The multiple outcomes bias in antidepressants research

Abstract

Despite the widespread use of antidepressant medication, there are no signs that the burden of depression and suicide is decreasing in the industrialised world. This is generating mounting scepticism on the effectiveness of this class of drugs as an approach for the treatment of mood disorders. These doubts are also fuelled by the increasing awareness that the literature on antidepressants is fundamentally flawed and under the control of the pharmaceutical companies. This article describes systematically for the first time what is probably the most insidious and misleading of the biases that affect this area of research: the "multiple outcomes bias". Most trials on the effectiveness of antidepressants, instead of first establishing a hypothesis and then trying to demonstrate it, following the scientific method, start instead "data mining", without a clear hypothesis, and then select for publication, amongst a multitude of outcomes, only the ones that favour the antidepressant drug, ignoring the others. This method has obviously no scientific validity and is very misleading, allowing the manipulation of the data without any overt fraudulent action. There is the need to generate new research, independently funded and with clear hypotheses established "a priori ". What is at stake is not only the appraisal of the balance between benefits and potential damage to the patients when using this class of medications, after the realisation that they are not as harmless as believed. It is also to establish whether the research on antidepressant medication has gone on a "wild goose chase" over the last half century, concentrating almost exclusively on molecules that modify the monoaminergic transmission at synaptic level and virtually ignoring any other avenue.