July 2006, Vol. XXIII, No. 8
Psychiatric diagnosis is currently based on a system of categories, much like the Linnaean classification in biology. However, a seemingly opposite diagnostic system is emerging, usually dubbed the “dimensional view,” in which related psychiatric conditions such as major depressive disorder (MDD) and bipolar disorder (BD) are seen as end points of a spectrum.
With this expansion of the concept of BD to include intermediate forms (ie, manifestations between clearly unipolar and clearly bipolar cases), concerns have arisen about the potential for overdiagnosis of BD1 and the blurring of our understanding.2 However, diagnostic error thus far has been skewed in the direction of underdiagnosis.3
The concept of overdiagnosis or underdiagnosis presumes some “natural types” that we diagnosticians are correctly or incorrectly classifying. Indeed, the term “classify” implies that we are looking at native classes. This is, for example, the process used by an ornithologist to identify a particular bird: scientists use visible traits such as size, color, beak shape, wing bars, eye rings, and other field signs. They may also observe preferences such as habitat (ground or high trees), interactions (flocking or pairing), diet (seeds, worms, or fish), and even behavior over time (eg, migration or wintering-over).
The DSM system is like that of the ornithologist. The purpose of the diagnostic interview is to gather observations, for which the DSM serves as a “field guide,” organizing the observed findings by category of illness. But just as in ornithology, over time an experienced clinician relies less and less on the specific DSM rules and progressively more on pattern recognition. A good bird watcher distinguishes a woodpecker from a hawk not so much by specific field signs but from an overall gestalt—rapid synthesis of a whole complex of findings (eg, location in the tree, position on the branch, size, shape, color, behavior, call). Likewise, an experienced clinician might strongly suspect that a patient has BD based on observations on an inpatient unit, even before an interview.
Unlike birding, however, experience gives clinicians the ability to recognize intermediate cases that blur the boundaries of the classification system. Indeed, this lack of discrete boundaries has been studied quantitatively in at least 2 research settings, both of which showed a continuum of symptoms in patients with DSM-diagnosed MDD and BD. Some patients with MDD have no symptoms suggestive of BD, others have a few such symptoms, and still others have many with no gaps in this progression that might serve as a natural point of cleavage between the conditions.4
The same result was obtained by Benazzi5 in a study designed specifically to look for such gaps. Similarly, Mackinnon and Pies6 recently presented an elegant model for rapid-cycling and mixed-state BD, demonstrating that mood, energy, and creativity/speed-ofthought symptoms fluctuating asynchronously can create a continuum of forms of the illness. Although this model has not yet been subjected to rigorous study, it accords well with clinical impressions, just as Jamison7 summarized over a decade ago: “The clinical reality of manic-depressive illness is far more lethal and infinitely more complex than the current psychiatric nomenclature . . . would suggest.”
IS THE DSM INCORRECT?
Is the DSM system incorrect, then? Is BD really just one end of a mood spectrum continuum? The key word here is really: notice that again we are presuming to model reality. We are assuming there is a correct way of explaining, such that there could be overrecognition and underrecognition of bipolarity. Yet we still lack biologic underpinnings for these models. A truly valid nosologic system will ultimately be anchored by understanding etiology—from genetic susceptibilities through molecular differences to anatomic and/or physiologic changes. We are witnessing exciting progress in this regard. For example, recent work by Tsankova and colleagues8 on regulation of brain-derived neurotrophic factor transcription by histone methylation and acetylation in response to social defeat stress further calcifies neuroplasticity, an important bone in the skeleton of mood disorder pathology.
