The failure of placebo-controlled studies. ECNP Consensus Meeting, September 13, 1997, Vienna. European College of Neuropsychopharmacology.

In recent years an increasing number of clinical trials to test the efficacy of new potential treatments have failed to demonstrate a difference from placebo for either the new treatment or for an established reference drug. A rise in the response rate to placebo observed in a range of psychiatric disorders has not been paralleled by a rise in the response to drug and small effect sizes make it difficult to establish significant differences. A number of factors are thought to contribute to the rising placebo response or the smaller effect sizes. These include differences over time in the populations studied, changes in investigator behaviour, and failures of trial design. The inclusion of a greater number of patients with mild disorder or whose disorder has a fluctuating course is thought likely to increase the placebo response rates. Close attention needs to be paid to patient selection in terms of diagnosis, severity and absence of confounding comorbidity such as alcoholism, personality disorders or brief depression. The rising placebo response is associated with an increasing variability of placebo response seen in some centres. The ability of some centres to select appropriate patients for studies to demonstrate a separation of reference treatment and placebo and the inability of other centres suggests that a more careful selection of investigators is important. Selection should be based on their experience, their record from previous studies, and their aptitude for being trained. The inclusion of a reference treatment arm provides a useful means to judge the performance of individual centres. The exclusion of eccentric centres that fail to reach predetermined performance criteria, such as a failure to separate reference treatment from placebo, may be considered. Trial designs need to qualify adequately the study population and pay sufficient attention to diagnosis, minimum severity and comorbidity at entry. Greater care is needed in excluding concomitant overt or covert psychotherapy and in reducing the unnecessary therapeutic contact that has been increased unwittingly by some protocols. Identifying patients with prior stability of illness, with clear disability, and with a minimum severity at entry is likely to lower the placebo response substantially and increase the effect size and power of the study. The possible influence of comedication should also be considered. The inclusion of a placebo run in period is considered unhelpful and the use of better statistical techniques should be adopted to maximise the sensitivity of the study and increase the chances of testing efficacy.