draw definitive conclusions” (Section 11.27) [ 199 ]. We found that the earliest SR on acupuncture for fibromyalgia was published in 1999 (7 studies, 3 RCTs [ 136 ]) and concluded that there was limited evidence supporting the use of acupuncture for fibromyalgia but this was based on only one high-quality study. Subsequently, SRs published in 2007 (5 RCTs [ 135 ]) and 2009 (6 RCTs [ 134 ]) concluded that acupuncture had no symptomatic benefit, and in 2010 were inconclusive (7 RCTs [ 133 ], and 25 studies, 12 RCTs [ 132 ] respectively). In 2013, a Cochrane review conducted by Deare et al. (9 RCTs [ 131 ]) found low-quality evidence that acupuncture might be superior to no acupuncture or medication, and moderate-quality evidence that acupuncture was not superior to sham. Non-Cochrane SRs published in 2013 (16 RCTs [ 130 ]) and 2014 (9 RCTs [ 129 ]) were inconclusive. In 2019, two SRs have produced evidence that acupuncture was superior to sham but the evidence status was downgraded due to high levels of heterogeneity and inadequate sample sizes (10 RCTs [ 128 ]; 12 RCT [ 127 ]). Evidence from SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for fibromyalgia pain. There is no NICE guidance on the treatment of fibromyalgia. We found one overview of SRs on acupuncture for primary dysmenorrhoea which was published in 2018 and concluded that the evidence was inconclusive (5 SRs [ 138 ]). We found a number of SRs on acupuncture and associated therapies for primary dysmenorrhea, although all report a high-risk of bias leading to evidence that is inconclusive (30 RCTs [ 147 ]; 4 RCTs [ 146 ]; 16 RCTs [ 144 ]; 20 RCTs [ 145 ]; 23 RCTs [ 142 ]; 60 RCTs [ 139 ]). In 2008, a SR investigating acupuncture for pelvic and back pain during pregnancy was inconclusive (3 RCTs [ 152 ]), and in 2010, a SR described RCTs findings as ‘promising’ but inconclusive for primary dysmenorrhea (27 RCTs [ 151 ]). A Cochrane review by Zhu et al. in 2011 on acupuncture for endometriosis included one low-quality RCT and was inconclusive [ 150 ]. A follow-up non-Cochrane review in 2017 including 10 RCTs was still inconclusive [ 143 ]. We found five SRs on acupuncture for chronic prostatitis and/or chronic pelvic pain, and despite promising RCT findings, all reviewers concluded that the evidence was inconclusive (9 RCTs [ 149 ]; 35 RCTs [ 148 ]; 7 RCTs [ 141 ]; 4 RCTs [ 140 ]; 4 RCTs [ 137 ]). Evidence from the SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for primary dysmenorrhea or chronic pelvic pain. There is NICE guidance on endometriosis (NG73) [ 200 ] but this does not recommend any form of Chinese medicine for this type of pelvic pain, although acupuncture is not specifically mentioned. In 2018, an overview of SRs concluded that acupuncture has minimal or no impact on joint pain associated with rheumatoid arthritis (7 SRs, 20 RCTs [ 153 ]). We found that the earliest SR on acupuncture for pain associated with inflammatory rheumatic diseases was published in 1997 and found insufficient high-quality evidence to make a judgement on efficacy. Subsequently, a Cochrane review published in 2005 (2 RCTs [ 159 ]) and various non-Cochrane SRs published in 2008 (8 RCTs [ 158 ]; 8 RCTs [ 157 ]), 2013 (10 RCTs [ 156 ]) and 2016 (28 RCTs [ 155 ]) have been inconclusive. The most recent SR on acupuncture for rheumatoid arthritis reported that RCT findings were tentatively positive but inconclusive (13 RCTs [ 154 ]). Evidence from the SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for pain in inflammatory arthritis. There is a NICE guideline (NG100) [ 201 ] for the treatment of rheumatoid arthritis but this does not recommend acupuncture. The earliest SR on acupuncture for neurological symptoms was published in 1997 and reported that findings were positive for alleviation of symptoms associated with lumbar disk herniation (38 studies, 7 RCTs [ 171 ]). The majority of SRs have been conducted on peripheral neuropathy of various aetiologies, but all had methodological shortcomings resulting in inconclusive evidence (chemotherapy-induced peripheral neuropathy, 6 studies, 3 RCTs [ 168 ]; various peripheral neuropathies, 15 studies, 13 RCTs [ 166 ]; and diabetic peripheral neuropathy, 14 RCTs [ 164 ]). There are two SRs on acupuncture for trigeminal neuralgia (12 RCTs [ 170 ]; 33 RCTs [ 162 ]), two SRs on acupuncture for carpal tunnel syndrome (6 RCTs [ 169 ]; 12 RCTs [ 23 ]) and one SR on acupuncture for post-herpetic neuralgia (7 RCTs [ 165 ]). None were able to judge efficacy with any degree of confidence due to insufficient high-quality RCTs. Evidence from the SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for neuropathic pain or neuralgia. There is NICE guidance (CG173) [ 202 ] on the management of neuropathic pain, but acupuncture is not included in the list of recommended/not recommended treatments. In 2002, a Cochrane review found insufficient high-quality RCTs to determine the efficacy of acupuncture for lateral elbow pain (4 RCTs [ 194 ]). In 2005, a Cochrane review found insufficient high-quality RCTs to judge the efficacy of acupuncture for