Must we wait for the rest of the skeleton to form to see how bipolar and unipolar mood disorders really are related? (Is there truly a joint in there somewhere?) One hopes not—critical clinical decisions rest on these distinctions. Many patients do not respond, or do not sustain their response, to antidepressant treatment. What percentage of them have an underlying bipolar component as the basis for that failed treatment? Who is at risk for a dangerous reaction to antidepressants? How can we recognize subthreshold bipolarity if, as shown in the most recent iteration of the United States National Comorbidity Survey, it can cause as much disability as diabetes or asthma?9
Many have suggested that the DSM should include a broader range of bipolarity, including Klerman’s 10 categories I through VI; Akiskal and Pinto’s11 I through IV, including 1½, 2½, 3½; and the recent bipolar spectrum disorder proposed by Ghaemi and colleagues.12 A review for the International Society for Bipolar Disorders’ position paper series on bipolar diagnosis13 concludes that 3 groups of patients should be considered for such an extension of the categoric system, including:
- Subthreshold cases, as suggested by the National Comorbidity Survey Replication Study (NCS-R)9 and the Zurich cohort.14
- Patients with multiple soft signs of bipolarity, as defined by Ghaemi and colleagues,12 even if they lack detectable hypomania.
- Patients with antidepressant-induced hypomania or mania.
Or, might it be better to view bipolarity as an illness dimension, one that can be present to varying degrees? This approach has been adopted, at least as a working approach awaiting further study, by the bipolar clinic at Harvard’s teaching hospital, where instead of asking, “Does this patient have bipolar disorder or not?” clinicians are taught to ask “How much bipolarity might this patient have?”15 Among other diagnostic tools (available at www.manicdepressive.org), the clinic uses the Bipolarity Index, a 100-point scale with 20 points for hypomania or mania, but an additional 80 points for other features (eg, family history, age at onset, course of illness, response to medications).16
What problem do we need to solve right now? When should antidepressants be used?
As we consider revisions to the current diagnostic system, we should remember the problem we are trying to solve. At present, the DSM system works well at the extremes of diagnosis: mania identifies BD well when recognized as such. (Numerous factors can impede this recognition, such as lack of insight on the part of the patient or lack of sufficient search on the part of the clinician; neither is the fault of the DSM system.) Conversely, an adult with a single episode of major depression following severe psychosocial stress with no family history of BD will be widely recognized as having MDD. These fully expressed, recognizable cases—shown in the Figure (See Psychiatric Times, July 2006, p. 76) as points E and A—will be treated with mood stabilizers and antidepressants, respectively.
Point D in the Figure (See Psychiatric Times, July 2006, p. 76) represents identifiable BD II, with fully expressed hypomania. The DSM system of diagnosis works reasonably well for such a patient. Our problem comes with intermediate forms. Point C represents subthreshold cases per the NCS-R. Some psychiatrists would call this BD not otherwise specified (BD NOS; ie, bipolar but not meeting BD II criteria). However, others would regard this as a form of depression, perhaps atypical but not bipolar because DSM criteria for hypomania are not met.
The same disagreement arises—and perhaps with more strength—at point B, where only a few bipolar-minded clinicians might persist in calling this BD NOS, and most psychiatrists would call it MDD. Obviously, from a spectrum viewpoint there are almost an infinite variety of other intermediate cases (but let us not digress into Zeno’s paradox!).
Some clinicians are already using a spectrum view of diagnosis, but many are not. The resulting division of opinion was demonstrated dramatically in a 2005 Webcast presented by Harvard faculty, with an audience of hundreds of psychiatrists.17 The program focused on a patient with clear symptoms of depression, as well as anger, anxiety, sleep problems, and alcohol(Drug information on alcohol) use. His case was clearly selected to present features of MDD yet suggests possible BD II. His symptoms were presented in considerable detail to focus on this differential. After the Harvard experts concluded in favor of a unipolar diagnosis, the audience was asked to voice their opinion— bipolar or unipolar? Two thirds chose bipolar. Not only did they disagree with the experts, they were substantially divided among themselves. What an embarrassment for our field.
Notice that on this disagreement hinges a critical, relatively immediate decision. Suppose a patient presents with depression. Suppose the important elements in a standard differential diagnosis have been excluded: substance use or hypothyroidism and other suborganic causes such as Parkinson disease or B12 deficiency (some would include personality disorders, but that is a complex issue18). Suppose the patient strongly prefers a medication approach to psychotherapy and is unlikely to adopt an exercise program as the primary intervention. If the patient is electing a pharmacologic approach for managing depression, should the clinician begin with an antidepressant or consider a mood stabilizer with antidepressant clout?
