The Mood Disorders Unit,University of New South Wales

The Mood Disorders Unit (MDU) was established at Sydney's Prince Henry Hospital in 1985. I had been recently re-graded by the University and appointed Clinical Director of the Hospital's Division of Psychiatry, a model introduced by our Foundation Dean to facilitate links between clinical and research activities. I was keen to create some ‘clinical program’ modules joining clinical and research endeavours in an iterative way. ‘Depression’ was an obvious module, building on the seminal and distinctive subtyping studies and clinical interests of Leslie Kiloh, our University's Foundation Professor of Psychiatry.

The first task was to form a group of chief investigators (i.e. Philip Boyce, Henry Brodaty, Philip Mitchell and Kay Wilhelm), and then to determine an appropriate model and set research priorities. We judged that our research studies should be nested within a new clinical service. Structurally, this involved establishing an inpatient unit of 15 beds, an outpatient assessment service, and a name, with ‘Mood Disorders Unit’ selected on patients' preferences. The model allowed for patients being referred for review of diagnostic symptoms or to address management issues, predictably leading to disproportionate numbers of those with severe and/or treatment-resistant disorders, together with patients from the catchment area. For the ‘tertiary referral’ patients, we determined that the MDU would focus on assessment, and that we would only take management responsibility if facilities or treatments (e.g. electroconvulsive therapy or ECT) were not available to the referring doctor, or if assessment required hospitalisation. Thus, we envisaged a State-wide clinical resource.

The clinical service was rapidly established under the smooth administrative directions of Henry Brodaty and, subsequently, Philip Mitchell. It was rewarding to the consultant psychiatrists, who were stimulated by the challenge of such patients, with discussions of all outpatient and many inpatient referrals by our consultants, with their differing viewpoints enriching our competency. As has been recorded in Journal publications, referrers and patients rated the service highly. Over the first decade, patients came from across New South Wales, as well as from interstate and occasionally overseas. The large numbers allowed many clinical studies to be initiated readily.

The iteration between clinical and research work is worth emphasising. Clinical exposure generates hypotheses that can then be tested by formalised research. Studies are often more profitable when the hypothesis ‘fails’ and has to be reformulated by re-focusing clinical observations. One example, described shortly, is our failure to demonstrate any support for ‘reactive’ or ‘situational’ depression. Similarly, research findings shape assessment and treatment, enhancing clinical standards. Other advantages of the meld have emerged. When researchers have clinical responsibilities, this increases patients' willingness to take part in research studies. However, wishing to ‘please’ their psychiatrist, patients may agree to take part in a study against their intrinsic interest. Such nuances, which must be conceded and addressed by researchers and not merely left for formal consideration by ethics committees, have required attention by our team.

Assessment has often been onerous for patients. For one major study, they completed a pre-assessment baseline questionnaire, were interviewed by a research assistant for an hour, underwent a consultant assessment and research ratings for 2 h, were assessed by a psychologist, and completed other questionnaires while waiting for a final formulation from the consultant after group discussion. We sought to reduce the demanding nature of the process by relying on staff who clearly informed patients about assessment processes, ensured tea, coffee and food breaks, and otherwise supported the patients on assessment days. Such a lengthy assessment was rarely viewed as an imposition, with many patients appreciating its rigour and depth and particularly valuing the consultant's feedback following his or her discussion with colleagues.

Many patients have come to the MDU as if they were coming to ‘Mecca’. This raising of expectations is doomed to disappointment. In our feedback to patients and referrers, we try to ensure against inflated expectations (e.g that our recommended treatment will necessarily work), and thus reduce the chance that referral might be counter-productive or cause management difficulties for the referrer. Predictably, the most difficult subgroup remains as patients who state that the MDU is their ‘last hope’.

This exposition of clinical issues highlights the advantages to iterative clinical and research activities. While this represents only one model for a research centre, its efficient clinical base has provided an enriching experience for staff of all disciplines.

The Mood Disorders Unit research model

From inception, our main focus has been to develop an improved model for subtyping and diagnosing depressive disorders. Assessment (inpatient or outpatient) has involved four serial baseline assessment schedules over 14 years, each designed to address our objective but having ‘add-on’ components.

