The extended debate as to whether the depressive disorders are best conceptualized as comprising one or more distinct disorders warrants overview. The ‘unitarian’ model presupposes one depressive disorder, varying essentially by severity. The strict ‘binarian’ view postulated two separate types (i.e. ‘endogenous’/‘psychotic’ versus ‘neurotic’/‘reactive’). There were a number of ascriptions to the ‘endogenous’ (now termed ‘melancholic’) type. Firstly, as indicated by its naming, its determinants weighted genetic and other biological factors rather than exogenous psychosocial factors. Secondly, that it had a distinctive pattern of (‘endogeneity’) symptoms and signs—noted shortly. Thirdly, that it showed a preferential response to physical treatments (e.g. antidepressant drugs and ECT) and less responsivity to psychotherapy. By contrast, the second ‘neurotic’ or ‘reactive’ depressive type was viewed as more reflecting depression emerging as an interaction of a predisposing personality style and precipitating life-event stressors.

The debate was strongly influenced by Lewis’s clinical study⁽⁴⁾ finding no clear demarcation between depressive types, examined both cross-sectionally and longitudinally, thus delivering support to the unitarian view. The introduction of multivariate statistical approaches led to the debate being reactivated in the 1960s, with the so-called Newcastle School arguing strongly that their analyses supported a binary view. In a representative paper, Kiloh and Garside⁽⁵⁾ used a factor-analytic strategy to argue for separate ‘endogenous’ and ‘neurotic’ depressive conditions. However, factor analysis is not ideal for developing a typology, in that it produces dimensions (here of symptoms) rather than groupings of patients. Subsequently, more appropriate strategies have been used, such as cluster analysis⁽⁶⁾ and latent class analysis,⁽⁷⁾ and with those studies providing some support for separate classes. Critics suggest, however, that such classes or subgroups could still be determined by severity or, even if sub-types can be identified, question whether sub-classification has any management importance.^{(8, 9 and 10)}

This latter challenge is fundamental, taking us to the heart of any consideration of the diagnosis and classification of the depressive disorders. To the unitarians, as depression essentially varies only by degree, treatment choices (e.g. electroconvulsive therapy (ECT), antidepressant drugs, psychotherapy, or cognitive behavioural therapy) are commonly decided on the basis of severity. The opposing argument—for conceding sub-types—was well put by Kendell,⁽¹¹⁾ who drew on historical analogies. For example, he noted that distinguishing between cardiac and renal forms of ‘dropsy’ allowed prediction of those who would respond to digitalis.

Thus, if there are valid depressive sub-types, the contribution of putative psychosocial and biological risk factors may vary across each, and exert differential responses to differing treatment modalities. If the sub-typing model is valid, forcing homogeneity by creating dimensionally based categories such as ‘major depression’ will ensure muddied results. As noted by Hickie,⁽¹²⁾ numerous studies of patients with DSM-defined ‘major depression’ have failed to demonstrate any coherent pattern of neurobiological changes, replicate key biological correlates, and demonstrate any specific pattern of treatment response outside inpatient treatment settings.

How then have the official classificatory systems addressed such a substantive issue? In developing the DSM-III system,⁽¹³⁾ the working group was required to make a decision on the competing unitarian or binarian models. While the binarians were at the door, they had, until then, failed to prove their case and the DSM-III committee chose a compromise. Thus, DSM-III depression classification was predicated on an initial dimensional component (i.e. ‘major’ versus ‘minor’ disorders). If criteria for a major disorder were met, second-order and more categorical decisions about the
presence of melancholia or psychotic depression were specified. This model proved unsatisfactory for melancholia. For example, Zimmerman and colleagues⁽¹⁴⁾ noted that the DSM-III melancholia criteria set, unlike the definition provided in the predecessor (DSM-II), ‘did not predict treatment response’. Thus, the DSMIII-R⁽¹⁵⁾ criteria set for melancholia was revised to include complete recovery after previous episodes, previous good response to somatic treatments, and no significant personality disturbance, to overcome the lack of predictive validity by building into the definition some of the ‘givens’ held by many clinicians about melancholia. However, the criteria set for melancholia developed for DSM-IV returned essentially to the DSM-III set, with limitations considered below. The contrasting system, ICD-10, is essentially based on a stricter dimensional or unitarian view of the depressive disorders—comprising ‘severe’, ‘moderate’, and ‘mild’ disorders.

During the extended debate as to whether a categorical and more ‘biological’ type of depression exists—it was variably termed ‘endogenous’, ‘endogenomorphic’, ‘autonomous’, and ‘melancholic’ depression. The last is probably preferable as numerous studies have quantified few or no differences in the likelihood of those with ‘endogenous’ and ‘non-endogenous’ depression reporting antecedent life events, so arguing against any term weighting ‘internal’ or ‘external’ causes.

