Editorial

In this issue we publish something rarely seen in medical journals – a ‘failed’ or ‘negative’ randomised, controlled trial; a trial in which neither the test drug nor the active comparator separated statistically from placebo. Steve Potkin and his team compared three doses of lurasidone with haloperidol at 10 mg a day and placebo in a six week study of acute exacerbation of schizophrenia. They found all treatment arms equally effective with only minor, non-statistical, numerical advantage over placebo observed for the highest dose of lurasidone and for haloperidol. The trial was sponsored by the manufacturers of lurasidone, Sunovion Pharmaceuticals.

In recent years, particularly since the wholesale adoption of the concept of evidence-based medicine, many have called for the publication of all trials whatever their results. With meta-analysis seen as the highest form of evidence, the inclusion of all trials ever concluded in a particular therapeutic area is critical to interpretation of outcomes and so to promoting the use of the most efficacious treatments. However, so-called publication bias blights almost all meta-analysis because positive studies (those supporting the sponsor’s drug) are much more likely to appear in print and so be included in meta-analysis. Manufacturers are often blamed for not submitting the results of failed studies for publication, their vested interest being to present their product in the best light. This certainly happens, but another factor is also important: the reluctance of journal editors to publish these trials, which they see as uninteresting and, more importantly, unlikely to enhance their journal’s impact factor. In contrast, we are proud to publish this study in our journal to promote the concept of full declaration of trial outcomes and as an example to others.

How does one interpret a study in which neither haloperidol nor lurasidone, both effective antipsychotics, fail to show superiority over placebo? We should first appreciate that in many mental health conditions placebo is a powerful treatment. Subjects receive far more attention in trials than in clinical practice and the number of ‘visits’ for assessment is closely correlated with outcome (the more the better). Expectation of improvement also plays a key part, especially in trials like this one where the likelihood of being randomised to placebo is very small (something which affects assessors, too). In some countries, trial subjects receive financial rewards for their participation and this will have a major effect on their perceptions of treatment alongside perhaps a desire to ‘please’ their assessors. Moreover, assessments are more subjective than we would like and this leads to phenomena such as ‘inflation of baseline severity’ where assessors unconsciously or otherwise overestimate a subject’s symptom severity so that they may be included in a trial. At the subsequent assessment all subjects will be scored as being improved even if their condition has not changed, the imperative to inflate severity having been reversed.

No sensible person would conclude on reading the report of this study that haloperidol and lurasidone are ineffective. Nonetheless, these results deserve publication so that the problems with clinical trial methods in mental health can be fully appreciated and, most crucially, so that the published database for these drugs tends towards completeness.