Any research unit of this type requires a focus. Ours respected the School of Psychiatry's history, defined research priorities, appeared achievable and worthwhile, and linked the researchers to a common goal despite disparate individual research interests. There was less concern about choosing our first objective (distinguishing melancholic and non-melancholic depression clinically), and more attention paid to how we might proceed when others had foundered. Monthly research meetings were lengthy, involving psychiatrists, research assistants, and infrastructure staff, were a forum for vigorous debate, and galvanised the team. Heady thinking underpinned discussions about research design. A divergent ‘Let a thousand flowers bloom’ philosophy influenced the initial database, with everyone's favourite clinically differentiating feature included.

The divergent hypothesis-generation phase, subject to the capacity to implement or force later convergent phases, has a number of advantages, as ‘fishing expeditions’ sometimes do catch fishes. In recent years, designs have become more convergent, and for several reasons. First, the chief investigators have all ‘matured’ (Editor's term, not mine), and lost some of the breathless enthusiasm and group dynamics which energise a new endeavour. Second, our prioritised studies have generated relatively clear-cut findings, and thus led to more hypothesis-testing rather than hypothesis-generating investigations. Third, individual investigators have had their own interests, and showed more generosity than might have been expected to initial corporate objectives. Now that the base camps have consolidated, each feels freer to climb their own hill. From the corporate position, it has been proper to ensure that activities ‘bud off’ and develop their own independent existence. It respects evolutionary principles, while limiting the tendency to aetiolation and the intellectual incest that can occur in any organisation.

The current dilemma in evolving the ‘best model’ requires judging the need for central united activities, or alternatively, abandoning them, and having research proceed via independent satellites as occurs in many successful institutes. Two disadvantages to becoming a ‘shelf’ institute are the loss of ‘cut and thrust’ discourse and a formalised ‘generativity’ component, whereby junior researchers are exposed to the wisdom of the elders. As it is time for the MDU to disseminate applications of research findings to clinicians and for wider educational purposes, the centre may well expand from coordination of research studies to providing a resource for teaching and education. If that happens, it will not promulgate treatment guidelines based on pseudo-categories such as ‘major depression’ or ‘dysthymia’ (for reasons noted later).

Research has spanned many domains (e.g. biological, psychological, social), pursued many aetiological factors, and evaluated many treatments. To list them would be tedious. Instead, I will overview a number.

Depression subtyping

Our initial objective was to distinguish endogenous/melancholic from ‘neurotic’/‘reactive’ depression clinically, as we subscribed to a binary view (i.e. two distinct types await clinical discrimination).

Our exploratory study (CORE-I) involved assessment of a vast number of symptoms and signs. After several hundred patients were assessed, data were handed over to our scientific officer, Dusan Hadzi-Pavlovic, and a pristine solution requested by afternoon tea. Commonly recommended analytic strategies failed to deliver an acceptable solution. After months of wrestling with the data, Dusan appeared, modestly offering a possible nugget. Using techniques relatively untried in psychiatric research (latent class and mixture analyses), he had identified a putative melancholic group on the basis of observable, but not reported, psychomotor disturbance. A sign from above! Validation studies supported the interpretation and indicated that delineation of a melancholic depressive ‘type’: (i) could be achieved by observer-rated psychomotor disturbance; (ii) was superior to using the so-called ‘endogeneity symptoms’; and (iii) showed discrimination superior to other systems, including the then current DSM-III-R decision rules.