shoulder pain. In 2011, a Cochrane review by Smith et al. found insufficient evidence to judge the efficacy of acupuncture or acupressure for labour pain (13 RCTs [ 187 ]). Our search found an additional 27 reviews for a variety of other pain conditions, including dental/facial pain, osteoporosis and upper extremity pain of various aetiologies, although none of these reviews provides sufficient high-quality evidence to make a judgement about the efficacy of acupuncture ( Table 1 ) [ 21 , 172 , 173 , 174 , 175 , 176 , 177 , 178 , 179 , 180 , 181 , 182 , 183 , 184 , 185 , 186 , 187 , 188 , 189 , 190 , 191 , 192 , 193 , 194 , 195 , 196 , 197 ]. Evidence from SRs suggests that there are insufficient high-quality RCTs to judge the efficacy of acupuncture for a variety of other painful conditions, including lateral elbow pain, shoulder pain and labour pain. There is no guidance available from NICE on the treatment of any of these conditions.

Our evidence synthesis reveals long-standing and continued uncertainty about the clinical efficacy of acupuncture to alleviate pain, despite a high volume of published research. We have revealed a raft of SRs with inconclusive findings due to persistent methodological shortcomings in RCTs contributing to a high risk of bias and downgrading of evidence. These shortcomings include inadequate statistical power, uncertainty about adequacy of acupuncture technique and dose, and inappropriate design of ‘placebo’ acupuncture controls. These contribute to methodological and clinical heterogeneity, deterring systematic reviewers from pooling data for meta-analyses. When meta-analyses are conducted, substantial statistical heterogeneity results, markedly reducing confidence in findings and inferences [ 18 , 203 , 204 ]. The high financial cost of continuing to undertake research that produces inconclusive evidence is of concern and demands reconsideration of the methodological design and delivery of future RCT design. We will discuss three common challenges to the design of RCTs of acupuncture that emerge from our evidence synthesis: adequate sample sizes, adequate acupuncture intervention, and adequate placebo controls. RCTs with small sample sizes are associated with an overestimation of treatment effects. Dechartres et al. [ 18 ] found that treatment effects were, on average, 48% larger in trials with fewer than 50 patients. Overestimation of treatment effects occurs in studies with sample sizes of 100–200 participants per treatment arm, suggesting that at least 200 participants per treatment arm is necessary to achieve a low risk of bias. Roberts [ 205 ] argued that the production of fewer but broader reviews that exclude underpowered trials would increase the validity of review findings and create a more trustworthy evidence base. Turner et al. [ 203 ] examined the distribution of statistical power within meta-analyses published as part of Cochrane reviews and argued that the results of meta-analyses that contain at least two adequately powered studies are not influenced to any significant degree when underpowered studies are omitted. At present, the inclusion of underpowered studies in meta-analyses is at the discretion of reviewers. Funding constraints that prevent the use of larger sample sizes in RCTs is likely to continue into the future. Thus, strategies to reduce statistical heterogeneity associated with high variance in pain data in RCTs need consideration. Often, pain data used as the primary outcome within RCTs is a continuous variable, such as pain intensity measured on a visual analogue scale (VAS) and expressed as an average. Averages of pain intensity data from VAS can be misleading because averages may obscure good and poor responders to acupuncture [ 206 , 207 ]. There is a likelihood that scores of pain intensity produce U-shaped rather than bell-shaped distributions, with some participants experiencing large reductions in pain and others not. Thus, pain intensity data from acupuncture responders may be diluted by data from non-responders [ 208 ]. For this reason, the Pain and Palliative Support and Care group of the Cochrane collaboration recommends the use of primary outcome responder rates of participants reporting relief of 30% or greater (i.e., at least moderate pain relief) or 50% or greater (i.e., significant pain relief) expressed as frequency (dichotomous) data. Acupuncture RCTs can assess two aspects of the active ingredient of treatment: effects associated with needling acupuncture points and effects associated with needles piercing the skin. Thus, two common controls used in RCTs of acupuncture are: inserting real needles into the skin at non-acupuncture points and using ‘sham’ needles which touch but do not penetrate the skin. It is important that SRs and RCTs emphasise exactly which outcome is being assessed at the outset, and ideally include this in the title and aim of the report. Controls that involve inserting needles into the skin at non-acupuncture points can be used to determine the influence of needling discrete points of the skin on outcome. If administering