Large Reviews of Acupuncture: Individual Patient Data Meta-Analyses
What the biggest pooled analyses and Cochrane reviews of acupuncture found, how individual patient data meta-analyses work, and why large reviews can still disagree.
Content last updated . This page has not been medically reviewed.
Why reviews matter more than single trials
One trial can be small, lucky or biased. A systematic review gathers all the trials on a question; a meta-analysis pools their numbers. An individual patient data (IPD) meta-analysis goes a step further: the researchers obtain the raw results for each participant and analyze them together. This page covers the biggest reviews of acupuncture and what to make of them. Labels: established, promising but limited, mixed, insufficient evidence. It is information, not advice about your health.The Vickers IPD meta-analyses (chronic pain)
The first analysis used data from 29 of 31 eligible trials, 17,922 patients, of back and neck pain, osteoarthritis, chronic headache and shoulder pain. [1] Only trials with unambiguously adequate allocation concealment were included. [1] Acupuncture beat sham and no-acupuncture control. Against sham, pain scores were 0.23, 0.16 and 0.15 standard deviations lower for back and neck pain, osteoarthritis and chronic headache. [1] The authors called the differences relatively modest, with factors beyond needling contributing. [1] They reported that results held across sensitivity analyses, including those for publication bias. [1]The 2018 update added 13 trials, for 20,827 patients from 39 trials. [2] Differences were close to 0.5 standard deviations against no acupuncture and close to 0.2 against sham. [2] The effect decreased only about 15% by one year. [2] The authors found that variation between trials was driven mostly by what the control group received, not by acupuncture technique. [2] Sham trials using a penetrating needle for the sham had smaller effects. [2]
Cochrane reviews of specific conditions
Cochrane reviews of acupuncture, by condition:| Review | Size | Main finding | Label |
|---|---|---|---|
| Chronic nonspecific low back pain (2020) | 33 studies, 8,270 people [3] | Low-certainty evidence that pain relief versus sham was about 9 points on a 100-point scale, below the 15-point threshold used; moderate-certainty evidence of clinically important relief versus no treatment [3] | Mixed |
| Neck disorders (2016) | 27 studies [4] | Moderate-quality evidence of better pain relief than sham at short-term follow-up; not sustainable long-term [4] | Promising but limited |
| Tension-type headache prevention (2016) | 12 trials, 2,349 people [5] | 51% versus 43% with at least a 50% headache reduction against sham; moderate or low quality, mainly for lack of blinding [5] | Promising but limited |
A 2025 Cochrane protocol exists for acupuncture in preventing episodic migraine; a protocol has no results yet. [6] NCCIH, which summarises many 2016 to 2021 reviews, reports for back or neck pain that data from 12 studies (8,003 participants) showed acupuncture beat no treatment. [7] It adds that the difference from no treatment was greater than the difference from sham. [7]
Reading the numbers
- Standard deviations (SD) express the difference between groups in units of how much people vary. They are hard to translate to daily life. The Cochrane and BMJ reviews also give points on a 0 to 100 pain scale, which are easier to picture.
- Certainty (GRADE) is how confident reviewers are in the result, separate from the size of the result.
- Side effects reported in the Cochrane back pain review included insertion-point pain, bruising, bleeding and worsening of back pain. [3] Fuller coverage: side effects and precautions.
What reviews cannot do
Reviews cannot fix flaws in the underlying trials, and many trials of acupuncture are hard to blind. Publication bias, where positive results are more likely to be published, is a general worry; the 2012 Vickers analysis reported results robust to sensitivity analyses for it. [1] For the topic in Traditional Chinese Medicine research, see publication bias in TCM research. Reviews also do not tell you whether acupuncture suits you personally. For the condition-by-condition summary, see Does Acupuncture Work? and acupuncture for chronic pain: what large reviews found.Frequently asked questions
What is an individual patient data meta-analysis?
It pools the raw results for each participant across trials rather than only the published summaries. The Vickers analysis pooled 20,827 patients from 39 trials. [2]
Do the biggest reviews say acupuncture works?
For chronic pain, the Vickers analyses say yes, with a modest advantage over sham and larger over no acupuncture. [2] Cochrane back pain reviewers were more cautious about the sham difference. [3]
Why do reviews give different answers?
Differences in included trials, shams and the threshold for "clinically meaningful" all matter. [2][8]
Do effects last?
In the Vickers update, only about 15% of the effect was lost at one year. [2] The Cochrane neck pain review said its effect did not seem sustainable over the long term. [4]
Are there reviews for other conditions?
Yes, and NCCIH summarises many. Results range from supportive to negative; see Does Acupuncture Work?. [7]
More details: evidence, limits and cautions
The details behind this page: cautions, limits of the evidence and where sources disagree. Open any note to read it.
When large reviews disagree
- Which trials were included (allocation concealment rules, outcomes).
- Whether a difference of this size is "worth it."
- How the sham was designed. [2]
- Blinding: the Cochrane back pain review noted most studies had high risk of performance bias because the acupuncturist could not be blinded. [3]
More on these design limits is on sham acupuncture and placebo controls.