As our descriptors were imprecise and had insufficient agitation items, we added ‘agitation’ to the ‘retardation’ and ‘cognitive processing’ items, wrote detailed descriptors, and determined anchor points distinguishing between the presence or absence of a feature and, if present, its severity. The CORE-II study was then undertaken to test the utility of this new measure and its capacity to distinguish ‘melancholic’ from what we now viewed as a heterogeneous residue of ‘non-melancholic’ disorders (rather than a pure second type of ‘neurotic’ or ‘reactive’ depression); thus, a modified binary view. The results and early validation studies, described in our 1996 Cambridge University Press Monograph, exceeded our expectations. As measured by the final 18-item CORE measure, observable psychomotor disturbance above the defined cut-off score went a long way to meeting rigorous validation criteria of being ‘necessary and sufficient’ (i.e. present or necessary in all those with true melancholia, and sufficient to make such a diagnosis without additional features). Validation was supported against a range of ‘givens’ for melancholia (e.g. older age, older age at first episode, decreased chance of antecedent life events or of a disordered personality style, dexamethasone non-suppression, impaired cognitive functioning, and superior response to physical treatments). The CORE measure has proved superior to DSM-III-R and DSM-IV criteria for subtyping melancholia, and with the latter diagnostic systems shown in passing to be non-differentiating of depressive type as most of their criteria symptoms measure depression severity rather than ‘melancholia’.

We view observable psychomotor disturbance as a marker of an underlying neuropathological process; this points to the likely disrupted biological processes. Respecting the neural circuit model, where disruption of those circuits connecting the basal ganglia and the prefrontal cortex result in three broad disturbances (i.e. depression, cognitive impairment and psychomotor disturbance), our finding advances theorising about circuits likely to be functionally and/or structurally impaired in melancholia, leading us to structural and functional imaging studies, and optimally allowing for treatments to be developed more rationally.

We similarly explored the status of ‘psychotic’ or ‘delusional’ depression. In comparison to the DSM model, which views it as a subtype of major depression, our analyses suggest that it is a subtype of melancholia. While analyses indicated the specificity of delusions and hallucinations, observable psychomotor disturbance is generally more severe than in non-psychotic melancholia, often so severe as to serve as a proxy marker (in mute patients) of such a disorder. While ‘endogeneity symptoms’ such as non-reactivity and a non-variable mood may be more severe than in melancholia, they are not specific to psychotic depression. Thus, our CORE-II study (subsequently replicated in two independent databases) allowed three depressive classes (i.e. psychotic depression, melancholia and a heterogeneous residue of non-melancholic depression) to each be distinguished by two specific features, suggesting a hierarchical model. Specifically, all three conditions have a lower-order mood disorder component, for they are all depressive disorders. Despite the mood state being more severe in melancholic than non-melancholic, and even more severe in psychotic depression, its commonality ensures that it lacks capacity to differentiate depressive subtypes and explains limitations to DSM and other severity-based models.

In proceeding from the non-melancholic disorders to melancholia, our model thus ignores the greater mood state severity and argues for the necessary presence of observable psychomotor disturbance. While the psychomotor disturbance and mood state are overall more severe in psychotic depression, our hierarchical model weights the specific presence of psychotic features.

The specificity model does not hold for the commonest class, comprising the heterogeneous residue of non-melancholic conditions, and here we have been required to consider and test other models. We rejected a common strategy of grouping on the basis of severity (e.g. major depression, minor depression and, say, subsyndromal depression). We now argue that the data best allow a ‘spectrum model’, linking Axis II personality/temperament with Axis I symptomatology. The model argues for neurobiological processes influencing temperament style which, when the individual is exposed to a ‘depressogenic stressor’, function like a lens to shape the surface expression of the non-melancholic disorder. We are currently delineating two principal non-melancholic spectrum disorders. The first involves those who are by temperament ‘anxious worriers’ and ‘act in’ when stressed, have a family history of anxiety and a high rate of lifetime anxiety disorders, develop their first episode of depression at an early age, and have multiple frequent episodes, in which anxiety symptoms predominate. The second includes those who are more likely to have a volatile personality style (having particular difficulty in maintaining relationships), ‘act out’ under stress and, when depressed, are ‘irritable’ and ‘hostile’.

As noted, we have not found support for a non-melancholic ‘reactive’ or ‘situational’ subclass. While many patients attribute their depression to a life event, our research suggests that onset and persistence of their depression is more due to ‘vulnerability factors’ such as anxiety and personality style.

