Acupuncture beats no acupuncture by 0.5 SD and sham acupuncture by 0.2

Acupuncture reduces chronic pain by close to half a standard deviation compared with no acupuncture, and by close to a fifth of a standard deviation compared with sham acupuncture [s1]. Both differences are statistically robust. The distance between them is where the entire argument about acupuncture lives.

The estimate comes from an individual patient data meta-analysis — the strongest form of pooling, in which raw data are obtained from the original investigators rather than effect sizes lifted from published abstracts. The update covered 20,827 patients from 39 randomised trials in non-specific musculoskeletal pain, osteoarthritis, chronic headache and shoulder pain, searching MEDLINE and the Cochrane Central Register of Controlled Trials to 31 December 2015 [s1]. Trials were included only where allocation concealment could be unambiguously determined to be adequate [s1].

Acupuncture was superior to both sham acupuncture and no-acupuncture control for each of the four pain conditions, all at P < .001 [s1]. The differences were close to 0.5 standard deviations against no acupuncture and close to 0.2 standard deviations against sham [s1]. The effect persisted, with roughly a 15% decrease in treatment effect at one year [s1].

What each comparison is actually measuring

The no-acupuncture comparison measures everything that happens when someone receives acupuncture rather than nothing: the appointment, the practitioner's attention, the ritual, the expectation, the needles, and the natural course of the condition over the trial period. Half a standard deviation is a substantial effect on that reading — and it is also the effect a patient experiences, which is why it is not a trivial number.

The sham comparison strips out everything the two arms share and leaves whatever is specific to needling real points at real depths. That residue is 0.2 standard deviations [s1] — small, but not zero, and reproduced across four conditions in 39 trials with adequate allocation concealment.

The authors' own conclusion holds both halves at once: acupuncture is effective for chronic musculoskeletal, headache and osteoarthritis pain, the effects persist, and they cannot be explained solely in terms of placebo effects — while factors in addition to the specific effects of needling at correct point locations are important contributors to the treatment effect [s1].

The sham is not a neutral instrument

The most methodologically useful finding in the analysis is about the controls rather than the treatment. In secondary analyses, effect sizes showed no obvious association with the characteristics of the acupuncture given — how many needles, which points, how many sessions — but were strongly associated with the type of control group used [s1]. Effect sizes were smaller in sham-controlled trials that used a penetrating needle for the sham, and smaller in trials where the control arm itself involved a high intensity of intervention [s1].

That is a serious problem for the field, and an underappreciated one. A "sham" that penetrates the skin is not an inert placebo; it is a weaker version of the treatment. If the sham is partly active, the sham-controlled effect size is an underestimate of the treatment's specific effect, and comparing two acupuncture trials with different shams is comparing different questions.

The reverse reading is equally available: if superficial needling anywhere produces most of the benefit, then the theory of specific point locations is not doing the explanatory work its proponents claim. The data as they stand are consistent with both readings, and the meta-analysis does not adjudicate between them.

Where it lands in guidance and in pooled comparisons

The American College of Physicians recommends acupuncture among the first-line non-pharmacological options for chronic low back pain, at moderate-quality evidence, and for acute or subacute low back pain at low-quality evidence [s3].

In the largest network meta-analysis of conservative treatments for chronic low back pain — 551 studies, 71,126 patients — acupuncture produced one of the largest short-term effects on back pain intensity, a mean difference of −20.91 points on a 0–100 scale (95% credible interval −24.00 to −11.95), and cleared the clinically important threshold for disability as well, at −10.52 (−11.84 to −6.59) [s2]. Those figures are against no treatment, not sham, so they carry the same interpretive load as the 0.5 SD figure above.

Two caveats travel with them. That analysis graded 98.6% of its evidence very low certainty across interventions and time points [s2]. And in the long term, at twelve months or more, none of the statistically significant differences it found reached clinical significance [s2].

What a reader can conclude

Three things, in descending order of confidence. Patients who receive acupuncture for these conditions report meaningfully less pain than patients who receive nothing, and that persists for about a year [s1]. Some part of that — around 0.2 standard deviations — survives comparison with a sham procedure, so it is not entirely context and expectation [s1]. And the size of that surviving part depends on what the sham was, which means the honest confidence interval around "how much of acupuncture is acupuncture" is wider than any single trial suggests [s1].

What to watch

Whether the field converges on a non-penetrating sham standard. Until it does, every sham-controlled acupuncture trial is measuring a slightly different thing, and the pooled 0.2 will keep meaning different things in different studies.

This article is informational and is not medical advice.