DSM categories
The DSM categorizes illnesses into discrete entities, using 5 axes to assign severity and plausible cause. Axis I is a description of the presumed clinical disorder. Many conditions share comorbidities of 70% to 90% (eg, ADD, oppositional defiant disorder, conduct disorder, and antisocial personality disorder), thus blurring diagnostic boundaries. Conditions change from one form to another depending on when the patient is seen (eg, OCD, generalized anxiety disorder, panic disorder, atypical affective disorder).

Because physicians are limited to a few tests to define a disorder (serology for syphilis, blood glucose for diabetes, and neuroimaging for dementias come to mind), the diagnosis takes on the hallmarks of a Rorschach projective test, based on the judgment and experience of the observer.

Conversely, when psychiatrists use a system based on a checklist of symptoms, disorders that are etiologically related are viewed as different. Tucker¹⁹ gives the example of Huntington disease, a condition in which the cause is well defined but the expressions of the illness meet the criteria for dementia, psychosis, depression, and antisocial personality disorder.

Data validity
EBM places heavy emphasis on the results of large, multisite, double-blind studies. This being the case, it is important to know where and how these studies are conducted.

For the most part, studies are commissioned by industry consultants and are conducted either at institutions that do clinical trials for profit or at academic institutions where such studies are considered of low merit and the actual data collection is done by nurses, social workers, or narrowly trained raters with no clinical experience. Participants are often enrolled from databases of patients who have participated in previous studies or who are "borrowed" from clinics. It is not uncommon for patients to participate in multiple studies. Participants become "professional" research patients who take part in several studies to collect the stipend. These populations are not representative of the patients in clinical psychiatric practices.

Assuming the best of circumstances--that is, the participants represent a cross-section of patients with a defined psychiatric illness--the validity of double-blind, placebo-controlled evidence must still be assessed. Ideally, neither the patient nor the investigator should be able to distinguish the active substance from placebo; however, numerous experiments show that the "blind" condition is easily broken.

In studies of antidepressants, physicians correctly distinguished the active drug in 73% to 89% of cases, and patients correctly distinguished the treatments 64% to 75% of the time. In crossover studies, physicians successfully identified the active drug 100% of the time and patients were correct 93% of the time.²⁰

The failure to report negative findings is most egregious in assessing efficacy. In addition to the reluctance of editors of journals to publish negative findings, industry project managers "seal" negative results and neither file the data with the FDA nor allow the investigators to report the results.²¹ For every study of a psychotropic drug that reports positive results, there are often 4 to 8 unpublished studies that fail to show superiority to placebo.²² This "file-drawer phenomenon" explains, in part, why drug company-supported research is 4 times as likely to produce a positive response as studies from other sources.²³ If the results of the studies that did not show positive results were reported, it is conceivable that they would alter the conclusions of the meta-analysis.⁷

Newer agents that may be patented are preferentially studied, diluting evidence that supports the effectiveness of older treatments and encouraging the wider use of newer and more expensive treatments.²⁴

Dosage problem
Clinical trials underestimate the effective dose of medications; they may also overestimate, but this is unusual. The "evidence" is then used by insurance companies and government agencies to accept or reject claims for payment for the higher doses often needed in clinical practice, depriving patients of adequate treatment. The Table illustrates the differences between the originally suggested doses of atypical antipsychotics and what are now accepted as the actual clinical doses.²⁵

Application to clinical practice
It takes considerable time for a clinical observation to be validated by a large-scale study.^2,26 In practice, a medication is first approved for marketing as an effective treatment for one entity and then clinical observations reveal that it has other uses. If we assume that it will generally take up to 4 years for a clinical trial to be designed and executed, EBM will always be several years behind clinical practice. If the standard of clinical practice is limited to evidence-based studies, practitioners will be reluctant to attempt novel treatments or to vary dosing regimens. The evaluation process reduces innovation.

EBM is favored by insurance companies and national health services.^27,28 This method is, by its very nature, conservative, and it promotes the use of lower doses of medication and discourages polypharmacy. Such restrictions and standards reduce costs, but they also reduce effective treatments. The problem is compounded when we consider that EBM criteria are commonly offered as the basis for malpractice litigation.

Conclusion
Critics of present practices call for narrow diagnostic criteria in patient selection, better monitoring of the patients selected for studies, longer periods of study, and better criteria for outcome. Some relatively simple measures that would be helpful are monitoring the participants by their social security numbers (to minimize the use of "professional patients"), tracking placebo response percentages, and ensuring better reporting of adverse events. The common practice of competitive enrollment, which encourages research sites to enroll as many patients as possible, should be discouraged because it magnifies the contributions made by specific research installations and skews results even further.

EBP is a system that is based on unreliable data. Its unreliability results from an imprecise and poorly founded diagnostic system, inaccurate data collection, and obfuscation of experience by the pharmaceutical industry and compliant academic leaders. Whatever the reasons, the drive to adopt EBP as a standard of practice is best discouraged.