Sham Acupuncture and Placebo Controls: Why Interpreting Trials Is Hard

Why this matters

When a trial reports that acupuncture "worked," the next question is: compared with what? A needle can be felt, and the person giving it knows what they are doing. That makes a fair placebo hard to build. This page explains how researchers handle that problem and how to read the results. For the same issue across Traditional Chinese Medicine more broadly, see sham acupuncture, placebo and blinding in TCM trials. For what the trials added up to, see Does Acupuncture Work?. Evidence labels: established, promising but limited, mixed, insufficient evidence.

Common kinds of comparison

NCCIH gives one example of a sham as a device that pokes the skin but does not penetrate it. [1]

The sham is not a neutral zero

In the update of an individual patient data analysis, effect sizes were smaller for sham-controlled trials that used a penetrating needle for the sham. [2] That suggests some shams are not truly inert, or at least that the choice of sham shapes the result. A BMJ review of 13 three-armed pain trials (3,025 patients), however, found no association between the type of placebo acupuncture and the effect of acupuncture. [3] The studies do not agree, which is part of the problem.

What blinding can and cannot do

In the 13 trials of the BMJ review, the clinicians giving the treatment were not blinded in any of them. [3] In the Cochrane review of chronic low back pain, most studies were judged at high risk of performance bias because the acupuncturist could not be blinded. [4] Participants can sometimes be fooled: in a 150-person migraine trial using non-penetrating sham, the share of patients who believed the needle had penetrated was similar in both groups (79% real versus 75% sham). [5] Participants who cannot tell which they received is a better design, but it does not remove the effect of a practitioner's own expectations.

How big is the "placebo" part?

The BMJ review found a small difference between acupuncture and placebo acupuncture, about 4 mm on a 100 mm pain scale, which the authors said seems to lack clinical relevance and cannot be clearly distinguished from bias. [3] It found a moderate difference between placebo acupuncture and no acupuncture. [3] A Cochrane review of placebos in general concluded they do not have important clinical effects overall, but that larger effects were linked to physical placebos such as sham acupuncture. [6] It added that it is difficult to separate real patient-reported placebo effects from biased reporting. [6]

The largest pooled data tell a more favorable story: a difference close to 0.2 standard deviations over sham, and the authors say placebo effects alone cannot explain the improvement. [2] Note the difference in aim: one asks whether acupuncture has any effect beyond the sham, the other whether the effect is clinically large enough to matter.

The ritual and the relationship

One trial split the placebo response into parts in 262 adults with irritable bowel syndrome. Adequate relief of symptoms was reported by 28% on a waitlist, 44% with placebo acupuncture alone, and 62% when the practitioner added warmth, attention and confidence. [7] The authors concluded the patient-practitioner relationship was the most robust component. [7] This does not make acupuncture "fake"; it shows that care and context can produce real, measurable improvement.

Do sham and real acupuncture look different in the body?

Imaging studies suggest they can. In a PET study in fibromyalgia, real acupuncture increased opioid-receptor binding in several pain regions and the sham group showed small reductions. [8] In an fMRI study where high expectation produced equal reported pain relief with real and sham needles, real acupuncture produced a greater decrease in pain-related brain signal. [9] These are single studies and do not settle whether the difference is clinically useful. See brain imaging and neuroscience research.

How to read a headline

1. Ask what the comparison group was.
2. Ask whether patients could tell which they received.
3. Look for the size of the difference, not only whether it was "significant."
4. Remember that a sham-controlled result in one condition says little about others.