Why Evidence-Based Medicine Cannot Be Applied to Psychiatry
Why Evidence-Based Medicine Cannot Be Applied to Psychiatry
Evidence-based medicine (EBM) is rapidly becoming the norm. It is taught in medical schools and is encouraged by both government agencies and insurance plan providers. Yet, there is little proof that this model can be adapted to fit psychiatry.
EBM supposedly allows the clinician to offer the most effective treatment for each patient.1,2 This goal is laudable, but the model is not appropriate for psychiatry because precise and stable diagnostic criteria are lacking in our specialty. Treatment outcomes in psychiatry are not defined by remission or cure. Instead, fractional reductions in the number and severity of symptoms are accepted, as measured by rating scale scores. Evidence-based psychiatry (EBP) is an untested hypothesis; for this theory to be either useful or valid, 3 basic assumptions must be examined.
• Is the diagnostic system valid?
• Are the data from clinical trials assessing efficacy and safety sound?
• Are the conclusions in a form that can be applied in clinical practice?
The paradigms that define EBM (and EBP) are based on data in published clinical studies. Each study is assessed according to the methods used to collect the data and a value placed (by the reviewer) on its quality. EBM uses 3 types of evidence3:
• Grade A: Randomized clinical trials, homogeneous populations, placebo-controlled.
• Grade B: Randomized clinical trials, heterogeneous populations, not placebo-controlled.
• Grade C: Observational studies, case collections, open clinical trials.
The strength of the evidence decreases from Grade A to Grade C, with more weight given to Grade A than to Grade C studies.
The diagnostic system
The DSM represents diagnostic groupings developed through discussion and consultation. These groupings are not based on experimental evidence. The manner in which the DSM creates diagnoses assumes that psychiatric illnesses can be divided into separate categories and that each illness is unique. The process is circular. It begins by assuming that discrete categories exist and produces a document that divides psychiatric illness into discrete categories.
The illnesses in the DSM are delineated by phenotypic features, with a contribution from the patient's recall of the course of the illness. The separation of classes based on these criteria is, by its nature, imprecise. Although these criteria are intended to separate clinical entities, their descriptions are overlapping. The judgments introduced by the clinician's need to decide which of several conditions best meets the diagnostic criteria are subjective, putting the system in doubt. It is not surprising that overlaps are common.
Major psychiatric disorders overlap
The committee-driven DSM classification divides psychiatric diseases into discrete entities based on cross-sectional symptoms and signs. These disorders exist as syndromes and not as specific illnesses. Psychosis, for example, may result from drug toxicity, neurological illness, trauma, or as a feature of delusional depression and isolated delusional states. The genetic predispositions for schizophrenia and affective illness overlap.4,5 Kendall6 has demonstrated that most patients have characteristics of both groups and that our diagnostic concepts are based on the extremes of what is better visualized as a continuum.
EBP relies on categories that call for specified algorithms for treatment.7 However, in clinical practice the emphasis is not on treating syndromes or diseases but on applying empirically derived prescriptions for symptoms and symptom complexes. Psychiatry lacks antischizophrenia, antidepression, or antianxiety disorder medications. What exist are medications that symptomatically treat psychosis, depressed mood, and anxiety. The severity and associations of the symptoms vary greatly depending on genetic traits, environmental influences, and duration. The genetic traits are particularly subtle; the substitution of a single nucleic acid markedly alters symptoms.8-10 Variations in genetic polymorphism influence phenotypic presentations of illnesses over a spectrum of clinical syndromes. It is difficult to reconcile these observations with the current diagnostic system's separation of these conditions into distinct categories and subcategories.
Several conditions exhibit such high rates of comorbidity that one must be skeptical of the idea that we are dealing with discrete entities. The overlap between attention-deficit disorder (ADD), oppositional defiant disorder, and conduct disorder is an example.11-13 Separating these syndromes into distinct entities obscures the possibility that they may be expressions of a single genotype. This is also the case with obsessive-compulsive disorder (OCD), generalized anxiety disorder, panic disorder, and major depressive disorder, in which patients may display symptoms of each condition, with the dominant symptom changing with time and circumstance.14-17
This relationship frequently causes problems when the nosological diagnosis obscures the possibility of a common biological substrate. A common example of this occurs in patients with OCD in whom manic symp- toms develop when they are taking antidepressants.18
If a nosological approach were reliable, we would expect greater homogeneity in treatment response. That is, if depressive mood disorders were a single entity, we would expect that a single treatment agent would be effective in all or almost all patients who receive this diagnosis. Such uniformity of diagnosis and treatment response is expected, for example, in bacterial infections and in diabetes. However, this is not the case in psychiatric disorders. Lacking effective predictors or tests, the clinician searches for clues for treatment selection in personal and family history or engages in multiple drug trials, augmentation strategies, and polypharmacy.
Syndromes that are separated by DSM criteria as single entities often respond to the same pharmacological therapies. OCD, panic disorder, generalized anxiety disorder, and depressive disorder respond to SSRI agents. Schizophrenia, bipolar disorder, tox-ic psychosis, and major depressive disorder with psychosis respond to atypical antipsychotic agents. Are we to assume that the treatment agents have very broad effects, and that they are effective for different disorders? Or should we assume that the different disorders have a common biological underpinning that responds to the singular effects of specific medications? Either assumption casts doubt on the present diagnostic schema.
Similarly, diagnostic categories are added or deleted in each iteration of the classification based on fashion and political correctness. Examples include the rejection of homosexuality, unipolar mania, and melancholia, and the addition of caffeine and nicotine addiction. When fashion, rather than scientific evidence, dictates diagnoses, the entire system should be questioned.