Diagnosing bipolar disorder: Defining thresholds and setting boundaries

If there were reason for these miseries, Then into limits could I bind my woes

—William Shakespeare (1564–1616), Titus Andronicus, III. i.

Diagnosis is essential to identify those who are suffering and in need of treatment; ideally it also reflects underlying disease processes, informs management and predicts prognosis. However, the absence of a pathophysiological foundation leaves psychiatrists dependent on phenomenology to make diagnoses. In psychiatry, these are called ‘disorders’, a term used to describe clusters or syndromes of symptoms; they are not diseases (Malhi, 2013a). Hence, in a manner akin to divining the existence of subterranean water by using a dowsing rod, psychiatrists sift through elicited symptoms to identify patterns that best serve as ‘pointers’ to psychopathology. These are then grouped into syndromes (Nemeroff et al., 2013).

The perils of this analytical process are aptly exemplified by Homer’s mythological dilemma in negotiating the Strait of Messina, which involved having to steer his ship between Scylla, a six-headed monster, and Charybdis, a deadly whirlpool. Similarly, psychiatric diagnosis, in practice, usually requires carefully having to evaluate competing contingencies, each of which is associated with potential risks. For example, a hastily arrived at diagnosis risks inaccuracy and may lead either to the incorrect labelling of an individual as having a psychiatric illness, or to diagnosis of the wrong illness. Such premature diagnosis or misdiagnosis has obvious harmful consequences, including the commencement of unnecessary treatment. On the other hand, not making a psychiatric diagnosis can result in missing an opportunity to initiate timely treatment, which might benefit the patient, who instead continues to suffer and moves towards a poorer outcome. This is the daily challenge that psychiatrists face; namely, having to carefully optimize the diagnostic process to achieve a balance between sensitivity and specificity. In recent times, the diagnosis of bipolar disorder, perhaps more so than any other psychiatric diagnosis, has brought this into sharp relief (Malhi, 2013b).

The threshold of mania

Remarkably, almost half a millennium before Christ, Hippocrates of Cos and members of his school had described the symptoms of melancholia and attributed their origins to organic brain dysfunction. These early philosophers and physicians rejected the attribution of disease to divine or supernatural forces, but their reasoning was not as enlightened as it may first appear because it was based on an equally fantastic humoral theory (Angst and Marneros, 2001). In this schema, which dominated medical thinking for many centuries, illness arose because of an imbalance between four humours: blood, phlegm, yellow bile and black bile. These corresponded to temperaments, and had commonality with theories related to various elements (air, water, fire and earth). For example, in humorism, mania was thought to occur because of an excess of yellow bile (a choleric temperament), whereas depression was attributed to an excess of black bile (a melancholic temperament).

In practice, depression and mania were regarded as separate conditions and it was not until the detailed observations of Aretaeus of Cappadocia (2nd Century) that mania and melancholia were linked. He viewed mania as a variety of melancholia, and described an alternating pattern of illness in which an individual could alternate between the two states. He also described cyclothymia and suggested that personality could contribute to the development of manic symptoms (Goodwin and Jamison, 2007). Despite these brilliant insights, deliberations remained largely dormant for many centuries until the 19th century, when two French physicians almost simultaneously described a single illness comprising both mania and depression (priority should be assigned to Falret, both because his definition resembles more closely present-day manic-depressive illness and bipolar disorder and because he conceived the idea first (Pichot, 2004)). Jean-Pierre Falret (1794–1870) used the term la folie circulaire (circular madness), by which he meant a regular continuous succession of depression and mania (Falret, 1854); whereas Jules Baillarger (1809–1890) described la folie à double forme (double insanity) to define a single illness consisting of both mania and depression (Baillarger, 1854). Others also contributed to the definition of circular disorders including, in particular, Wilhelm Griesinger, who essentially echoed Aretaeus’ thinking, and Ludwig Kahlbaum and Ewald Hecker, who laid down the foundations for present-day cyclothymia (Benazzi and Akiskal, 2006; Malhi and Berk, 2014). The task of synthesis and simplification ultimately fell to Emil Kraepelin, who, at the dawn of the 20th century, separated manic-depressive illness from dementia praecox and created a model that continues to serve as a foundation for our current nosology in both the Diagnostic and Statistical Manual of Mental Disorders (DSM) and International Classification of Diseases (ICD) systems. However, an important caveat is that modern-day ‘bipolar disorder’, a term coined by Karl Leonhard, is quite separate from manic-depressive illness: the latter essentially emphasized recurrence, whereas in the former it is polarity that holds salience.