The diagnostic term ‘endogenous depression’ lost relevance when it was demonstrated that precipitating life events were commonly found. Thus, it appears reasonable to conclude that life events, or their absence, are not a valid criterion to derive subtypes. It appears preferable to view them as precipitating agents in those who carry some vulnerability (e.g. genetic diathesis in melancholia, and personality style or high anxiety in non-melancholic depression).

Why subtype the depressive disorders, and why has this strand been such a major activity?

Since the introduction of DSM-III, the dominant model has been unitarian, favouring a single entity. As patients may talk about having ‘cancer’, rather than particular types of cancer with varying aetiologies, treatments and outcomes, patients (and psychiatrists) commonly view depression as one entity. We argue for differing substantive depressive types, and where determinants, natural history and responses to differing treatments may differ categorically or vary in relevance. Such ‘types’ are not well captured by dimensionally crafting subcategories of major depression, minor depression and now subsyndromal depression. Since DSM-III, research into neurobiological underpinnings of depressive disorders has failed to produce clear-cut findings, and has illustrated how severity-based models confound interpretation about underlying mechanisms. Treatment implications are equally important. Data suggest that psychotic, melancholic, anxious non-melancholic and irritable/hostile non-melancholic disorders show quite differing responses to differing antidepressant treatments and particularly to varying antidepressant drug classes. Such gradients are less evident when depression is classified on a severity basis. Kendell's analogies have been used in many of our papers, that until the pox had been subtyped (i.e. smallpox and chicken pox) it was impossible to predict who would live and who die; and similarly that until renal and cardiac forms of ‘dropsy’ had been distinguished, we could not predict who would respond to digitalis.

Thus, current research plans include refining measurement of the non-melancholic spectrum disorders, developing algorithms for clinical and research purposes which effectively subtype the principal depressive conditions, and pursuing aetiological and treatment questions.

It is salutary to recognise that we have re-discovered previously identified clinical markers of depressive subtypes, although we have had to identify the best ways to model the disorders and develop improved measures. As described in our monograph, the importance of psychomotor disturbance to melancholia was well recognised centuries ago, while rich descriptions of ‘anxious’ and ‘irritable/hostile’ non-melancholic depression dominated the literature in the two decades prior to DSM-III. This is encouraging, not only in suggesting that our statistically derived solutions are less likely to be idiosyncratic, but in indicating the importance of respecting clinical observation. At the ‘macro’ level, we would like to see such research return classification to a more categorical model.

We now argue for the two differing operative models: one hierarchical, clinical feature-specific model for distinguishing three major depressive classes; and a ‘spectrum model’ for refining non-melancholic subgroups. We do so by the usual processes of publishing research evidence in journals and in textbooks, and by presenting at international meetings. Convincing the DSM and ICD Titanics to change course will not be easy, but we fancy the role of the iceberg.

Developing measures

Psychiatry deals with vague concepts and constructs. Much research is confounded by failure to operationalise and measure constructs with minimal extraneous ‘noise’. The development of the Parental Bonding Instrument (PBI) was a useful model in recognising the importance of measuring salient constructs and then undertaking application studies (e.g. can poor parenting styles induce all psychiatric disorders known to DSM?) on a sounder footing. We have, therefore, developed a number of measures (e.g. Philip Boyce's Interpersonal Sensitivity Measure; Kay Wilhelm's Intimate Bond Measure), an abbreviated PBI (i.e. the Measure of Parenting Style or MOPS) and measures of behavioural inhibition and worry. We are developing measures of depression in the medically ill, and in Asian groups where ‘depression’ is commonly denied or expressed somatically.

Exploring sex differences in depression

In the late 1970s, it was widely held that more women suffered depression than men, suggesting that anatomy was destiny. Even so, there were exceptions, with sex differences absent in bipolar disorder, and in university and college student populations. In 1977, Kay Wilhelm initiated a study, enrolling a group of college students then without sex differences in depressive experience. It was hoped that by assessing on several baseline measures and then following them up at regular intervals, determinants could be identified when sex differences emerged.