treatment at any point on the skin produced equivalent benefits and harms when compared with needling specific points, this would challenge the need for anatomical acupuncture charts and prescribed acupuncture practitioner training. Controls that use ‘sham’ needles which touch but don’t penetrate the skin are often labelled as placebo controls. The purpose of a placebo control comparison is to isolate the effect of the act of receiving a treatment from the active ingredient of the treatment. Placebo controls are usually operationalised using fake or sham interventions and enable measurement of non-specific treatment effects associated with expectations, conditioning, anxiety and social context (i.e., therapist/patient interaction and theatrical elements of the treatment) [ 209 , 210 ]. It has been argued that the reason why some RCTs fail to detect differences in treatment effects between real and sham acupuncture is that sham needling techniques are not physiologically inert, and this may have contributed to an underestimation of acupuncture effects in the evidence base [ 211 ]. This argument is valid but can be misleading if taken at face value. The purpose of a control intervention is not to be physiologically inert but rather to control for outcomes associated with non-specific effects of the act (theatre) of receiving the treatment. No placebo control (including a sham needling) is ever physiologically inert because it instigates changes in physiological (and psychological) state. The human body evolved to detect and respond to disturbances in the internal and/or external environment (i.e., stimuli) from physical, physiological, social and/or environmental change. This is the premise of homeostasis. Placebo controls are research tools that enable isolation of effects associated with the active ingredient(s) of the treatment. Thus, a comparison of effects during real needling versus sham needling, whereby needles touch but do not penetrate the skin, enable investigators to isolate the magnitude and incidence of effects associated with needles piercing the skin per se (i.e., the ‘acu’ and ‘puncture’). If puncturing the skin with needles produces equivalent benefits to touching without puncturing the skin, then it may be safer not to puncture the skin in clinical practice, providing that the sham needles do less harm. Interestingly, a system of evaluating the physiological effects of sham needling has been proposed to assist researchers [ 212 ]. The term ‘placebo’ is used extensively in research and clinical literature, although it lacks scientific precision and has become emotive. We would prefer precise statements of purpose and method when describing control interventions. For example, a control group that uses fake needles that do not puncture the skin would be used to isolate effects associated with needles puncturing the skin. We would also encourage a shift away from assessing patient ‘blinding’ using questions such as ‘Do you think the intervention was a placebo?’ to questions assessing the ‘credibility’ and ‘functioning’ of interventions using questions such as ‘Do you think the intervention was credible?’ and ‘Do you think the intervention was functioning correctly?’, as has been suggested for other non-pharmacological interventions such as transcutaneous electrical nerve stimulation (TENS) [ 213 ]. Acupuncture practitioners argue that acupuncture is a complex intervention that should not be standardised but instead tailored to each individual patient, based on principles of practice and the experience of the clinician. Components of needling include type, number, and location of needles, needling technique (e.g., thrusting, rotation, flicking, pecking), duration of needle insertion, regimen of treatment and philosophical paradigm. Debates about optimal technique are long-standing and there are evidence-based principles underpinning optimisation of technique for acupuncture treatment [ 214 ]. Delivering identical acupuncture prescription to all participants runs the risk of some participants receiving sub-optimal dose. Often, acupuncture interventions used in RCTs are grounded in principles of Western acupuncture with flexibility to individualise treatment at the discretion of individual practitioners. Individualising acupuncture treatment increases between-subject variability in treatment (e.g., needling number, location, technique, duration). At face value, this may appear to conflict with classical RCT methodology that aims to standardise methodology and treatment intervention under strictly controlled conditions. However, standardisation can be based on the principles of optimising treatment per individual, as is the case when titrating drug dosage to therapeutic window. What constitutes adequacy of acupuncture technique and dose has been a matter of much debate [ 56 , 214 , 215 , 216 ]. In trials of pharmacological agents, dose is crucial, and it should be no different in studies investigating the efficacy of acupuncture. The Standards for Reporting Interventions in Controlled Trials of Acupuncture (STRICTA) were