Setting boundaries

Bipolar disorder is, by its very nature, a complex illness (Malhi et al., 2012), partly because of the many domains it affects. It is therefore not surprising that it generates considerable diagnostic uneasiness. Some of this uncertainty applies to the whole field of psychiatry, but some aspects are unique to bipolar disorder and are a consequence of how it has been defined. Psychiatric research worldwide relies predominantly on the American Psychiatric Association (DSM), within which a definition of bipolar disorder only appeared relatively recently, in its third revision (American Psychiatric Association, 1980). Later still, subtypes such as bipolar II disorder gained formal recognition as separate diagnoses in DSM-IV (American Psychiatric Association, 1994), published two decades ago.

The DSM diagnosis of bipolar disorder hinges on the occurrence of hypomania/mania and, even though clinically the syndrome of mania is arguably more stereotypical than any other in psychiatry, its separation from normalcy, variants of personality and partitioning into subtypes remain both problematic and contentious. For example, in DSM-5 (American Psychiatric Association, 2013), mania is defined on the basis that an individual has mood disturbance that is sufficiently severe to cause marked impairment (in social or occupational functioning), requires hospitalization (to prevent harm to self/others), or there are psychotic features. At first pass these criteria may seem reasonable, but upon closer scrutiny it is questionable how consistently these features can be applied clinically. For instance, although the evaluation of impairment as ‘marked’ is somewhat arbitrary and not clearly operationalized, it is used in DSM-5 to differentiate mania from hypomania. Similarly, hospitalization is a poor proxy of illness severity: in reality, although hospitalization is a potential consequence of being unwell, whether or not an individual is admitted to hospital is determined by a combination of social, economic and cultural factors, including the availability and desirability of inpatient care. It is somewhat puzzling therefore that hospitalization is regarded as a diagnostic parameter of bipolar disorder; indeed, in DSM-5 it separates mania and hypomania.

Hypomania is also poorly demarcated from the normal vicissitudes of mood, and briefer and less severe manic syndromes (Malhi et al., 2011). The minimum duration criterion for hypomania of 4 days has been shown to lack diagnostic purpose. Episodes shorter than 4 days’ duration essentially have the same features as those that are longer (Angst et al., 2012; Benazzi, 2007). Specifically, longer episodes of hypomania (4 or more days) do not seem to be more severe than those of shorter duration (1–3 days) (Angst et al., 2003) and demographic features, such as age and employment, also seem to be similar (Bauer et al., 2011), suggesting that the cut-off of 4 days does not reflect a true trend break in frequency or severity of symptoms, and that the division is unlikely to represent an underlying biological schism. Rather, a shorter duration cut-off of 2 days may be able to differentiate brief periods of hypomania from major depression on the basis of illness course and family history (Angst et al., 2012; Benazzi, 2007). This distinction between bipolar disorder and major depression is clinically important because, though bipolar disorder is defined on the basis of mania or hypomania, typically, in at least three-quarters of individuals, the illness presents initially with depression. Furthermore, there is a considerable delay before bipolar disorder shows its hand, such that depressive symptoms usually precede the emergence of mania by 2 years (Berk et al., 2007). In practice, the likelihood of detecting hypomania is further diminished by the fact that people quite understandably fail to recognize a period of ‘profound wellbeing’ as mental illness and don’t seek treatment for ‘feeling wonderful’.

