Large, well-controlled, randomized clinical trials of new interventions in psychiatry, as in other fields, often yield contradictory results (1). When combined by metaanalysis and translated into treatment guidelines, they too often fail to improve recovery rates of individual psychiatric patients (2). What can be done to improve the transformation of efficacy research (mean group difference) into clinical effectiveness research that is meaningful to individual patients?

Some of the major problematic issues are that a) populations studied in trials are not representative of most patients in the community, b) protocols adhered to in experimental trials are not usually followed in general practice, and c) the expertise of  study investigators is not necessarily reflected in the skills or experience of the ordinary community of practitioners.

Clinical trials, by their very nature, are conducted in academic settings on selected samples of hand-picked subjects. The volunteers who choose to participate in trials consent to severe constraints that most patients in clinical practice would not readily tolerate. Research subjects agree to undergo a double-blind, randomized trial, which means that neither they nor their clinician know the treatment they are

Manuscript received May 2000, revised, and accepted January 2001.
¹Professor, Department of Psychiatry, University of Toronto; Staff Psychiatrist, Centre for Addiction and Mental Health, Toronto, Ontario.
Address for correspondence: Dr MV Seeman, Centre for Addiction and Mental Health, 250 College St, Toronto, ON  M5T 1R8
e-mail: mary.seeman@utoronto.ca

receiving. They do not know, until the end of the trial, whether they have been given a state-of-the-art intervention, a placebo, or a long-proven treatment that could just as well have been prescribed outside the confines of the trial. Subjects who consent are known to constitute a special breed (3,4). As a minimum, those who enter such trials need to be very trusting of the medical system. In mental health populations, especially among the more severely ill, this requirement of basic trust excludes a large segment of individuals for whom the intervention is intended. The exacting nature of the research enterprise also eliminates the apathetic and the unmotivated, another large proportion of mental health consumers. Subjects in clinical trials are expected to comply with the prescribed study regimen—although they do not always do so (5). In the community, it is generally accepted that over 40% of schizophrenia patients, for instance, will be noncompliant with doctors’ orders (6).

The differences between study subjects and patients from the general population necessarily impact on the generalizability of results (7). Frequently, there is an age limitation for trials so that the very young and the very old are not represented. Pregnant women, or even potentially pregnant women, are automatically excluded if the intervention is a drug. To be able to provide informed consent (8) and fill out complicated questionnaires, subjects must be fluent in the applicable language, which rules out the illiterate, the poorly educated, and most recent immigrants. Participants in many trials must not suffer comorbid conditions—medical conditions other than the one under study. In mental health studies, this means they must be physically healthy and have no mental disorder other than the one for which the intervention is intended.