In the Hippocratic tradition, clinicians should treat diseases, not symptoms. Diagnosis thus becomes extremely important, which entails careful attention to be as scientific as possible in our diagnostic system. As discussed elsewhere, the DSM approach has failed in this regard for the past three decades. One cannot, therefore, solely rely on DSM diagnoses in the practice of clinical psychopharmacology. A more scientific approach is needed.
The solution is scientific research on clinical diagnoses, an approach that is based on a century of medical research, and has been well established and accepted psychiatric research for the past half-century. This approach involves four classic diagnostic validators.
The key insight as to the importance of diagnostic validators is the fact that clinical symptoms by themselves are inadequate, insufficient, and misleading for medical diagnosis. This is an obvious fact, and widely accepted in medical research and practice. Pain in the chest is insufficient to diagnose coronary artery disease, since it can also be caused by musculoskeletal problems of the ribs or by gastrointestinal problems, among other possibilities. Over a century ago, in the late 19th century, clinicians like William Osler and others showed that clinical diagnosis by means of signs and symptoms at the bedside needed to be validated by an independent source of evidence. The best source, they found, was pathology at autopsy. If a patient had yellow skin and a fever, it was determined that that clinical syndrome reflected hepatitis by microscopic examination of the liver at autopsy. Some patients might have yellow skin but no fever, and autopsy would show no abnormality of the liver. This information could then be taken back to the bedside, and used to draw clinical judgments about the relevance of yellow skin as a symptom to diagnoses of hepatitis versus other possible illnesses. Symptoms alone could not differentiate different illnesses. Pathology was the gold standard. This was called the classic clinical pathological approach to medical diagnosis.
In psychiatry there is no gold standard. A century of research was unable to associate abnormalities in the brain at autopsy with the major psychiatric illnesses of schizophrenia and manic depressive illness. There was a need for an alternative approach, which is how the concept of multiple diagnostic validators was developed.
Let’s revisit the central insight behind the need for diagnostic validators besides symptoms:
A patient has pneumonia with cough and headache. Another patient has pneumonia without a cough and headache. Are these two different “disorders”? Most clinicians would say no, because the symptoms of cough and headache do not represent a different disease.
A patient has depression with mania. Another has depression without mania. Are these two different “disorders”? These days most clinicians would say yes, because that is what DSM has taught. Mania is a different set of symptoms than depression, one might argue. But cough and headache are symptoms, which clinicians agree do not represent a different disease.
That is the intuition behind the concept of diagnostic validators. It is not enough to say that symptoms differ and thus we have different “disorders”. One has to show that those different symptoms represent some kind of different conditions or diseases.
How? Not by simply referring back to the different symptoms: that would be tautologous. There has to be some different line of evidence, separate from symptoms, that represents a different illness. In the case of pneumonia, clinicians have access to pathology: tests can show evidence of inflammation in the lung, whether or not you have a cough and fever. So those symptoms do not represent a different disease. In the case of psychiatry, there is no access to pathology (usually). So what should the independent lines of evidence be?
This is where a classic article from 1970 revolutionized psychiatry. Eli Robins, who had trained in Boston, was chairman at the Washington University in St Louis. In that era, the major US cities were dominated by psychoanalytic thinking. Robins was influenced by Emil Kraepelin, the late 19th century German psychiatrist who taught that “diagnosis is prognosis”, that the course of illness tells you which symptoms represent different diseases. Robins left Boston to go to the smaller city of St Louis, and from there, he trained a series of researchers who produced the change in US psychiatry which led to the third edition of DSM (DSM-III) in 1980. With his colleague Samuel Guze, Robins articulated four other diagnostic validators that, along with symptoms, should be used to identify if groups of patients differ from each other enough to justify seeing them as having different diagnoses (their article focused on schizophrenia, but they later applied these principles to all diagnoses).
Those validators are as follows:
2. Course of illness
4. Treatment effects
5. Laboratory tests or biological markers
The most important validator is course of illness, Kraepelin’s key criterion. Some conditions are chronic, and symptoms are present all the time (like schizophrenia); others are episodic, with symptoms coming and going (like manic-depression). The next most important validator is genetics: if diagnoses are genetic, you’ll find evidence in family members. Next are laboratory tests or biological markers, which are useful in research more than in clinical practice, but can be useful clinically sometimes, as in white matter abnormalities on MRI of the brain in the diagnosis of vascular depression. Robins and Guze also referred to “delimitation from other disorders”, which meant that symptoms were specific to one condition rather than another. This is not always the case, since many symptoms, like anxiety, can occur in many conditions. Since that classic study, instead of delimitation, the diagnostic validator of treatment effects has been used, although it should be used cautiously, since many drugs are nonspecific in effect, and some, like amphetamines, are even effective in normal individuals. Treatment effects can also be seen as a proxy for biological markers, but only if treatment effects are specific to an illness (like antidepressant-induced mania).
Unfortunately, these diagnostic validators have been suppressed by the evolution of DSM. Originally, these diagnostic validators were the basis for scientific justification for diagnoses in psychiatric research, leading to the original Research Diagnostic Criteria (RDC) that Robins’ St. Louis group created. The RDC identified about two dozen scientifically valid diagnoses. DSM-III started with those diagnoses, and added about 270 others. In almost all cases, although the other diagnostic validators were used to justify diagnoses, only symptoms were used in the DSM criteria definitions (an exception is schizophrenia, where there is a course criterion of 6 months or longer for psychosis). Now we have about 400 diagnoses in DSM-5, and clinicians are used to only looking at symptoms for definitions. This leads to the perennial arguments: Is the attentional problem ADD or bipolar disorder? Is the anxiety part of major depressive disorder or generalized anxiety disorder? Is the sexual impulsivity mania or a paraphilia or borderline personality? These debates will never end as long as they are conducted on the single dimension of symptoms.