Mixed modelling

A further area of considerable uncertainty in the diagnosis of bipolar disorder is that of mixed states, which are defined as coterminous depression and mania. The diagnostic criteria for mixed states have altered considerably from DSM-IV to DMS-5. In the former, threshold mania and depression were required simultaneously, a very uncommon occurrence and one seldom encountered in practice. Consequently, most instances of ‘actual mixed episodes’ were rarely captured in clinical settings, and in research studies mixed episodes were often coded alongside manic episodes. Therefore, the true prevalence of mixed episodes and how they are best managed remains unknown (Berk et al., 2005). To address this problem, DSM-5 has adopted a more flexible approach and created mixed features specifiers, which indicate the occurrence of symptoms from the opposing pole manifesting in depression, mania or hypomania. This ‘mixed model’ has solved one problem but created others. First, it broadens the concept substantially without providing any additional clues as to causality. Second, it is unknown what implications mixed features specifiers will have for treatment. Perhaps the most contentious element of the mixed features specifier is its potential applicability in unipolar major depression. For example, it is most unclear whether a mixed state within major (unipolar) depression indicates a potential bipolar diathesis and, if so, whether it should prompt modification of therapy towards bipolar treatments or whether it should be seen as an agitated ‘subtype’ of major depression and treatment adjusted accordingly. Furthermore, the minimum duration of mixed features is ambiguous and unspecified. In the context of mania, a period of 1 week of mixed symptoms would suffice, but in the context of depression mixed symptoms need to be present for at least 2 weeks. A further difficulty is the poor specificity of many of the constituent symptoms of a mixed state (Malhi, 2013c). Intriguingly, DSM-5 does not include distractibility and psychomotor agitation as mixed features, and yet these may be the defining elements of mixed states (Malhi et al., 2014). The lower threshold for mixed features means that even transient perturbations of mood may be incorrectly labelled as ‘mixed features’. For example, in the context of bipolar depression irritability is a common symptom that often reflects underlying anxiety or personality factors. The lack of diagnostic precision in defining mixed features in DSM-5 risks further inflating the diagnosis of bipolar disorder and once again fails to accurately capture the true nature and extent of mixed states in bipolar disorder.

Coming of age

One of the most contentious diagnoses in all of psychiatry in recent times is that of paediatric bipolar disorder, which at its peak was being diagnosed 40 times more often in the US than in other parts of the world (Moreno et al., 2007). The reasons for this were not immediately apparent and researchers and clinicians only became aware of this ‘epidemic’ after it had taken hold. The putative origins of the increase can be traced back to a number of key centres in the US, which actively promoted the diagnosis.

Clinically, there are two main reasons for this marked inflation in the diagnosis of paediatric bipolar disorder. First, diagnosing bipolar disorder in children and adolescents is inherently even more complex than it is in adults because the detection of symptoms requires pristine separation of these from the many emotional, cognitive and behavioural changes that normally take place during labile and formative years. Therefore, detection and diagnosis is naturally difficult and rarely possible with confidence, even with extensive experience. Second, the criteria for diagnosing bipolar disorder have been gradually loosened in children and adolescents to the point of losing diagnostic significance. For example, some authorities have argued that paediatric bipolar disorder is not characterized by typical adult cyclicity but instead has more ‘mixed’ features that occur more chronically. But prodromal research in first-episode mania cohorts, as well as in cohorts of offspring of bipolar subjects, has thus far failed to identify evidence for a prototypical paediatric bipolar phenotype (Conus et al., 2010; Duffy, 2009; Zeschel et al., 2013). This has made this diagnosis difficult to apply with any meaningful specificity (Cahill et al., 2007).

Thus, the likelihood of misdiagnosing bipolar disorder in children and adolescents is high, but the key concern is the medicalization of childhood behavioural disorders and the wrongful initiation of medication once a diagnosis is made. This is particularly worrying in children and adolescents because of the enduring nature of therapies with a tight risk/benefit ratio, such as lithium (Malhi et al., 2012). Bipolar medications such as lithium that have been used in adults have been extrapolated to children and adolescents, despite far poorer evidence of efficacy (Dickstein et al., 2009; Geller et al., 2012). This could be dangerous because almost every bipolar medication has potentially serious adverse effects. For example, many neuroleptics can lead to extrapyramidal side effects and metabolic syndrome, and even commonly used antidepressants may worsen mood instability and increase the likelihood of ‘switching’. Ultimately, the effects of prevailing bipolar treatments on the developing brain are unknown, and the long-term consequences are uncertain; hence diagnosis, which inevitably leads to treatment, generates unfathomable risks.

