On the evaluation of the clinical effects of acupuncture

This paper reviews some of the recent randomised trials on acupuncture published in the literature, with particular reference to the definition of placebo (control), sham acupuncture and real acupuncture. Response rates of 30, 50 and 70% of placebo, sham and real acupuncture respectively, are suggested for the groups of patients studied. The statistical methodology necessary in the design of acupuncture trials, in order to detect differences of this order of magnitude, is discussed. If the postulated response rate had been envisaged at the onset of the trials here reviewed, then the majority of these trials have very low power at a conventional 5% level of significance. It is emphasised that one cannot necessarily conclude from trials which produce statistically non-significant results that acupuncture (when compared with placebo for example) is ineffective. The paper suggests that meaningful comparisons between alternative treatments may be made by using 'pain free intervals' with subsequent analysis using life table techniques. Tables are provided for determining appropriate patient numbers.