Antidepressant or antidepressant plus placebo effect?

Since the first observations of the therapeutic effect of imipramine, an extensive armamentarium of effective antidepressant medications has been developed, but the level of efficacy achieved appears to be considerably less today than might have been predicted from the discoveries of more than 40 years ago. Public opinion is currently suspicious as to the value of antidepressants in treating depressive illness. Khan and Brown’s (1) thoughtful elucidation of possible factors involved helps to clarify if there has been a change in the way studies to establish efficacy are conducted, a change in the patients coming forward for treatment, and whether we should change our expectations on treatment efficacy.

The rise in the response to placebo reported in clinical studies, which certainly makes the demonstration of efficacy more difficult, has been attributed by some to pressure from the pharmaceutical industry to find a rewarding outcome in efficacy studies. The comparison of Khan et al (2) of the symptom reduction data reported in non-pharmaceutical industry studies with the data submitted by pharmaceutical companies to the U.S. Food and Drug Administration (FDA) finds few substantive differences. Differences in assessments between the two datasets appear to be related to size rather than direction, and when the non-pharmaceutical investigators were unaware of the study design the assessments came even closer to those of the FDA data. Industry-sponsored studies submitted for regulatory approval are designed in discussion with the regulatory authorities and have to meet their strict criteria; the similarity of the data from the two sources is reassuring.

Clinicians recognize the powerful contribution of placebo response in depression as in many other conditions. The task of distinguishing the pharmacological from the placebo response has become more difficult over the years, as the proliferation in regulatory requirements to address efficacy, safety issues, particular patient groups, integrity of the study population, etc. has led to an increase in complexity of study design. The assessments now required to address all these aspects of treatment take considerable amounts of time to complete, and the therapeutic benefit of time spent with the patient is well recognized: an increase in the placebo response and a decrease in the separation of active medication from placebo is to be expected. We should remember that efficacy studies in psychiatry use a combination treatment paradigm: putative antidepressant plus placebo effect vs. placebo.

Khan and Brown (1) rightly point to the need for a reappraisal of the response measures used in determining efficacy. The bar to declaring efficacy differs between the U.S. and the European Union. In the European Union, efficacy has to be not merely established but also shown to be clinically relevant and found in placebo-controlled studies to persist in long-term treatment. Long-term treatment studies tend to be more consistently significant than short-term studies in demonstrating efficacy. Khan and Brown refer to their meta-analysis of response in continuation treatment (3), on the basis of which they consider that response on placebo is persistent. However, the methodology was flawed in that discontinuations from causes such as administrative dropouts or dropouts from side effects were censored and not taken into account. In a more specific study, response to placebo was not found to be persistent (4).

It has to be remembered that the efficacy of antidepressants was originally established in depressed patients, often hospitalized, with clear and relatively severe symptoms. As could be expected, on the basis of this success, the use of those and subsequent antidepressants has been extended to a much wider patient population suffering from a broader range of severity of illness. Antidepressants appear to be less effective or ineffective in patients with mild depression, and the global assessment of the efficacy of antidepressants would be diminished by the inclusion of this patient group. There is also a risk that, instead of identifying those patients who would best benefit from antidepressant treatment, under the pressure of the therapeutic imperative, some members of the “worried well” group receive the diagnosis of depressive illness rather than appropriate reassurance. This group has a high placebo response and should not be included in efficacy studies. The possible inclusion of large numbers of the worried well or of mild depression in U.S. clinical efficacy studies may have contributed to the increased difficulty in demonstrating efficacy of an antidepressant compared to placebo over the years. We would do well to heed the warning of the past chairman of the DSM-IV task force, in relation to the developments in DSM-5, that mild major depression is not major, not depression and not a disorder (5).

In the present climate, where both regulatory authorities under pressure from politicians and institutional review bodies require extra scales and restrictions, it is difficult to design a study which will allow the drug to show efficacy. Candidate antidepressants now need to be more effective in order to separate from placebo in a population with a higher placebo response. Studies carried out in an assay sensitive population cannot be compared with those carried out in a population with a high placebo response rate, and comparison of earlier and current studies is invalid. We should emphasize the success of the effect size achieved in current studies of antidepressants (0.31-0.33), which lies in the same range as many accepted treatments in general medicine.