This classic paper reminds us that symptoms only go so far: like cough and fever, clinicians need to look elsewhere to know which symptoms matter diagnostically and which do not. They need to look to course of illness, genetics, and treatment effects. Clinicians should learn to use these four modified diagnostic validators in clinical practice, as will be emphasized in the chapters that follow.
Applications to clinical practice
The application of these ideas to clinical practice would entail paying much less attention to symptoms in the context of clinical diagnosis, and instead obtaining information from family history and course of illness especially, and sometimes treatment effects, to make a diagnosis more probable than others. Symptoms are the main clinical features used for diagnostic purposes by the DSM system. Even though the other diagnostic validators-like genetics and course of illness-are used sometimes to support the diagnostic concepts included in DSM, those same diagnostic validators are not allowed into DSM criteria sets. For instance, sexual abuse is a clear validator of diagnosis for borderline personality, but this course criterion is not allowed in the definition of DSM criteria for borderline personality disorder. The rationale given by DSM leadership for this approach is that they want to exclude any claims to etiology, but unfortunately this approach can lead to overdiagnosis of borderline personality in some individuals who do not have this important predictor of the diagnosis. Many other examples could be given.
The upshot of this discussion is that clinicians should look into course of illness much more carefully, as well as family history, and include that information in addition to symptoms, equally or even more importantly, for diagnostic purposes. Some examples will be given here, and will be repeated in the following chapters.
A typical scenario is a patient who presents with the clinical syndrome of a full depressive episode, and does not have any prior manic symptoms that meet full definitions for hypomanic or manic episodes. The DSM diagnosis would be “major depressive disorder.” The patient would have a mother with diagnosed and confirmed bipolar illness type I. In the DSM system, this family history will be ignored completely for diagnostic purposes. In this scientific non-DSM-based approach that incorporates all diagnostic validators, the family history would indicate that the patient’s affective illness is related to bipolar illness, and not different from it. The technical diagnosis of bipolar illness using DSM definitions could not be made, but the DSM-based diagnosis of MDD, as different from bipolar illness, also would be rejected. If this patient also had an early age of onset of the first depressive episode at age 15, with multiple brief episodes lasting three months or less, the bipolar-like course of illness would be confirmed as more relevant to diagnosis and treatment, rather than the opposite approach of ignoring course and genetics entirely, as the DSM system would do in insisting on the diagnosis of MDD. In the non-DSM approach, this patient would be treated with agents that often are used for bipolar illness, like second messenger modifiers or dopamine blockers, and would not be limited to treatment with standard monoamine agonist antidepressants, as would be the practice when focusing on the DSM diagnostic label of MDD.
Another scenario is a patient who presents with marked inattention and executive dysfunction, and is diagnosed with adult ADD. The patient denies past depressive or manic episodes or severe anxiety states. The course of illness indicates that the patient reports having had similar symptoms in childhood, although he never received clinical tension or diagnosis at that time. The patient is high functioning as a lawyer, is successful in his life, but feels that his cognitive symptoms keep him from functioning even better. Family history identifies schizophrenia in his uncle. Since ADD is not genetically associated with schizophrenia, his family history would throw major doubt on the diagnosis of ADD. In fact, if the underlying biology of the patient’s cognition is genetically related to schizophrenia, treatment with amphetamines would only worsen that underlying biology, and could worsen some of his clinical symptoms. Further investigation of the patient’s history will be warranted, and if there was evidence of paranoid features, for instance, other diagnostic possibilities will be raised, such as schizotypal personality, which is genetically associated with schizophrenia.
A third case scenario could involve a patient in her mid-30s with multiple mood swings and unstable interpersonal relationships, diagnosed with borderline personality. Assessment of the course of illness found that these symptoms were present for the past three years, but had not been present for five years before that time. Hence, the course of illness is not consistent with the long-term chronicity associated with the concept of personality disorders. Family history identifies a sister who committed suicide, and an uncle who was institutionalized with the diagnosis of manic-depressive illness in the 1950s. This family history is consistent with affective illness, not borderline personality. The course of illness also is not consistent with borderline personality. This evaluation would make the likelihood of affective illness more probable than borderline personality, once the other diagnostic validators are taken into account.
In summary, clinical diagnosis cannot be, and should not be, limited to symptoms primarily, as is common practice today. The DSM system encourages this emphasis on symptoms mainly, which leads to never-ending controversies. Many of these clinical dilemmas can be resolved by taking other diagnostic validators seriously, especially family genetics and course of illness.
Decker HS. The making of DSM-III: A diagnostic manual’s conquest of Americna psychiatry. Oxford: Oxford University Press, 2013.
Ghaemi SN. Toward a Hippocratic psychopharmacology. Can J Psychiatry. 2008;53(3):189-96.
Shorter E. Before Prozac: The troubled history of mood disorders. Oxford: Oxford University Press, 2003.
Wootton D. Bad medicine: Doctors doing harm since Hippocrates. Oxford: Oxford University Press, 2007.