Professor Edzard Ernst examines a new systematic review on the use of acupuncture for back pain, knee pain and headache.
Acupuncture is often used for pain management. But does it help? I know many GPs who are confused by the often contradictory evidence. Perhaps a new systematic review brings clarity?
This analysis (1) aimed to 'synthesize the evidence gathered from systematic reviews on the pooled data of high-quality RCTs comparing acupuncture to sham acupuncture for chronic pain'. Essentially the authors looked at 8 meta-analyses of acupuncture for three indications: back pain, knee pain or headache. They found that acupuncture was consistently better than sham. Therefore they conclude that there is now 'sufficient evidence to consider acupuncture to be more than a placebo' and argue that researchers should 'shift research priorities from asking the placebo-related question to asking more practical questions about whether the overall benefit is clinically meaningful and cost-effective'.
I beg to differ. First, the authors selected only three chronic pain conditions. I don't see how this permits them to extrapolate to chronic pain or even all ailments treated with acupuncture. Between 2000 and 2009, more than 50 systematic reviews of acupuncture for pain were published. To select just 8 seems very odd indeed.
Second, the meta-analyses were based on more or less the same, often flawed, primary studies. Thus aggregating them in this way may simply have amplified false positive results.
Third, the authors clearly think that sham-controlled trials serve a purpose (this is why they did this review of sham-controlled trials in the first place). But then they seem to say 'let's quickly stop our research here, before anyone comes up with negative findings that might destroy our convenient results'.
Having done my fair share of acupuncture research (and being partly responsible for some of the meta-analyses included in this review), I think that we urgently need more and better sham controlled trials. Why? Because the better we are able to control for placebo-effects in acupuncture trials – e.g. by using non-penetrating sham devices – the smaller the effect size. The authors of the review already had to admit that the pooled effects are small. Once we use more credible sham procedures, the effect might disappear altogether.
And that, I think, would be well worth knowing.
Professor Edzard Ernst
