Chronic pain

The largest analyses of acupuncture for chronic pain are the individual patient data meta-analyses of the Acupuncture Trialists’ Collaboration, an international group of trial authors whose analyses Vickers and colleagues published. Instead of pooling the averages each trial published, they obtained the record of every participant from the trialists and reanalysed them together. The first analysis, in 2012, covered 29 trials and 17,922 patients; the 2018 update covered 39 trials and 20,827 patients. Both found acupuncture better than sham and better than no acupuncture for each pain condition they studied, with the difference from no acupuncture about two and a half times the difference from sham. NICE, which writes clinical guidance for the NHS in England, drew on a separate review of 27 studies when it recommended in 2021 that a single course of acupuncture be considered for chronic primary pain.

NICE defines chronic pain as pain that persists or recurs for more than three months; the Collaboration required a current episode of at least four weeks for musculoskeletal pain. Sham acupuncture explains the controls that imitate treatment, and why a comparison with sham and a comparison with no acupuncture answer different questions.

The 2012 analysis

The Collaboration admitted only trials in which allocation concealment was unambiguously adequate, meaning nobody enrolling patients could foresee or influence which group the next one joined. The trials covered four conditions: non-specific back and neck pain, osteoarthritis, chronic headache and shoulder pain. Results are standardised differences, in units of the standard deviation of pain scores, so that trials using different pain scales can be combined.

In the primary analysis, which included every eligible trial, acupuncture beat both sham and no acupuncture for each condition, each with P below 0.001. In that analysis the trials by Vas and colleagues were clear outliers against sham: their neck pain trial had an effect about five times the pooled estimate. With those trials set aside, the effect sizes became similar across conditions, as the table shows, and shoulder pain had too few trials left to pool; Shoulder pain covers that condition.

The authors translated the figures into a worked example. A typical trial might start with pain at 60 on a 0 to 100 scale. Follow-up scores might then be 43 with no acupuncture, 35 with sham and 30 with acupuncture, and if a responder is someone whose pain halved, about 30%, 42.5% and 50% of each group would respond. They concluded that acupuncture is more than a placebo and a reasonable referral option, and that the modest difference from sham means factors other than the needling itself contribute much of the effect.

The 2018 update

The update searched to the end of 2015 and added 13 trials. Its main estimates were close to the first analysis: differences of about 0.5 standard deviations against no acupuncture and about 0.2 against sham, all statistically significant. The effect fell by about 15% at one year. The update found no obvious association between a trial’s result and the characteristics of the acupuncture given. It did find one with the control group: trials whose sham pierced the skin, and trials whose control arm received more intensive care, found smaller effects. The authors concluded that variation between trials came mainly from what the control groups received.

How long the effect lasts

MacPherson and colleagues used the Collaboration’s 2012 dataset to follow pain scores after treatment ended, in a 2017 analysis in Pain. Twenty trials with 6,376 patients had longer-term follow-up. Against no acupuncture, the effect shrank by a statistically non-significant 0.011 standard deviations every three months, which projects to about 90% of the benefit remaining at 12 months. Against sham, it shrank by 0.025 every three months, about half of the difference gone at 12 months.

A different reading of the sham comparison

Madsen, Gøtzsche and Hróbjartsson in Copenhagen took another route in a 2009 review in the BMJ, using only trials that had all three groups: acupuncture, placebo acupuncture and no acupuncture. From 13 trials with 3,025 patients, after excluding one outlying trial of 70, they found acupuncture beat placebo acupuncture by a standardised mean difference of 0.17, about 4 mm on a 100 mm pain scale, and placebo acupuncture beat no acupuncture by 0.42, with wide variation between trials. The acupuncturists were not blinded in any of the trials. The authors concluded that the small difference from placebo seemed to lack clinical relevance and could not be clearly told apart from bias, and that whether needling at acupuncture points, or anywhere, reduces pain independently of the treatment ritual was unclear.

The two groups measured similar differences from sham, about 0.2 standard deviations in the Collaboration’s analyses and 0.17 in Madsen’s, and read them differently. The Collaboration took a difference from sham as evidence of more than placebo; Madsen’s group took a difference of that size as too small to matter and too close to the possible bias to be trusted.

NICE and chronic primary pain

NICE’s 2021 guideline on chronic pain, NG193, defines chronic primary pain as pain with no clear underlying cause, or pain whose impact is out of proportion to any observable injury or disease. Fibromyalgia is one type. NG193 is to be used alongside NICE’s guidelines for specific conditions, and where an underlying condition adequately accounts for the pain, that condition’s guideline applies; the guidelines for low back pain and osteoarthritis both advise against acupuncture.

Recommendation 1.2.5 says to consider a single course of acupuncture or dry needling, within a traditional Chinese or Western acupuncture system, for people aged 16 and over, but only if it is delivered in the community by a healthcare professional on NHS pay band 7 or lower with appropriate training, and uses no more than 5 hours of that professional’s time, or else costs the same or less some other way. The committee’s rationale reports that 27 studies showed acupuncture reduced pain and improved quality of life for up to three months against usual care or sham, and that the committee read the benefit over sham as a specific treatment effect despite the difficulty of blinding. Its own economic model found acupuncture likely to be cost-effective. The limits on setting, grade and hours came from that modelling, and the committee recorded that it was uncertain whether the benefit would last.

What the analyses could not settle

The analyses agree on the direction of the comparisons and differ on what the sham comparison means. The trials available could not settle it: in none of the trials in Madsen’s review was the acupuncturist blinded. The Collaboration’s finding that control groups drive most of the variation between trials means that a pooled effect size describes the trials’ choice of comparison as much as the acupuncture. Long-term evidence is thinner than short-term: MacPherson’s persistence estimates came from 20 of the 29 trials, and NICE found too little evidence beyond three months. The analyses cover four conditions, so they say nothing directly about chronic pain elsewhere; the condition pages in this section report what each body of trials found, and Reading an acupuncture trial covers how to weigh a single trial. Temporomandibular disorders reports the trials in chronic jaw pain, which none of the four analyses covered.

AcuiQ’s chronic pain page lists the protocols individual studies prescribed, each with its citation, and the musculoskeletal pain and myofascial pain pages do the same for the terms those studies used.