There is an organized movement by governmental, academic and commercial interests to make evidence-based practice the standard of care in the United States. There is little proof that this model can be adapted to psychiatry. We examine the diagnostic system, the validity of the data from clinical trials and how these are applied to clinical practice. The discipline of psychiatry relies on imprecise and unstable diagnostic criteria. It divides psychiatric disorders into discrete categories based on discussion and consultations among designated experts in the field. The diagnostic system is based on consensus and not experimental evidence. In fact, psychiatric disorders are not discrete. High co-morbidities between disorders and the propensity of one condition to change into another makes the present diagnostic system extremely questionable. Outcomes of clinical trials are defined by fractional reductions in the number and severity of symptoms measured by rating scales and not remission of illness. The data obtained from clinical trials are flawed in design, execution and the selective reporting of outcomes. There is substantial evidence to indicate that both investigators and patients can distinguish between active treatment and placebo in double blind studies. In addition, negative outcomes are frequently not reported. Such evidence impacts not only on the specific study, used as evidence, but invalidates the value of meta analyses. Financial considerations lead to the inclusion of inappropriate subjects into studies and favor newer, patented treatments. When the conclusions derived from evidence-based psychiatry are applied to clinical practice they have little to offer and often produce poor treatment outcomes. In fact, when the data used to support the principles of evidence-based psychiatry are examined, they are unsound. The system itself is best considered an untested hypothesis. The diagnostic system, the manner in which data are gathered, and financial factors combine to produce a system that is misleading and indeed, dangerous.