Mental illness does not fit distinct phenotypes.
The idea that a diagnostic system can support drug development started with the transition from the Diagnostic and Statistical Manual of Mental Disorders (DSM) I-II to DSM III-IV. We rejected a system based on explanatory dynamics and adopted one based on phenotypic similarities. Our hope was that the new system would reduce biologic, genetic, and pathophysiologic variability. By narrowing phenotypes, we aspired to match biologic targets to pharmacologic solutions.
For a diagnostic system based on phenotype to yield a biologically explainable system, however, three assumptions must be valid. The first is that psychiatric illnesses are real elements and not convenience-driven, man-made categories.
Second, we assume we can categorize mental illness into distinct, well-defined phenotypes. We presume that conditions don't overlap, that there are real points of neutrality between schizophrenia, dissociative disorder, and bipolar disorder. Third, and most important, is the assumption that distinct, well-defined phenotypes are caused by distinct, well-defined phenomena.
Is it possible to categorize psychiatric illnesses into distinct phenotypes? I think not.
Phenotypic categorization uses nonobservational, unverifiable information. I interview the patient or the family. There is no way to verify what they tell me, and I am subjective in my interpretation. Even trained observers vary in how they score observable information. Psychiatric diagnosis is contextual and judgmental.
As for distinct phenotypes, most known risk factors apply to multiple disorders. Socioeconomic status, stress, immigration, and cannabis abuse are risk factors for schizophrenia, depression, substance abuse, and personality disorders.
The main benefit of switching to a phenotypic system was to facilitate communication among mental health professionals. What I call schizophrenia in the United States is what people call schizophrenia in Japan. That doesn't say anything about the validity of our terms.
Likewise, its reliance on phenotype does not necessarily aid drug development. Biologic markers reduce the biologic variability of the phenotype, so that clinical trials can recruit the most homogenetic cohort, objectively assess a drug's effects, and help clarify its mechanism of action.
The paradox of modern drug development is that trials provide evidence of efficacy and safety at usual doses in a general population. In clinical practice, we use much lower or higher doses to treat individuals who vary widely in their response to drug therapy.
Historically, the Food and Drug Administration approved psychotropic drugs for a diagnosis or a syndrome. It has revised its approach, however, so that drugs are indicated for well-defined aspects of a disease or syndrome if supported by valid trials. For example, olanzapine and ziprasidone IM were approved for agitation in schizophrenia and clozapine for suicidality in schizophrenia. The FDA also accepts claims for nonspecific symptoms that present across diseases. Pain and fever are the best examples.
We don't have to change the clinical classification system, but we should look differently at the classification of psychotropic drugs. I would probably begin by targeting what are called today endophenotypes. You have a better chance of succeeding if you target the biologic process of hallucination than if you target the social and biologic complexity in schizophrenia.
DR. DAVIDSON is director of psychiatry at Sheba Medical Center and chairman and professor of psychiatry at the Sackler School of Medicine in Tel Aviv.
MICHAEL DAVIDSON, M.D.
Targeting phenotypes has not led to new drugs. …