Whether psychotic (or delusional) depression is a ‘severe’ form of melancholia or a separate entity also remains problematic. DSM-III had a category ‘major depression with psychotic features’ for use when delusions or hallucinations are present or when there is ‘depressive stupor (the individual is mute and unresponsive)’, thus viewing ‘psychotic depression’ as a sub-type of the generic ‘major depression’ category rather than a sub-type of melancholia. While ‘depressive stupor’ may be a useful marker or proxy for the condition, this criterion was not retained in DSM-III-R or DSM-IV, but is included in ICD-10. Two points argue for psychotic depression as a distinct entity: the presence of psychotic features, and its poor response to antidepressant medication alone and to neuroleptic medication alone in comparison to high responsiveness to their combination.⁽¹⁶⁾

A strict interpretation of the ‘binary’ view would place the non-psychotic, non-melancholic depressive conditions in a pure second class. However, rather than view this as a pure ‘type’, this class is best regarded as a heterogeneous residue category (i.e. nonmelancholic depression), with its heterogeneity expressed widely— across aetiological factors, clinical expression, and natural and treated history.

Both ICD-10 and DSM-IV have multiple conditions and specifiers. The ICD-10 system allows mild and moderate depressive episodes (with or without a ‘somatic syndrome’ conceptualized as reflecting ‘melancholic’ features), and severe depressive episode (with or without psychotic symptoms). There are separate codes for a similar set of ‘recurrent’ disorders, while several ‘persistent’ mood disorders (including cyclothymia and dysthymia) and residual conditions are listed. DSM-IV has two principal ‘stem’ disorders (major depressive episode and dysthymia), with the first having a number of optional specifiers including ‘with’ melancholic, catatonic, psychotic, or atypical features, as well as including disorders showing longitudinal patterns of rapid cycling or a seasonal pattern. Both systems have categories for affective disorders secondary to organic disease, while DSM-IV includes mood disorders due to a general medical condition or substance use, or occurring in the post-partum period. Both classificatory systems include adjustment disorders with depression.

Formal classifications are therefore built principally on severity, features of current episode, patterns of disorder expression over time, as well as persistence and recurrence. Few diagnoses are consistent across the ICD-10 and DSM-IV systems and, while each provides definitions that allow a ‘shared language’ to be used by clinicians and researchers, the extent to which their severityweighted groupings capture ‘meaningful’ depressive sub-types remains problematic.

For example, and as detailed elsewhere,⁽¹⁷⁾ ‘major depression’ has come to be viewed as an entity, sufficient in and of itself for testing antidepressant therapies and to generate treatment recommendations. Limitations to such a model become apparent if we consider the analogy of ‘major breathlessness’, which could be a transient consequence of acute exercise, or reflect quite differing pathological processes (e.g. asthma, pneumonia, or a pulmonary embolus) benefiting from quite differing treatment approaches. Thus, a diagnosis of ‘major depression’ or ‘clinical depression’ is, in reality, a first-level domain diagnosis, and benefiting from secondary specification. The latter tends to proceed on the basis of severity, but alternative and more categorical models have long been proposed as considered elsewhere in this chapter.

The DSM-IV definition effectively requires an initial or previous manic episode for bipolar I disorder, while bipolar II disorder requires hypomanic episodes and one or more previous episode of major depression.

To meet DSM-IV diagnostic status, manic episodes must have lasted 7 days and hypomanic episodes 4 days. Both ICD-10 and DSM-IV have course specifiers for bipolar disorder containing 10 and 4 subgroups, respectively. In addition to the number of subgroups, differences include a greater emphasis on distinguishing bipolar I and II in DSM-IV, and cyclothymia being listed as a ‘bipolar disorder’ in DSM-IV as against being a ‘persistent’ mood disorder overlapping with a personality style in ICD-10. Distinguishing ‘hypomania’ and ‘mania’ is regrettably imprecise in the formal classificatory systems. Both DSM-IV⁽¹⁸⁾ and ICD-10⁽¹⁹⁾ disallow a diagnosis of hypomania if psychotic features are present but, conversely, do not require psychotic features for a diagnosis of mania. DSM-IV lists essentially similar clinical criteria (and criteria number cut-off) for hypomanic and manic episodes, but distinguishes mania by the presence of marked impairment in social functioning (risking subjective judgement), requirement for hospitalization (which is logically more a consequence than a defining criterion although it may have some proxy value), and the presence of psychotic features in a manic episode. As noted earlier, ICD-10 views hypomania as ‘an intermediate state without delusions, hallucinations, or complete disruption of normal activities’.

Thus, formal ‘cleavage’ between bipolar I and II (and constituent manic and hypomanic states) is largely dimensional in relation to clinical features and with some logical fallacies. Further, duration
criteria (i.e. at least 7 days for mania and at least 4 days for hypomania in DSM-IV) do not appear sustainable. In recent years there have been many studies⁽²⁰⁾ indicating that clinical definition of bipolar disorder is not dependent on the duration of the highs, and that imposing DSM-IV duration criteria for both mania and hypomania may exclude a significant percentage of those with true bipolar disorder.

As detailed earlier, there are intrinsic difficulties in classifying depression according to any single model when it is a term encompassing normal mood states through to possibly categorical diseases, and when any imposition of a severity-based model raises problems about how to differentiate meaningful groups (e.g. ‘cases’ from ‘non-cases’). A personal mixed model is now detailed for consideration—one shaped by clinical experience and supported by research findings. It is described in line with the ‘reasoning steps’ that a clinician might employ in assessing a potential depressive disorder.