Methodology for the scientific evaluation of complementary and alternative medicine

It has been suggested that CAM research should establish efficacy before examining mechanism. This paper shows that the efficacy-mechanism distinction is a false one, as any test of efficacy assumes a particular mechanism and is a test of the theory underlying that mechanism. The term RCT is currently used in medicine for two different sorts of study. The randomised controlled trial (RConT) requires an experimental manipulation that can 'control' for the mechanism under consideration, and therefore tests the efficacy of that mechanism. The randomised comparison trial (RComT) requires only an experimental manipulation creating a therapeutically relevant comparison, and tests the effectiveness of that therapy. The ability to achieve control coupled with an assumed implausibility of hidden moderating variables characterises drug therapy and some CAM therapies where the RConT can be used. However, other CAM researchers assume a variety of holistic mechanisms, where control is necessarily poor and the hypothesis of complex interactions suggest the existence of multiple moderators. In these cases other experimental (e.g. RComT), quasi-experimental or non-experimental designs are needed to evaluate therapeutic practice. Researchers from both communities should make explicit their underlying assumptions and the mechanisms they seek to evaluate when carrying out empirical studies. Research design needs to be appropriate for the mechanism under test. (C) 2003 Elsevier Science Ltd. All rights reserved.