The cohort has been studied every 5 years (four occasions). Sex differences for depression have been either non-existent or trivial, suggesting that, in homogeneous groups, homogeneity of social factors may restrict sex-dimorphic factors that operate in the wider community. Similar negative findings have emerged in studies involving adults selected on the basis of similar occupations or belonging to socially homogeneous cultural groups (e.g. a large sample of Jewish adults in London). Women in our cohort had higher neuroticism scores and higher anxiety disorder rates than the men, suggesting ‘primers’ for possible differentiation following exposure to ‘depressogenic stressors’. In addition, the design has allowed artefactual influences on sex differences to be identified (e.g. men tend to forget or minimise previous depressive episodes, while women ‘remember’ or exaggerate their severity).

Most of the cohort members are now in their mid-40s. While the study continues to allow quantitative issues to be addressed in a non-clinical group (e.g. the capacity for anxiety disorders to predispose to depression; and the influence of anxiety and personality style on non-melancholic depression), it also provides a unique opportunity for qualitative research, as illustrated by Vaillant's cohort of former Harvard students.

Primary care research

Ian Hickie has studied psychological disorders, including depression, in primary care settings. We now develop and test algorithms to determine and subtype depressive disorders in a general practice setting and test the utility of the Personality, Anxiety and Life Events (PAL) paradigm as a model in this context.

Fatigue research

Following clinical and laboratory work on patients with chronic fatigue syndrome (and pursuing links with, or independence from the depressive disorders), Ian Hickie has begun to examine a ‘post-infective’ cohort, post-malignancy fatigue syndromes, and genetic influences on fatigue and depressive syndromes.

Electroconvulsive therapy and transcranial magnetic stimulation research

Philip Mitchell has undertaken several ECT studies, while he and Colleen Loo have pursued the effectiveness of transcranial magnetic stimulation (TMS), in collaboration with Perminder Sachdev and Simon Gandevia, a neurophysiologist. This procedure has excited imagination, given the antipathy of many to ECT, as TMS involves neither an anaesthetic nor a convulsion. Their group undertook a sham-controlled study, reported in the American Journal of Psychiatry, which did not find TMS effective, and is now examining bilateral TMS in depression, and its potential role for other conditions such as obsessive–convulsive disorder and schizophrenia. These studies are important in addressing key questions provoked by any new treatment (e.g. is it effective, is it safe?).

Neurocognitive studies

Marie-Paule Austin has demonstrated patterns of cognitive impairment, largely restricted to those with melancholic depression, at least when ‘melancholia’ is defined by the CORE system and not by DSM criteria.

Endocrine and molecular biological studies

Philip Mitchell has been interested in biological underpinnings of unipolar and bipolar disorder, reporting a blunted prolactin response to fenfluramine in melancholic patients and a correlation between serum cortisol levels and psychomotor disturbance. Molecular genetic studies into bipolar disorder have been undertaken in collaboration with Peter Schofield, a molecular biologist at the Garvan, and Jennifer Donald, a genetic statistician at Macquarie University, replicating previous reports of linkage to chromosome 21q and finding strong linkage to a novel susceptibility locus on chromosome 4q35. Molecular biological studies of postmortem brain tissue from bipolar patients have been initiated by Kathryn Lovric.

Neuroimaging studies

Ian Hickie, in association with Philip Ward, took responsibility for a series of structural imaging studies in collaboration with assessment of vascular risk factors, thus advancing our model of late-onset depression. Identifying observable psychomotor disturbance as a marker of melancholia encouraged a study assessing differences between melancholic and non-melancholic depression across a range of measures and investigatory techniques. This study, central to our initial Program Grant application, allows understanding of the neurobiology of melancholia. Structural (i.e. magnetic resonance imaging or MRI) and functional (i.e. single photon emission computed tomography or SPECT) brain scans were undertaken with images coregistered to identify regions of interest, together with neuroimmunology, neuroendocrine and neurocognitive testing; the imaging objective being to identify perturbed circuits and structural regions in subtypes. Assessment across multiple parameters has the potential to identify linked and independent domains across depressive types; we anticipate important findings from this ‘Convergence of Evidence’ (COE) work.