We can debate whether the line between unipolar and bipolar should be drawn farther to the left on the continuum. We can debate whether there really is a continuum. These debates have persisted for years. (The term “soft bipolar spectrum” was first used in 1987.19) Perhaps in a few years the DSM-V will include one more intermediate form (eg, BD III). Yet here we sit in the field of psychiatry with a system of diagnosis that forces us to disagree with one another in public. This system may even be doing harm by forcing misclassification of patients as unipolar, leading to dangerous adverse effects from antidepressants. All of this, because we persist in clinging to a categoric system for a continuum of symptoms.
To address this dilemma now, instead of waiting for the DSM-V, we could reframe our current division of opinion in new, more manageable terms. Thus, when considering a medication approach to depression, which findings warrant changing our current default assumption from antidepressants to mood stabilizers? This reframing alone will not solve our problem; some psychiatrists would require clear-cut mania or hypomania, setting the bar for this shift in strategy between points C and D in the Figure (See Psychiatric Times, July 2006, p. 76). Others would extend use of mood stabilizers as far as C and others even to B. Several well-known researchers have even suggested that some patients at Point A (ie, no observable signs or history of hypomania) might warrant consideration for mood stabilizers. The bipolar spectrum disorder proposed by Ghaemi and colleagues12 suggests that even patients who have never had hypomania, only multiple soft signs—such as a first degree relative with BD, depression onset between ages 15 and 19, and subsequent repeated brief episodes—may still have enough bipolarity to present some risk if treated with antidepressants alone.
By comparison, another research group has shown up to 6 months of improvement treating BD II with antidepressant monotherapy, which would appear to contradict the concerns about this practice. However, Amsterdam and Shults20 did indeed see an increase in mania scores (Young Mania Rating Scale [YMRS]) in patients treated with fluoxetine(Drug information on fluoxetine). Although they dismiss it as “not clinically meaningful,” the YMRS is not very sensitive to subtle hypomania; thus finding any increase with this instrument may warrant concern.
In general practice, however, most physicians are “lumpers”—when using a medication approach, they treat depression with antidepressants. Indeed, physicians verge on lumping even bipolar depression into the same approach: as of 2003, as many as 80% of community psychiatrists’ patients with BD were taking antidepressants.21 So perhaps a more operational reframing of the diagnostic dilemma might be summarized by a single question: presuming, as discussed above, that a proper differential diagnosis of depression has been conducted and a medication approach is now to be initiated, how many antidepressants should be tried in a patient lacking DSM indications of BD (and, therefore, “unipolar”), before trying a mood stabilizer?
In current practice, we routinely try multiple antidepressants, do we not? But for a patient with a family history of BD, perhaps the transition in strategy should come earlier. And for a patient with many soft signs of bipolarity, perhaps the transition should come after a single antidepressant trial— or even before. This is the principle value of the spectrum concept: it makes you think. It adds bipolarity to the differential of apparently unipolar depression, even when a thorough search for hypomania has yielded little. (Credit for this concept goes to Jack Katzow, MD, my colleague on the International Society for Bipolar Disorders’ spectrum subcommittee.)
Should we simply expand the range of patients for whom one can rationally consider mood stabilizers? All of the atypical antipsychotics have been used with some evidence of efficacy,22 but few formal randomized trials have been conducted. Interestingly, 2 important randomized trials were not published. Data from these studies show that lamotrigine(Drug information on lamotrigine) failed to improve the course of treatment-resistant patients with unipolar disorder,23 even though one would think such a group would represent an enriched sample of soft bipolarity. These data obviously give us pause in considering the breadth of applicability of lamotrigine. Furthermore, olanzapine(Drug information on olanzapine) in this population caused the same weight gain problems we know all too well.24
Clinicians also consider medication risks when choosing among options. In my opinion, we may be underestimating the risk of antidepressants relative to those of mood stabilizers (a very broad and complex group, none of which meet the most rigorous definition of this term25). When, for a given patient, antidepressants lack efficacy as evidenced by a series of failed but presumably wellconducted treatments, what should we do next (besides reconsidering the differential yet again)? In my opinion, clinicians—and patients—may be slow to consider mood stabilizers because of the greater perceived risks of such medications. This deserves study, certainly. In the meantime, let’s exam-ine what we know so far about these relative risks.
