What Kind of Control Condition should we use in Trials of Acupuncture?

Abstract

Disappointingly little has been achieved by literally hundreds of attempts to evaluate acupuncture for chronic pain. Major methodological flaws are apparent in the vast majority of studies. Controlled studies have shown positive findings for low back pain, and equivocal results for migraine and asthma. Nevertheless larger scale studies are warranted for all these disorders, though other types of musculo-skeletal pain, tension headache and arthritis are also possible candidates.Controlled trials of any treatment have become an immensely difficult and technical undertaking. They are expensive, time-consuming and ideally require the collaboration between practitioners and researchers and consultation with a statistician. It is not really possible for acupuncturists in private practice to mount such trials, and it is very difficult for a professional association or college. However, it is clear that there is no longer any point in conducting small, preliminary studies of acupuncture treatment. There are dozens of such studies, with some encouraging findings. The only way acupuncture can gain full acceptance as a valid form of treatment is through good controlled trials that are large enough to answer the questions they pose.Ter Riet's list of criteria as a good starting point for anyone designing a trial of acupuncture. Specific points I would emphasize after reviewing the existing research on acupuncture are:1. Trials should be single blind; it is not feasible to conduct double-blind trials. Some trials are nevertheless incorrectly described as double blind.2. A range of outcome measures should be used, preferably with some independent assessment. An adequate follow-up is essential.3. Considerable care needs to be taken in the choice of control group, especially with placebo controls. For a placebo I suggest a form of acupuncture treatment that is designed to have minimal effects. It will be the option that is the closest match to the true treatment and avoids the difficulties inherent in randomizing patients to a non-acupuncture treatment.4. It is very useful to check the adequacy of any control treatment with a measure of credibility, or similar assessment, as the choice of control is frequently a matter for criticism.5. Trials have generally been too small to permit firm conclusions. Ter Riet (Ter Riet, Kleijnen et al. 1990) implies that 50 patients per group are needed. This may not be necessary. However preliminary calculations of the necessary size for a reasonable power need to be carried out.