Editor—In their paper on the use of placebos in clinical practice, Nitzan and Lichtenberg say that they were unable to find more than one other study on the use of placebos in a clinical context.1,2 Eight similar studies are indexed in PubMed (see bmj.com).
In the accompanying editorial, Spiegel points out that the Cochrane review on the placebo effect probably underestimated the placebo effects of treatments.3,4 Spiegel gave some methodological explanations for this underestimation but did not mention a much more important reason.
A problem with the Cochrane review, as mentioned in three letters by Lilford and Braunholtz, Kuppers, and Shrier,5 is that the included studies were done in a setting completely different from the situation in clinical practice. The included studies are three armed studies, in which patients are randomly allocated to a supposedly active treatment, to a placebo, or to no treatment. The placebo effect is then defined as the difference in effect in the patients receiving placebo compared with those receiving no treatment.
Obviously, neither the patient nor the doctor in such a trial will have any substantial belief in the (placebo) treatment or consider it particularly meaningful. This situation is completely different from clinical practice, where the patient and the doctor believe in the therapeutic powers of a treatment that they probably consider meaningful. The difference between randomised trials and clinical practice is always a problem, but much more so in the study of placebos and related phenomena. This is one of the main reasons why the Cochrane review does not exclude the existence of strong placebo effects in some situations of clinical practice.