Personality disorders

A consistent focus has been to model (and then measure) variable constructs. We seek to extend that focus by studying personality disorders (PDs). If we are to model depressive spectrum disorders, we need to measure the temperament and personality components. A large database, accrued with the assistance of clinical colleagues, allows us to determine: whether PDs are best modelled dimensionally or categorically, whether they should be deconstructed into separate components of personality style and of disordered function, what the basic PD ‘building blocks’ are, and how these are best assembled to create meaningful dimensions.

Research funding

The MDU has received regular funding from the National Health and Medical Research Council (NHMRC) since 1987. In 1995, we were awarded the first NHMRC Psychiatric Program Grant in Australia and this was re-awarded in 1999. In 1997, we received an inaugural grant from the New South Wales Department of Health to support research and educational activities. Benefactors have provided additional support. We heartily thank them all.

Current directions

Several studies have been planned for the next 5 years. The main strand will focus on refining the spectrum disorder model across the non-melancholic class, and pursuing aetiological determinants and identification of differential treatment response across subclasses. The neuroimaging node will undertake studies to identify regions of functional disturbance in psychotic and melancholic depression. The ECT and TMS studies will continue. The teachers' cohort study will undertake a 25-year follow-up. We will pursue the development of depression case finding measures in several groups, including the medically ill, the elderly, and in general practice and postnatal samples. Several measures assessing depression severity in subtypes, and as rated by nurses and by family members, are under development, and daily rating scales refined to examine speed of action of antidepressant drugs more precisely. We will model the personality disorders, particularly distinguishing personality style and disordered functioning components. Molecular genetic and postmortem studies will pursue neurobiological determinants of bipolar disorder.

Conclusions

Some of the strengths (and pleasures) of the MDU have been noted earlier, while its success as a research centre is best judged by others. There can be no definitive model for a psychiatric research facility, for design components have as many axes as does DSM-IV, being influenced by the researchers (as researchers and as people), the environment (hospital-based vs university-affiliated vs stand alone), financing (direct vs indirect; amount), ‘support’ from those who control resources (whether at hospital, area, State or Commonwealth levels), peers (both local and international, the ‘gate-keepers’ of science), the structural and functional model, enthusiasm and creativity, and the capacity to set and work to short-term and long-term objectives.

Key strengths of the MDU can be traced to the staff. The imagination and interests of our consultants have been noted. We have been blessed by equal strengths, but differing skills, of those who sit in its ‘engine room’, supporting the research psychiatrists, and ensuring quality control, as well as by a number of managers. Many research assistants have brought energy, dedication and their own ideas to the enterprise, well illustrated currently by Gemma Gladstone and Kay Roy. Yvonne Foy copes gracefully and meticulously with the multiple secretarial demands. Christine Taylor, our data manager, has tenacity and an eye for detail that adds to the primary quality control mechanisms, the deceptively important ‘soft belly’ that can compromise research organsiations and their component studies.

Two people provide a third quality control level. A Chief Investigator on our new Program grant, our Scientific Officer, Dusan Hadzi-Pavlovic takes responsibility for statistical analyses, and consistently delivers a high-quality product (and some unequivocal ‘break through’ results). He studies key papers, imposes rigorous standards, and gives detailed advice. As our administrator, Kerrie Eyers, has attended to a plethora of tasks, editing papers technically, but also detecting errors or inconsistencies of logic or argument, and usually suggesting a superior way of addressing such issues. Both go beyond their job description in also acting as ‘gate keepers’ of the MDU's objectives and product by putting considerable effort into addressing systemic and corporate issues.

This long-serving infra-structural support has been value-added ‘glue’, with dedication to the ‘cause’ by so many being the most important factor in our development. I view the MDU as a neat example of a psychiatric research unit and clinical facility. It is a common non-psychiatric model in Australia's teaching hospitals, but rare and insufficiently appreciated for its potential to address psychiatric disorders by similarly establishing ‘a critical mass’ of clinician researchers.

Acknowledgements

My thanks to MDU colleagues for assistance in preparing this paper.

For copies of our new subtyping manual, a list of the MDU's publications, an annotated bibliography, the new CORE training video, or other educational and research material, please contact Kerrie Eyers on [phone] or by Email: [e-mail]. More information is available on our website: