How large are the nonspecific effects of acupuncture? A meta-analysis of randomized controlled trials
Linde et al. · BMC Medicine · 2010
Evidence Level
STRONGOBJECTIVE
To investigate the size of nonspecific effects of acupuncture by comparing true acupuncture, sham acupuncture, and no-acupuncture groups
WHO
37 trials with 5,754 patients across diverse conditions: chronic pain, surgical procedures, and other varied conditions
DURATION
Trials varied: short duration (less than 3 days) up to 12 weeks of treatment
POINTS
Varied widely across studies — included traditional acupuncture, electroacupuncture, auriculotherapy, and diverse points
🔬 Study Design
True acupuncture
n=1733
Traditional acupuncture at appropriate points
Sham acupuncture
n=1733
Simulated acupuncture at nonspecific points or non-penetrating needles
No acupuncture
n=1420
Control group without any acupuncture intervention
📊 Results in numbers
Nonspecific effect (sham vs no acupuncture)
Specific effect (acupuncture vs sham)
Total effect of acupuncture
Between-study heterogeneity
Percentage highlights
📊 Outcome Comparison
Standardized mean difference
This study found that simulated (sham) acupuncture produces moderate effects compared with doing nothing, suggesting that important psychological and contextual effects are associated with the ritual of acupuncture. This means that part of the benefit of acupuncture may come not only from specific points but also from the therapeutic experience as a whole.
Article summary
Plain-language narrative summary
This important meta-analysis, conducted by Linde and colleagues, investigated a fundamental question in acupuncture research: how large are the nonspecific (placebo) effects associated with acupuncture interventions? The researchers analyzed 37 randomized controlled trials that included three comparison groups: true acupuncture, simulated (sham) acupuncture, and no-acupuncture controls, totaling 5,754 participants.
The methodology was rigorous, with comprehensive searches of databases (MEDLINE, Embase, and Cochrane Central) through April 2010. Included trials spanned a wide range of conditions: 14 trials in chronic pain (3,369 patients), 8 short-duration trials related to surgical procedures (522 patients), and 15 trials in other varied conditions (1,863 patients). The analysis used standardized mean differences with a random-effects model.
The main results revealed surprising findings about the nature of acupuncture effects. The comparison between sham acupuncture and no-acupuncture groups showed a standardized mean difference of -0.45 (95% CI -0.57 to -0.34), indicating moderately large nonspecific effects. This effect was more pronounced in chronic pain trials (-0.53) compared with short-duration trials (-0.23) and other conditions (-0.42). The comparison between true acupuncture and sham showed smaller specific effects (-0.37), while the total effect of acupuncture versus controls was substantial (-0.77).
A particularly interesting finding was the inverse correlation between specific and nonspecific effects: trials that reported larger effects of sham acupuncture over controls tended to find smaller differences between true and sham acupuncture. This suggests that potent nonspecific effects may mask smaller specific effects. Exploratory analyses indicated that nonspecific effects tended to be larger in trials with fewer co-interventions and when sham acupuncture did not involve skin penetration.
The clinical implications of these findings are significant. The results suggest that sham acupuncture interventions are often associated with moderately large nonspecific effects, which may make it difficult to detect small additional specific effects. Compared with inert pharmacological placebos, the effects associated with sham acupuncture may be larger, with considerable implications for the design and interpretation of clinical trials.
The authors emphasize that, for a suffering patient, it may not matter whether relief is due to specific or nonspecific effects. The total effect of acupuncture, including both specific and nonspecific components, was at least moderate in magnitude across several conditions studied. This raises important questions about how the clinical efficacy of complex nonpharmacological interventions should be evaluated.
Study limitations include strong heterogeneity across trials regarding patients, interventions, outcome measures, and methodological quality. In addition, comparisons between sham acupuncture groups and no-treatment controls are not blinded, introducing potential bias. Patients randomized to no-treatment groups may experience nocebo effects or provide excessively negative assessments.
This study has profound implications for future acupuncture research. The findings suggest that sham-controlled trials investigating small specific effects (standardized mean difference of 0.2) would need to recruit about 800 patients to achieve 80% statistical power, indicating that nearly all available trials comparing true and sham acupuncture would be statistically underpowered. The results also raise fundamental questions about when evidence of clinically relevant total effects from non-blinded comparisons is sufficient to consider a treatment effective, even if specific effects due to the postulated mechanism of action are smaller or nonexistent.