Getting personal

A further area of uncertainty is the ‘boundary’ between personality and mood disorder. Clearly, this is dependent on the definition of both. Many phenomena occurring within bipolar disorder are mirrored by personality disorders, including, for example, mood instability, irritability, impulsivity and dramatic behaviour. These symptoms reside in an ill-defined zone that spans personality and bipolar disorder. For example, ‘affective instability’ is often confused with symptoms of a mood disorder and wrongly used to identify bipolar disorder. Conflation of these symptoms has probably contributed to the over-diagnosis of bipolar disorder and misdiagnosis of personality disorders. The key differential of bipolar disorder, especially bipolar II, is borderline personality disorder (BPD), which has been documented as the most common personality comorbidity. Both clinically and biologically the two disorders are distinct but can naturally co-occur (Malhi et al., 2013). Difficulty in disentangling them arises when borderline personality disorder, which is driven predominantly by psychosocial stressors, occurs alongside modest or subsyndromal changes in mood (Ghaemi and Dalley, 2014). A history of childhood trauma and self-harm is usually common to both syndromes, which also seem to respond to overlapping psychological interventions and pharmacotherapy, but it is the pattern of onset and the specific phenomenological profile, along with the longitudinal clinical course, that differentiates the two disorders (Bassett, 2012; Coulston et al., 2012). In practice, eliciting the necessary information can prove to be a challenge and may require several assessments over a period of time. This aspect, and in particular the importance of longitudinal appraisal of patients with bipolar disorder, has been generally under-emphasized. Instead, self-report and cross-sectional use of screening using Internet measures, for instance, has been encouraged and, although this has assisted in increasing awareness of bipolar disorder, it has also promoted premature and inaccurate diagnosis. Given the dimensional nature of personality traits and the continuous nature of mood changes, a cut-off and pristine distinction between the two is unlikely to emerge from clinical measures alone.

So what have we divined?

The diagnosis of bipolar disorder is clearly anything but clear. Its boundaries on all fronts are difficult to delineate and its current status as a disorder that reflects a core or unitary underlying disease is increasingly in doubt as its definition is progressively diluted. The distinction of bipolar disorder from major depression is pivotal, both in terms of early detection and initiation of treatment. How many manic symptoms and how much mania is required to cross the threshold into bipolar disorder from major depression is debatable. Separating bipolar depression from major depression is proving even more difficult, with few reliable distinguishing features. Similarly, characterizing symptoms as bipolar as opposed to anxiety driven or a consequence of personality factors continues to complicate diagnosis, but perhaps even more perplexing than difficulties of circumscribing the illness are the problems in partitioning the illness itself into meaningful subtypes. Both lower and upper thresholds of hypomania, which defines bipolar II disorder, are questionable, and subsyndromal symptoms, which span the bipolar spectrum as it melds into normalcy, are poorly characterized and understood. This state of affairs has been compounded by a lack of diagnostic precision in the introduction of some poorly validated specifiers.

From a practical perspective, diagnosis has many benefits for the individual and is essential for understanding and researching an illness. Invariably it also leads to the initiation of treatment and this is where the difficulties multiply. The process of diagnosis can be dimensionalized to some extent on the basis of grading severity and numbers of symptoms and, in the case of bipolar disorder, a spectrum model may fit, but treatment decisions are necessarily categorical and greater certainty is needed as to the benefits and risks of treatment. This complexity of making a categorical diagnosis within a spectrum of syndromes is exemplified by the efficacy of medication within that spectrum: we know, for example, that the further one strays from the diagnostic core of the disorder, the lower the efficacy of the primary psychotropics, such as lithium. Perhaps this knowledge should be used to ‘define’ primary bipolar disorder?

Clinically, the diagnosis of bipolar disorder is a life-changing event for all concerned and carries with it the risks of stigmatization and illness behaviour, as well as the fact that most treatments have limited efficacy and poor tolerability. Therefore, diagnosis needs to be considered with caution and assigned following careful deliberation. Extant taxonomies of psychiatric disorders are a sad reminder of the failings of the field to distil meaningful diagnoses tethered to pathophysiology. This is not a criticism of the authors of our classifications; we lack the tools to do better at present. Bipolar disorder is not a disease. It is not even a homogeneous entity. But it is a serious mental illness and a disorder that impacts the lives of many. It therefore warrants investigation and better treatment and the first step towards achieving this is to develop a better definition – derived, ideally, from an understanding of neurobiology.