Drs Campbell and Jainer touch on an area of controversy in the design and conduct of psychiatric drug efficacy studies, arguing that the Declaration of Helsinki ‘clearly discards the use of placebo… when a “ proven” treatment exists’. We disagree with their interpretation of the Declaration on several grounds, and note the broad support for careful use of placebo in psychiatric trials (Temple & Ellenberg, 2000) (including the support of the multiple independent ethical review boards that approved the protocol for our study). There are abundant data that non-specific interventions can have marked beneficial effects, albeit on average less than active drugs. Non-drug therapies are often offered to patients as an alternative to drug therapies, and the absence of risk related to adverse drug effects can offset the potential for lesser efficacy. In our trial, both treatment groups had marked improvement from baseline. In this regard, placebo is not ‘no treatment’.

Drs Campbell and Jainer suggest drawing conclusions about drug efficacy based solely on comparisons of active agents. Unfortunately, in many trials a drug previously shown to be active is not superior to placebo despite adequate powering and the use of standard trial designs. Such trials are often referred to as ‘failed’ and in anxiety and depression are extremely common. A comparison of a new agent against a drug previously shown to be active without a placebo comparator is uninterpretable unless one agent is superior to the other. Concluding that a drug is efficacious without a placebo comparison can lead to an incorrect assumption of drug-specific effects if neither the investigational drug nor the active drug was, in that trial, any better than placebo would have been if included. Introducing a drug into therapeutic use on the basis of such a trial would expose patients to a compound with no greater benefit than placebo but all the risks of a pharmacological intervention (Temple & Ellenberg, 2000). Placebo is also critical in the assessment of safety, as it provides a base rate for determining which adverse events are truly related to the investigational drug. For these reasons, placebo-controlled trials are almost universally demanded by regulatory bodies to demonstrate efficacy for new pharmacological interventions.

Drs Campbell and Jainer also assert ‘no new treatments should be introduced into medicine unless they have been shown… to be superior to existing treatments… [or] cheaper and safer’. This absolute statement reflects several misconceptions and confounds the investigation of a drug with its introduction into general use. There is no general agreement about how to define or demonstrate equivalent or relative efficacy — precisely the reasons why most regulatory bodies will not consider relative efficacy claims in labelling. Furthermore, clinical trials provide information about group responses. Individual patients may not respond to or tolerate a particular drug, yet benefit from a different drug that is not, on average, more efficacious or safer than the first agent — it is in patients' interest to have several choices. For example, using Campbell and Jainer's procedure, the selective serotonin reuptake inhibitors, now proven to be safer and better tolerated than tricyclic antidepressants, might well not have been introduced into practice.

Finally, price is not an issue of trial design or science, but determined by the value that patients and purchasers put on a drug, based on evidence about the drug and experience with it (effectiveness as well as efficacy). Whether new drugs for panic or other psychiatric disorders should be ‘introduced into medicine’ and how they should be priced are decisions made on the basis of assessments of data about safety, efficacy and potential place in the therapeutic armementarium — decisions that cannot be made before the data are collected. Campbell and Jainer may feel that the results of this trial do not warrant further investigation of the use of fluoxetine for panic disorder, although we would disagree. It is, however, wrong to suggest that the trial as designed should not have been performed or published.