developed from the consolidated standards for reporting trials (CONSORT) [ 217 ] to encourage accurate reporting of the acupuncture intervention [ 218 ]. STRICTA recommend that six items should be included: rationale, details of needling (e.g., points used, depth, angle, needle thickness, number of needles), treatment regimen, co-intervention, practitioner background and control interventions [ 219 ]. The impact of using STRICTA has been positive with improvements in reporting quality of RCTs on acupuncture [ 220 , 221 , 222 , 223 ]. In 2008, White et al. published a meta-analysis that provided evidence that better outcomes in comparisons of acupuncture with non-acupuncture controls were achieved when noted that greater numbers of needles and treatment sessions were used [ 214 ]. In 2019, Sun et al. conducted a systematic review of eight RCTs (2106 participants) to determine whether the effect of acupuncture is dose-dependent for symptom management in knee osteoarthritis [ 56 ]. Sun et al. proposed a scoring system whereby +1 score was awarded if ≥9 points needled, if de qi was present, if ≥2 treatment sessions a week and if ≥8 treatment sessions in total. A score of −1 was awarded to each of these parameters if they were below these thresholds. The sum of scores was taken and high dosage categorised for total between 1 and 4, medium dosage for a score of 0 and low dosage for scores from −4 to −1. Sun et al. categorised one RCT as low dose, one RCT as medium dose and 6 RCTs as high dose and concluded that higher dosage of acupuncture was associated with better pain relief and functional improvements. It is becoming common for journal editors to require STRICTA in RCTs of acupuncture and this will improve comparison and assessment of adequacy of acupuncture dose in systematic reviews. What is less common, however, is the inclusion and reporting of ‘run-in phases’ in RCTs, whereby optimisation of technique and dosage is titrated over a period of weeks prior to randomisation into real and placebo acupuncture. It has been argued that enriched enrolment with randomised withdrawal (EERW) study designs are of value for treatments influencing symptoms but not necessarily the course of the underlying disease or pathology, as is the case for acupuncture in the management of chronic pain [ 224 ]. The potential for using such designs in the assessment of pharmacological agents has been recognised [ 225 ], although EERW designs are rarely used to assess non-pharmacological interventions. The EERW trials consist of (i) an observational ‘open-label’ phase with all participants receiving active treatment (acupuncture), during which treatment technique and dosage would be titrated and optimized, followed by (ii) a RCT phase, whereby participants who had potential for response were enrolled (i.e., an enriched sample) and randomised to receive either experimental (real needling) or control interventions (sham needling). Selection of participants for the enriched sample of the RCT is based on the findings from phase one and would exclude participants who did not wish to continue treatment or experienced non-manageable adverse events, although their data from phase one would be analysed. Trials with EERW designs increase sensitivity to detect treatment effects by enriching the sample of participants enrolling into the randomised controlled phase of the trial, thus reducing the need for large sample sizes [ 207 ]. To our knowledge, there have not been any published studies of acupuncture using the EERW design, although it has been used to determine the efficacy of drugs for chronic pain conditions. Given the shortcomings in classically design RCTs on acupuncture, it would be interesting to observe the results of studies using an EERW design. A limitation of this synthesis is that it does not contain granular quantitative analyses. It could be argued that there is a case for an all-encompassing SR and meta-analysis of all RCTs on acupuncture for pain conditions, but this would be a considerable undertaking with the possibility that it not produce any meaningful information due to the relatively poor quality of RCTs resulting in amplification of heterogeneity.

We hope that our evidence synthesis of systematic reviews and meta-analyses of RCTs of acupuncture for chronic pain conditions serves as a reference tool for practitioners, researchers and commissioners. Our evidence synthesis reveals a long-standing unresolved debate about the clinical efficacy of acupuncture to alleviate pain that is grounded in a high volume of inconclusive RCT evidence. If healthcare providers and commissioners are to be able to make informed choices on the role of acupuncture for chronic pain, it is essential that the quality of clinical trials of acupuncture is improved. Our evidence synthesis has revealed three methodological challenges that have faced investigators of RCT of acupuncture for decades. We have argued that enriched enrolment with randomised withdrawal trial designs may provide a way forward. We hope that our review catalyses further debate on this issue.