Meta-analysis of randomised controlled trials (RCTs) involving acupuncture for labour pain shows acupuncture to be more effective than comparison treatments in several significant circumstances

Authors


Sir,

Recently, Cho et al.1 published what appears to be a thorough systematic review of acupuncture for pain relief during labour. Their abstract’s concluding statement was surprisingly and inappropriately negative: ‘The evidence from RCTs [randomised controlled trials] does not support the use of acupuncture for controlling labour pain.’

This negative statement is a gross misrepresentation of the findings presented in the previous ‘Results’ section of the same abstract, where four of the six results were highly significantly in favour of acupuncture for the provision of relief of pain during labour when compared with various control conditions. These six results were, in turn, a selected subset of the 17 outcomes summarized in Figure 3, where, in fact, nine of the 17 results were significant, and six of these were significant at better than P < 0.003. Their concluding negative statement is also a gross misrepresentation of the results and conclusions presented in the text of the article.

Their reasons for rejecting their own summary of the evidence are unclear. They appear to do so because of inadequate blinding, study heterogeneity and effect size. In the case of blinding, virtually all of the studies would have failed, although most were at least partially blinded. This tough criterion of rigour begs the question: why bother to perform the meta-analyses in the first place? In the case of heterogeneity, although some comparisons suffer from this, Figure 3 clearly demonstrates that many of the mini-meta-analyses show extremely low heterogeneity, with eight analyses showing heterogeneity ranging from 0 to 7%. Apparent attempts to denigrate their findings because they represented ‘only’ a 4–11% improvement are clinical judgments not appropriate in this type of analysis, and certainly not properly justified.

Among a number of other problems, acupuncture and electroacupuncture were inappropriately mixed in the meta-analyses involving the ‘no treatment’ control condition. The section should more accurately be labelled, ‘Electroacupuncture vs no treatment’, as only one in four of the included studies involved nonelectrified acupuncture and only at hours 1 and 2 after the start of treatment. Removal of the nonelectroacupuncture study would have resulted in even more impressive evidence for the effectiveness of electroacupuncture.

Despite their conclusion of no effects of acupuncture, we would argue that a more moderate, consistent and accurate statement summarizing their results would be as follows: ‘Effectiveness of acupuncture varied as a function of type of acupuncture and the comparison conditions.’ This is clearly demonstrated in Figure 3, but effectively ignored in the text summary and abstract summary. Furthermore, in all conditions, acupuncture tended to be most effective at reducing pain during the first hour of labour, but typically continued to be helpful, although less so, in the second and third hours. Electroacupuncture was clearly superior to no treatment, but was only marginally superior to placebo electroacupuncture. Standard (nonelectrified) acupuncture was slightly but not significantly superior to minimally penetrating acupuncture at off-meridian points. Finally, acupuncture was very effective at reducing the need for conventional analgesia.

In summary, based on the findings reviewed by Cho et al.,1 there is substantial evidence that acupuncture is effective in reducing labour pain (as well as other clinical indices of positive response). More research is needed to clarify the circumstances under which it is most likely to be effective.

Ancillary