Acupuncture for Chronic Pain: What the Guidelines Actually Say

Acupuncture occupies an unusual position. It is the alternative therapy that most thoroughly crossed over — recommended in mainstream pain guidelines, covered by Medicare for chronic low back pain, offered inside academic medical centres — while the mechanism proposed by its own tradition, the movement of qi along meridians, has no anatomical correlate anyone has been able to demonstrate.

Both of those things are true at once, and the resulting literature is more interesting than either the enthusiasts or the debunkers usually allow.

What the guidelines say

A 2025 systematic review of clinical practice guidelines on acupuncture for chronic musculoskeletal pain identified seventeen guidelines carrying thirty-five relevant recommendations. Sixty percent of those recommendations supported using acupuncture.

That figure travels well in marketing copy. The rest of the breakdown travels less well: of the supportive recommendations, 5.7 percent were strong, and 54.3 percent were weak or conditional.

A weak or conditional recommendation is a specific object in guideline methodology. It means a panel judged the intervention reasonable to offer while explicitly acknowledging that the evidence is uncertain, that the effect may be small, and that a different patient with different preferences might reasonably decline. It is not a shrug, but it is not an endorsement either.

What the trials say

The underlying trial evidence is where the conditional language comes from.

A 2025 systematic review and meta-analysis of acupuncture versus usual care for chronic low back pain, searching the literature through late 2024, found effects on pain and disability — and rated the overall certainty of that evidence low to very low. The stated reasons were risk of bias in the constituent trials, inconsistency between them, and suspected publication bias.

Publication bias deserves a moment. It means the review's authors saw a pattern in the distribution of published results suggesting that trials finding nothing were less likely to have been published than trials finding something. Where that pattern is present, the pooled estimate is inflated by an unknown amount.

The picture is somewhat better in specific territories. A meta-analysis of eighteen randomised trials in chronic neck pain found that acupuncture used as an adjunct therapy produced pain relief that persisted at three and six months after treatment ended — durability being exactly what short-follow-up trials usually fail to demonstrate.

The honest summary is not that acupuncture works, nor that it does not. It is that the effect is probably real, probably small, and measured through instruments that struggle with this kind of intervention.

The sham problem

Trials of acupuncture face a methodological difficulty that shapes everything downstream. The standard control is sham acupuncture — retractable needles, or real needles inserted shallowly at non-meridian points. Across a large body of work, real acupuncture tends to beat no treatment or usual care clearly, and to beat sham acupuncture only slightly.

There are two readings of that, and they are not reconcilable by wanting one of them. The first is that most of the benefit is contextual: the appointment, the attention, the ritual, the expectation, the therapeutic relationship, and the general tendency of pain to fluctuate. The second is that sham acupuncture is not inert — that pressing needles into skin anywhere produces real physiological effects, in which case the sham comparison systematically understates the true effect.

Both readings have serious researchers behind them. What is not defensible is citing the real-versus-usual-care result while ignoring the real-versus-sham one.

What this means for a patient

For chronic low back pain, chronic neck pain, and knee osteoarthritis, acupuncture is a reasonable thing to try, with a safety profile that is consistently favourable when practitioners are licensed and use single-use sterile needles. Serious adverse events are rare and are concentrated in practice that departs from those basics.

The reasonable expectation is modest improvement, not resolution, possibly not durable, and not obviously superior to other conservative options with comparable evidence. That is a legitimate offer. It is simply a smaller one than the field's public presentation of itself suggests.

This article is general information, not medical advice. See our Editorial Standards.