Strengths
- 1Comprehensive meta-analysis with a large number of participants
- 2Rigorous methodology with systematic search
- 3Analysis of three different comparisons providing a complete picture
- 4Innovative investigation of the correlation between specific and nonspecific effects
Limitations
- 1Substantial heterogeneity across included trials
- 2Absence of blinding in sham vs control comparisons
- 3Possible publication bias in some analyses
- 4Limited search of Chinese databases
Expert Commentary
Dr. Marcus Yu Bin Pai
MD, PhD · Pain Medicine · Physical Medicine and Rehabilitation · Medical Acupuncture
▸ Clinical Relevance
This meta-analysis by Linde and colleagues, with 5,754 participants across three comparison arms, offers a quantitative framework that any clinician who prescribes acupuncture should have internalized. The total effect of acupuncture versus no-intervention control reached a standardized mean difference of -0.77 — a magnitude that, in musculoskeletal pain practice, corresponds to clinically perceptible improvement for the patient. The critical point is that this total effect comprises two measurable vectors: a robust nonspecific component (-0.45) and an additional specific component (-0.37). For the clinician in the office, this decomposition matters because it validates prescribing acupuncture as a complete therapeutic intervention — the ritual, the bond, and the context are not noise to eliminate; they are part of the dose. Populations with chronic pain benefit especially prominently, with an estimated nonspecific effect of -0.53 in that subgroup, reinforcing the indication in low back pain, neck pain, and myofascial syndromes refractory to pharmacological treatment alone.
▸ Notable Findings
The inverse correlation between specific and nonspecific effects is the most provocative finding of this analysis. Trials that documented greater response to sham tended to record smaller differences between true and sham acupuncture — suggesting that, in clinical settings with strong therapeutic alliance and elaborate ritual, the 'ceiling' of improvement is reached even before specific point stimulation. This has direct implications for how we interpret negative trials: a trial that does not separate acupuncture from sham is not necessarily evidence of absence of clinical effect; it may be evidence of a nonspecific effect so robust that it obscures the additional delta. Another notable datum is that the nonspecific effect was smaller in trials with sham involving skin penetration, suggesting that any nociceptive stimulus — even minimal — already recruits descending pain modulation pathways, blurring the distinction between verum and control.
▸ From My Experience
In my practice in the pain and rehabilitation service, these numbers reflect very well what we observe longitudinally. I commonly see a subjective response within the first three to four sessions in patients with chronic musculoskeletal pain — which likely already incorporates much of this nonspecific effect quantified by the Linde group. After eight to twelve sessions, most responders are at a plateau of improvement, and we move to a monthly maintenance protocol. The profile that responds best, in my experience, is the patient with moderate-intensity chronic pain, without a predominantly central neuropathic component, who establishes a good alliance with the acupuncturist physician — exactly the scenario in which contextual effects are maximized. I routinely combine acupuncture with motor physical therapy and, when indicated, with duloxetine in pain with a functional component. I do not indicate acupuncture alone when there is an untreated decompensated psychiatric component, since the window of nonspecific response is compromised. The finding of a more pronounced nonspecific effect without co-interventions made me reconsider the practice of overloading the initial session — sometimes, less is more.
Full original article
Read the full scientific study
BMC Medicine · 2010
DOI: 10.1186/1741-7015-8-75
Access original articleScientific Review
Marcus Yu Bin Pai, MD, PhD
CRM-SP: 158074 | RQE: 65523 · 65524 · 655241
PhD in Health Sciences, University of São Paulo. Board-certified in Pain Medicine, Physical Medicine and Rehabilitation, and Medical Acupuncture. Scientific review and curation of every entry in this library.
Learn more about the author →Medical disclaimer: This content is for educational purposes only and does not replace consultation, diagnosis, or treatment by a qualified professional. Some information may be assisted by artificial intelligence and is subject to inaccuracies. Always consult a physician.
Content reviewed by the medical team at CEIMEC — Integrated Centre for Chinese Medicine Studies, a reference in Medical Acupuncture for over 30 years.
Related articles
Based on this article’s categories